Experimental study of the operation of a two-stroke machine: with 3 tables and 47 drawings

V. L. Maleev.

Experimental study of the operation of a two-stroke machine: with 3 tables and 47 drawings

V. L. Maleev.

Submitted 1912 | SovietRxiv: ru-191201.07258 | Translated from Russian

Pre-reform orthography. The original is printed in the Russian spelling used before the 1918 reform (ѣ, і, ъ).

Full Text

V. L. Maleev.

EXPERIMENTAL STUDY

OF THE OPERATION

OF A TWO-STROKE MACHINE.

With 47 drawings and 3 tables.

Pp. 1–200.

PREFACE.

Feci, quod potui, faciant
meliora potentes.

Among the persons who, being interested in the questions touched upon in the present work, will read it, there may perhaps be some who will reproach the author with the fact that, as the result of extensive investigations and numerous tables, he gives comparatively few general conclusions and new indications concerning the improvement of existing two-stroke engines.

To this the author permits himself to observe that, although in undertaking the indicated work he did indeed expect to obtain more new data for practice, he nevertheless hopes that this work, even in the form in which it is now printed, is of sufficient interest to the appropriate specialists, or to those studying this question, and that the labor and considerable time required for its production*) have not been spent in vain.

Of course, for the majority of technicians it is especially interesting to obtain indications that can immediately be used in practice; but it seems that the description of new methods in carrying out experiments and in processing them—and the exposition of such methods occupies a considerable part of the proposed book—also presents interest, since it advances the important question of the experimental study of engines and, in particular, facilitates the performance of similar experiments by other persons who, having at their disposal a more suitable engine, will continue them. Then, the experimental confirmation of already known theoretical propositions also has substantial significance.

) The decision to undertake the present topic arose for the author as early as 1902, during the testing of a 6-horsepower Avance engine, described by the author in the Bulletin of the Polytechnic Society, 1902, pp. 269–276; then it was finally strengthened during the study of N. Güldner’s book, published in 1903, Entwerfen und Berechnen der Verbrennungsmotoren*. However, owing to various circumstances, it proved possible to begin the preparatory work only in December 1906, when the engine intended for the experiments was delivered to the Laboratory of Heat Engines of the Tomsk Technological Institute. Its installation, and that of the auxiliary machines and instruments, as well as the manufacture of various appliances, took more than a year. In the spring of 1908 it was finally possible to begin the preliminary investigations and the adjustment of the experiments themselves, and from July 3 of the same year, the experiments themselves. Their execution and processing took another three years.

Preface.

As regards its scientific and laboratory interest, the description of experiments in the course of which it was necessary to struggle with various difficulties and to devise special appliances and methods, and then, lacking the possibility—owing to circumstances that became clear, in part even after the completion of the experiments—of using the material obtained in its raw form, as is done in the majority of cases, it was necessary first to subject it to verification and partial reworking; the description of such experiments, even if they did not yield especially significant new general conclusions, is in a certain respect of greater interest than the account of experiments which, although they gave entirely new results, were carried out smoothly, by generally known methods.

Thus the author hopes that the present book will prove useful not only to persons interested in internal-combustion engines, or even in particular in two-stroke engines, but also to anyone interested in and engaged with questions of experimental investigation of power and, in general, heat installations.

In conclusion, the author considers it a pleasant duty to express gratitude to his chief assistant, mechanical engineer V. G. Karpenko, who took an active part in carrying out the experiments, and also to the students of the T. T. I.: N. N. Dosser, I. Z. Sapozhnikov, and P. V. Cherepanov, who helped in preparing various reports, and to the laboratory mechanic A. Ya. Tkachenko, who carried out with great care all the apparatus needed for the experiments.

Tomsk. November 1911.

V. L. Mal’ev.

	Page
4. Machine	11
Principal dimensions of the machine. Diagram of its operation	11
Actual execution of the scavenging	12
Adjustment. Ignition	13
Change of compression. Change of the volume of the pump’s harmful space	14
Non-return valve. Suction of air. Tightness	15
Determination of the compression chamber of the machine and of the pump	16
5. The whole installation	17
Air clocks	17
Air vessel and air-pressure regulator	18
Air blower	20
6. Measuring instruments and appliances	20
Brake	20
Revolution counters. Tachometers	23
Determination of the number of flashes	24
Indicators	25
Indicator drives	26
Measurement of pressures. Measurement of temperatures	29
Measurement of the consumption of kerosene and of the circulating water	31
Measurement of the consumption of cooling water	31
Determination of the consumption of oil for lubricating the cylinder	32
7. Taking samples of exhaust gases	32
Sampling of gases from the silencer and at the cylinder itself	32
Differentiator	33
Drive for the motion of the differentiator slide valves	36
Aspirator	38

		Page
8.	Absorption of gases by liquids	40
	Experiments on absorption by water	40
	Experiments on absorption by aqueous solutions of chloride salts	42
	Choice of a solution of common salt	44
	Shortcomings of the saline solution	45
9.	Gas sampling	45
	Experiments with sampling in various places	45
	The phenomenon of exhaust in two-stroke machines	47
10.	Experiments with Junkers’ calorimeter	48
	Inconstancy of the temperature of the outgoing water	50
	Best conditions for operation of the instrument	54
	Accuracy of the calorimeter readings	56

Conduct of the Main Experiments

		Page
11.	General instructions	57
	Variable factors	57
	Beginning and duration of individual experiments	58
	Load	58
	Setting the misfires	59
	Operation with a blow lamp	60
12.	Measurement of various quantities	60
	Taking indicator diagrams	60
	Determination of the scales of diagrams	62
	Barometric pressure	65
	Number of revolutions. Number of misfires	65
	Measurement of temperatures	65
	Spark advance	66
	Kerosene consumption	67
	Air supply	67
	Water consumption	68
	Taking gas samples	68
13.	General survey of the conduct of the experiments	68
	Number of participants and distribution of duties	68
	Specimens of tables filled in during the experiment	70
14.	Gas analyses	72
	Analysis with Hempel apparatus	72
	Analysis by combustion	73
	Arrangement of instruments for analysis by combustion	73
	Measurement of the gas supply	75
	Course of the analysis by combustion	77
	Specimen record for this analysis	81

Auxiliary Experiments

		Page
15.	Investigation of kerosene	82
	Heating value of kerosene	82
	Composition of kerosene	83
	Flash temperature	83
16.	Operation of the flywheel spokes	83
	Purpose of its determination	83
	Data in the literature	84
	Experimental determination	85
	Accuracy of the results obtained	87
17.	Air leakage through the blow lamp	88
	Detection of the leakage	88
	Experimental determination of the correction	88
	Accuracy of the correction	89

General instructions — 99
Accuracy and finding mean errors — 99
Criticism and correction of records — 100
Calculation of works — 104
Effective work — 104
Mean error of its determination — 104
Indicator work — 105
Mean error of it — 105
Indicator work of the air pump — 107
Mean error of it — 107
Work of the resistance — 108
Mean error of the work of piston friction — 109
Mechanical efficiency — 109
Mean error of it — 110
Heat balance — 111
Total available heat — 111
Mean error of it — 111
Heat equal to the indicated work — 111
Mean error of it — 111
Heat carried away by the cooling water — 112
Mean error of it — 112
Heat equal to the work of piston friction — 112
Mean error of it — 112
Heat carried away by the combustion products — 112
Choice of the value of the heat capacity — 113
Composition of the expression for the heat carried away by the combustion products — 115
Mean error of this heat — 118
Heat lost owing to incomplete combustion — 121
Mean error of it — 122
Heat lost to radiation — 123
Mean error of it — 124
Heat corresponding to the living force of the gases — 125
Mean error of it — 126
Discrepancy — 127
Correction of the heat balance — 128
Finding the burned kerosene from $CO_2$ — 129
Mean error of it — 130
Finding the burned kerosene from $O_2$ — 130
Mean error of it — 131
Finding the kerosene consumption from the heat balance — 132
Degree of compression — 134
Compression in the working cylinder — 134
Compression in the pump — 136
Degree of accuracy of the compression values — 138
Air and exhaust velocities — 138
Expression for their calculation — 140
Air-suction velocity — 141
Mean error of it — 145
Air-inlet velocity — 145
Mean error of it — 148

Chapter IV.

Results of the Experiments.

	Page
26. Operation of the air pump	161
Change in volume $V_3$	161
Influence of $V_3$ on $\lambda_0$ and $\lambda$	166
Influence of $V_3$ on $p_i'$ and on $\xi$	167
Influence of $V_3$ on $L_0/N_i$	167
Influence of the number of revolutions $n$	167
Influence of the gaskets $r$ (window opening)	168
Influence of the check valve	168
Experiments with air blowing	169
27. Charging of the working cylinder	170
Influence of volume $V_3$	173
Influence of pressure $p_c'$	174
Duration of scavenging	175
Relation between scavenging and exhaust	175
Influence of the number of revolutions $n$	175
Influence of $n_2$	176
Influence of gaskets $r$	176
Influence of compression $\varepsilon$	176
Operation with air blowing	177
28. Compression and combustion	178
Influence of compression $\varepsilon$	178
Influence of water injection	179
Ignition	180
Fuel nozzle	182
Influence of the number of revolutions $n$	183
Influence of compression $\varepsilon$	183
Relation between the pressures $p_c$, $p_3$, and $p_i$	183
Temperature of the exhaust gases	184
Content of $CO_2$ and $CO$	185
Data of analysis of soot	187
Water injection	187
Use of heat	188
Influence of air blowing	188
Exhaust	188
Pressure after exhaust	193
Calculation of exhaust velocity	195
Duration of exhaust	195
Influence of shadow phenomena	196
30. Additional remarks on the operation of the machine	196
Relative magnitude of the charging work	196
Mechanical coefficient of efficiency	197
Operation of rotation of the flywheels	197
Piston friction	197
Kerosene consumption	197
Economic coefficient of efficiency	197
Loss from incomplete combustion, from radiation, and kinetic energy of the exhaust gases	198
31. General conclusions	198
Data on the operation of the tested machine and its improvement	198
Improvement of the construction and operation of machines of this type	199
Clarification of the principal questions of two-stroke operation	199
Directions for carrying out tests of internal-combustion machines	200

Noticed Misprints

Page	Line	Printed	Should be
16	22 from bottom	$\sqrt{(0.02\cdot 2)^2+(0.04)^2}=\pm0.06\%.$	$0.02\cdot 2+0.04=\pm0.08\%.$
17	17 from top	$19.167$	$19.467$
38	table 1, 1st line from bottom	$511+22$	$511+63$
66	4 from bottom	$^{8})$	$^{31})$
83	16 from top	$18.06$	$18.016$
86	drawing 36, 1st line from top	$lo$	$a. s.$
86	drawing 36, 1st line from bottom	$400$	$400$ rev./min.
89	drawing 37, 1st line from top	$mm.$	$mm.^3$
89	drawing 37, 1st line from bottom	$mm.$	$mm.$ water column
92	7 from top	heat-containing	thermal
92	10 from top	for	at
105	18 from bottom	$F_p:(Hn-i)$	$F_p:H(n-i)$
110	14 from top	However, we preferred	However, for comparing the experiments with one another
114	table 16, 6th line from top	$0.275\quad 0.286\quad 0.277$	$0.275\quad 0.276\quad 0.277$
114	table 16, 9th line from top	$0.0000254$	$0.000025$
114	table 16, 10th line from top	$0.0000214$	$0.0000414$
114	table 16, 11th line from top	$0,\ 38$	$0.38$
114	table 16, 14th line from top	$+0.0001347$	$+0.000134\,t-$
114	table 16, 14th line from top	$-0.732512$	$-0.07325\,t^2$
114	table 16, 15th line from top	$0\ 478$	$0.478$
114	table 16, 17th line from top	$-0.00001687$	$-0.0000168\,t+$
114	table 16, 17th line from top	$+0.\!4472$	$+0.0744\,t^2$
114	table 16, 18th line from top	$0.487\mid 0.484\mid 0.484\mid$	$0.487\mid 0.485\mid 0.484$
115	10 from bottom	$Q^{\prime\prime}_{\mathrm{p}}$	$Q_{\mathrm{pr}}$
115	10 from bottom	$G^{\prime\prime}_{\mathrm{p}}$	$G_{\mathrm{pr}}$
118	3 from bottom	$L,$	$L_o.$
119	6 from bottom	$G^{\prime\prime}$	$G_{\mathrm{v}}^{\prime\prime}$
119	5 from bottom	$\pm1.37\%.$	$\mp1.37\%.$
122	11 from top	$Q^{\mathrm{ng}}$	$Q_{\mathrm{ng}}$
122	5 from bottom	$\mathrm{H_2O}:V_{\mathrm{l}})$	$(\mathrm{H_2O}:V_{\mathrm{l}})$
132	16 from bottom	, we	$H_n$, we
137	table 21 (experiment 50)	$,\ 3.60\mid 1.146\mid 1.105$	$4.86\mid 3.82\mid,\mid,$
141	8 from bottom	quantities	quantities
143	drawing 45, 1st line (left)		$\mathrm{cm}^2$
144	table 22 (experiments) 50–53, 41		$43.5$
144	column 2, 41		$43.5$
144	column 2, 41		$43.5$
144	column 2, 41		$43.5$

Noticed Misprints

Page	Line	Printed	Should be
145	3 below	lichny	lichiny
147	8 above	kanal'	kanal
151	16 below	davlenie.	davlenie
151	1 below	$(15,23\ 0,125)^2$	$(15,23.0,125)^2$
152	3 above	smeshchen-	smeshchen-
154	2 above	100.	100,
162	9 above	imevshikh	imevshim
164	15 above	opredelit'	opredelit'
166	2 below	$\lambda^\circ$	$\lambda$
166	1 below	$\lambda$	$\lambda^\circ$
169	14 above	$N_1'$	$N_i'$
169	20 above	$p_1'$	$p_i'$
170	11 above	poluchaem $L/k_1$.	poluchaem $L/k_4$.
170	13 below	$Q/N_1^{\,*}$	$Q/N_i$
171	(6th column)	ch.p. $N_1$	kgp. $N_i$
171	(14th column)	$N_1$	$N_i$
173	1 below	3,377	3377
182	7 above	vody	vod'
192	6 above	uglov, $0^\circ$,	uglov', $0^\circ$,

Table	Experiment	Column	Printed	Should be
table 26	4	5	2,714	2,520
table 26	51--53	4	15,5	20/5
table 26	51--53	5	2,761	2,520
table 26	64 and 65	6	88,09	46,94
table 26	63	29	98,5	98,5*
table 26	68	29	111,3*	111,3
table 27	16	28	7,0	77,0
table 27	49	21	141	113
table 27	75	3	4,5	4,9
table 28	(heading)	21	$q_{\mathrm{nch}}$	$q_{\mathrm{ng}}$
table 28	(heading)	23	$q_{\mathrm{n}}$	$q_{\mathrm{zh}}$

Abbreviations and Symbols Used

General abbreviations:

abs.—absolute.
mm., mm.^2, mm.^3—millimeters, square mm., cubic mm.
cm. (cm.^2, cm.^3)—centimeters (square, cubic).
m. (m.^2, m.^3)—meters (square, cubic).
ltr.—liters.
g.—grams.
mg.—milligrams.
kg.—kilograms.
sec., ″—seconds.
min., ′—minutes.
h.—hours.
h.p.—horsepower (indicated).
d.h.p.—actual (brake or useful) horsepower.
—; ÷; /—division signs.

Heat:

$t^\circ$—temperature in degrees Celsius.
$T^\circ = t^\circ + 273^\circ$—temperature in absolute degrees.
$T_n$, $t_n$, $\tau$—normal temperature ($+288^\circ$ abs., $+15^\circ$ C.).
t.u.—thermal units (kg-calories).
$H_n$—calorific value (useful).
$c_p$—heat capacity at constant pressure (1 kg).
$C_p$—heat capacity at constant pressure (1 m.^3).
$Q$—heat in t.u./hour.

Pressure:

atm.—atmosphere = 1 kg/cm.^2 = 737.4 mm Hg.
w.c.—water column in mm.
Hg col.—mercury column in mm.
$p$—pressure in kg/cm.^2.
$p_i$—mean indicated pressure in the engine, kg/cm.^2.
$B$—barometric pressure in mm Hg.
$B_n$—normal barometric pressure = 737.4 mm.

Machine:

$D$—cylinder diameter (in mm.).
$H$—piston stroke (in mm., in m.).
$F$—piston area in cm.^2.
$V_x$—volume described by the piston (in ltr.).
$V_c$—volume of the compression chamber.
$V_z$—volume of the rear (harmful) space of the pump.
$\varepsilon$—degree of compression in the machine.
$\varepsilon_{\mathrm{II}}$—degree of compression in the air pump.
$\lambda$—pump delivery.
$N_i$ ($N'$)—indicated work of the machine.
$N_e$—useful work in d.h.p.
$R$—work of resistances in h.p.

Other quantities:

$l$—length (of the indicator diagram in mm.).
$m$—scale of the indicator diagram in mm./atm.
$V$—volume in m.^3.
$V_l$—volume in ltr.
$v$—specific volume in m.^3/kg.
$f$—area in cm.^2.
$w$—velocity in m./sec. (of gases).
$\eta$—efficiency in %.
$m$—magnitude of the mean error in named units.
$\nu$—magnitude of the mean error in %.

Literature abbreviations:

Ann. d. Phys.—Annalen der Physik.
Berl. Ber.—Berichte der Berliner Akademie.
Elektr. Kraftbetr. Bahn.—Elektrischer Kraftbetrieb und Bahnen, Wien.
Gasmot.—Die Gasmotorentechnik.
Hütte.—Hütte, des Ingenieurs Taschenbuch.
J. Gasbel.—Journal für Gasbeleuchtung und Wasserversorgung.
Schweiz. Bauz.—Schweizerische Bauzeitung.
Verh. V. Gew.—Verhandlungen des Vereines zur Beförderung des Gewerbefleisses.
Z. phys. Chem.—Zeitschrift für physikalische Chemie.
Z. V. d. I.—Zeitschrift des Vereines deutscher Ingenieure.

INTRODUCTION

1. The Significance of Two-Stroke Machines

All modern internal-combustion engines operate either according to the so-called four-stroke method or according to the two-stroke method; the other methods, such as the six-stroke, the four-stroke with unequal strokes, etc., have passed into the realm of history.

Both of the indicated methods continue to exist and to develop in parallel. True, if one judges by the number of factories engaged in the construction of these and other machines, and by the number of machines produced annually, the preponderance will be on the side of four-stroke machines. However, as has been pointed out more than once, this is explained not by the real advantages of the four-stroke method, but rather by the personal sympathies of designers and manufacturers, i.e., by accidental circumstances, and partly, perhaps, also by the well-known conservatism of customers, who prefer the old, familiar, more widespread thing to the less tested one.

Be that as it may, despite the considerable perfection—both in the sense of economy and in the sense of reliability of operation—rapidly attained by modern large four-stroke machines, two-stroke machines still have not disappeared and, in their turn, are developing and being perfected, contrary to the sharp but insufficiently justified verdict of Prof. A. Riedler, who in his report at the annual meeting of the Society of German Engineers in 1904 expressed the conviction that the large gas engine would again return to the four-stroke method¹). The words of a person so authoritative in machine building could, of course, not remain without consequences and, probably, more than one customer who had been leaning in favor of a two-stroke machine, after these words changed his mind and gave preference to a four-stroke one. And nevertheless two-stroke machines continue to be built!..

What, then, are their chief advantages, or at least their important merits?

First of all, other conditions being equal, for the same power the working cylinder is obtained, at least theoretically, with twice as small a volume as in a four-stroke machine; in other words, the power coefficient of two-stroke machines in comparison with four-stroke—

¹) Z. V. d. I. 1905, pp. 273, 321.

Matzén.—Study of a Two-Stroke Machine.

Introduction.

§ 1.

with $K = 2$. True, a two-stroke engine has two additional pump cylinders—for air and gas; however, owing to the very low and moderate maximum pressures in these pumps, their cylinders are comparatively very light and, consequently, not expensive.

Furthermore, because of the rapid admission of the fresh mixture, its temperature, in contrast to four-stroke engines, does not have time to rise appreciably, as a result of which the so-called charging coefficient of the working cylinder, $\lambda$, is greater.

A further increase in $\lambda$ occurs because the fresh charge enters with a certain excess pressure, 1.05 to 1.10 atm., instead of the suction rarefaction of four-stroke engines, which gives only 0.95 atm., and often still less. The excess of scavenging air in engines with an independent air pump acts in the same direction.

All these three circumstances increase the power of the engine; that is, in essence, they reduce to the first-mentioned advantage.

But there is also one further essential advantage of two-stroke engines: greater uniformity of rotation under otherwise equal conditions, or a lighter flywheel for the same degree of non-uniformity, as well as better utilization of the crank-and-connecting-rod mechanism.

These latter advantages are retained also by high-speed two-stroke engines operating without a special air pump and predominantly on liquid fuel. Their power coefficient $K$ is undoubtedly considerably less than 2, but, owing to the absence of additional pumps, they come out considerably cheaper than four-stroke engines even when $K$ is only slightly greater than 1.

Finally, in two-stroke engines the discharge of the incandescent exhaust gases takes place through ports opened and closed by the edge of the working piston. Although many authors point out that the use of such ports instead of valves is forced by the insufficient time available for exhaust, nevertheless anyone familiar with the difficulties encountered in maintaining the exhaust valves of large engines—which are their weakest point—will unconditionally recognize their absence in two-stroke engines as a great advantage.

It is not without reason that similar ports have recently begun to be used for the exhaust of steam in steam engines!..

Of course, two-stroke engines also have their shortcomings: the possibility that a certain amount of fresh mixture may leak away during the period of simultaneous opening of the intake and exhaust; the necessity of keeping to a comparatively moderate number of revolutions in order to avoid strong vortex formation, which entails the indicated leakage; a rather considerable—

§ 1. Advantages and disadvantages of two-stroke machines.

the work of charging and the resulting low mechanical efficiency of the machine; a certain loss of area of the indicator diagram owing to the necessarily early advance of the exhaust; and, finally, a rather strong noise in the muffler and its shaking owing to the very considerable velocities of the escaping products of combustion.

However, as practice shows, all these shortcomings are either quantitatively insignificant, such as the leakage of the fresh charge and the reduction in the area of the diagram, or they can be reduced by rational design, such as the work of charging or, still more, the noise and shaking from the exhaust.

Thus the advantages, evidently, have the upper hand over the disadvantages and fully justify efforts toward the further improvement of these machines. One of the means toward this is to clarify, by properly arranged experiments, the obscure questions concerning the phenomena occurring during the operation of these machines.

The proposed experimental investigation pursues precisely this aim.

CHAPTER I.

2. Questions to be resolved.—Already at the first acquaintance with the scheme of operation of two-stroke engines, one is struck by the complexity of the phenomena taking place during charging, when the spent gases are removed—so to speak, displaced—by air and fresh mixture entering the cylinder under a certain pressure. One need only recall such phenomena as the ease with which vortices are formed during the motion of gases and their diffusion, and it at once becomes clear that, for the satisfactory performance of this displacement, which is commonly called scavenging, a whole series of conditions must be observed, partly in the design of the engine and partly subsequently, during its operation.

Two-stroke engines, similar to the modern types of small engines, appeared on the market around 1890; larger engines, of Oechelhaeuser, in 1896, and Körting’s in 1898. But the first to dwell in detail on the obscure phenomena, requiring clarification, that occur during the charging of two-stroke engines was only Güldner, in his book—now almost classical—which appeared in 1903: Entwerfen und Berechnen der Verbrennungsmotoren. He pointed out that the following principal questions confront the designer:

What pressure of the scavenging air should be used—low or high?
What duration of scavenging should be adopted—short or long?
What channels for admitting air and gases should be made—narrow or wide, axial or radial?
Where should the scavenging air be admitted—at the bottom of the piston or at the head?
Is an intermediate vessel needed for the scavenging air, and of what capacity—large or small?
How great should the surplus of scavenging air be?

In posing these questions, Güldner at the same time notes that there are as yet no answers to them in the literature. Factories and designers, resolving them one way or another each for themselves, evidently do not wish to make public the corresponding experiments; engineering laboratories at higher educational institutions, meanwhile, do not wish to take up these questions.

More than eight years have already passed since the appearance of the first edition of Güldner’s book, yet the situation has not changed: as before,

there is no information in the literature even about attempts to resolve the indicated questions by experimental means. Strictly speaking, there is nothing especially surprising in this: as regards factories, although many of them continue to work intensively on the question of two-stroke machines—as is evidenced, for example, by the appearance four years ago of machines from the Koerting and Wittkowitz works with considerable changes and improvements to the Koerting type, and then, almost simultaneously, the recent appearance at several works of Diesel two-stroke machines—nevertheless, in view of the particularly sharpened competition, it has become frankly even dangerous for factories to publish the results of their investigations. As regards engineering laboratories, such experiments could not be carried out in them because of the absence of suitable machines, since gas two-stroke machines of the Koerting or Oechelhaeuser type are not built smaller than 200, or even 500 h.p.; that is, they are too large and expensive for a laboratory, while the laboratories evidently did not wish to acquire small-power machines operating without a special air pump, considering them of little interest.

In view of this state of affairs, but not sharing the latter considerations, we decided to carry out the corresponding experiments, even if on a small machine, and even if of imperfect design—without an independent air pump. With the appropriate additional equipment, such a machine was to make it possible to elucidate at least some of the questions indicated above, namely questions 1, 2, and partly 5 and 6; questions 3 and 4, with only one machine and without major alterations, cannot be resolved.

In setting up our experiments, we expected, at least partly and approximately, to elucidate such important questions, which had so long remained without attention, and also to learn incidentally something more detailed about the operation of a machine without a special air pump—for example, the degree of volumetric filling of the pump, the expenditure of work on charging, the conditions of combustion, and so on.

Moreover, in the very setting up and processing of the experiments, new questions were bound to arise, and the answers to them were to compensate for the impossibility of resolving the principal questions to the desired extent.

3. Literature on the present question.—Despite the obvious absence in the literature of information about experiments on the questions outlined by us above, we considered it necessary, before proceeding to our own experiments, carefully to review the periodical and corresponding book literature, with the aim of finding at least a theoretical elucidation of the questions of interest to us, or some factual data, even in the form of descriptions of acceptance tests, which might provide material for preliminary considerations and hypotheses.

Among the persons who, in one way or another, carried out experiments with two-stroke machines and published them in greater or lesser detail, we found only A. Wagener and E. Meyer. Both the one and the other worked with Oechelhaeuser machines.

Vagener, who was the first to publish² some experimental material concerning the Echelheiser machine, dating from 1896, carried out his experiments at that time, however, only with the aim of setting up operation with a fuel then new—blast-furnace gas. The article under consideration contains no general indications of any kind, apart from the fact that the velocity of discharge of the combustion products may reach 365 m/sec. Further on, the author mentions in it also experiments with a second machine of the same system, but gives no data concerning them. The experiments were of a commissioning character and were of very short duration.

Subsequently³ Vagener carried out experiments already of a theoretical character, precisely for elucidating some of the questions we had outlined. However, he performed only a small number of experiments, interrupting them owing to certain circumstances beyond his control, and the center of gravity of his work lies in the theoretical investigation of the phenomena of scavenging and charging. The formulae derived by him, nevertheless based on his experiments, give the dependence between the area of the port openings, the mean scavenging pressure, and the dimensions of the intermediate vessel. He explains this dependence also by a graphical method, very clear and convenient for practice, and indicates the extension of his method, originally derived for machines of only one type—Echelheiser’s, i.e. with ports opened and closed by the edge of the piston, both for exhaust and for intake—to other machines with valves for admitting the fresh charge, of the Koerting type, and also, in general, to valve timing.

While giving many valuable indications for the designer, Vagener’s work nevertheless does not at all answer the questions concerning the most advantageous operating conditions—the pressure of the scavenging air, its quantity, velocity, etc.

Meyer’s experiments⁴ pursued chiefly a practical aim—the evaluation of the economy of the machine’s operation. True, on the basis of the indicator diagrams obtained for the working cylinder and the pumps, the author makes several general conclusions concerning the magnitude of the work of charging the machine, but he has no answers, even indirect ones, to the questions posed above, nor can they be extracted from his figures. Meyer also established the magnitude of the thermal efficiency, as well as the resistances of the machine he investigated. In view of the thoroughness of the experimental arrangement, as is also testified by the name of the investigator, known for his precise experiments with internal-combustion machines, and also in view of the general absence of good experimental data, Meyer’s article is very valuable, although, in particular, for setting up our experiments it could give us no indications or guiding threads.

² Z. V. d. I. 1900 S. 1517.
³ Gasmot. 3, 1903 S. 69.
⁴ Z. V. d. I. 1905 S. 324.

§ 3. Literature on the Present Question.

One may also mention the article by G. Davin¹, who confines himself, however, merely to a description of the construction of the Junk machine and of its testing by the American professors Carpenter, Diederichs, and Hitchkop. The numerical material reported by him is insufficiently complete, relates only to a definite type of machine—namely, with compression of the air in the rear cavity, but with an additional pump—and, most importantly, arouses suspicion as to its correctness: according to these data the Junk machine is essentially not inferior to the Diesel machine in respect of fuel utilization, and one can only wonder why it has not become widespread and is in practice entirely unknown.

The remaining works known to the author are predominantly theoretical in character, or even rather descriptive and speculative, and deal chiefly with the question of comparing the operation of two-stroke machines with that of similar four-stroke machines. These articles are based, of course, also on practical data, but data obtained not from direct experiments, but mainly from observations of the operation of machines under normal conditions. In chronological order these articles are as follows.

A note by I. Kerting² concerning attacks on two-stroke machines by F. Alberti³ is chiefly polemical in character. In it the author mentions the very important question of the formation of vortices during the exhaust of the fresh charge and of the influence of these vortices on fuel consumption, but confines himself merely to the rather categorical assertion that, according to his numerous experiments, no formation of vortices occurs in Kerting machines.

Next comes the article mentioned above by A. Riedler[^1] on large gas engines, in which, speaking in favor of four-stroke machines and against two-stroke machines, he illuminates in detail all the shortcomings of the latter and indicates, in part, the direction in which the development of two-stroke machines should proceed from the constructive standpoint. Despite insufficient impartiality, the article can give a number of valuable indications to the designer of two-stroke machines, but it provides nothing for clarifying the questions that interest us.

The article by P. Schwem⁴ analyzes in considerable detail the operating conditions of two-stroke machines and compares them with the operation of four-stroke machines. The author is somewhat biased in favor of two-stroke machines, being evidently well acquainted with them, but he does not at all touch upon the questions that interest us and gives no new material for their resolution.

Chapter I. § 3.

It gives nothing new, as does also the small polemic between Shwem and Tuckerman⁵.

Then follow two articles by Heydorff⁶, in which the author, confining himself exclusively to machines of the Körting type, presents several considerations for and against these machines, and also gives interesting material concerning their most recent constructive design.

An article by A. Wilmer⁷, interesting by its title, “The Charging and Regulation Phenomena of the Körting Two-Stroke Machine,” is only a description of the modified design of the distributing organs of the air and gas pumps, resembling Rider cylindrical slide valves and placed under the direct action of the governor.

A small polemic that arose over this article between its author and Heydorff⁸ points to the existence of a whole series of questions—such as: the influence of pressure oscillations and the appearance of waves in the gas and air conduits between the pumps and the cylinder of the machine; the influence of the capacity of the intermediate vessel and of the size of the inlet ports—questions whose resolution is possible only by experimental means. An exchange of opinions of a speculative character does not advance these questions toward resolution.

The article by A. Wimpfflinger⁹, devoted to the important question of the amount of excess air necessary for flushing out the products of combustion in two-stroke machines, and appearing to give an expression for the theoretical determination of pump dimensions, in essence gives only a formula for comparing them as a function of the different degree of compression of the air and the composition of the gas. Of interest is the table compiled by the author, according to a special formula, for comparing the magnitude of the excess of air in machines that have been built. This excess turns out to be very close for machines very different in system, power, and fuel. For the four machines cited by Wimpfflinger this excess fluctuates only within the limits of 50 to 70% of the useful volume swept by the piston. Whether this number is accidental, or whether the designers consciously strove for it, having found from their own unpublished experiments that it is the most advantageous value, remains unexplained to this day.

Finally, rather categorical answers to the questions 1, 3, 5, and 6 indicated by us are given by A. Dussmann¹⁰. However, his propositions he ni-

3. Literature on the Present Question

which proves nothing; apparently they are merely a repetition of Güldner’s opinions, which will be discussed in detail below.

Turning to the literature in book form, one may note at once that almost none of the authors of the numerous, and often very bulky, manuals on internal-combustion engines—such as Witz, Mathot, Shettler, Göring, Heder, Donkin, Hutton, Martens—have even concerned themselves with the questions indicated above as being important for the existence of two-stroke engines. Strange as it may seem, even Clerk, who for a long time was himself engaged in the construction and improvement of two-stroke engines, does not at all touch upon these questions. As was already stated above, the first author to give them his attention was Güldner. To do him justice, he not only indicates the most important of these questions, but also tries to give answers to them, relying partly on his own experiments and partly on theoretical considerations. Of the articles analyzed by us above, however, he makes use only of Wagener’s works.

Of the propositions set forth by Güldner, the most important are as follows: the pressure of the scavenging air must be low, 0.05 to 0.20 atm.; between the pump and the engine an intermediate vessel of the greatest possible capacity is desirable; in order to avoid the formation of vortices and the leakage of fresh gas, the velocity of entry of the scavenging air and of the fresh mixture must be as small as possible; the pressure in the cylinder at the moment when admission of the scavenging air begins must, as far as possible, be equal to atmospheric pressure.

All these propositions Güldner derives from the requirement of the most perfect possible scavenging, i.e. the replacement of the exhaust gases by fresh air with the least leakage of it, the least possible expenditure of work on the pumps, and a low consumption of fuel.

All these propositions are so obvious to anyone familiar with the operation of two-stroke engines, or even simply with anyone who has thought about it, that one cannot argue against them. Nevertheless, they still need experimental confirmation and, chiefly, the clarification of those numerical limits—of pressure, velocity, and volume—within which the designer must keep if he wants his engine not only to operate, or even to operate satisfactorily, but also to be as economical as the four-stroke engines of competitors.

To this may also be added Wagener’s interesting observation that Güldner’s propositions, for all their naturalness, are by no means applicable to all engines, or at least not fully so. Thus, Wagener believes that the excess pressure in the pumps may be raised approximately to 0.6 atm. without worsening the scavenging or increasing the work of the pumps, since the latter is affected not so much by the magnitude of the absolute pressure as by the mean pressure in the intermediate vessel. He further indicates that, in the case of [[unclear: continuation cut off at bottom of page]]

Chapter I. § 3.

operation with a comparatively cheap fuel, such as blast-furnace gas, and, given the requirement of very uniform running of the engine, Güldner’s stated demand for the greatest possible capacity of the intermediate vessel is even incorrect. The smaller this capacity, the sooner the action of the governor will make itself felt when the load changes.

A fairly large amount of space is devoted to the question of charging two-stroke engines in the recently published book by Dubbel¹¹ on large gas engines. In it the author gives rather detailed instructions on how to design a new engine. However, all his statements are based on the material given by Güldner and partly in the articles mentioned above by Wagner, Handorf, and Wimplinger. Dubbel has no new data.

Thus, from a survey of all the literature we may draw the following conclusions:

the views of different authors on the essence and significance of two-stroke engines are highly contradictory and partly indefinite;
the questions indicated as important for two-stroke engines have only been outlined and, even if in part they have been resolved on the basis of known considerations, they nevertheless require experimental confirmation;
in general, as regards the experimental investigation of two-stroke engines, very little has been done—or at least published—up to the present time;
whatever the supporters of four-stroke engines may say, two-stroke engines have many advantages, and the work spent on improving them will not be wasted.

This state of affairs prompted us to undertake the experiments described below, despite the fact that the engine at our disposal obviously could not yield material directly applicable to modern large gas engines. This same state of affairs subsequently gave us, in our opinion, the right to publish the material obtained, despite the fact that the experiments cannot be regarded as fully completed, and that they have certain shortcomings which did not depend on the way in which they were carried out, as will be seen in the course of their analysis.

CHAPTER II.

Arrangement of the Experiments.

4. The machine.—We shall begin with a description of the machine itself: it is a horizontal two-stroke machine, of the so-called Avant type, built by the L. Nobel works in St. Petersburg, by special order of the author, specifically for experiments in the Heat-Engine Laboratory of the Tomsk Technological Institute. The machine is built primarily to run on kerosene, but it can also run on any other liquid fuel—alcohol, gasoline, crude oil—and, if the cylinder head is replaced by another one with a special injector and a special heating device, it can even operate on petroleum residues. However, for lack of time, and partly also because of other circumstances, the entire series of experiments described below was carried out with kerosene.

The principal dimensions of the machine, found by careful measurement with the aid of instruments equipped with verniers, are as follows: cylinder diameter $D = 222.26$ mm, piston stroke $H = 251.0$ mm. The degree of accuracy with which $D$ was determined is $\pm 0.05$ mm, i.e. $\pm 0.02\%$, and for $H$—only 0.1 mm, or $\pm 0.04\%$.

The normal number of revolutions of the machine, according to the catalogue, is $n = 340$ per minute, at which $n$ the machine freely develops 10 h.p. and permits an overload of up to 20%, but for some reason the works supplied it with a plate bearing the inscription: “norm. no. of rev. 270 min.”; taking this into account, the machine should more correctly be called an 8-horsepower one.

Despite the rather wide distribution of this type of machine, there is no description of it in any of the more substantial manuals; it can be found only in journals. We shall therefore briefly describe the construction and operation of our machine, Figs. 1–5, Table I.

The scheme of its operation is as follows: during the forward stroke, compression of air takes place in the working cylinder; shortly before the dead position, kerosene is injected into the cylinder by the small pump $a$, and is atomized by the injector $b$, the droplets striking the projecting tongue $c$ in the cylinder, with the incandescent ignition ball $d$; the kerosene vapors ignite, an explosion occurs, and an impulse is imparted to the piston, which moves back. On the other side of the piston, in the crank chamber, during the forward stroke a rarefaction occurs, under the influence of which, when the piston, near the outer dead position, opens the port $e$ with its rear edge, air from the outside atmosphere rushes into the crank chamber.

Chapter II. Arrangement of the Experiments. § 4.

cavity. When the piston, having passed the outer dead point, again closes port e, compression of the air begins in the crank chamber.

After passing about 0.8 of the stroke, the piston begins to open the ports: first, with its lower edge, the exhaust port f, and soon thereafter, with its upper edge, port g, through which the working cylinder communicates with the rear cavity, which serves as an air pump.

The process of charging, i.e., of filling the working cylinder with fresh air, takes place in the following manner: the spent gases, which at the end of expansion still have a pressure of up to 3 abs. atm., and sometimes even more, rush, expanding, through port f into the muffler and thence into the atmosphere. The pressure in the cylinder at the moment port g opens falls, or at the very least should fall, almost to atmospheric; then the air, compressed to approximately 1.1 abs. atm., bursting into the working cylinder, drives out the remainder of the spent gases and itself fills the cylinder.

In order that the air, instead of driving the spent gases out, should not rush directly into the muffler, the piston has a projection h, which is intended to deflect the air in such a way that the latter, moving in the upper part of the cylinder, is first directed toward the opposite end of it, and only from there fills the lower part as well.

This theoretically very simple and ingenious method of removing the products of combustion and filling the working cylinder with fresh air was, in the present machine, actually carried out rather unsatisfactorily; that is, at the end of the scavenging process, the working cylinder contains not fresh air, but a mixture of air and combustion products, and moreover one containing a great deal of the latter.

The causes of the unsatisfactory scavenging, which may be indicated in advance, are the following:

the volume of air entering the cylinder, owing to the low volumetric delivery of the pump (the rear cavity)—which is inevitable if the design of this pump is taken into account—is considerably less than the volume which it has to fill;
the entry of air into the working cylinder is inevitably accompanied by the formation of vortices, which mix the air with the products of combustion, so that the displaced products of combustion carry with them part of the already insufficient quantity of air;
part of the air may, even apart from vortex formation, rush from port g directly into port f, so to speak along the shortest path;
finally, it is possible that, at the moment of exhaust, the products of combustion still contain unburned particles, carbon monoxide, methane, acetylene, or, finally, even simply kerosene vapors, which, upon encountering fresh air, burn out, reducing the oxygen content in the new charge.

§ 4. The Machine. Regulation. Ignition.

Clarifying the influence of these 4 possible causes was also included among the tasks before us, alongside the attempt to find answers to the fundamental questions indicated above, 1, 2, 5, and 6, p. 4.

As for the operation of the machine itself, it must be noted that its regulation is effected by omissions in the following manner: the kerosene pump $a$ is actuated by the impacts of tooth $k$ against the intermediate part $i$, against which the end of its lever bears; tooth $k$ is fastened to the weight $l$ of the governor; this weight receives a reciprocating motion from an eccentric seated on the crankshaft; $l$ slides along the inclined plane $m$ and, when the number of revolutions increases above normal, rises somewhat; tooth $k$ then slips past slot $n$—an omission results. In order that the omissions should not be too frequent and the running of the machine sufficiently uniform, the supply of kerosene may be varied by changing the length of stroke of pump $a$, which is accomplished by turning nut $o$.

The governor described has one very important advantage for a machine intended for experiments: it makes it possible to change, while running, the number of revolutions within wide limits and very quickly. To do this it is only necessary to move, by turning handwheel $p$, the inclined plane $m$, bringing it closer to the crank of the machine in order to increase the number of revolutions and moving it away in order to decrease them. In our machine it is possible in this way to change the number of revolutions from $n = 180$ to $n = 370$; at a smaller number of revolutions pump $a$ ceases to operate, since tooth $k$, striking from below, merely bends back slot $n$, and even the mass of the flywheels proves insufficient; we were afraid to allow a larger number of revolutions, since at $n = 370$ the peripheral speed at the rim of the flywheel already reaches 25 m/sec., and a further increase of it may cause rupture of the flywheels.

As for ignition, as has already been mentioned, it occurs owing to the contact of the combustible mixture with the incandescent walls and, chiefly, with the tongue of the ignition ball $d$, which must have a temperature of red heat, i.e. 600° to 800° C. When starting the machine, this temperature is attained by heating the ball with a special kerosene lamp with a Primus burner system. During operation the heat developed inside the cylinder is sufficient to maintain the indicated temperature of the ball, at least at normal and down to half load. With a further decrease of the load and a corresponding decrease in the total consumption of kerosene, the temperature of the ball begins to fall, so that it becomes necessary to heat it again with the lamp from outside. In our experiments, which were carried out exclusively at loads close to normal, this did not have to be done.

As a certain peculiarity of this machine, it is necessary to point out the injection of water into the working cylinder. This injection is intended to prevent premature flashes and to make it possible to increase the deg—

sion pressure, and with it the economy of operation of the machine. Injection is carried out at the moment of scavenging, when a certain rarefaction occurs in the cylinder as a consequence of the inertia of the gases flying out. In our machine, for the purpose of measuring the quantity of water injected, the latter is taken not directly from the water jacket, but comes from a special small tank, of about 0.25 liter capacity, into which it is poured by hand. The tank is located at a height of about 700 mm above the cylinder and is attached to the pipe serving to discharge the hot water from the jacket. The admission of water into the cylinder is effected through an ordinary oil-cup, with a check valve and with the delivery of the water by drops visible; the delivery can be regulated in the usual way by hand, by screwing in or unscrewing the locking cone; it can thus be varied within the limits from 0 to 5 liters of water per hour.

In addition to the devices indicated for varying the kerosene feed and the number of revolutions, devices which every machine of the given system has and which in our case were only improved in the sense that the required nuts were provided with convenient appliances for turning them, so that the corresponding setting is achieved without the aid of wrenches, the machine described has several more devices allowing the conditions of its operation to be varied, which is so important for experiments.

The machine has, first of all, a device allowing the volume of the compression chamber $V_c$, i.e. the compression ratio $\varepsilon$, to be easily varied. This is achieved by the fact that between the connecting rod and the bearing embracing the crankshaft one may insert a steel shim $r$, Fig. 2, and, by thereby increasing the length $L$ of the connecting rod, decrease the value of $V_c$. There are 4 such shims $r$, with thicknesses of 5, 10, 15, and 20 mm.

Next, attention was paid to the fact that when $L$ is changed, not only $V_c$ changes, but, owing to the displacement of the piston edges, the moments and duration of opening of the ports $e$, $f$, and $g$, as well as the area of opening of these ports, also change. As is immediately evident, with an increase of $L$ the duration of opening and the area of port $g$ for the inlet of compressed air and of port $f$ for the exhaust of the spent gases decrease, while port $e$ for the suction of air, on the contrary, increases. Thus, in changing $\varepsilon$, we involuntarily change such important factors as the resistance to suction and exhaust and, correspondingly, the velocities of the air and of the spent gases. In order to determine the influence of these factors, it is necessary, on the contrary, to vary them at constant $\varepsilon$. In the given machine this was easy to accomplish by placing under the cylinder cover a ring $s$ of the same thickness as the shim $r$; then, by as much as the volume $V_c$ is decreased by the shim $r$, by the same amount it is increased by the ring, and $\varepsilon$ remains unchanged.

Finally, in order to vary the volume $V_z$ of the harmful space of the air pump, namely to increase it, two special cast-iron cap-rings $t_1$ and $t_2$ were cast, each having a capacity somewhat greater than twice the volume $V_z$,

§ 4. Check Valve. Air Intake. Tightness.

described by the piston. To reduce $V_3$, these caps can be opened and closed, respectively, directly by the covers $u_1$ and $u_2$.

Further, mention may also be made of attempts to increase the volumetric delivery of the pump by fitting window $e$ with a check valve. However, the factory found it difficult to install it there, since the frame model would have had to be somewhat altered, and therefore the valve $q$ was placed on the intermediate bottom of the frame. This valve consists of a light steel plate $a$, Drawings 6 and 7, and a stop $b$; $a$ descends under the action of its own weight, without any springs, which were discarded so as not to increase the resistance to suction.

Of course, when operating with a check valve, the opening $w$ in the intermediate bottom, Drawings 2 and 4, must be closed with a special plug.

In conclusion, it should be noted that, unlike machines of the usual type, in which air is drawn in directly from the atmosphere through a series of openings in the lower frame, the machine in question was provided with a bottom $v$, to which a pipe leading from the air (gas) meters was fitted. For the accuracy of measuring the quantity of air drawn in, it is very important that this bottom be completely airtight. In view of this, the factory was ordered to cast it integrally with the lower frame. Unfortunately, the factory refused to do this, citing the difficulty of casting, and made the bottom wooden, of thick boards planed into tongue-and-groove joints. However, in time the boards dried out; attempts to fill the cracks with asphalt varnish did not give satisfactory results, and therefore, when beginning the experiments, we filled the bottom with a layer of mortar made of Portland cement and sand; the thickness of the layer was about 25 mm; on top, the cement was painted with several coats of enamel paint. Experience showed that this method had been chosen successfully; one year after the filling, the tightness remained complete, and not a single crack appeared in the cement.

Drawings 6 and 7.

With regard to tightness, attention also had to be paid to the joint between the upper and lower frame and to the main bearings. The tightness of the joint was easily achieved by coating it with thick enamel paint and tightly drawing up the bolts. The work of attaining tightness in the main bearings was somewhat more painstaking. Since their liners are solid, tubular, it is true, closely embracing the shaft and always filled with oil, but having no device for tightening—

Chapter II. Experimental Setup. § 4.

then all attention had to be concentrated on the tight fit of the cheeks of the shaft, which have the form of disks, to the covers $x,x$. To facilitate fitting, between the cheeks of the shaft and the covers $x,x$, over the bronze washer $y,y$, which rotates freely on the shaft, four planes had to be scraped in from each side (on the cover $x$, the cheek of the shaft, and both sides of the washer $y$) and a firm flange at the cover $x$ had to be obtained, so that the clearances between the separate rubbing surfaces were reduced to fractions of a mm and, moreover, were always filled with oil. By observing the air pressure it was possible to establish that, although air leakage did remain, it had been reduced to an insignificantly small magnitude. Moreover, such leakage in the crank chamber merely reduces the volumetric delivery of the pump, but does not affect the accuracy of measuring the air consumption, since approximately the same amount of air as leaks out of it during compression enters, bypassing the gas ports, during suction.

Next we shall dwell a little longer on the determination of the quantities necessary for calculating the compression ratio $\varepsilon$, both in the working cylinder and in the air pump. These quantities are the volume of the compression chamber $V_c$, the volume swept by the piston $V_x$, and the volume of the pump clearance space $V_z$.

The volume $V_x$ was calculated from the above-mentioned $D$ and $H$ and was found to be $V_x = 9.738$ liters; since the quantity $D$ was determined with an accuracy of $\pm 0.02\%$, and $H$ with an accuracy of $\pm 0.04\%$, the value of $V_x$ computed from them is correct with an accuracy

\[ \nu = \sqrt{(0.02 \cdot 2)^2 + (0.04)^2} = \pm 0.06\%. \]

The volume $V_c$ was measured by directly filling the compression chamber with machine oil through the nozzle opening. Oil was preferred to water because, owing to its viscosity, there was no danger of leakage through the piston rings, which could easily occur when filling with water and would have given an exaggerated $V_c$. The amount of oil poured in was determined by weighing on precision balances, and the specific weight of the oil was measured with an accurate areometer, and, for checking, once again by a hydrostatic method with the aid of a Riemann weight and precision analytical balances. During filling, all measures were taken so that not a single air bubble remained in the cylinder. This, however, was not difficult to achieve, in view of the outlines of the compression chamber and the location of the nozzle opening at the upper point. During filling, the cavity of the ignition tube was isolated by means of a plug, and its volume was determined by a separate filling of the tube, for which purpose the cylinder head, of course, had to be dismantled.

The filling of the compression chamber was repeated 9 times; after each filling the cover was removed, at least $1/2$ hour was allowed for all the oil to drain off, and then the walls were again carefully wiped.

The volume $V_c$, without any gaskets under the connecting rod and the cover, found as the mean value of all observations, was $V_c = 3.106$ liters. All the individual measurements gave very close figures, which may

It follows from this that the mean error of the mean value, calculated by the well-known formula from the squares of the differences, is only 0.0074 liters, i.e. $V_c$ was found with an error $\nu < \pm 0.3\%$, which must be regarded as more than satisfactory.

When the above-mentioned gaskets $r$ are inserted, the volume $V_c$ decreases for every 10 mm of the thickness $r$ by 0.388 liters; when the ring $s$ is inserted, the volume $V_c$ increases by the same amount.

In the same way, by filling with machine oil, the volume $V_z$ was determined. In view of the smaller percentage error $\nu$ in measuring so large a volume, and of the closeness of the figures obtained, we decided to limit ourselves to three fillings.

On the average of these three measurements, $V_z = 46.545$ liters, likewise without any gaskets $r$ or caps $t_1$ and $t_2$. The mean error of this mean value, calculated in the same way, proved to be $\pm 0.085$ liters, i.e. $\nu < \pm 0.2\%$.

In the same way were also found the volumes $V_1$ and $V_2$ of the additional caps $t_1$ and $t_2$: $V_1 = 19.167$ liters, and $V_2 = 21.174$ liters. In this case the mean error of the mean value proved equal to $\pm 0.036$ liters, i.e. $\nu < \pm 0.2\%$.

5. The entire installation.—Since one of our principal tasks was to measure the air delivery under various conditions, the remaining part of the installation consisted precisely of devices intended for measuring and regulating this delivery.

In drawing 8, Table II, is shown part of the laboratory engine room; $a$ is the machine, $q$ the air meters.

The meters—a gas meter with a drum of capacity 1.50 m$^3$, i.e. with a throughput of up to 150 m$^3$ per hour—were purchased from the firm C. Elster in Berlin. Despite the undoubted desirability, in carrying out scientific experiments, of checking all instruments used in the experiments, we did not check these meters for the following reasons:

our experiments were intended chiefly to determine the change in the air delivery as a function of certain factors; the absolute value of the delivery, although of interest in view of the complete absence in the literature of corresponding data for machines of the given type, was for us a secondary question;
the unavoidable influence on the absolute value of the delivery of such additional factors, not changing during the experiments, as the resistance in the pipeline, the influence of leaks in the bearings of the crankshaft, irregularities in the ports of the machine—all this taken together can, after an approximate estimate, be considered approximately equivalent to the possible error in the readings of the meters;
carrying out a check with sufficient accuracy was very difficult, since the only one available in Tomsk, selected ...

the control gas meter from the Junkers calorimeter has a capacity of only 3.0 liters; for one revolution of the drum of the large meters it would have to make 500 revolutions, whereby an error equal to the error of the meters could easily accumulate; checking by means of a gas holder or an air vessel, of which more will be said below, would have been too complicated;

finally, the meters were entirely new, supplied by a first-rate firm, so that it was possible with sufficient confidence to rely on the direct readings; the firm guaranteed an accuracy of $\pm 3\%$.

All these considerations taken together induced us to abandon the checking of the meters; instead, we paid great attention to ensuring that the meters did not change their readings in the different experiments. For this it is necessary that two conditions be satisfied: first, the water level in the meters must always stand at one and the same height indicated by the factory, and, second, the water level must, during operation, be as nearly as possible completely at rest.

Since, when air passes through the meters, the water, saturating it, gradually evaporates, its level falls, and the meters begin to indicate less than the true value. The loss of water must be replenished. The best method of replenishment consists in continuously adding water to the meter during operation, while the excess flows out through a special overflow. These meters were provided with such a continuous supply of water and overflow of King’s system.

It was more difficult to satisfy the second requirement: piston machines always take in air in pulsations, causing a constant oscillation of the water level in the meters. Although this machine, in contrast to four-stroke ones, takes in air at every revolution, so that the number of revolutions is comparatively large, i.e., with a rather long, about 10 m, and large-diameter, 125 mm, air pipe, a continuous current of air of moderate speed, less than 0.4 m/sec, should result, and although the volume of air taken in at each revolution, compared with the capacity of the meter drum, is small, only 4–6 liters, and, finally, although the “load” of the meters was small, only from 50 to 100 m$^3$/hour, i.e. from one to two thirds of the throughput capacity,—nevertheless the very first trial showed, contrary to theoretical expectations, the complete impossibility of working without a special device to moderate the oscillations of the level, which reached 20–30 mm, i.e. without the inclusion of some pressure regulator.

To maintain a constant pressure we resorted to two means: first, we included in the air line a large iron riveted vessel $c$, Fig. 8, of 4.5 m$^3$ capacity, whose direct purpose in the laboratory is to serve as a receiver for compressed air supplied by a special experimental compressor.

The water level in the tanks thereafter became calmer, but not entirely so. Almost complete immobility of the level was attained after installing, at the tanks themselves, the pressure regulator $d$, shown separately in Drawings 9 and 10. This is a wooden box, 1020 mm high and with a base area of $500 \times 900$ mm. Its two side walls, $1020 \times 900$ mm, are replaced by rubber 1.5 mm thick.

Drawing 9 and Drawing 10: pressure regulator

Drawing 9. Drawing 10.

Taking the difference in air pressure necessary for passage through the tanks as 5 mm of water column, as was shown by the U-shaped water manometer $m$, connected to the ring between the tanks and the pressure regulator, it is not difficult to calculate that the amount by which this regulator pulsates, i.e., changes its volume (see the dotted position of the rubber walls), when the pressure fluctuates by 1 mm, is approx. 5 liters, i.e., quite sufficient.

Needless to say, all measures were taken so that both the pressure regulator and the entire pipeline, assembled from cast-

of 5-inch pipes, partly on flanges and partly on couplings with sockets filled with a special putty, were perfectly tight, which, however, was not difficult to achieve at the given slight vacuum. For the sake of the indicated tightness, the rubber walls of the regulator were fastened in the following manner: the rubber was laid over the edges of a wooden box, painted with oil paint, and tightened with screws through an intermediate frame made of the same wooden bars. The distance between the screws was taken as 60–70 mm.

Despite the apparently calm operation of the air clocks, the air, in passing through them, carried along so considerable a quantity of water that the latter, separating out during motion through the rather long air duct, after several hours of operation noticeably constricted its cross-section. To facilitate the draining of this water, and also for its possible separation, by making use of centrifugal force when changing the direction of the air motion, the elbow connecting the air duct with the bottom of the pressure regulator was replaced by a special cast-iron box $d$ with an opening $f$ for the drain cock. To attain the indicated purpose, the air duct, of course, had to be given an inclination toward the water separator $d$.

In addition to experiments with the normal supply of air by means of its suction by the piston of the machine, experiments were also carried out with an increased air supply. The increase was achieved by having the air not be sucked in, but enter under a certain pressure from vessel $c$, Fig. 8, into which it was forced by a Roots blower $f$, driven by a belt from a 4-horsepower electric motor $e$. To obtain the required air supply, it was possible to vary, within the limits from 300 to 800, the number of revolutions of the electric motor by means of a shunt rheostat $g$, located under the common distribution board $h$ of the laboratory.

However, such regulation of the supply proved insufficiently sensitive and, moreover, at small $n$ of the electric motor the rheostat became too hot. A second method of regulating the air supply was then added, consisting in covering, to a greater or lesser extent, the Ludeau slide $i$, which, during operation of the blower, separates the suction part of the pipeline from the discharge part. The excess compressed air is thereby discharged back into the suction part. This method proved remarkably convenient and accurate in the sense of establishing in vessel $c$ the required pressure.

6. Measuring instruments and appliances.—We shall describe here the principal instruments, except those used for determining the composition of the exhaust gases, which, in view of their importance, are treated separately.

Brake.—In designing the brake, two principal requirements were kept in mind: the possible elimination of load oscillation, so unpleasant with the ordinary Prony brake, and ease of handling.

From these considerations, having abandoned the use of a self-adjusting brake, which is more convenient during operation but less accurate, we settled on the type in which the end of the lever presses on the platform of a decimal balance, Drawings 11–13, p. 22.

To satisfy the first requirement it was necessary to obtain, as far as possible, a constant circumferential force, which in an ordinary brake changes owing to fluctuations in the coefficient of friction, depending on the temperature and the condition of the surface (lubrication) of the brake pulley. In order that the temperature of the brake pulley should remain constant and low, the pulley was provided with continuously acting water cooling: the water, the inflow of which from the water-supply system is regulated by the cock $a$, enters through the tube $b$ and is led back through the tube $c$, being caught by a flat scoop-funnel $d$. The tube $c$ runs under the floor and comes out into the sink $l$, set into the floor, Drawing 8, and connected with the common sewer. Further, on the brake there is also a continuously acting drip oiler $e$. Finally, the smoothness of the brake’s operation is greatly aided by the fact that the wooden blocks pressing on the pulley are fastened to the rubber belt $f$, which embraces and tightens the pulley. This belt, consisting of three rows of rubberized canvas with rubber interlayers, is flexible, can stretch slightly, and at the same time gives no residual elongation, either under the action of tension or from accidental wetting with water.

Ease of handling and accuracy of brake adjustment are achieved by the use of a special tightening device: the more tightly stretched end of the belt is fastened rigidly to the wooden lever $g$, or rather to the iron bracket $l$ attached to $g$; the less tightly stretched end is fastened to the bracket $h$, the ends of which pass through the lever $g$, are provided with a fairly fine screw thread, and enter two nuts; these nuts form one whole with worm wheels and thus receive simultaneous rotation by means of the worm $k$, coupled with them and turned by the handle $i$. Since the nuts make one revolution for 15 revolutions of the worm, and the pitch of the thread in them is 2.14 mm, then, taking the smallest fully determinate amount as $1/4$ revolution of the handle $i$, the length of the belt $f$ can be changed with an accuracy of up to 0.04 mm, and thereby the smallest change in its tension can be monitored.

It should be noted that, in order to maintain a constant load during operation, it is almost never necessary to release the brake, but only, from time to time, to tighten it slightly. In this case the pointer of the decimal balance $n$ scarcely oscillates about its position of equilibrium. The balance $n$, Drawing 8, which we used in the experiments, has a lifting capacity of 150 kg and, having a runner with divisions of 0.05 kg, with some practice allows the weight to be determined with an accuracy up to 0.02 kg; its sensitivity is about 0.01 kg.

V. L. Maltsev.

Figs. 11–13.

In conclusion, it may be mentioned that, for greater accuracy, the wooden support $m$ between the lever $g$ and the platform of the scales $n$ is provided at its ends with steel, hardened knife-edge shoes, of which the lower rests on the cast-iron platform of the scales, and the upper in a special hardened-steel socket $o$ let into the wood.

The actual weight of the brake itself—let it also be denoted by $g$—which in this arrangement must be subtracted from the readings of the scales when calculating the circumferential force, was found by direct weighing as follows: the brake was placed on the pulley, but the belt was completely slackened, so that the brake did not touch the pulley but lay on the edge of a triangular prism placed on the pulley exactly above the shaft axis. The other end of the lever $g$ rested on the platform of the scales through the support $m$, as during operation. Having balanced the scales, we obtained the weight $g = 1.53$ kg.

In order not to complicate the calculations of the effective work unnecessarily, in all the experiments the movable weight on the scales was set directly at 1.53 kg, and only the weights placed on the scale pan were taken into the calculation.

At the same time the length of the brake arm was found—let it be denoted by $l$—by measuring the distance between the recess of the socket $o$ and a vertical line drawn through the shaft axis of the machine; the result was $l = 909$ mm.

Revolution counters. The number of revolutions of the machine, the exact measurement of which is necessary for calculating the effective and indicated work, was determined by a continuously acting counter made by Schäffer and Budenberg, applicable, unlike so-called anchor counters, up to the very highest numbers of revolutions per minute.

This counter consists of a worm, fixed rigidly to the shaft of the machine and meshing with two flat worm wheels arranged on one axis, one above the other. One wheel has 100 teeth, the other 99; a stationary pointer attached to the counter casing indicates units and tens of revolutions, while the pointer showing the relative displacement of the two wheels indicates hundreds and thousands. This counter is very simple, absolutely accurate, does not break, and requires no maintenance.

In order to set quickly the required number of revolutions and, during operation, to observe whether the established motion is being maintained, we also attached a centrifugal tachometer by Schäffer and Budenberg, No. 18931, with a large dial, driven by a belt passed over the flywheel hub instead of a small pulley.

In view of the fact that, at a large number of revolutions and with a small diameter of the driven pulley, the belt slips badly and often comes off, we replaced it with ordinary insulating tape used for electrical wires. The results were excellent: the tape,

pressing against the pulleys with its sticky side, does not slip and does not jump off; and, in view of the extraordinary flexibility both of the tape and of the joint, the tachometer operates very steadily.

The tachometer was especially useful in the event of faulty operation of the engine: without it, it would have been impossible to notice sufficiently promptly the threatening stoppage of the engine, during which, if the brakes were not released in time, the latter, by their lever, when the pulley rotated backward—owing to the inability of the flywheels to overcome the compression in the cylinder—would strike the surrounding instruments and might break them.

Determination of the number of misfires. At first the idea was to attach to the engine a special instrument for counting the number of misfires, more accurately the number of strokes with kerosene admission. Such an instrument can be made, for example, in the form of an armature counter, which is actuated together with the rod of the kerosene pump and records when the regulator slips past this rod, i.e., at each misfire. However, the unsatisfactory results obtained by us in preliminary tests of armature counters at a small stroke with a considerable number of strokes per minute, and the small and comparatively insignificant error which, on the contrary, is obtained when the number of misfires is counted by eye, as will be described in detail below, compelled us to abandon the construction of a special instrument.

It is true that at the present time there already exist special counters of flashes, actuated by the pressure of the gases at the flash. Such, for example, is the instrument manufactured by the firm Dreyer, Rosenkranz and Droop, in which the mechanism summing the number of flashes—a small armature counter—is actuated by an ordinary indicator piston with the same indicator return spring. The spring is made somewhat stiffer than corresponds to the compression pressure; then the latter, in the event of a miss, i.e., the absence of a flash in the cylinder, is not able to raise the piston sufficiently and change the digit. If a weaker spring is inserted, the instrument turns into an ordinary stroke counter.

Still more interesting, although exactly twice as expensive, is the counter of the Grechem system, manufactured by the firm Schäffer and Budenberg. In this instrument the armature counter is likewise actuated by the pressure of the gases at the flash, by means of a special piston, a knee-jointed lever, and a return spring. Its action is the same as that of the preceding one. The other half of the instrument, completely symmetrical but with a weaker spring, is an ordinary stroke counter actuated by the compression pressure.

We were kept from purchasing and installing one of these instruments, despite their apparent convenience, by the following considerations:

Installing the instrument on our engine would have been rather difficult, since it has only one indicator opening; to install, however,

so-called double indicator cock was not wanted, since with it the passage connecting the cylinder with the indicator is lengthened, which has an unfavorable effect on the accuracy of the indicator diagrams, and, moreover, with the installation of such a cock the arrangement of the indicator drive would become more complicated;

2, the instrument on our machine would not always give correct readings: thus, on the one hand, premature flashes observed in it sometimes occur before the admission of kerosene that has not been atomized and has not had time to evaporate during the preceding stroke and therefore did not then burn; if the preceding flash was especially strong, then this stroke may be precisely an omitted one, while the instrument would record it as a flash, i.e. as a working stroke; on the other hand, on our indicator diagrams there are often observed outlines with such a delayed and slow combustion that it gives only a small pressure, almost or even entirely not exceeding the compression pressure; such outlines, on the contrary, the instrument would not record;

3, although we did not work with these instruments, on the basis of our experience with indicators we believe that prolonged communication of the instrument with the cylinder of the machine during selection should cause such strong heating of the piston of the instrument and its walls that no lubrication would withstand it; under these conditions the piston could easily seize, and in any case the instrument would allow gases to pass and would wear out very quickly;

4, finally, the cost of the instrument, even without taking into account its possible defects and inconveniences, is comparatively very considerable, especially if one considers that it would have to be used very rarely, in view both of the scarcity at present and the continuing decrease in the number of machines regulated by omissions.

Indicators. In the experiments we used simultaneously 3 indicators: one for the working cylinder, a second for the air pump, and a third for the muffler, which stood in the machine room.

Of the 16 indicators in the laboratory, of the most varied designs and sizes, after a series of tests we settled upon 3 indicators of the firm Dreyer, Rosenkranz and Droop, as the most accurate and convenient in operation. All 3 indicators had to be of different designs, depending on the peculiarities of their work.

On the working cylinder there was indicator No. 7418, of small size, with an external cold spring working in tension; this indicator is especially convenient in view of the speed with which the spring can be changed in it; for this it is sufficient to unscrew the upper nut and the spring itself, without disassembling any other parts.

Of the advantages, recognized by everyone, of external cold springs when working with internal-combustion engines at the present time there is no longer any need to speak. The laboratory also has another indica-

of the same firm, with an external spring, but working in compression. To change the spring in this indicator one must unscrew the upper nut and both ends of the spring, i.e., do the work—or, more accurately, spend the time—one and a half times as much. In addition, a spring working in compression, especially with so long a transmission mechanism as in the indicator with an external spring, can always cause the piston to jam from skewing, which does not occur with a self-centering spring working in tension.

When working with indicator No. 7418 we used a piston with $d=20$ mm, and springs with $m'=3.5$ and 4 mm for working diagrams and $m''=8$ and 10 mm for the scavenging period.

On the rear cavity, the air pump, at first we tried to install ordinary indicators with $d=20$ mm, but they, having a spring of only $m=30$ mm, gave diagrams that were too small; the Makhayk indicator, which had a weaker spring with $m=60$ mm, proved unsuitable owing to excessive piston friction. The pressure corresponding to the friction in the instrument¹² proved in it to be greater than the distance between the upper and lower lines of the diagram!...

Beginning with experiment No. 23 and to the end, indicator No. 7590 was mounted on the air pump, with piston diameter $d=40$ mm, designed specially for air blowers. In view of the large area of the piston, it gives large ordinates with comparatively stiff springs; we used springs with nominal $m=15,\ 20,\ 25$, and 40 mm, which gave ordinates on the scale 60, 80, 100, and 160 mm $=1$ kg/cm².

We did not take this indicator from the very beginning because we feared unsatisfactory results owing to the comparatively large mass of the moving parts and of the drum, especially taking into account our rather considerable number of revolutions. But, as the experiments showed, with a proper choice of the indicator spring and of the tension of the drum’s return spring, this indicator can be used even at $n>300$ rev/min, though, of course, when using the diagrams one must take into account the effect of the inertia of its parts, as will be indicated below.

On the muffler there stood a small-model indicator No. 7280 with an internal spring, since diagrams from the muffler were taken only once or twice during one experiment, and the indicator therefore did not heat up excessively. For it we always took a spring with $m=30$ mm.

Indicator drives. In view of the large number of diagrams taken, special attention was paid to the design of the indicator drives.

The drive for ordinary diagrams of the working cylinder and the rear cavity is shown in drawings 14–16: it is a cylindrical rod-slide $a$, receiving a reciprocating motion with a stroke,

equal to 76 mm, from the eccentric $b$, mounted on the crankshaft. The eccentricity lies, of course, in the same plane as the crankshaft. A loop from the cord of the working-cylinder indicator was hooked onto the hook $c$, and a loop from the cord of the air-pump indicator was placed on the tooth $d$. The first indicator was coupled to the drive without any additional rollers, by means of a very short cord, which is highly desirable at a large number of revolutions, since it eliminates the possibility of distortion of the diagram as a result of stretching of the cord.

Drawings 14–16

Drawings 14–16.

The air-pump indicator was coupled by means of a cord about 1000 mm long, which, in order to obtain the proper direction, was passed over the intermediate roller $e$. In order, as far as possible, to reduce the stretching of the cord, in this drive the braided hemp cord was left only on the indicator drum and in the middle part, where it bent around the roller, for a total length of only about 170 mm; the remaining part—between the indicator and the roller and between the roller and the end with the hook—was replaced by iron wire with $d = 1.0$ mm, which in practice gives no elongation.

To obtain so-called shifted diagrams, i.e. diagrams shifted by $90^\circ$ relative to the crank and giving, on an expanded scale of abscissas, those parts of the diagram corresponding to the phenomena occurring at the moments when the piston passes through the dead centers, another drive was used, drawings 17 and 18.

Since displaced diagrams are used not to determine the area of the diagram, but only to clarify certain phenomena, so to speak, from the qualitative side, this drive may also not be precise; that is, the abscissa points of the diagram may be spaced somewhat differently from one another than they would be if there were a strict proportionality between them and the paths of the piston. But the ratio $l/r$, the length of the connecting rod to the radius of the driving crank, must be known as accurately as possible, as must also the magnitude of the angle $\alpha$ between the cranks of the drive and of the machine, if it differs from the value normal for displaced diagrams, $\alpha = 90^\circ$. The point is that if one wishes to use a displaced diagram for any calculations, it must in any case be reconstructed, taking time—or, what is the same thing, the angle of rotation of the crankshaft—as the axis of abscissae; for the accuracy of the construction the nature of the kinematic connection between the motion of the machine piston and the indicator drum is immaterial, provided only that this connection is known.

Drawing 17–18

Drawings 17–18.

In the drive shown in Drawings 17 and 18, pin $a$ serves as a crank forming an angle of $90^\circ$ with the direction of the crankshaft knee, and, instead of a connecting rod, a portion of cord from pin $a$ to roller $b$ is used; the length of this cord is variable and oscillates from $l_1 = 300$ to $l_2 = 360$ mm. In view of this, the ratio of the length of this cord to the radius of rotation of crank $a$ is not only not equal to the value $L/R$ of the machine, but also varies, namely from 10 to 12.

The special feature of the pin, consisting in the fact that its end is bent toward the shaft axis by a smooth transition, is necessary in view of the high-speed nature

of the machine; to hook the small hook onto an ordinary crank pin at $n>200$ rev./min. is almost impossible, whereas onto such a pin, the end of which stands motionless, a ring tied to the end of the cord and itself sliding into place, i.e. onto the groove on the pin, is put on directly.

In the drive described above, only part of the drive, about 90 mm long, thrown over roller $b$, is made from braided cord; the other two pieces are made of iron wire with $d=1.0$ mm.

It is appropriate to mention here that, in order to connect and disconnect, when necessary, the loops at the end of the indicator cord with the ends of the wire, it proved more convenient to do this not with the aid of the usual indicator hook, but with the aid of a loop likewise formed by the end of the wire. The end of the wire, bent back and twisted with the wire itself along a helical, not especially steep line, is unwound by hand, the loop of cord is inserted, and the end is again twisted by hand along the old turns. Such a lock is very reliable, is quickly connected and disconnected, and scarcely increases the mass of the moving parts.

Further, from this same displaced drive the indicator on the silencer is also set in operation. Its cord, about 3200 mm long, was, of course, made of the same iron wire with the same loop-locks at the ends.

Measurement of pressures. The pressure, more precisely the rarefaction of the air in the knee between the clockwork and the pressure regulator and in the air vessel $c$, Fig. 8, was measured by means of U-shaped water manometers; and, in the case of forcing air into vessel $c$ by means of the blower, its pressure was measured by means of the same U-shaped manometer, but with longer legs, up to 400 mm, and filled with mercury.

The atmospheric pressure was measured by means of a mercury barometer with divisions of 0.5 mm.

Measurement of temperatures. The temperature of the combustion products was measured by means of a thermoelement composed of silver and constantan wires, 0.5 mm in diameter. To insulate the wires from each other, a thin glass tube was put over the silver wire. The thermoelement was inserted into a glass tube sealed at one end and clamped in stuffing box $a$, Figs. 19 and 20, by means of packing made of asbestos cord $c$; this tube was pushed into the exhaust pipe to such an extent that the soldered end of the thermocouple, resting against its sealed end, came approximately at the center of the exhaust pipe. The two other, so-called cold junctions were immersed in a vessel with water, the temperature of which was measured by a special mercury thermometer and added to the reading of the millivoltmeter.

Fig. 10 and 20.

A precision millivoltmeter was used, of the d’Arsonval system, from the firm Siemens & Halske, No. 062479, with two scales placed side by side—in millivolts and °C. This temperature scale has divisions of 5°, but readings can be made by eye with an accuracy of up to 1°.

In order that the thermoelement should not indicate an overestimated temperature of the combustion products, it was inserted into the exhaust pipe at the very point where it joins the cylinder of the machine; moreover, the entire exhaust pipe inside the machine room was wrapped with a one-inch layer of asbestos.

In view of the cases that occurred of a brief, though incomplete, short circuit in the thermoelement circuit, in which the latter indicates 100–150° less than the true temperature, an additional iron fitting was screwed into the exhaust pipe, into which a mercury thermometer with divisions up to 450° was inserted. This control thermometer usually indicated, of course, less than the readings obtained with the thermoelement: on the one hand, because of the influence of the protruding column of mercury; on the other, because the contact between the bulb of the thermometer and the fitting was not especially tight, since, owing to the high temperature, neither mercury nor mineral oil could be poured into the fitting, and it was undesirable to fill it with tin, which melts already at 232° C., in view of the possibility, for example, in the event of an unexpected stoppage of the machine, of forgetting to remove the thermometer, which, when the tin cooled and solidified, might be damaged. When, however, the readings of the mercury thermometer sometimes proved higher than the readings of the thermoelement, this meant that a brief short circuit had occurred in the latter; it was immediately found and eliminated.

In conclusion, it may also be mentioned that, in order to spare the millivoltmeter as much as possible, current was passed through it only at the time of reading; for this purpose one of the conductors was cut and a special switch was included. This switch consisted of a small glass jar into which mercury had been poured and one end of the conductor immersed; the other was attached to a glass tube-handle and was lowered into the mercury at the moment of reading.

Readings of all the remaining temperatures were made with mercury thermometers; the latter were taken of the best grade, of Jena borosilicate glass 59ʺ, from the firm Arno Haak in Jena.

In order that the thermometer readings should lag as little as possible behind the true temperature, all thermometers were in direct contact with the corresponding medium. The thermometers with which the air temperature in the pump and in the channel leading into the cylinder was measured were, owing to the insignificance of the air pressure, inserted directly by means of rubber stoppers with an opening. The thermometers for the cooling water were inserted by means of clamps according to drawing 21: c is a rubber ring, both providing the required tightness and at the same time holding ...

the thermometer from being forced out by the pressure of the water. On the pipe with hot water the thermometer is placed as close as possible to the cylinder, and, moreover, the pipe is coated with a half-inch layer of asbestos.

This pipe is only $3/4''$ in diameter and runs vertically; the small branch pipe $a$, Drawing 21, is soldered in obliquely, making an angle of about $40^\circ$ with the vertical. Owing to this, not only the whole bulb with mercury, but also the lower part of the thermometer is in the hot water, whereby the magnitude of the correction for the projecting column of mercury is reduced.

Drawing 21

Drawing 21.

Measurement of the consumption of kerosene and of the injected water. The consumption of kerosene was measured by direct weighing on decimal table scales with a lifting capacity of 25 kg, indicating with an accuracy of up to 0.005 kg, which, for a consumption of 1.3 to 2.5 kg, gives an accuracy of more than $\pm 0.4\%$. To measure the consumption of kerosene, the wall of the tank from which the kerosene goes to the pump was provided with an indicating glass, communicating at the bottom with the tank and open at the top; into this glass a fairly thick wire spiral, wound from iron wire, was inserted; the lower end of the wire was sharpened. When the kerosene level fell during operation of the machine, the instant at which the level passed the point was very easy to observe with great accuracy, since the kerosene wets the wire, and the breakaway of the adhering drop from the point is seen very clearly. To reduce the error in determining the consumption of kerosene, the cross-section of the tank was deliberately made as small as possible, only $180 \times 180$ mm, so that a half-hour kerosene consumption in the tests corresponded to a column of kerosene of 50 to 100 mm. In view of this, accidental inaccuracies in recording the instant at which the point breaks away, which can correspond only to small fractions of a millimeter, in no case more than 0.5 mm, cannot give an error in the kerosene consumption greater than 1%.

To determine the consumption of the injected water, a wire with a point was also inserted into the vessel from which the water went to the cylinder, but in this case the point projected upward, near the upper rim of the vessel, and the instant at which the point appeared was noted.

Measurement of the consumption of cooling water. The quantity of cooling water was determined by directly weighing it on decimal scales with a lifting capacity of 250 kg, giving readings with an accuracy up to 0.05 kg. On the scales stood an iron riveted tank $m$, Drawing 8, with a capacity of 230 liters; water was poured into it in a free stream from a canvas hose, which was fitted onto the end of the discharge pipe. This pipe runs upward from the cylinder of the machine; then, at a height somewhat greater than a man’s height, the pipe passes horizontally to the wall and ends in a downward-bent elbow. At the moments of weighing the tank, and generally whenever the water did not have to be collected, the end of the canvas hose was taken out from behind the rim of the tank and then hung over the sink $l$, Drawing 8, made ...

directly into the field. Water is likewise discharged there from tank m through a 2-inch branch pipe, fitted with a cock, attached at its very bottom.

Determination of the oil consumption. The oil consumption was determined only for the cylinder. In the first tests, when lubrication was supplied from a drip lubricator, the consumption was determined by directly weighing, before and after the test, the tin vessel from which the oil was poured; in the remaining tests, when lubrication was carried out by means of Malrup’s device, the oil consumption was calculated from the volume displaced by the piston of the device, depending on the number of teeth through which the worm of the device was turned by means of the ratchet.

A rotation by 1 tooth at 100 r.p.m. of the machine corresponds to an oil delivery of 0.0064 kg/hour. Since the smallest possible turn in our case was equal to 3 teeth, and the largest to 7, then, taking the limits of the number of revolutions as 190 and 360, we obtain the smallest delivery, 0.037, and the largest, 0.163 kg/hour. On the other hand, taking into account that, of the amount of oil delivered to the cylinder, not more than $1/3$ can burn, while the remaining $2/3$ passes into the rear cavity—as we repeatedly observed during the tests—we find that, with an average hourly consumption of kerosene of about 2.6 kg at $n=190$, or 4.8 kg at $n=360$, the quantity of oil burning in the cylinder amounts to approximately 0.5 to 1.5% in comparison with the kerosene consumption, i.e., to a quantity that may quite well be neglected.

7. Taking samples of the exhaust gases.—For the majority of the tests, samples for gas analysis were taken exclusively from the muffler b, Fig. 8, where the mixing of the gases as a result of the periodic impulses may be considered to give a sufficiently uniform mixture, close to the average composition.

The samples were drawn off by means of a gas tube screwed into the cover of the muffler, about 300 mm long, and an ordinary, well-ground gas cock attached to its end, into ordinary-type aspirators consisting of a pair of glass vessels, of capacity from 1 to 15 liters, with tubulures at the top and at the bottom.

In view of the doubt that arose as to whether the composition of the products of combustion taken from the muffler corresponds—when they contain unburned particles—to the composition of the products of combustion at their exit from the cylinder of the machine, and whether afterburning does not occur on the way to the muffler and in the muffler itself, altering, of course, the composition of the products of combustion, in some tests samples were also taken at the very beginning of the exhaust pipe.

In order to eliminate the possibility of afterburning, the tube a, Figs. 22 and 23, through which the sample was taken, was coiled into a spiral and immersed in a tin vessel into which cooling water flowed; at the lower point of the spiral a branch pipe b was made for draining off the water condensed from the products of combustion, and this branch pipe is closed by a rubber tube with a glass stopper. The small apparatus described is connected to the fitting

on the exhaust pipe by means of a union nut c; for the sake of simplicity, and also of greater reliability with respect to tightness, the other end of the tube a, instead of a cock, is provided with a rubber tube with a clamp.

Differentiator. Of greatest interest is the instrument that was designed by the author for the purpose of taking separate samples of the discharged exhaust gases at various successive moments of the exhaust. By means of this instrument it was intended to clarify the phenomenon of scavenging, i.e., to what extent the compressed air actually displaces the products of combustion or, perhaps, mixes with them and itself, to a considerable extent, immediately flies out.

Drawing 22 and 23

Drawings 22 and 23.

This instrument will hereafter be called the differentiator, since its purpose is to divide the stream of issuing gases into several separate streams corresponding to successive moments of the exhaust.

The idea of this instrument consists in making, at the very beginning of the exhaust pipe, a row of openings and opening them for an instant at the following moments: the first opening—at the beginning of the exhaust, then the second, the third, etc., and the last one just before the very cessation of the exhaust. Such successive opening and closing can most simply be obtained by moving a slide valve with a small opening in front of a row of ports communicating with the exhaust pipe. Since scavenging occurs near the dead point, the motion of this slide valve must be imparted from a drive similar to the drive for taking displaced diagrams. However, it is impossible to confine oneself to a single slide valve, since on its return motion it will open the ports a second time and thereby connect the aspirators taking samples with the exhaust pipe at certain other moments, when the exhaust pipe is disconnected from the working cylinder and filled with products of combustion corresponding to the end of the exhaust. To prevent this from happening, measures must be taken so that, by the time of the return motion of the slide valve, the ports are closed, i.e., another slide valve must be placed between the first slide valve and the pipe; it will play the role of a cut-off slide and move somewhat in advance of the first; then, having, instead of an opening, a narrow, elongated slot, despite the advance it will, on the forward stroke, leave open—

open, and during the return stroke to keep closed precisely those against which, at the given moment, the opening of the main slide valve is located.

This differentiator is shown in Figs. 24 and 25: $a$ is the main slide valve, $b$ the cut-off slide valve, both of brass; $1, 2, 3, 4$, and $5$ are copper tubes, in the order in which they are opened, onto which rubber tubes are fitted, connecting them with the aspirators.

Figs. 24 and 25

When designing the differentiator, the following principal requirements were kept in view:

the greatest possible tightness of the slide valves, i.e., isolation of the separate channels both from the external atmosphere and from one another;
the smallest possible capacity of the channels, especially those communicating with the various openings, i.e., channels $c$, $d$, and $e$, and of the recesses in the slide valves;
a forced, strictly determined motion of both slide valves.

To satisfy requirement 1, all 8 rubbing surfaces on the slide valves and mirrors were carefully lapped, and subsequently, during the experiments, were from time to time checked again and lapped. So that the channels would not communicate with one another, the openings on the front mirror were spaced sufficiently far apart. Then all 5 plates, 3 fixed ones and 2 slide valves, were constantly pressed against one another by means of 4 spiral spr—

valves $f$. In order that the pressure might be applied uniformly, these springs were to press not directly, but through the plate $g$, the protrusion on which was to press on the cast-iron piece $h$ in the middle plane.

Then, finally, the rubbing surfaces were lubricated. At first, in view of the high temperature, up to $200$—$300^\circ$, which the apparatus assumes during operation, we wished to use cylinder oil as the lubricant; but it proved too viscous, and offered too great a resistance to the motion of the slide valves. After a series of trials we settled on lubrication with graphite; the surfaces were slightly greased with vaseline oil and then sprinkled with a barely perceptible layer of the finest graphite.

To satisfy requirement 2, all the channels were made of minimal dimensions: only 4 mm in diameter; it was impossible to make them still smaller for fear of increasing too much the resistance to the delivery of the gases; the lengths of the channels were also made as small as possible.

Finally, to satisfy requirement 3, it was necessary—bearing in mind that the drive has to be made of cord—to reduce as much as possible the resistance to the motion of the slide valves and the length of the cord, which inevitably gives a certain stretch.

The resistances to motion may be divided into two groups: some depending on the speed, i.e., on the number of revolutions of the machine, and others not depending on the speed.

The former are the inertia of the moving masses. Their reduction was achieved by reducing as much as possible the dimensions of the slide valves. It was intended, in the extreme case, to make them even of aluminum, if the copper ones proved too heavy, but practice showed that this was unnecessary. Aluminum slide valves themselves seemed undesirable in view of their easy scoring, especially at high temperature.

As for the resistances independent of speed, they arise from the force of pressure of the springs $f$ and from the viscosity of the lubricant. Practice showed that well-ground-in surfaces require almost no pressure; therefore, after the very first tests, the springs were replaced by softer ones, only slightly tightened; if desired, their pressure could be adjusted by means of the nuts $i$.

To determine the influence of the lubricant, and also of the pressure, we measured with a stopwatch the time required to move the slide valve from one extreme position to the other under the action of a constant force—a weight of 2 kg, attached to a cord passed over a pulley. With lubrication by cylinder oil this time, depending on the pressure of the springs, ranged from 6.8 to 18.4 sec.; with lubrication by graphite with vaseline oil and weak pressure it was only 3–4 sec.; with lubrication by graphite alone, as a dry powder, this time was so short that it could not be measured with a stopwatch.

As for the actual setting in motion of the slide valves, it is shown in Drawing 26: in front, the slide valves move, being pulled—the main one $a$ by cord $k$, passed over pulleys $l$ and $m$, and the cut-off one $r$ by cord $n$, passed over pulley $o$. The pulleys were taken from indicators, so that they could easily be mounted in two planes, since, as may be seen, for example, from the photograph in Drawing 27, the planes in which the slide valves move form a certain angle with the plane in which lie the parts of the cords from the eccentric $p$ to the pulleys $m$ and $o$.

Drawing 26

Drawing 26.

The eccentric $p$, which serves specifically to operate the kerosene pump, lags behind the crank of the main shaft by an angle of $75^\circ$. The angle formed by the direction of the upper part of cord $k$ with the horizontal is about $15^\circ$; thus the motion of slide valve $a$ occurs as if from a crank making $90^\circ$ with the crank of the machine shaft. The angle between the directions of cords $k$ and $n$ is about $45^\circ$; by this amount the cut-off slide valve leads the main one. Thus, using only 3 pulleys, it is possible, comparatively simply, to obtain from one eccentric the required different motions of both slide valves.

It goes without saying that, in order that the motion of the slide valves, even at a larger number of revolutions, should remain the same as during slow rotation, the drives are made of iron wire and the cords are inserted only at the places where they bend around the pulleys.

The return pull of the slide valves is produced with the aid of springs $s$ and $t$. The calculation of these springs presents certain difficulties: on the one hand, the springs must be sufficiently strong that, overcoming the inertia of the slide valves, they keep cords $n$ and $k$ always taut—

...stretched; on the other hand, if the force of the springs is excessive, the drawing-out of the cords will increase uselessly, and even their breaking may occur.

We first calculated these springs theoretically in the following manner: let the mass of the slide valve be $m$, and with sufficient accuracy one may take $m=0.1\,g$, where $g$ is the weight of the slide valve in kg; let the path through which the slide valve must be moved when acceleration is imparted, i.e. half of its travel, be $h$, and the time of one half-stroke $t$; then the force $P$ of the return spring may be found from the equation:

\[ P > \frac{2mh}{t^2}=\frac{0.2Gh}{t^2}. \tag{1} \]

For our slide valves, direct weighing gave the weight of the main one as $g_1=0.081$ kg, and the weight of the cut-off one as $g_2=0.071$ kg; $h=0.03$ m. For $n=360$ we find $t=0.041$ sec. Thus we obtain:

\[ P_1 > 0.35\ \text{kg}. \]

If even 100% is added for friction, then even so we obtain the necessary force of only about 0.7 kg.

Practice, however, showed that the apparatus works well only with springs giving, on the average, a tension of about 2 kg.

In view of the rather considerable $h$, we took a rather large number of turns in the springs, up to 40, so that the stress, decreasing as the slide valves return to the initial, right-hand position, would not change too greatly. In the present case, however, the decrease in tension is still permissible, since the springs perform the greatest work during the first half of the reverse motion of the slide valves.

In conclusion it may be noted that, despite the absence of experimental data for the construction of such an apparatus as our differentiator, its design, carefully thought out and, as far as possible, calculated in advance, immediately proved quite satisfactory and required no alterations. The only thing that did not work at once was the drive; at first it was made entirely of ordinary indicator cord—it stretched terribly; the cord was left only on the pulleys—it began to tear; an attempt was made to replace the cord with a flexible electric conductor made up of the finest wires—and, despite replacing the metal blocks with wooden ones of large diameter, about 60 mm, the little wires very quickly broke from the large number of bends.

Having already almost lost hope of setting up the apparatus, at least for a considerable number of revolutions, we resorted to the last means—we ordered from abroad, from the firm Dreyer, Rosenkranz and Droop, an indicator cord twice as thick as the normal one, i.e. four times as strong. It turned out that the firm had such a cord ready and sent it at once.

The very first trial with a drive made of wires with a new thick cord on pulleys gave excellent results. The drive does not stretch, and the cord lasts quite a long time. Moreover, so that neither the differential indicator spools would operate needlessly nor the cord would wear through, the instrument was put into action during each experiment for as short a time as possible; for the rest of the time it stood motionless, for which it was sufficient to unlatch the springs $s$ and $t$.

The external appearance of our differential indicator together with all its drives is seen in the photograph, Fig. 27, Plate III.

To characterize the action of the differential indicator, Table 1 gives the moments of opening and closing of the separate ports obtained by direct measurement. In the following columns are indicated

Table 1.

Port No.	Crank angles: opening	Crank angles: closing	Crank angles: duration of opening	% of piston travel: opening	% of piston travel: closing	duration of opening at $n=340$, in sec.
1	$-69^\circ$	$-47^\circ$	$22^\circ$	36.4	18.4	0.011
2	$-37^\circ$	$-16^\circ$	$21^\circ$	11.7	2.3	0.010
3	$-9^\circ$	$+10^\circ$	$19^\circ$	0.6	0.8	0.009
4	$+23^\circ$	$+48^\circ$	$25^\circ$	4.6	19.1	0.012
5	$+22^\circ$	$+87^\circ$	$25^\circ$	30.3	52.8	0.012

these same moments, recalculated in terms of piston travel in %, and in the last column is indicated the duration of opening of the corresponding ports in fractions of a second. As we see, the opening time at $n=340$ amounts to only about $0.01''$, increasing at the lowest number of revolutions $n=190$ to approximately $0.02''$.

Of course, the quantities indicated in Table 1 must be regarded only as averages. The unavoidable change in the length of the drive when replacing a worn cord shifts the moments of opening and closing; these shifts are very small for the middle port 3, for ports 2 and 4 amounting to $\pm 5^\circ$ relative to the quantities indicated in Table 1, and for ports 1 and 5 already $\pm 10^\circ$.

The aspirator. In conclusion it remains to say a few words about the aspirator used for taking gas by means of the differential indicator.

In view of the cramped space and inconvenience of handling several aspirators of the usual type described above under the hot cylinder of the engine, a special apparatus was constructed, Figs. 28 and 29: these are six bottles of white glass with the bottoms broken off, all fastened to frame $a$, by means of which they can be raised and lowered, being secured on uprights $b$ by latches $c$ at the desired height—

bottles. The bottles are closed at the top with rubber stoppers, through which glass tubes $d$ are passed for drawing in the gas; rubber tubes are put onto the tubes $d$, by means of which the individual vessels of the aspirator are connected with the branch pipes at the openings of the differential indicator.

So that the box $c$, in which the sealing liquid—water saturated with table salt—is contained, should not rust, it is made of galvanized iron and painted with oil paint. Before the taking of gases, the frame $a$ is raised to the upper position; water is drawn through the rubber tubes so that no air at all remains in the bottles, and the tubes are clamped with clamps. The resulting column of water, 70–100 mm high, gives a rarefaction for drawing in the products of combustion, which, generally speaking, are expelled from the cylinder under a certain, sometimes rather considerable, pressure.

Figs. 28 and 29.

Figs. 28 and 29.

When samples were taken, usually 3 bottles were connected to the differential indicator, to openings 3, 4, and 5; from opening 2, in which its own pressure was often not only absent, but even a rarefaction was observed, the sample was taken by an ordinary aspirator from two separate bottles giving a greater rarefaction. Opening 1, which is closed, as it subsequently proved after the measurements according to Table 1, almost until the beginning of exhaust, was usually not used at all.

In order to transfer the collected samples into burettes for analysis, the frame $a$ is lowered, and the gas in the bottles is then under sufficient pressure.

V. L. Mal’ev.

Preliminary Experiments.

8. Absorption of gases by liquids.—Anyone who has had occasion to analyze products of combustion, whether from the flue of a steam boiler or from the cylinder of an internal-combustion engine, knows how different the results obtained are depending on whether the analysis is made immediately after taking the sample or after several hours, and also on whether the aspirator by means of which the sample was taken contained water that had been in use for a long time, or fresh water. In a word, the water with which aspirators are usually filled absorbs gases, but not to the same extent: some, such as $\mathrm{CO}_2$, more; others, such as $\mathrm{O}$ and $\mathrm{N}_2$, less. If there is a mixture of gases, then, owing to the unequal absorption of the individual constituent parts, the composition of the mixture in percent changes, and often very substantially.

Since in our experiments it was necessary to take a very large number of samples, the analysis of which would have required much time, and, in addition, one of the methods of analysis we used, by combustion, required that the products of combustion remain in the aspirator for a long time, we, striving for the greatest possible accuracy and bearing in mind the especially great importance of gas analysis in the forthcoming experiments, set ourselves in advance the aim of taking measures to reduce, if not eliminate, the absorption of gases by the sealing liquid.

Strange as it may seem, this question is not touched upon at all in the corresponding manuals, or it is merely stated that one should use an aspirator filled with water saturated with gas of approximately the same composition as the sample being taken. However, this indication is very indefinite: on the one hand, the absorptive capacity of water depends very strongly on its temperature and on the pressure of the gas; on the other, special measures must be taken so that the water proves to be fully saturated with gas. Moreover, it is enough to draw into an aspirator, in which there is water fully saturated with gas of a definite composition, a gas of another composition for the water, as a result of the change in partial pressures, to begin acting on the new gas, absorbing those of its constituent parts which have a greater proportional pressure than in the former gas mixture and, conversely, releasing gases whose content in the new mixture is smaller.

To determine the influence of absorption by water, we carried out a series of experiments that yielded rather interesting results. The experiments were arranged as follows: from an aspirator containing spent gas from a kerosene engine, samples of 20–30 cm³ were taken into three graduated Hempel burettes with stopcocks at the top and bottom. To prevent exchange of gas with atmospheric air through diffusion, the lower stopcock of the burette was also closed, so that the gas was in contact with a precisely determined volume of water. The sealing ...

the liquid in the burettes was water, which had been saturated with the same gas by passing it through for a long time and shaking it with it.

At the same time the composition of the gas in the aspirator was determined at the moment it was taken into the burettes, and the barometer height and the temperature of the gas (in the room) were also recorded.

After 24 hours had elapsed, by connecting the burette (graduated) with tubes for bringing the levels together, the remaining volume of gas was determined, recalculating it, like the first, to 737.4 mm and $15^\circ$ C. The gas in one of the burettes was then analyzed.

During the next 24 hours the same procedure was followed with the two remaining burettes, and the composition of the gas in the second burette was determined; finally, after another 24 hours, the volume and composition of the gas in the third burette were again determined.

As an example, Table 2 gives the results of one of the experiments. These results finally convinced us that water absolutely cannot be used when gases remain for a long time in the aspirator, no matter what precautions are taken.

Table 2.

Gas composition	Initial, %	After 24 hours: composition	After 24 hours: change	After 24 hours: in % of initial	After 48 hours: composition	After 48 hours: change	After 48 hours: in % of initial	After 72 hours: composition	After 72 hours: change	After 72 hours: in % of initial
$\mathrm{CO_2}$	8.0	6.4	−1.6	−20.0	6.1	−1.9	−23.8	6.0	−2.0	−25.0
$\mathrm{O_2}$	9.2	9.4	+0.2	+2.2	9.5	+0.3	+3.3	9.6	+0.4	+4.3
$\mathrm{CO}$	0.6	0.4	−0.2	−33.3	0.3	−0.3	−50.0	0.3	−0.3	−50.0
$\mathrm{N_2}$	82.2	83.8	+1.6	+1.9	84.1	+1.9	+2.3	84.1	+1.9	+2.3
Volume in cm$^3$	30.0	29.3	−0.7	−2.3%	29.1	−0.9	−3.0%	29.05	−0.95	−3.2%

What, then, should replace water? The only indication we found in the relevant manuals is in Hempel¹³, who advises using, instead of pure water, a saturated solution of $\mathrm{MgCl_2}$. Having become generally interested in the question of the absorption of gases by solutions of various salts in water, we expected to find ready figures in reference books¹⁴; unfortunately, we found neither figures nor references to suitable literature, specifically with regard to the gases of interest to us—$\mathrm{CO_2}$, $\mathrm{O_2}$, $\mathrm{CO}$, and $\mathrm{N_2}$—and to solutions intended to reduce the absorption of gases to a minimum.

Of the literature cited there, only Mackenzie’s paper^19 gives the coefficient of absorption of CO₂ in a saturated NaCl solution that we need; Mackenzie gives no experiments at all with MgCl₂ solutions, and with CaCl₂ solutions he carried them out only with weak, unsaturated ones. Others among the cited authors, Setschenow^20 and Bohr^21, also worked only with unsaturated solutions, and moreover only with NaCl, and examined the question exclusively from a theoretical point of view. Thus we were unable to find either comparative data on the absorption of gases in saturated solutions of different salts, or data on gases other than CO₂. In view of this, we decided to carry out our own experiments. Of course, our aim was not to investigate the purely scientific question of the coefficient of absorption of gases by various solutions, but only to obtain certain data for direct practical use. In accordance with this, we settled on the simplest, perhaps even insufficiently scientific, instruments and methods, and limited ourselves to a comparatively small number of observations.

Assuming, on the basis of general physicochemical properties, that saturation of water with related but more active salts, CaCl₂ and NaCl, should have approximately the same influence as the salt MgCl₂, we decided to compare the solubility of individual gases in saturated aqueous solutions of the three salts indicated, and, to check our method, also in pure water.

The experiments were conducted as follows: into a Hempel burette with stopcocks, filled with one or another of the indicated solutions, about 50 cm³ of gas was admitted at known barometer pressures and temperatures, which were recorded; the admitted volume was determined from the graduations of the burette and recorded. Then the lower stopcock of the burette was also closed, and the latter was brought into a horizontal position, secured on a special stand; the horizontal position was adopted in order to increase the contact surface between the gas and the liquid.

After 24 hours the burette was again brought into a vertical position and, with the lower stopcock open, liquid was added from the burette without graduations until atmospheric pressure was again established in the graduated burette; the remaining volume of gas was read off and, after the stopcock was again closed, the burette was moved to a horizontal position.

After 24 hours the operation was repeated, and so on until the volume of gas ceased to decrease. The loss of gas, divided by the quantity of liquid that was in the measuring burette at the end of the experiment, represented the coefficient of absorption of the given gas in the given solution.

^19 Ann. d. Phys. 1, 1877 S. 438.
^20 Z. Phys. Chem. 4, 1889 S. 117.
^21 Ann. d. Phys. 62, 1897, S. 644; 63, 1897 S. 500.

Study of a Two-Stroke Machine

In order, as far as possible, to weaken the influence of temperature, the experiments, which occupied two months, were conducted in a room with one window facing north. The temperature in the room was very constant; the minimum and maximum thermometers over the 3–4 days during which a single experiment lasted differed by no more than $1/2$--$2^\circ$.

Maintaining one and the same pressure, despite changes in the height of the barometer, presented no difficulty: when taking readings it was necessary only to bring the levels in the burettes not to coincidence, but, when the barometer stood above normal, to bring the level of the liquid in the burette with the higher pressure above the level in the other burette by the height of the column of liquid equivalent to the difference in the barometer heights; when the barometer pressure was low, the reverse was done.

Since the purpose of our experiments was not to study the question of the absorptivity of gases in general, but only to compare the absorptivity of gases by different liquids, we limited ourselves, on the one hand, to only the two gases most important among the combustion products, $CO_2$ and $O_2$; the first was taken from Kipp’s apparatus, and the second from a gasometer, into which it was admitted from an oxygen cylinder. On the other hand, in order to approach the conditions under which the gases are found in the aspirator, the experiments were carried out in two series: in one, an effort was made to bring the liquid into the closest contact with the gas, shaking the burettes both before each reading and, in general, several times a day; in the other series, which more closely corresponds to the drawing of gases into the aspirator, the burettes, on the contrary, were turned as carefully as possible. It is self-evident that the results of repeated experiments of this series II, depending on accidental conditions, differed among themselves much more than in the experiments of series I, since in this series I complete saturation of the liquid with the gas, of course, did not occur. However, strange as it may seem, the absorption of gas after 3–4 days ceased, at least within the limits of accuracy of our measurements.

Table 3.

Series of experiments	Gas	$H_2O$	$H_2O + 0.35\,NaCl$	$H_2O + 0.74\,CaCl_2$	$H_2O + 0.54\,MgCl_2$
I	$CO_2$	0.875	0.265	0.205	0.124
II	$O_2$	0.028	0.008	0.006	0.004
I	$CO_2$	0.565	0.135	0.100	0.072
II	$O_2$	0.015	0.004	0.003	0.002

Table 3 gives the absorption coefficients found for $B = 737.4$ mm and $t = 20.0^\circ C$. The first two rows give coefficients close to complete saturation, but, in all probability, owing to imper-

V. L. Mal'ev.

of our instruments, somewhat reduced; in the last two lines are given coefficients that have significance only for our experiments, i.e., when gas is drawn into the aspirator without shaking, and, of course, are only approximate, exemplary values.

The coefficients of absorption of $CO_2$ and $O_2$ in water agree sufficiently well with the coefficients found by other authors in exact investigations$^{22}$, as does the coefficient of absorption of $CO_2$ in a saturated solution of $NaCl$, for which Mackenzie, in the work mentioned above, found, at $22^\circ$ and for gas reduced to $B = 760$ mm and $t = 0^\circ$ C, $\alpha = 0.263$. Thus, despite the primitiveness of our experiments, the coefficients obtained by means of them may be considered quite satisfactory for the known practical purposes, and all the more so for our purposes—comparison of the solutions with one another.

Having Table 3, it is now easy to understand why Hempel advises, if one wishes to attain greater accuracy, using as the sealing liquid a saturated solution of magnesium chloride: this solution has an absorption coefficient almost 8 times smaller than that of pure water, and more than 2 times smaller than that of a saturated solution of sodium chloride.

Nevertheless, in our experiments we decided to fill all the aspirators with a saturated solution of common salt, proceeding from the following considerations: magnesium chloride is very expensive, both absolutely and comparatively, since a rather large quantity of it is needed to saturate water; calcium chloride, which occupies a middle position both in absorption coefficients and in cost, is inconvenient chiefly because its saturated solution is thick and syrupy; this results partly from the very great solubility of $CaCl_2$ in water, which makes, despite the moderate cost of the $CaCl_2$ itself, the cost of its saturated solution rather considerable. Bearing in mind that in all we used 12 aspirators in the experiments, to fill which up to 60 liters of liquid were required, it was impossible not to take account of the cost of the salt.

At the same time it should be noted that, for the sake of cheapness, we took not pure $NaCl$, but commercial inexpensive common salt, containing rather appreciable quantities of $CaCl_2$ and $MgCl_2$, which for us was even advantageous in view of their properties. In order to have a solution actually saturated with salt, and, if possible, a transparent liquid, the solution was prepared in advance in a large tank and poured into the aspirators by means of a glass siphon, so that all the turbidity, which had settled to the bottom and partly floated on the surface, did not get into the aspirators.

Of course, despite replacing water with saline solution, we nevertheless tried to saturate the liquid in the aspirators in advance with the exhaust gases.

$^{22}$ Land. u. Börnst. S. 598, 602.

To compare the practicality of using a saline solution, Table 4 gives the results of an experiment on the change in gas composition when the latter is stored in an aspirator with the indicated solution.

The experiment was carried out in exactly the same way as the one whose results were given above in Table 2. In view of the closeness of the gas compositions, Tables 4 and 2 make it possible to make a direct comparison.

Although, judging from Table 3, one might have expected somewhat better results, those obtained, for example, in Table 4 are sufficiently satisfactory, all the more so since in the cited example the solution had previously been insufficiently saturated with gas, and during the experiment the burettes were shaken, which, of course, one tries not to do with aspirators.

Table 4.

Gas composition	Initial %	After 24 hours: composition	After 24 hours: change	After 24 hours: in % of initial	After 48 hours: composition	After 48 hours: change	After 48 hours: in % of initial	After 72 hours: composition	After 72 hours: change	After 72 hours: in % of initial
CO₂	8.0	7.3	−0.7	8.8	7.1	−0.9	−11.3	7.0	−1.0	−12.5
O₂	9.4	9.5	+0.1	+1.1	9.5	+0.1	+1.1	9.5	+0.1	+1.1
CO	0.5	0.4	−0.1	20.0	0.3	−0.2	−40.0	0.3	−0.2	−40.0
N₂	82.1	82.9	+0.8	+1.0	83.1	+1.0	+1.2	83.2	+1.1	+1.3
Amount, in cm³	40.0	39.5	−0.5	−1.2%	39.3	−0.7	−1.7%	39.3	−0.7	−1.7%

In conclusion, however, it is necessary to note also certain unfavorable phenomena observed when using a saline solution.

Salt acts on rubber, and the rubber tubes rather soon begin to crack and tear;
glass stopcocks seize up, and they have to be opened with effort and precautions, while being thoroughly washed with clean water;
salt, crystallizing out, appears through the slightest openings and cracks, for example around a tightly driven-in rubber stopper, and also through the pores of an ordinary wooden stopper; the latter circumstance was especially unpleasant in the aspirator for analysis by combustion, in which glass rods were passed through a large stopper, and the stopper was covered on top with Mendeleev putty. Salt, emerging from the stopper in the form of the finest crystals, somehow passes through the putty and causes leaks in the stopper; true, these leaks are easily eliminated by local heating of the putty, for example by applying a heated iron rod, but they are nevertheless unpleasant, requiring frequent checks.

9. Taking a gas sample.—Above, Figs. 22 and 23, we have already indicated an apparatus for taking samples of combustion products which, being connected to the exhaust at the very cylinder of the machine, ensures obtaining

samples of exhaust gases with a composition not altered by afterburning of the combustible constituents. In order to determine how useful this apparatus is for our investigations, a series of experiments was carried out with it, from which the following became clear:

afterburning undoubtedly occurs and has a strong effect on the composition of the products of combustion taken from the muffler;
the indicated apparatus does indeed make it possible to obtain samples without afterburning;
when working with a two-stroke engine, samples cannot be taken directly from the pipe at the beginning of the exhaust, so that in this case the indicated apparatus is, it would seem, inapplicable.

Proof of proposition 1 is easily obtained by comparing analyses of exhaust gases taken simultaneously by means of the indicated apparatus at the cylinder itself and in the usual way from the muffler, to which a pipe 2 m long leads from the exhaust.

Table 5 gives the most characteristic of our experiments:

Table 5.

Composition	Experiment 1: at the cylinder	Experiment 1: from the muffler	Experiment 2: at the cylinder	Experiment 2: from the muffler	Experiment 3: at the cylinder	Experiment 3: from the muffler
CO₂	7.6	8.0	4.9	5.5	5.0	4.5
CₘH₂ₙ	traces	—	—	—	traces	—
O₂	8.5	8.0	12.7	11.4	13.2	12.9
CO	1.3	1.0	0.4	0.3	0.4	0.2
N₂	82.6	83.0	81.8	82.9	81.6	82.4

In experiment 1 the engine was operating with a deficiency of air; in experiment 2 the nozzle functioned poorly, there was a very large consumption of kerosene, but part of it passed through the engine without taking part at all in the combustion.

In experiment 1, per kg of kerosene there were in all 13.2 m³ of air. Although theoretically only about 11.5 m³ is required, practice has shown that, since the igniting mixture consists not only of air and kerosene vapor, but also contains a considerable proportion of products of combustion, in reality for complete combustion there must generally be no less than 16 and up to 22 m³. As became clear from our later experiments, in this engine it is necessary to have an especially considerable excess of air; complete combustion occurs, and then only under favorable conditions, only at $L/q = 20$, and sometimes even higher.

As can be seen from Table 5, in experiment 1 the composition of the products of combustion in the muffler contains considerably fewer unburned constituents than at the outlet from the engine. The same conclusion can be drawn

and from consideration of the results of experiment 2. The greater decrease in the content of O₂ than can be calculated from the increase in CO₂ must, of course, be ascribed to the formation of H₂O, which, with our method of analysis, using Hempel apparatus with water, cannot be taken into account. This same circumstance also explains the change in the percentage content of N₂, the absolute quantity of which, during afterburning, of course does not change.

The correctness of proposition 2 is clear, first, from the very construction of the apparatus: the length of the path from the exhaust port to the entrance into the cooled part of the spiral of the apparatus is only about 200 mm, which, at a gas-outflow velocity of from 33 to 70 m/sec, corresponds to only 0.006 to 0.003 sec, an interval of time far too short for afterburning to take place to any appreciable extent at the existing pressures, only slightly above 1 atm. abs., and temperatures from 300° to 400° C. Of course, the cooling of the copper spiral by water must still further reduce the possibility of afterburning.

Secondly, the results of the experiments also testify to the same thing: in gases taken with the aid of our apparatus at the very cylinder itself, combustible substances were almost always found, even sometimes when the gases in the muffler did not contain such substances or contained very little of them.

Finally, the correctness of proposition 3 follows from a direct consideration of the phenomenon of exhaust.

In two-stroke engines, during the exhaust period there occurs a change in pressure, which first falls, then rises, and then falls again. Such a wave-like fluctuation of the pressure was known earlier, but was established with particular clarity only by our experiments. Our experiments have proved indisputably that the composition of the products of combustion is different at different moments of exhaust: at first they contain more CO₂, and then they are diluted with air. Consequently, a sample taken from the exhaust pipe itself, and moreover at its beginning, when the gases have not yet had time to mix thoroughly, can by no means be regarded as an average one; in this sample the gas that issues at the moments of the greatest exhaust pressure will predominate, and the gas that issues at moments when the pressure is below atmospheric will be entirely absent. Since the periods of fluctuation of the exhaust pressure depend on many factors, such as the number of revolutions, the degree of compression, the ignition pressure, etc., the moments of greatest pressure also vary, and accordingly the composition of the sample taken near the exhaust must differ more or less from the average.

That this is indeed so was confirmed by a whole series of experiments. As an example, Table 5 gives the results of one, experiment 3; according to this experiment it appears as though, on the way from the cylinder to the muffler, part of the carbon dioxide disappears somewhere; this, obviously, cannot be the case; in this experiment, when the sample was taken at the cylinder, gas with a larger ...

with a lower content of CO₂ than the average sample. Of course, the decrease in the CO₂ content could also have resulted simply from air being drawn into the aspirator, owing to a leak in the bottle or in the cock at the muffler; but then the O₂ content should also have increased, which we did not observe.

It may be noted, incidentally, that in experiment 3 the phenomenon of afterburning is nevertheless revealed: in the sample from the muffler the heavy hydrocarbons disappear and the content of CO and O₂ decreases.

Once proposition 3 has been proved—i.e., that incorrect gas sampling at the exhaust itself, in the case of investigating a two-stroke machine, can occur—the question may arise: was it worthwhile, in view of this, to dwell at such length on these experiments? In our opinion, it certainly was, since although these experiments, like some others, cannot give precise figures, cannot, so to speak, elucidate the phenomenon quantitatively, they nevertheless elucidate it qualitatively, and from two points of view: first, they proved the phenomenon of afterburning in the pipe, proposition 1; and second, they proved afterburning of the composition of the outgoing exhaust gases at separate moments of the exhaust. Subsequently this change in composition will be clarified in greater detail, so to speak quantitatively, by means of figures obtained with the aid of the differential apparatus described above; but this phenomenon was first discovered by us in this way.

In view of these two circumstances, we considered it possible and interesting to dwell in greater detail on these preliminary experiments, despite the fact that they will not figure in the main experiments.

10. Experiments with Junkers’ calorimeter.—Frequently resorting to the Junkers calorimeter, we noticed that the figures for the heat-producing capacity of one and the same sample obtained with its aid sometimes differ rather considerably from one another. On examining the operation of this instrument, one may observe that in it, too, various working conditions can be varied greatly, and these undoubtedly affect its readings. Since in the relevant literature we found no instructions as to how one should work with the Junkers calorimeter in order to obtain the most precise results for liquid fuel, we decided to conduct ourselves a series of experiments whose task was to determine the influence, in the calorimetry of liquid fuel, of the following factors:

the difference in temperature of the incoming and outgoing water, or, what is the same thing, the flow rate of the cooling water;
the air pressure in the lamp, or, what is the same thing, the consumption of liquid fuel per unit time;
the vertical position of the burner in the apparatus;
the position of the throttle valve at the outlet of the burner products.

Study of a Two-Stroke Machine

Our aim was to find the most advantageous conditions for obtaining, as far as possible, accurate results, and also to ascertain what accuracy, in general, may be demanded of this instrument in calorimetry of liquid fuel when it is handled properly.

The results of our experiments are presented in Table 6. In explanation of the table it should be noted as follows: all the experiments were carried out with kerosene taken from one barrel, so that it may be assumed to have been all one and the same sample. Each time 10 g of kerosene was burned.

Table 6.

1 No.	2 Cooling-water consumption $G$, kg	3 Temp. difference $t_2 - t_1\ ^\circ\mathrm{C}$	4 Condensate $g$, g	5 Heat productivity $H_k$	6 Heat productivity $H_n$	7 Heat productivity $H_n'$	8 Pressure $p$, mm	9 Valve position	10 Pr.-valve position
1	3,01	37,12	11,60	11160	10436	10427	275	h.	open
2	7,25	14,97	12,00	10850	10100	10117	500	l.	”
3	7,38	15,29	12,02	11280	10530	10547	550	h.-m. p.	”
4	8,51	12,88	11,49	10960	10243	10227	400	l.	”
5	8,66	12,96	11,50	11220	10502	10487	325	h.	”
6	8,68	12,77	12,05	11080	10328	10345	350	mid.	”
7	10,43	10,59	11,58	11060	10337	10327	450	h.-m. p.	”
8	10,55	10,76	11,44	11360	10645	10627	350	h.	”
9	10,69	10,57	11,58	11300	10577	10567	500	h.-m. p.	”
10	12,89	9,73	12,04	11260	10505	10527	200	”	”
11	6,88	15,62	(11,00)	10750	10063	—	600	mid.	ajar
12	8,84	10,47	(11,40)	9260	8548	—	450	h.	closed
			$\mathrm{avg}_{10}=11,73$	$\mathrm{avg}=11205$		$\mathrm{avg}=10482$

The figures in columns 2—5 of each experiment represent the mean from several individual experiments carried out under identical conditions and giving results close to one another. In view of this, the product $G(t_2 - t_1)$, of the figures in columns 2 and 3, of course, does not give quite exactly $H_k$ in column 5, i.e. the upper limit of heat productivity. As regards the lower limit of heat productivity, or, as we shall call it hereafter, the useful heat productivity $H_n$ in column 6, it was calculated as the mean of the figures of the corresponding group of experiments, found from the corresponding quantities of collected condensed water $g$. Since, for one and the same composition of sample, as in our case, differences in the magnitude of $g$ must be explained exclusively by inaccuracy either of the instrument or of the conduct of the experiment, the true value of $g$ should properly be taken as the mean of the first 10 normal experiments, $g' = 11,73$. To this $g'$ there corresponds the latent heat of vaporization, counting from $+15^\circ$ C and referring it to 1 kg of kerosene,

\[ g'r = g'(i'' - i') = 1,173(639,7 - 15,0) = 733\ \text{heat units}. \]

By subtracting these 733 heat units from $H_k$, the $H_n'$ values of column 7 were obtained. The figures in column 8 indicate the mean air pressure in the lamp in mm Hg. The marks in column 9 denote: “h.” the highest possible

position of the burner, “mid.” = middle, “low.” = lower; in this case the higher position was attained by placing the scales on a wooden support, a board about 50 mm thick; the middle position was obtained with a support about 25 mm thick, and the lower one without any supports. In addition, in the higher position of the burner the experiment was conducted so that for almost the entire time the arm of the balance with the lamp was raised, whereas in the lower position, on the contrary, it was lowered.

Experiments of groups 3, 7, and 9, for which the notation “small swing” appears in the column “notes,” were carried out in such a way that, when weights were placed on the pan in order to arrest it, the balance beam was raised only slightly, so that it could make only a small swing; that is, during the entire experiment the lamp flame was displaced very little in height. The swing with such an incompletely raised beam was 40 mm; normally, however, it is about 60 mm.

The notes in column 10 mean: “open” = open, i.e. the throttle valve turned along the tube; “closed” = closed, and “covered” = partly closed, i.e. turned by 45°.

Already after the completion of the series of experiments summarized in Table 6, during one of the subsequent calorimetric determinations we decided to pay attention to the following phenomenon, which is, of course, familiar to everyone who determines the heat productivity of a liquid fuel by means of the Junkers calorimeter, but on which, as far as is known from the literature, no one had yet dwelt: when, at the beginning of the calorimetric experiment, the pointer of the balance again passes through the middle position and the lamp swings into the upper position, the temperature of the outgoing water immediately rises; then, when the experimenter again moves the lamp to the lower position by placing a 10-gram weight on the pan suspended under the lamp, the temperature of the outgoing water immediately falls by several tenths of a degree.

Wishing to investigate this phenomenon, we carried out a new, additional series of experiments, the results of which are presented in Table 7.

Table 7.

1	2	3	4	5	6	7	8	9	10	11
No.	$G$ kg	$t_2-t_1$ °C	$H_k$	$\Delta t_{\max}$	$\Delta t_{\min}$	$\dfrac{\Delta t_{\max}}{t_2-t_1}$ %	$\dfrac{\Delta t_{\min}}{t_2-t_1}$ %	$p$ mm Hg	burner position	lowering time
13	3,08	36,31	11180	2,27	3,83	6,2	10,5	275	high	slow
14	3,08	35,77	11020	2,32	2,53	6,5	7,1	325	”	rapid
15	3,19	34,80	11100	2,51	2,24	7,2	6,4	400	”	slow
16	3,21	33,33	10700	1,82	2,63	6,1	7,9	325	low	”
17	5,06	20,10	10170	0,69	0,28	3,4	1,4	500	”	rapid
18	5,84	18,65	10900	0,65	0,30	3,5	1,6	570	”	slow
19	5,98	18,54	11080	1,35	0,69	7,3	3,7	400	high	rapid
20	7,13	15,58	11110	1,17	1,03	7,5	6,6	300	”	slow
21	7,71	14,31	11020	0,78	0,42	5,5	2,9	475	”	rapid
22	8,83	12,58	11100	0,85	0,95	6,8	7,5	325	”	slow
23	12,35	8,78	10930	0,97	0,98	11,0	11,2	200	”	rapid

In it are given the results of individual single experiments; altogether a somewhat larger number of experiments was performed, but some of them, carried out under very similar conditions and therefore giving similar results, have been rejected by us, since in general all these experiments were set up only for a qualitative clarification of the question.

The special feature of these experiments is that they were conducted with a fully established temperature state; in this the temperature $t_1$ of the fresh water was constant to an accuracy of $0.01^\circ$, while the temperature $t_2$ of the outgoing water slowly and uniformly fell, owing to the slow decrease in the consumption of kerosene per unit time—the result of the inevitable gradual fall of the pressure $p$ of the air in the lamp. When such a slow fall of temperature had become established, weights were again placed on the pan in such a way that, after the next $1$—$1 \tfrac{1}{2}$ min., the pointer of the scales again passed through the middle position; this moment was recorded; then in some experiments, marked “п.” in column 11 of Table 7, late, i.e. before the end of the experiment, in others, marked “р.”, considerably earlier, the lamp was shifted again into the lower position, placing the corresponding weight on the pan raised under the lamp; this moment was also recorded in the notes. Then, of course, the reverse passage of the pointer through the middle position, the end of the experiment, was recorded; but the readings of the temperature, which after this immediately rose, were continued for a time, until its slow fall again set in.

For greater clarity, instead of numerical records of the readings of the temperatures $t_2$, made every 15 sec., we give them graphically, Fig. 30; time is plotted on the abscissas, and temperature in °C on the ordinates.

A black triangle on each line depicting the course of the temperature $t_2$ marks the moment of placing the weight, i.e. of shifting the lamp downward; a small circle denotes the moment of completion of the calorimetry, i.e. the collection of the cooling water and, consequently, the second shifting of the lamp into the upper position. For some experiments the course of the temperature is represented by a dotted line; this has been done only for clarity, in order to distinguish some lines from others.

It should further be noted that these experiments were conducted with kerosene from another barrel, of somewhat different heat-producing capacity. Then, for the sake of comparing the course of the temperatures, in experiment 15 actually 15 g were burned, and in experiment 16 even 20 g, whereas in the others 10 g were burned; in general it is hardly advisable to burn more than 10 g, since this changes the pressure $p$ of the air in the lamp too much.

As can be seen from Fig. 30, most experiments normally last about 4 min.; an experiment may be prolonged to 7—8 min. when more than 10 g of kerosene is burned, experiments 15 and 16, or when the pressure $p$ in the lamp is low, experiment 23, or, finally, as a result of slower combustion when the burner is clogged.

Since these experiments were conducted exclusively in order to compare the influence of the position of the lamp, we determined only the calorimetric heat output $H_k$, while the quantity of condensate $g$ was not measured.

Drawing 30.

Finally, in explanation of Table 7 one may add the following:
$\Delta t_{\max.}$ in column 5 indicates the difference between the highest temperature of the outflowing water in the given experiment and the mean $t_2$, calculated,

of course, also only from the moment of the first passage of the pointer through the mean position and to the moment of its secondary passage in the given experiment; $\Delta t_{\min}$, column 6, is the same difference between the lowest and the mean $t_2$; in columns 7 and 8 these quantities are given in % of the corresponding $(t_2 - t_1)$. The remaining columns, 1, 2, 3, 4, 9 and 10, are the same as in Table 6.

Let us first dwell somewhat on the indications of experiments 13—23. The upward jump of the temperature when the lamp is shifted to the upper position must be explained partly by a decrease of heat loss by radiation, and partly, so to speak, by the additional heating of the portions of water rising past the neck. In the same way the downward jump of the temperature when the lamp is shifted to the lower position must be explained partly by an increase of heat loss by radiation, and partly by the passage of a certain quantity of water past the neck owing to the downward motion of the lamp.

That this is so is apparent both from Table 7 and from Draw. 30: experiments 16—18 with the lower position of the lamp gave $H_k$ less than $H_k$ in the other experiments; then, in all experiments with a close $G$, the experiments “P.” with late placing of the weight, i.e. with a longer presence of the lamp in the upper position, give a higher $H_k$ than the corresponding experiments with “R.”, i.e. with early release of the lamp and a longer presence of it in the lower position.

Draw. 30 and columns 7 and 8 of Table 7 show that one cannot speak here, strictly speaking, of a settled temperature, required in calorimetry. Nevertheless, except for experiment 21, in which the temperature of the outflowing water was subject to some incomprehensible fluctuation, the change of temperature proceeds everywhere very regularly. All attempts to eliminate the fluctuation of $t_2$ by one means or another gave no positive results.

Numerically, the influence of early lowering of the lamp reduces to a decrease of $H_k$ by 0.5 to 1.5%. Likewise, $H_k$, computed in the form of the mean value over 8 experiments with the high position of the lamp, is obtained as 11068 t. units, as against 11122 t. units if, from these experiments, only the 4 with late lowering of the lamp are taken; that is, again less by 1.37%.

Looking into the course of the temperatures $t_2$, we may note one more peculiarity: despite rather sharp fluctuations, the general direction, determined by the beginning of the experiment, by the triangle, i.e. by the point of shifting the lamp downward, and by the temperature established toward the end of the experiment, in almost all experiments is expressed by an inclined straight line, or by a very flat curve convex toward the axis of abscissas. In this case, the two areas bounded, on the one hand, by the course of the actual change of temperatures, on the other, by this imaginary inclined straight line, and lying, one in the first half of the experiment, above this straight line, the other, in the second half of the experiment, below the straight line, are approximately equal in magnitude, i.e. as though mutually balancing each other.

All this shows that, despite fluctuations of the temperature $t_2$, the mean thermal state of the apparatus may be considered fairly close to the established one, with a uniformly falling temperature as a consequence of the fall of the air pressure $p$ in the lamp.

With regard to estimating the accuracy of the calorimeter readings, the following must first be noted: when the observations are conducted properly—chiefly when the thermometers are read accurately and frequently, in our case every 15 sec.—the calorimeter may give a value of the heat production either close to the true one or, rather, somewhat too low, owing to unavoidable heat emission and possible losses during combustion, but hardly ever too high. It is true that Gramberg¹⁵ states that, according to experiments of the German Physico-Technical Imperial Institute, carried out with calorimetry of pure hydrogen, Junker’s calorimeter gives readings exaggerated by $+0.4\%$. Unfortunately, we have been unable to obtain the report on these experiments, and therefore we find it difficult to explain why this may occur. In any case, it may be assumed with sufficient accuracy that the operating conditions of the calorimeter are most correct when the apparatus gives the largest readings. An exception is the case indicated below, when, at a very low temperature of the cooling water, much air is passed through, leaving with a temperature considerably below the temperature in the room; this case can always be eliminated.

Proceeding from the considerations indicated above, on the basis of our experiments summarized in Table 6 and partly in Table 7, we can answer the questions posed above in the following way:

The accuracy of the apparatus depends almost not at all on the flow rate of the cooling water within the limits of its possible variation, i.e., it depends almost not at all on the temperature difference of the incoming and outgoing water. This is seen, for example, from comparison of experiments 1 and 3, on the one hand, and 5–10, on the other, or 13–15 and 19–23; all these experiments were conducted under otherwise equal conditions, if one does not count the pressure in the lamp, which, as will be seen below, is of no great importance. The change of $G$ from 12.89 kg to 2.95, i.e. by more than 4 times, remains without influence on $H_k$; this speaks in favor of the stability of the apparatus readings. Theoretically, one should expect an increase of $H_k$ with an increase of $G$, since with an increase in the water flow rate its velocity of motion also increases, which, as is known from numerous experiments¹⁶, favors the exchange of heat between gases and water; in addition, when the flow rate increases, the temperature difference increases, and this, as is known, causes an increase of the heat-transfer coefficient; finally, when the flow rate increases, other conditions being equal, the temperature of the outgoing water decreases, and, consequently—

...and the inevitable loss to radiation; the absence of the effect $G$ indicates a large reserve in the surface of contact between the gases and the water;

the air pressure in the lamp also has no substantial significance; this is evident, for example, from a comparison of experiments 1—3, 7—8, 9—10. From the theoretical point of view one might have expected better results at higher pressure, since in that case the same quantity of kerosene burns more quickly, the velocity of the products of combustion is greater, heat transfer is consequently better; in addition, the whole experiment proceeds faster and the loss to radiation is therefore smaller. In practice, however, one must recognize as most suitable for kerosene a pressure of about 400 mm Hg; higher pressure has the following inconveniences: the lamp burns unevenly, evidently because of the excessive velocity of the kerosene; slight clogging of the nozzle occurs of itself; moreover, it is much more difficult to attain proper tightness in the joints; as a result of this an experiment begun at a pressure of 600 mm usually ends already only at 500—450 mm, whereby the whole process somewhat loses the character of a fully established state, which is incomparably easier to approximate at a pressure of about 400 mm.
the position of the lamp in the apparatus is of substantial importance: the higher the flame is situated, the farther from the lower edge of the flame tube, the greater the readings; this is clearly seen when comparing experiments 1—2 or 4—5. This may be explained as follows: although, with the removal of the flame deeper into the apparatus, the length of the path and the time of contact of the hot products of combustion and the water decrease, the length of the path in the apparatus is evidently taken with such a reserve that the indicated decrease has no effect on the absorption of heat; on the other hand, the loss of heat by radiation from the flame itself is considerably reduced. In view of this, in the subsequent calorimetry we first used the support under the scales mentioned above, and later simply somewhat increased the length of the tube to the end of which the burner is attached at the lamp.
the position of the throttle valve, at least in the calorimetry of liquid fuel, has no great significance until, however, as a result of excessive closing of it, the inflow of air becomes insufficient, as, for example, in experiment 11, and especially 12 of Table 6. In this case products of incomplete combustion are obtained, the formation of which, fortunately, is accompanied by a characteristic odor, especially when kerosene is being burned.

In general, in order that the readings of the calorimeter be correct, one must see to it, by slightly closing the throttle valve, that the thermometer located in the exhaust channel shows a temperature of the products of combustion close to the temperature in the room; if their temperature is lower, which is often observed at a very low temperature of the...

V. L. Mal’ev.

of the entering cooling water, the readings $H_k$ will prove to be exaggerated, owing to the heating of the water partly at the expense of the heat of the air. In view of this, for precise experiments it is better not to use the water main directly, but to take the water from a special tank, where the water assumes room temperature.

Such a tank is also very useful because, with it, the water supply is very constant, which cannot be said of the municipal water main with its variable pressure.

In conclusion we shall also point out that, on the basis of the experiments cited, not only have the best methods of calorimetry of liquid fuel by means of Junkers’ apparatus been clarified, but it is also possible, to a certain degree, to establish the accuracy of its readings.

Indeed, if we take experiments 1, 3, and 5–10 as sufficiently correct and use, as more correct, column 7, then we obtain the mean calorific value $H'_n$ of the given sample as 10472 t. u. Calculating, by the method of the squares of the differences, the error of the mean value, we obtain for it $\pm 38$ t. u., i.e. $\pm 0.36\%$; for the 8 experiments of Table 7 we have $\pm 27$ t. u., or $\pm 0.26\%$. For the sake of reliability, let us take the round figure for the accuracy as $\pm 0.4\%$.

As regards the accuracy of determining the mean value of the quantity of condensate $g = 11.73$ g, calculating its error by the same method, we find it equal to 0.095 g, or $\pm 0.81\%$. If we assume that in practice one is limited to measuring the condensate with a graduated cylinder, i.e. with an accuracy of approximately 0.5 g, and accordingly round the values for $g$, then this does not affect the mean value of $g$: for the same 8 experiments we obtain $g' = 11.73$ g.

Of course, such agreement is accidental; nevertheless it shows that, with a sufficiently large number of determinations of $g$ by means of a graduated cylinder, the mean value will indeed be sufficiently accurate.

An error of 0.095 g corresponds to 6 t. u., or, relative to $H'_n$, only $\pm 0.06\%$. The total error is therefore

\[ \nu = \sqrt{(0.38)^2 + (0.06)^2} = \pm 0.38\% \]

or, as we have already said above, we shall take the round figure $\pm 0.4\%$.

Finally, if we assume that the Junkers calorimeter also gives, for liquid fuel, readings exaggerated by 0.4%, in accordance with the above-mentioned experiments of the German Physico-Technical Institute, then the accuracy of determining $H'_n$ may still be considered about $\pm 0.8\%$. If the indicated exaggeration actually remains in force also for liquid fuel—about which, unfortunately, not having a report on these experiments, we cannot judge—then the corresponding correction may, and even should, be introduced; and then the accuracy of the apparatus may be taken as $\pm 0.4\%$, which is quite sufficient not only for technical purposes.

On the other hand, summing up our experiments with the calori-

with Junkers’ meter, it must be admitted that, although we did manage to find more correct conditions of operation with this instrument, deviations from them, within the limits usual in practice, give values of $H_n$ differing from the true ones by no more than 4%, which for most practical cases may be considered satisfactory.

This property—the comparatively small dependence on the methods of operation and, so to speak, the universality of Junkers’ calorimeter, which can be used for almost any liquid and gaseous fuel—explains the wide spread of this instrument in practice and the gradual displacement of other calorimeters for gas; for liquid fuel, as far as we know, it is even the only instrument.

Conduct of the Main Experiments

11. General indications.—Factors varied. As has already been said above, the principal tasks of the experiments were the study of the scavenging phenomenon in the working cylinder and of the character and magnitude of the pump work. Proceeding from this assignment, in carrying out the individual series of experiments only those factors were varied, in order to ascertain their influence, which, as could be foreseen, have a predominant influence on the indicated aspects of the machine’s operation.

These factors are:

the number of revolutions $n$ of the machine, of interest insofar as, with its variation, the velocities of the air and of the exhaust gases change in passing through the corresponding ports;
the volume of the rear cavity $V_z$, i.e. the harmful space of the pump, of interest both in the sense of changing the degree of delivery of the pump and, chiefly, in the sense of changing the pressure of the scavenging air and, consequently, the initial velocity of its flow into the working cylinder;
the thickness of the gaskets $r$ under the connecting rod and $s$ under the cylinder cover, of interest partly in the sense of changing the compression ratio and the associated expansion pressure at the moment preceding exhaust, and chiefly in the sense of changing the magnitude and time of opening of the ports, i.e. again from the point of view of the velocities of the air and gases;
the air pressure in the large vessel $c$, obtained by forcing it in there with the aid of an air blower, of interest in the sense of changing the delivery of air, the velocities and the pressure of the scavenging air;
finally, the use of a check valve in the pump, from the point of view of its influence on the degree of delivery.

The indicated 5 factors we tried to vary within the widest possible limits and systematically.

True, in our experiments other factors also varied, such as: the load of the engine, the temperature and quantity of the cooling water, the relative number of misfires, and certain others; however, these factors were of only secondary importance for us, varied not within wide limits, and in part even beyond the will of the experimenter.

The beginning and duration of individual experiments. The experiment itself, or rather the production of the readings and observations, should begin only when a fully established state has set in. In internal-combustion engines, in contrast to steam engines, which require prolonged warming up, the established state comes rather quickly—20–30 minutes after the machine has received the corresponding load. A sure sign of the established state is the constancy of the temperature of the outgoing cooling water. This same sign is also indicated in Art. 7 of the rules for testing internal-combustion engines established in 1906 by the Society of German Engineers.²⁵ Of course, the remaining readings, such as, for example, the temperature of the combustion products, the color (temperature) of the ignition cone, etc., must also become established, but this usually occurs still sooner.

As regards the duration of an individual experiment, according to Art. 7 of the rules mentioned, for determining the fuel consumption the experiment should last, at full load, about 1 hour; and in the case where, in addition, the consumption at a lower load is also determined, then still less. In the case where the aim of the test is the determination of the mechanical output, the experiment, according to Art. 8 of the same rules, may last still less, provided only that 10 sets of indicator diagrams are taken, for which 15–20 minutes are sufficient. Thus, with a fully established state and with the production of numerous readings mutually checking one another, a half-hour duration of the experiment may be considered sufficient. The same opinion is held, for example, by such an authority on the given question as Prof. R. Mollier, who expressed this personally to the author; confirmation of this opinion may be found in printed works by his pupils, for example, Herberg²⁶; another of his pupils—Nagel, who himself has already become an authority on the question of testing internal-combustion engines, limits the duration of the experiment even to 20 min.²⁷

Load. Since we set ourselves the task of investigating the phenomena occurring inside the cylinder, the factor which, obviously, must—

²⁵ Z. V. d. I. 50, 1906 S. 1923.
²⁶ Gasmot. 3, 1904 S. 185–187.
²⁷ Z. V. d. I. 51, 1907 S. 1460.

...should remain constant, so as not to obscure the influence of other factors—the load of the machine, more precisely the indicated work developed by it; referred, of course, to one revolution.

As is known, it is rather difficult to determine in advance the magnitude of the indicated work of an internal-combustion engine; it depends on many accidental factors not dependent on the experimenter, such as the speed of ignition, the proper functioning of the nozzle, water cooling, etc. In view of this, although it was known in advance that the magnitude of the mechanical efficiency $\eta_m = N_e/N_i$ in the machine would fluctuate, and might do so rather considerably, it was necessary to rest content, in the tests, with maintaining constant the actual, or useful, work $N_e$, determined from the load on the brake. As regards the magnitude of this load, it was natural to choose the so-called “normal” load, i.e., that indicated by the factory when selling the machine.

Our machine is rated at 10 horsepower at 340 rev./min. Since the useful power when working with the Prony brake is expressed in the form:

\[ N_e=\frac{GL\pi n}{2250}, \tag{2} \]

or, with the length of the lever arm $L = 0.909$ m, more simply, with an accuracy to $0.06\%$,

\[ N_e = 0.00127\, nG, \tag{3} \]

then, with $N_e = 10$ and $n = 340$, we obtain $G_0 = 24$ kg. This load, or one close to it, $G = 25$ kg, more convenient for selecting weights on decimal scales, was the one we maintained in the majority of the tests. A smaller load, $G = 23, 22, 20$, and even 15 kg, was applied only a few times, either for comparison or in those cases when, for some reason, the machine could not carry the normal load. A larger load, $G = 30$ and 31 kg, was applied for comparison and only twice in all.

Of course, by $G$ we mean the net load on the pan of the scales; in order to balance the dead weight of the lever of the brake, the movable weight on the scales always stood at a division of 1.53 kg.

In order that the load should actually be constant all the time and precisely known, a special person watched the brake continuously during the test and, as the belt stretched, gradually tightened it, making sure that the tongue of the scales always stood, or rather oscillated, opposite the pointer.

Setting the misfires. At the beginning of each test, when the required number of revolutions had been established with the aid of the governor and the machine had been loaded by the brake, the number of misfires in 1 min. was set by turning nut $o$, drawings 1 and 2.

Theoretically, bearing in mind the investigation of the processes inside the working cylinder, it is desirable that all the indicator diagrams remain...

would be identical, i.e. so that the kerosene feed would be set in such a way that there was complete equilibrium between the load and the indicator work developed in the cylinder; in other words, so that there would be no misfiring at all.

However, in practice, in the absence of any action by the governor on the quantity of kerosene delivered by the pump in one stroke, operation without misfires would give an unstable equilibrium: it would suffice for the explosion in the cylinder, or the circumferential force on the brake pulley, or, finally, the resistance to the motion of the piston, to change even slightly, and the equilibrium would be disturbed; the engine would change its number of revolutions in one direction or the other, or would even stop altogether.

For our experiments, constancy of the number of revolutions during the test was especially important; it was therefore necessary to leave the engine under the action of the governor and to maintain some, if possible small, number of misfires. Depending on the constancy of the working conditions—chiefly the nature of the change in the resistances to the motion of the piston, the power reserve, etc.—the number of misfires was, in our case, from 3 to 7, sometimes up to 10% of the number of revolutions.

Operation with an air-blower. In work with an air-blower, special precautions were required in order to obtain a steady state. The experiment was conducted as follows: first the required number of revolutions of the engine was set; then the electric motor with the air-blower was started and, gradually closing the slide valve $i$, Fig. 8, the pressure in vessel $c$ was brought to the desired height; only after this was the brake applied and the required load given.

12. Measurement of various quantities.—Recording of indicator diagrams. A rather important question when taking diagrams is what paper to use: the so-called indicator paper, coated with a thin layer of lead salt (white lead), and a brass stylus, or ordinary writing paper and a graphite pencil? The advantage of the brass stylus is that it does not become blunt so quickly; but, on the other hand, the leaded paper gives greater friction when the stylus moves over it, owing to the looseness of the upper layer, and, under sharp pressure oscillations, it tears easily even with weak pressure. This was observed more than once in our first experiments, especially when the so-called premature explosions appeared. In view of this, in the subsequent experiments the diagrams from the working cylinder were taken on dense writing paper, the so-called office paper.

The actual recording was carried out as follows: every 5 minutes two diagrams were taken on one and the same sheet, one below the other, with a stiff spring with $m = 3,\ 3.5$, or $4$ mm, depending on the explosion pressure; and immediately afterward another above them, after shifting the paper, with a weak spring $m = 8$, sometimes $10$ m.

Despite the not entirely satisfactory superposition of the individual outlines, i.e., the so-called breaking up of the diagram, we ordinarily traced only 10 outlines, since otherwise diagrams are obtained that are too difficult to planimeter.

The described method of recording, on the one hand, speeds it up and, on the other, by not darkening the diagrams, makes it easier to compare the working diagram with the charging diagram.

A sample of the diagram obtained—so to speak, a facsimile of one of the diagrams of experiment No. 27—is presented in Drawing 31.

Drawing 31.

Immediately after recording such a pair of diagrams, displaced diagrams were recorded once—also in pairs, with a stiff and a weak spring, Drawing 32; and the next time, after 5 minutes, diagrams of the air pump. Thus both sets of diagrams were recorded every 10 minutes, which is quite sufficient both for the displaced diagrams of the working cylinder and for the pump diagrams, which, once the operation has become steady, remain identical.

The diagrams from the air pump were recorded on indicator paper. In view of doubts regarding the waves appearing on these diagrams—that is, whether they represent actual pressure oscillations or occur under the influence of the inertia of the masses of the indicator—alongside the recording of diagrams with the weak spring, $m = 160$ mm, diagrams with a stiffer spring, $m = 60$, or 80, or 100 mm, were recorded on the same sheet fairly often, though not in every experiment, in view of the rather considerable time required to change the spring. The comparison of diagrams taken with different springs was to serve,

as for checking the magnitude of their area, i.e., the accuracy of determining the work of the pump, and for clarifying the nature of the separate lines.

Drawing 32.

However, as subsequently became clear, the influence of the inertia of the moving masses of the indicator, taken into account by A. Fliegner’s method¹⁷, proved so considerable that all diagrams of the air pump had to be corrected accordingly, as will be indicated in more detail below, Part II.

A facsimile of such a diagram is given in Drawing 33.

Sometimes displacement diagrams were also taken from the air pump; at first no great importance was attached to them, but later they proved very useful in calculating corrections for the influence of the moving parts of the indicator.

Finally, at the beginning and at the end of each experiment displacement diagrams were taken at a scale of $m = 30$ mm from the silencer; the taking of normal diagrams from it is of no interest, in view of the character of the phenomenon, which occurs only at the moment when the piston is near the dead point.

A sample for the same experiment No. 27 is presented in Drawing 34, p. 64.

In conclusion, it should be added that all springs used were checked from time to time in their own indicators with the aid of a glycerine press, a modified Ruchholz instrument¹⁸, by direct loading with cast-iron weights. From the records obtained, the average scale of the spring was calculated for all diagrams, except

Drawing 33.

...of complete normal diagrams of the working cylinder, according to the formula

\[ m=\frac{s}{p}, \tag{4} \]

where $s$ is the path in mm traversed by the pencil under the action of the load

$p$ in kg/cm$^2$, which was chosen either equal to or the nearest greater value to the greatest pressure occurring in the given type of diagram. Of course, the value $s$ was taken here as the mean found from a threefold loading and unloading.

Drawing 34

Drawing 34.

The scale of the springs at 3, 3.5, and 4 mm for the working diagrams was determined by E. Meyer’s method, i.e., by the formula

\[ m=\frac{\sum s}{\sum p}, \tag{5} \]

where $\sum s$ is the sum of the successive paths of the pencil, during increases and decreases of load, all measured from the atmospheric line, and $\sum p$ is the sum of the corresponding loads¹⁹. Of course, $\sum s$ is also a mean value from several, 3 to 7, tests.

It is evident that, by means of formula (5), the influence of the scales in the lower pressure stages is taken into account more strongly, as should be done for calculating the area of the working diagrams of internal-combustion machines, in which the area of the lower part of the diagram is greater, and considerably so, than the area of the upper part.

Since the springs were new, and until then had hardly been in operation, the scales obtained differed comparatively little from the nominal ones given by the firm. Thus, for example, the spring with nominal $m_0=3.0$ gave $m=3.11$ mm; with $m_0=3.5$ it gave $m=3.62$; and with $m_0=4.0$, only $m=4.01$ mm. But, what is most important, repeated tests after several days, and then after several weeks of operation, gave the very same values of $m$. This is explained, on the one hand, by the good quality of the springs, and then by the comparatively easy working conditions of the springs—they remained

always quite cold, and, finally, the short duration of the individual experiments, i.e. the small number of diagrams for each experiment.

To estimate the accuracy of the quantities found from the indicated $m$, we computed for them, by the usual formula from the squares of the differences, the value of the mean error of the mean quantity $\nu$. For the different springs of the working diagrams, $\nu$ varies from $\pm 0.13$ to $\pm 0.34\%$.

For the springs of the indicator at the air pump, the nominal scales are: $m_0 = 160$, giving $m = 144.8$ mm; $m_0 = 80$, giving $m = 78.20$; $m_0 = 60$, giving $m = 58.8$ mm. The greatest mean error $\nu$ proved to be for the softest spring, with $m = 144.8$, and is equal to $\pm 0.46\%$.

In view of the indicated observed constancy of the $m$ found, their verification was carried out considerably less often than is usual when working with internal-combustion engines.

Barometric pressure was measured, before the beginning of each experiment, by means of the mercury barometer mentioned above. In view of the short duration of an individual experiment, checking the pressure after the end of the experiment was done very rarely. In no case did the pressure change appreciably during an experiment.

The number of revolutions was measured by the counter described above, connected to the shaft of the machine, readings being taken every 5 minutes with an accuracy of up to 1 sec. In the intervals the constancy of the number of revolutions was also monitored by tachometer.

The number of misses was measured with the aid of a stopwatch in the following manner: the stopwatch was started with the first miss, which, however, was not included in the count; then the misses were counted during an interval of about 1 minute, and with the last counted miss the stopwatch was stopped, and the number of misses and the corresponding number of seconds were recorded. This counting was done every 5 minutes, as were, in general, most of the other observations.

Although, when using the stopwatch, the eyes are not occupied with observing the watch hand and can watch the kerosene pump, it is very difficult to keep track of misses by eye with accuracy at a large number of revolutions and especially with any considerable number of misses. It is much easier to count the misses by touch, placing the index finger of the right hand under the pawl at the throw-off blade $n$, Figs. 1 and 2: during a miss the blade $n$ bends downward, striking the finger quite perceptibly.

Measurement of temperatures. All temperature readings at the machine, i.e. of the air in the rear cavity and in the scavenging channel, of the cooling water entering and leaving, and of the combustion products, were made every 5 min. The temperature in the room, i.e. of the air before the air—

hours, was measured before the start of the experiment; sometimes in winter, after an hour’s operation, it rose by 0.5 to 1°.

In addition to recording the readings of the mercury thermometers, it was always noted, in order to calculate the correction for the protruding column of mercury, to what division the corresponding thermometer had been inserted.

Great attention was paid to checking the thermometers. All mercury thermometers were checked by comparison with standards supplied with certificates from the German Physico-Technical Imperial Institute.

The comparison of thermometers indicating up to $+80^\circ$ was carried out by heating in a water bath.

The thermoelement used to measure the temperature of the combustion products was checked in the same form in which it stood during the experiments, i.e., together with the millivoltmeter and the soldered connecting conductors, whose resistance, however, was negligible. The check was made at fixed points: the melting of ice ($0^\circ$), the boiling of water (about $100^\circ$)²⁰, with, of course, a correction being made for the deviation of the barometric pressure from 760 mm²¹, the boiling of mercury (about $357^\circ$) and the boiling of sulfur (about $445^\circ$), both likewise with corrections for barometric pressure.

The mutual check of the correctness of the readings of the thermoelement and the mercury thermometer when measuring the temperature of the exhaust gases has already been discussed above, p. 30.

Radiation. In order to obtain an accurate heat balance, it seemed of interest to us to try to determine the heat lost by the engine through radiation. The magnitude of this loss obviously depends on the difference between the temperatures of the cylinder and the head, on the one hand, and that of the air in the room, on the other. Since the latter temperature was in any case measured, it was necessary additionally to measure only the temperatures of the cylinder and the head. However, through a series of measurements it was established that the temperature of the cylinder walls in the lower part is very close to the temperature of the incoming cooling water, and in the upper part to the temperature of the outgoing water. Thus the mean temperature of the cylinder walls may simply be taken as equal to the arithmetic mean of the temperatures of the incoming and outgoing cooling water.

Only the temperatures at various points of the head had to be measured directly; for this purpose, a mercury thermometer was applied to the corresponding point, covering and pressing it with a piece of asbestos cardboard.

As was to be expected, the temperature of the head proved to be highest at the very end, opposite the ignition ball, gradually decreasing in the direction toward the cylinder.

In view of the fact that the temperatures measured at various points of the head remained very constant during each experiment, depending, apparently, chiefly on the consumption of kerosene; further, in view of the known conjectural character of the formulas by which the magnitude of the loss through heat radiation had to be calculated; and, finally, in view of the comparatively small magnitude of this loss in the overall heat balance, the corresponding temperatures were measured only at the beginning and at the end of an experiment, and sometimes even only once, and not in all experiments.

Consumption of kerosene. When the operation of the machine had been adjusted and steady motion had set in, the point of the wire was brought to the indicating glass of the kerosene tank so that the point was slightly immersed. After a few seconds it became exposed, and at the moment when the lowering level of the kerosene separated from the point, the stopwatch catch was pressed. Then, into a small bucket, a certain quantity of kerosene was weighed out. Before the end of the experiment, i.e. after 25–30 min., kerosene was poured into the supply tank from this bucket until the point again was slightly immersed in the kerosene; and the moment of the second separation of the level was watched for, pressing at that instant the stopwatch catch, the hand of which stopped and indicated exactly, to 0.2″, the time during which a quantity of kerosene had been consumed equal to the difference in the weight of the bucket with kerosene before pouring it in and after it.

The second weighing of the bucket, with the remaining kerosene, was usually carried out after the experiment had already ended, so as not to be mistaken in the rush caused by the need quickly to take samples with the aid of the differentiator, remove the last diagrams, and make several final readings.

The moments of the beginning and end of the kerosene measurement were at the same time the moments of the beginning and end of the experiment; but since the other readings were begun in advance and were also finished somewhat later, for the calculations the periods of the readings were taken from the last corresponding reading before the beginning of the kerosene measurement and inclusively to the first reading after the end of the kerosene measurement; that is, time periods of individual readings several minutes longer than the duration of the kerosene measurement. With steady motion this is immaterial and rather even desirable, since it increases the accuracy of the mean values.

Air supply. Readings were made on the gas meters every 5 minutes, with an accuracy of up to 1 sec. Strictly speaking, it is sufficient to make these readings only twice, at the beginning of the experiment and at its end, noting precisely the interval of time between these readings.

Readings every 5 minutes are of interest from the standpoint of checking the completeness of the establishment of steady motion, and also in the respect that an accidental error in the first or last reading should not spoil the whole experiment. With readings every 5 min., it is easy to discover an error in these readings and to correct it, taking as the beginning of the experiment either the preceding reading or the following one. An error in one of the intermediate readings is of no significance.

As already mentioned, during the experiments, in order to maintain the water level at the required height, water issuing from the overflow in large drops was always passed through the gas meter.

Water consumption. The quantity of cooling water was determined by weighing in the usual manner: first the tare of the tank with the remaining water in it was determined; more exactly, the latter was first counterbalanced by placing on the pan of the decimal scales a cast-iron plate with shot; the discharge cock of the tank was then closed, and at the beginning of the experiment the end of the hose hanging over the sink was shifted so that the water began to flow into the tank.

After 5 min. had elapsed, with an accuracy of up to 1 sec., the hose was shifted back so that the water flowed directly into the sink. The quantity of water that had run in was weighed, as quickly as possible, in 1–1½ min., sometimes even in ½ min., and the hose was again shifted to the tank. Exactly after 5 min. the hose was again shifted, the tank weighed, and so on. Such weighing every 5 min. had, of course, solely the purpose of checking the steady motion.

Taking gas samples. Samples of the exhaust gases, by means of the differentiator, were taken either at the beginning of the experiment or immediately after its end, as already described above.

Samples for determining the average composition of the exhaust gases and for carrying out the analysis by combustion were taken from the first muffler into the 15-liter aspirator described below, with special divisions. In order that the composition of the gases should be as close as possible to the average, the withdrawal was carried out gradually throughout the entire experiment.

In the above-mentioned taking of samples from the exhaust pipe close to the cylinder itself, so that the composition should be as close as possible to the average, the sample taken into a small aspirator of 2–5 liters was likewise taken not all at once, but gradually, in several portions throughout the entire experiment.

13. General survey of the conduct of the experiments.—In conclusion, it may perhaps not be without interest to indicate how many persons took part in our experiments, and how the duties were distributed among them.

In all, usually 4 persons participated: one (the machinist’s assistant) stood at the brake and watched the scales, periodically tightening the brake as the belt stretched; he also watched the lubrication of the machine, filling the oilers as necessary, and helped in shifting and moving apart ...

indicators, in addition to taking the normal working diagrams; in this case both the indicator and the drive were at the disposal of the person taking the diagrams; the 2nd person measured the consumption of kerosene and cooling water, took gas samples for analysis of the mean composition and combustion, and assisted the author in taking samples by means of the differentiator; the 3rd person made readings from the revolution counter, of all the temperatures at the engine, i.e., from 6–8 instruments, and readings from the air clocks; the 4th person—the author himself—took all the indicator diagrams and counted the number of misfires; in addition, of course, he also watched the general progress of the test and made all the remaining, non-periodic measurements, such as: of the barometer, the temperature in the room, the temperature of the head for calculating the loss by radiation, watched the constancy of the supply of cooling water, regulating it with a cock according to the thermometer and 5-minute readings of its consumption, took samples of the combustion products with the aid of the differentiator, measured the consumption of cylinder oil and of the water injected into the cylinder in the corresponding tests, set the air blower in operation, watched the overflow of King’s meter at the air clocks, etc.

As we see, each person had more than enough work to do. If, during the test itself, something went out of order—for example, the indicator cord broke, or the pencil broke, or the drive at the tachometer jumped off—then, although the repair proceeded very quickly as a result of the acquired skill, the author often had to shift the next corresponding observation somewhat instead of making it exactly after an interval of 5 minutes. However, with steady motion this was of no essential importance.

In view of the incompletely satisfactory operation of the engine, chiefly in the sense of the piston seizing, about which more is said below, after almost every test the cylinder was inspected, the piston and its rings being removed and cleaned. Although by this the engine was placed under conditions differing from practical operation, on the other hand all the tests proceeded under more or less identical conditions, i.e., the very undesirable influence of a variable mechanical efficiency was weakened.

Finally, for completeness, Tables 8–13 present specimens of all the records that were kept during the test itself, and in the form in which they were actually made by us.

Each record, each table, was kept on a separate sheet. The entries that were made during the test itself are distinguished in the tables by italic type; the remaining data, obtained after the test in processing it, are shown in ordinary type.

Table 8

Number of revolutions. Experiment No. 32. 7.8.08

time	counter	difference	tachom.
11 h. 41 m.	2523		290
46	3994	1471	290
51	5467	1471	290
56	6939	1472	290
1 h. 1 m.	8416	1477	290
6	9889	1473	290
11	1264	1375	290
16	2734	1470	290
21	4204	1470	290
26	5663	1459	290
31	7125	1462	290

\[ \begin{aligned} 15663\\ -\;3994\\ \hline 11669\\ +\;\text{correction }100\\ \hline 11769 \end{aligned} \]

\[ \frac{11769}{40}=294.2\ \text{rev./min.} \]

Table 9

Temperatures. Experiment No. 32. 7.8.08

time	water, cold $t_5$	water, hot $t_6$	air 1 $t_2$	air 2 $t_3$	exhaust gases, mercury $t_4'$	exhaust gases, thermoelem. $t_4''$	exhaust gases, cold junction $\Delta t_4''$
	$0^\circ$	$+7.0^\circ$	$-11^\circ$	$+2^\circ$	$+200^\circ$	—	—
11 h. 42 m.	19,3	51	46,5	87	415	408	20,5
47	19,3	51	46,7	87	416	410	20,5
52	19,3	51	47	87	414	408	21,0
57	19,3	51,5	47	87	413	407	21,0
1 h. 2 m.	19,3	51	47	87	410	402	20,5
7	19,3	51	47	87	413	405	22,0
12	19	51	47,5	87	416	405	22,0
17	19	50,5	47,7	87	413	407	22,5
22	19	51	47,7	87	408	400	22,5
27	19	51	47,7	87	408	398	23,0
average	19,18	51,00	47,18	87,00	412,5	404,7	21,65
corrected	19,2	51,2	47,4	87,9	425	426	—

\[ \Delta t_5=0 \]

\[ \Delta t_6= \frac{(51-7)(51-22.5)}{6400} = \frac{44.285}{6400} =0.20^\circ; \]

\[ \Delta t_2= \frac{(47+11)(47-22.5)}{6400} = \frac{58.24.5}{6400} =0.22^\circ; \]

\[ \Delta t_3= \frac{(87-2)(87-22.5)}{6400} = \frac{85.64.5}{6400} =0.86^\circ; \]

\[ \Delta t_4= \frac{(412.5-200)(412.5-30)}{6400} = \frac{212.5.382.5}{6400} =12.7^\circ; \]

Table 10.

Air hours. Experiment No. 32. 7.8.08

time	reading	difference
12 h. 40 m.	2949,96	5,76
45	55,72	5,82
50	61,54	5,84
55	67,30	5,84
1 h. —	73,22	5,86
5	79,07	5,82
10	84,89	5,80
15	90,69	5,81
20	96,50	5,76
25	3002,26	5,71
30	07,97

\[ 3002,26 - 2955,72 = 46,54 \text{ in } 40 \text{ min.} \]

\[ \frac{46,54 \times 60}{40} = 69,81 \text{ m}^3/\text{hour.} \]

Table 11.

Kerosene consumption. Experiment No. 32. 7.8.08

start time	duration, min.	pail weight, kg	difference	consumption, kg/hour
12 h. 45 m.	38′40,8″ =	3,92	2,25	3,492
1 h. 28 m.	= 38,68′	1,67

From barrel No. 1.

\[ \delta = 0,820 \text{ at } 22,0^\circ. \]

Table 12.

Cooling-water consumption. Experiment No. 32. 7.8.08

time	difference	weight, kg	weight difference	consumption, kg/hour
12 h. 50 m. —			27,2	333,2
55 m. —	5	27,20	27,1
55 m. 30 s.
1 h. — 30 s.	5	54,30	27,2
1 m. —
6 m. —	5	81,50	28,0
6 m. 30 s.
11 m. 30 s.	5	109,50	28,2
12 m. —
17 m. —	5	137,70	28,6
17 m. 30 s.
22 m. 30 s.	5	166,30	17,0
23 m. —
26 m. —	3	183,30	(28,3)
total. . .	33 min.		183,3

Table 13.

Number of misfires.	Number of misfires.	Number of misfires.	Experiment No. 32.	7.8.08.
time	duration, sec.	number of misfires	$B = 746.7$ mm.	head $t$: $317^\circ \ \\| \ 208^\circ$
12 h. 44 m.	49	10	$t_1 = 22.5^\circ$	$320^\circ \ \\| \ 209^\circ$
47	67	10	$V_3 = 88$ liters (both vessels).
53	49	10	gasket 10 mm.
56	52	10	$G = 25$ kg on the brake.
1 h. 2 m.	51.6	10	dilution at the clock 3–4 mm, at the upper vessel 0–8 mm.
8	44	10	flash—rather moderate, normal.
12	38	10	ignition ball black; glowing below.
20	47.6	10
24	50	10	in 1 min. $\dfrac{90 \cdot 60}{448.2} = 12.5$ misfires.
total	448.2	90

14. Gas analyses.—First of all it may be noted that we tried to carry out all the analyses as quickly as possible, immediately after the conclusion of the experiment, since, in conducting the preliminary experiments described above, we had become convinced of the rather considerable absorption of gases by the sealing liquid of the aspirator, even when not pure water was used but water saturated with common salt.

Immediately after the end of the run with the machine, or of the experiments if several had been made in succession, we proceeded to the volumetric analysis of the exhaust gases with the aid of Hempel pipettes. This method was preferred to the Orsat apparatus, which is more commonly used in technical investigations, partly because of its greater accuracy, and partly because, for the absorption of oxygen, we used not pyrogallic acid but phosphorus. The point is that phosphorus gives more reliable results, since it allows one to judge precisely the end of the reaction by the cessation of the formation of a white smoke issuing from the pieces of phosphorus so long as the gas still contains even a small amount of oxygen. The inconvenience in using phosphorus is that it ceases to act in the presence of the slightest amount of heavy hydrocarbons. To absorb the latter, the gas was passed into a pipette filled with fuming sulfuric acid, which cannot be done in the ordinary Orsat apparatus with three vessel-pipettes.

After the volumetric analyses of the samples taken with the aid of the differentiator and of the mean sample had been carried out, we proceeded to the

to the analysis of an average sample by means of combustion. The purpose of the latter is to determine the amount of hydrocarbons, carbon monoxide, and hydrogen contained in the exhaust gases, i.e., to ascertain the loss in the engine due to incomplete combustion of the working mixture.

So far as is known from the literature, the determination of the loss from incomplete combustion in tests of internal-combustion engines has been carried out very rarely. Therefore, for the corresponding analysis there are neither strictly defined instructions nor even fully developed instruments. Using information on the corresponding experiments of F. Haber ^33), E. Meyer ^34), and F. Geiser ^35), the author, after several preliminary experiments, developed the following method, in which all measures are taken so that an experiment lasting several hours may proceed as uniformly as possible and not require overly attentive, tiring observation.

The entire apparatus is shown schematically in Fig. 35: $a$ is an aspirator, from which the gas under investigation issues under a certain pressure, passing first through the drying column $b$ with $\mathrm{CaCl_2}$, then through the wash bottle $c$ with a solution of $\mathrm{KHO}$, where it is freed from $\mathrm{CO_2}$; then the gas is again dried while passing through the tubes

Fig. 35.

Fig. 35.

^33) J. Gasbel. 1896 S. 83.
^34) Z. V. d. I. 1902 S. 948.
^35) Verh. v. Gew. 86, 1907 S. 437; Gasmot. 7, 1907 S. 134.

d, d, filled—the first with soda lime, and the others with CaCl₂; next the gas passes through the control U-shaped tube e with CaCl₂, weighed before and after each experiment. If the weight of tube e has not changed, everything is in order; if it has increased, this means that the absorption of H₂O in the preceding apparatus was incomplete. However, with a small increase, the experiment is not yet spoiled, since it may be assumed that tube e itself absorbed the remaining H₂O, serving, as it were, as an aid to the preceding apparatus. At first, in some experiments, a control potash apparatus was also placed directly after the washing bottle c, but practice showed that, when using Walter’s washing bottle with a coil, in which the gas bubble comes into especially close and prolonged contact with the KHO solution, after the washing bottle not even traces of unabsorbed CO₂ are ever detected. Therefore, in order to simplify the conduct of the experiments, to reduce the number of joints, which are always weak points of this method, and finally to reduce the resistance in the system—which is desirable in that, as the resistance decreases, the pressure in the system also decreases, which in turn reduces the likelihood of spoiling an experiment through gas leakage—in the majority of experiments the control potash apparatus was not installed. As a precautionary measure, instead of this, the absorbing liquid was changed more frequently: according to Hempel’s data²², 1 cm³ of KHO solution of strength 1 : 2 can absorb up to 40 cm³ of CO₂, i.e. about 72 mg; we, however, never absorbed more than 30–40 mg.

Next, after passing through all the devices listed, the gas, dried and freed from CO₂, enters tube f, filled with pieces of CuO in the form of oxidized pieces of wire, about 1 mm in diameter and from 2 to 5 mm long. Tube f, made of refractory glass, about 460 mm long, with an internal diameter of about 20 mm, is placed in a horizontal Heraeus electric furnace g and is maintained throughout the experiment at the temperature of red heat, i.e. about 600–700° C. On passing through the incandescent layer of CuO, the carbon monoxide, hydrogen, and various hydrocarbons contained in the gas burn, partly at the expense of the excess oxygen in the gas itself, partly by taking it from the CuO, forming CO₂ and H₂O, which are then captured—H₂O in the absorption apparatus h, a Winkler coil filled with strong sulfuric acid, and CO₂ in the potash apparatus k with the control U-shaped tube i, filled in its first half with soda lime and in its second with CaCl₂. The coil h and the potash apparatus k are weighed on analytical balances, in our case from the firm Rüprecht, with an accuracy of 0.1 mg, before and after the passage of the gas; the corresponding increases in weight show the quantities of newly formed water and carbon dioxide.

Knowing the quantity of gas passed through, and assuming that CO₂ is formed from free carbon, and H₂O from free hydrogen, it is not difficult to calculate the quantity of additional heat that could have been developed during the process of combustion in the machine under the condition of complete combustion, i.e., if the products of combustion contained no parts still capable of being oxidized.

In general, weight analysis by means of combustion, the so-called elementary analysis, is considered a rather difficult, laborious matter, accessible only to an experienced chemist. This is partly true with respect to the analysis of a solid or liquid substance. The analysis of gases, however, and moreover with the aid of our apparatus, is carried out very simply and conveniently.

One of the distinctive advantages of our arrangement is the use of an electric furnace: on the one hand, this furnace gives a very constant and uniform temperature over a considerable length; on the other, regulating the temperature in the presence of a resistance with small pear-shaped bulbs and an ammeter with sufficiently small divisions is very simple.

Before passing to the description of the actual performance of the experiment, it is not superfluous to mention the method we used for measuring the delivery of gas from the aspirator. The first thought was to use for this purpose the gas meter from the Junkers calorimeter; however, preliminary trials showed that working with the gas meter at a gas pressure of up to 200 mm water column, and, above all, setting the required pressure, is rather inconvenient. Moreover, the gas meter is inconvenient when the quantity of gas being passed is small—in our case only 8–12 liters; since, with a capacity in the aspirator of only about 15 liters, one cannot be sure that, after passing through the gas meter, whose drum capacity is 3 liters, 3–6 liters, we have completely removed from it the gas that was in it before, i.e., at the beginning of the experiment the gas may still be flowing with an admixture of the previous gas.

Closer to the purpose was the use of the aspirator from the Graefe calorimeter, i.e., a 5-liter jar with graduations every 1/2 liter directly on the glass. However, against this method of measurement stood its too low accuracy: it is impossible to estimate by eye the level of the water in the aspirator more precisely than to 1 mm, and with a jar diameter of 180 mm this corresponds to about 0.03 liter. If one takes into account that the errors at the beginning and at the end of the experiment may have different signs, i.e., the total error may be up to ±0.06 liter, this gives, for 5 liters, an error already up to ±1.2%, or even more.

After several attempts to improve the accuracy of this aspirator, we settled on the idea of using, for counting the moments at which definite divisions passed, the same principle as in measuring the consumption of kerosene—the appearance or immersion in the level—

...the level of the point. For this purpose we took two 15-liter bottles with wide necks and with a tube at the bottom, and, having connected the bottles through these tubes by a rubber tube and closed the neck of the bottle for taking gas with a stopper, through which two glass tubes were passed—one for admitting gas, the other for attaching a water manometer—we obtained the aspirator we required. Through the stopper with which the gas bottle was closed, we passed 16 pieces of iron wire, sharpened at the ends, of different lengths, the lower point being set near the bottom of the bottle and the upper near the stopper; the distance from one point to the next corresponds to 1 liter. The graduation, i.e., the setting of the points, was carried out with the aid of accurate scales, showing 0.5 g at the limiting load of 50 kg. When the wires had been arranged, the stopper, sunk approximately 10 mm below the edge of the neck, was sealed with Mendeleev putty. After this the vessel was checked again, determining the weight of water between the separate points. It turned out that no division was equal to 1 liter; some were larger, others smaller; however, this is of no significance. Repeated measurements, moreover carried out after some time, gave, for the same divisions, values differing among themselves by no more than 10, rarely 15 g. Having made 5 graduations and calculated, by the formula of the squares of the differences, the mean error of the mean value for those divisions which by chance gave the most divergent readings, we obtained an error from $\pm 0.7$ to $\pm 2.1\%$. And since, with our method of graduation, the errors of the individual readings do not affect the readings of adjacent divisions, i.e., the absolute magnitude of the error in liters remains unchanged for any number of divisions, then, at the smallest expenditure of 8 liters, the possible error proves to be less than $\pm 0.5\%$.

For convenience in using the divisions according to the mean values from 5 graduations, a table was compiled by which one can directly find the volume corresponding to the distance from any point to any other point lying above it.

The only conditions important for the accuracy of measurements with this aspirator are that the bottle should stand on a strictly horizontal surface, and that the water should not flow in too rapidly, i.e., that its level should be calm. Both of these conditions were fully observed in our experiments.

To finish with the aspirator, it should also be added that after the lapse of some time it was found that the wires, despite their tinning, had begun to rust, and their ends had as it were become displaced owing to the rust that had formed; moreover, the sealing liquid—water saturated with common salt—became cloudy, which also began to interfere with the accuracy of the readings. We replaced the iron wires with copper ones; the indicated unpleasant phenomena were weakened, but did not disappear completely. Quite good results were obtained only after replacing...

they were passed through glass rods, about 3 mm thick, with tips drawn out in the flame of a lamp. After this, graduation was carried out in the same way as described above, and, for use, the same new table of mean values was compiled.

Finally, at the connection of the gas jar with the jar for admitting water into the tube of the gas jar, a glass stopcock was inserted by means of a rubber stopper; its purpose was to regulate the rate of water discharge when gas was being drawn in, and, conversely, its admission when gas was being removed for analysis.

The analysis itself by combustion was conducted in the following manner: first the furnace was gradually heated; for this purpose the electric current was closed, with resistances inserted, so that at first there flowed only 2 amp.; every 10 min. a portion of the resistances was taken out, increasing by 2 amp., until the current reached 10 amp., at which the experiment itself was conducted. While furnace $q$ and the tube $f$ containing CuO$_2$ were being heated, the tightness of all the joints was tested; for this purpose, with the outlet end of tube $f$ closed, air from the zinc gasometer $i$, connected by means of the three-way cock $o$, was admitted into the system. If everything was tight, the air soon ceased to enter, which was detected by the cessation of the movement of bubbles in the washing bottle $c$; otherwise, the place of leakage was sought by wetting all the joints in turn with water, as when looking for a puncture in a bicycle tire, and the leak found was repaired. Such cases of leakage occurred, though rarely, and almost always resulted from cracks in the Mendeleev cement with which the ends of the calcium-chloride tubes $d, d$ had been sealed. Then the control tube $e$ was weighed, and, when tube $f$ was almost normally red-hot, air from the gasometer $i$ was begun to be passed through the whole system, connecting, however, to the outlet end of tube $f$ a simple calcium-chloride tube. The air was passed partly in order again to oxidize pieces of copper that might have remained unreduced from the preceding experiment, but chiefly in order, by passing dry and pure air through tube $f$, to make sure that there remained in it not even traces of moisture and carbon dioxide, which might have entered from the atmosphere after the preceding experiment.

During this operation the potassium apparatus $k$ and the coil $h$ were weighed, and the water level in aspirator $a$ was also brought into contact with the nearest point, and the gas pressure was adjusted by the water manometer $m$ to the required height of 170–190 mm. This last was achieved by first bringing, when raising the level by feeding water from jar $n$, the pressure to a somewhat higher value, and then releasing the excess, momentarily connecting the aspirator with the atmosphere by lightly pressing the clamp on the rubber tube.

When all these preparatory operations had been completed, the passage of air from the gasometer was stopped, the aspirator $a$ was connected to the system, in place of the safety calcium-chloride tube the coil $h$, the Kali apparatus $k$, and the control calcium-chloride tube $l$, connected with one another, were attached; then, opening the clamp at the aspirator $a$ and turning the three-way cock $o$ to the corresponding position, the passage of the gas to be analyzed was begun. At the same time it was recorded where the sharp point of the level stood, the pressure in the aspirator, the temperature of the gas (in the room), and the barometric pressure.

The rate of passage of the gas, important for the proper course of the reaction, was set so that it was easy to count the bubbles passing through the Kali apparatus. Gaber²³ says that in his experiments he kept the rate considerably below $3\ \mathrm{cm}^3/\mathrm{sec}$, i.e. $10.8\ \mathrm{liters/hour}$; we, however, became convinced that the experiment must be conducted so as to pass only $2$—$2.5$ liters, in any case not more than $3$ liters per hour; otherwise one cannot be sure that the absorption of $\mathrm{CO_2}$ and $\mathrm{H_2O}$ has taken place completely.

The necessary regulation of the rate was carried out in two ways: on the one hand, by increasing or decreasing the pressure in the aspirator $i$, for which purpose the cock for discharging the water is opened more or less; on the other, by closing more or less the cock at the outlet end of the calcium-chloride tube $l$. At first sight it would seem that the cock of tube $l$ is superfluous, and that the gas outlet can be opened completely; but practice has shown that, owing to the large capacity of the whole system and the periodicity of the gas supply through the washer $c$, the bubbles pass through the Kali apparatus $k$ unevenly: 5–6 bubbles slip through very rapidly, then there is a pause, then again 5–6 bubbles, and again a pause, and so on. If, however, the cock on the calcium-chloride tube $l$ is partly closed, then, although a somewhat higher pressure is established in the system, the gas bubbles pass with perfect regularity, and the total passage of gas even increases somewhat.

During the experiment itself the observer’s role is reduced merely to checking from time to time whether the established condition is being maintained, i.e.:

whether the gas bubbles are passing uniformly through the Kali apparatus $k$;
whether the tube $f$ has the proper color (temperature);
whether the pressure in the aspirator $a$ has changed.

If the bubbles begin to pass unevenly, i.e. periodically—which, generally speaking, is possible only if their passage was not properly adjusted at the outset—then the cock at tube $l$ must be closed somewhat more.

If tube $f$ begins to darken, the contact on the small rheostat must be shifted a little, increasing the current by $0.2$—$0.3$ amp.; if, on the contrary, its color begins to become too bright, then, in order to avoid excessive softening of the glass, whereby tube $f$ may easily be blown apart, the current must be slightly reduced, increasing the resistance.

Finally, if the pressure in the aspirator falls, i.e., if the speed of the gas has not yet been sufficiently regulated at once, it is necessary to increase slightly the supply of water from bottle $n$; if the pressure, on the contrary, rises, which may occur owing to an increase of resistance in the system—for example, from clogging of the potash apparatus by precipitated insoluble crystals, or from wetting of the first tube $d$ with caustic soda—then there is usually nothing to be done: the experiment must be interrupted in order to avoid blowing out tube $f$. If by that time not less than 5 liters have managed to pass, the experiment may simply be finished, bringing the level to the nearest point and then carrying out all the further manipulations and calculations. If less than 5 liters have passed, then enough gas still remains in the aspirator for a new experiment, which is then begun after removing the excessive resistance in the system. Disturbances of this kind, however, occur rarely.

As we see, all the observations are extremely simple and not tiresome. To control the rate at which the gas is passed through, it is also useful to note the time of passage of successive points and, if the speed is great, i.e., if 1 liter passes in 20 minutes or less, to reduce slightly the pressure in the aspirator.

Here it will be appropriate to note the advantages of a stopcock at the aspirator for letting in water, in the sense of regulating the pressure. Initially it was proposed to achieve constancy of the gas pressure by maintaining a constant difference of levels in the upper bottle $n$ and in the aspirator $a$. For this it would have been necessary gradually to raise bottle $n$ as the water in it diminished. The corresponding device—for example, in the form of a stool with a cover raised by a screw, or one rocking after the model of the apparatuses for generating carbon dioxide—turns out, for 15-liter bottles, to be rather cumbersome; moreover, it requires continuous observation and resetting. Automatic maintenance of a constant difference of levels, for example by means of something like scales, would be very complicated. We then came to the idea of making the flow of water constant, eliminating the influence of changes in level differences by making them offer very great resistance to the motion of the water. Experiments indeed confirmed this: by fitting a stopcock with an opening of 3 mm and opening this opening only to a very small extent, we obtained a flow of water that remained practically constant when the difference of levels varied within the limits from 1400 to 600 mm.; the pressure of the gas in the aspi-

the aspirator could then have from 170 to 200 mm water column, depending on the resistance to the outflow of gas. It may be that a slight regulating influence was exerted by an imperceptible change in the gas flow rate. In any case, once the tap had been set correctly at the beginning of the experiment, it no longer had to be touched; nevertheless, the pressure in the aspirator, despite the above-mentioned large change in the difference of levels, remained practically constant.

The analysis itself was considered completed when 8–10 liters had been passed, depending on the speed that could be established. After this, at the moment when the level came into contact with the nearest point of the turn of the three-way cock $o$, the passage of gas was stopped and air was admitted from the gas meter $i$, in order, by displacing the gas from the system, to force all of it to pass through the absorption apparatus; by this same passage of air, the pieces of copper, which are reduced when CO$_2$ and H$_2$O are formed, are again oxidized. In doing this, the tap at the bottom of the aspirator must, of course, be closed; otherwise, owing to the increasing pressure of the gas, the latter will break out, forcing the water out of the manometer $m$.

After a half-hour passage of air, the experiment may be considered completely finished. Then the coil $h$, the potash apparatus $k$, and the calcium-chloride tube $l$ must be removed, replacing them with a guard calcium-chloride tube, and, having sealed them with glass rods—stoppers—proceed to weighing. The weighing of each of the apparatuses $h$, $k$, and $l$ was carried out, in accordance with the advice of all analysts,^38 after removing all rubber tubes from it.

Then the control calcium-chloride tube $z$ was also weighed. At the same time the furnaces were allowed to cool, gradually reducing the current.

Of course, at the moment when the passage of gas was stopped, the pressures in the aspirator, the barometer, and the temperature in the room were again recorded.

As can be seen, with a well-adjusted system the entire experiment proceeds very simply and requires not so much skill as, chiefly, accuracy. The only circumstances that may still hinder, in certain cases, the spread in practice of such an analysis—whose importance, as we shall see below, is beyond doubt—are the rather high cost of the necessary instruments, chiefly the analytical balances and the furnace, if such are not already available for other purposes, and the considerable time required to carry out each analysis: from the moment the furnace begins to be heated until the complete completion of the experiment, i.e. the switching off of the current, 5–6 hours elapse.

For completeness, Table 14 presents a specimen record during one of the experiments. As in the preceding tables, the entries made during the experiment itself are set in italics,

^38 See, for example, Treadwell, Course of Analytical Chemistry, Odessa, 1906, vol. II, p. 272; Bliakher, Heat in Factory Work, Riga, 1905, p. 27.

Table 14.

Analysis by combustion. For experiment No. 32. 8.8.08

time	duration, min.	No. of sampling	corresponding volume, ltr.	aspirator pressure, mm	barom., mm	$t^\circ$ C.
11 h. 25 m.	30	2	14,125	160	747,2	24,0
55 m.	32	3		11,8 in. Hg
12 h. 27 m.	35	4			759
1 h. 02 m.	39	5		185
41 m.	44	6
2 h. 21 m.		7	9,120	302; 14,8 in. Hg	747,2; 762	24,0

gas consumption:

\[ \left(\frac{14{,}125 \cdot 759}{737{,}4} - \frac{9{,}12 \cdot 762}{737{,}4}\right) \frac{288}{297} = \]

\[ (14{,}53 - 9{,}43)\frac{288}{297} = \]

\[ = 4{,}945\ \text{ltr.} \]

Weights.

apparatus	before experiment, g	after experiment, g	increase, mg	per 1 ltr., mg
control (e)	37,0571	37,0570	−0,1
$H_2SO_4$ (g)	89,3347	89,3566	+21,9	$H_2O = 4{,}43$
$KHO$ (h)	38,4028	38,4079	+5,1 } 59,3	$CO_2 = 11{,}99$
$CaCl_2$ (i)	51,8352	51,8894	+54,2 } 59,3	$CO_2 = 11{,}99$

in ordinary type—calculations made after the completion of the experiment. The figures for $H_2O$ and $CO_2$ in the last column of the lower half of the table were obtained by dividing the increase in weight of the corresponding apparatus by the gas consumption, reduced to 737,4 mm Hg and 15° C.

V. L. Matveev.

Auxiliary Experiments

15. Investigation of kerosene. — In order to compile a complete heat balance for the operation of the machine it was necessary to know, with the greatest possible accuracy, the heating value of the kerosene and its composition; and it was also useful, for evaluating the combustion process, to know the flash point of the kerosene used.

Heating value of the kerosene. It was determined with the aid of Junkers’s calorimeter, for which purpose we made use, of course, of all the indications obtained by means of the preliminary experiments described above.

The experiment was carried out three times with each sample; 10 g of kerosene were burned each time. To burn more than 10 g at one time is difficult and inadvisable, since, on the one hand, this produces a very large quantity of cooling water, for which one bucket, used for collecting and weighing it, proves insufficient; while, on the other hand, with a longer duration of the experiment, namely toward its end, the air pressure in the lamp changes too much, and the requirement of a steady state cannot be fulfilled even approximately.

As the calculated quantity for the heat balance, the so-called useful heating value, $H_n$, was computed, i.e. after subtracting the heat corresponding to the latent heat of vaporization of the condensate obtained during calorimetry.

The actual determination of the amount of condensate was made not for each 10 g of kerosene, but for 30–50 g; for this purpose the condensate from all three determinations was collected, and sometimes, after the end of the calorimetric experiment proper, an additional quantity of kerosene was burned, continuing to collect the condensate. To increase the accuracy, the amount of condensate was determined not by volume, as is usually practiced, but by weighing, just as was the amount of cooling water passed through. The condensate was weighed on chemical balances with an accuracy of 0.01 g; the cooling water, on decimal balances with an accuracy of 0.005 kg.

A fairly important question is how to take the sample for calorimetry. The kerosene, being in a vessel which, as in our case a barrel, is not hermetically closed, evaporates, although slowly, which may affect both its composition and its heat-producing capacity, especially in the upper layers of it in the barrel. It is therefore important that the sample for calorimetry correspond exactly to the kerosene used during operation. We achieved this by taking the sample through the tap located at the bottom of the barrel, just as for the operation of the machine. Calorimetry was carried out twice: at the beginning, when the barrel had just been uncorked, and toward the end. However, both of these samples from each barrel gave the heating value and the composition of the kerosene within the limits of accura-

the accuracy of the measurement, are quite the same. And in general the heat-producing capacity of the kerosene, which was always supplied by the Nobel Brothers’ office, fluctuated very little.

Composition of the kerosene. For certain calculations it was necessary for us to know the composition of the kerosene—not, of course, in the sense of which and how many of the various hydrocarbons enter into it, but only how much hydrogen and carbon it contains by weight. In view of the fact that it was sufficient to know these numbers with moderate accuracy, we did not carry out a true elementary analysis, but calculated the content of $\mathrm{H}_2$ and $\mathrm{C}$ from the amount of condensate found in calorimetry, assuming that, apart from hydrogen and carbon, kerosene contains nothing else, i.e. neglecting the small content of oxygen and, perhaps, sulfur.

We find this composition from the following simple considerations: 1 kg of kerosene, when burned, yielded on the average 1.173 kg of condensate; taking the molecular weight of water $\mathrm{H}_2\mathrm{O}$ as 18.016 and that of hydrogen $\mathrm{H}_2$ as 2.016, we find the content of $\mathrm{H}_2$ by weight:

\[ \frac{1.173\, 2.016}{18.016}=0.1313, \]

i.e. 13.13%.

The carbon content $\mathrm{C}$ consequently proves, under the simplifying assumption stated above, to be 86.87%.

As a check we compared the figures obtained with those available in the literature²⁴; the agreement proved quite satisfactory.

Flash temperature of the kerosene, of interest only as a characteristic of the fuel used, was determined in the usual manner with the aid of Abel’s apparatus No. S & R 2883, provided with a certificate of the German Physico-Technical Institute. This temperature $t_{\mathrm{fl}}$ in most samples was about $34^\circ$C, but if the kerosene stood in an open bucket for twenty-four hours, then $t_{\mathrm{fl}}$ rose approximately to $40^\circ$; further standing of the kerosene apparently had no further effect on $t_{\mathrm{fl}}$. In one of the barrels $t_{\mathrm{fl}}$ proved equal to $39.5^\circ$, and from the bucket after twenty-four hours’ standing about $44.0^\circ$.

Likewise, the temperature $t_{\mathrm{fl}}$ rose by $3$—$4^\circ$ if a little water was poured onto the bottom of the little cup in which the kerosene is heated in the apparatus. This circumstance deserves special attention in view of the fact that in the given machine there is an arrangement for admitting water into the compression chamber. Unfortunately, the author has not yet been able to approach more closely the question of the influence of the presence of water on the process of ignition and combustion.

16. Work of rotation of the flywheels.—For the heat balance it is necessary to find the work of friction of the piston, which is converted into heat and carried

cooled by water. Knowing the total work developed by the combustible mixture in the cylinder, i.e. the indicated work of the engine, and the useful work—the load on the brake—we obtain, in the form of the difference between them, the total loss of work to resistances. The latter may be divided into 3 groups:

the work expended in obtaining compressed air for scavenging the cylinder;
the work expended in rotating the flywheels and eccentrics, for the kerosene pump and the indicator drive, and, finally,
the work of friction of the piston.

Since the work for obtaining compressed air is the indicated work of the pump, from which in our experiments diagrams were taken, i.e. it is known to us, then, in order to calculate the work of piston friction, it proves sufficient to determine, by one means or another, the work of rotation of the main shaft with the parts mounted upon it.

The greatest part of the loss of work in rotating the shaft is to be expected from the friction of the flywheels against the air. The work of this friction could be determined from ready-made formulae indicated by various authors.

Thus O. Möller $^{40}$ derives a formula which, after simplifying it, may be written more correctly in the form

\[ N_r = 0.25 \left(\frac{n}{100}\right)^3 \delta b r^4 m, \tag{6} \]

where $n$ is the number of rev./min., $\delta$ is a certain empirical coefficient, $b$ is the width of the spokes in the axial direction of the shaft in m, $r$ is the internal diameter of the rim in m, and $m$ is the number of spokes. The coefficient $\delta$ varies depending on the cross-section of the spokes; for an elliptical section Möller proposes taking $\delta = 0.6$; however, in order to take into account not only the resistance of the spokes but also that of the rim itself, it seems to us closer to reality to take $\delta = 0.8$. Substituting for our case $n = 300$, $b = 0.4$, $r = 0.53$, and $m = 6 \times 2$, we obtain $N_r = 0.21$ h.p.

In the literature, however, there is an indication $^{41}$ that in a known case this formula gives a value of $N_r$ exaggerated, and by as much as a factor of 3.

According to the latest experiments, processed by Weker $^{42}$, the resistance of air to the rotation of flywheels may be allowed for by the formula

\[ N'_r = v^{2.5} D^2 (1 + 5 b_0^2) 10^{-5}, \tag{7} \]

where $v$ is the circumferential velocity at the flywheel rim in m/sec., $D$ the diameter, and $b_0$ the width of the rim in m.

For our engine, at $n = 300$, $D = 1.255$, and $b_0 = 0.09$, expression (7) gives $N_r = 0.057$ h.p.

$^{40}$ Z. V. d. I. 1888 S. 283.
$^{41}$ Z. V. d. I. 1901 S. 1783.
$^{42}$ Electr. Kraftbetr. u. Bahn. 1907 S. 490.

In view of this disparity among the ready-made formulas, on the one hand, and the difficulty of calculating with any satisfactory accuracy the remaining losses connected with the rotation of the main shaft, such as friction in the main bearings, in the eccentrics with their rods, and the driving of the tachometer and the Malbrun oilers, we decided to proceed to determine the entire total loss during rotation of the shaft $R_v$ by means of direct experiments.

These experiments were arranged as follows: after uncoupling the connecting rod, the main shaft, with all the parts and instruments connected with it, was rotated by means of a small shunt electric motor through a belt drive. The number of revolutions of the electric motor could be varied in small steps over wide limits, from 190 to 780, which, owing to the transmission ratio between the pulleys of the electric motor and the machine, equal to 0.54, gave for the machine $n$ from 103 to 422, i.e. within limits even wider than we needed.

The experiments themselves were conducted in the following manner: setting the handle of the starting (which is also the regulating) resistance to a definite contact, we waited until the speeds of the electric motor and of the flywheels became steady; these speeds were measured by means of a hand centrifugal tachometer of the firm T. Thörne, and at the same time the readings were noted of accurate, calibrated voltmeters, connected at the very terminals of the electric motor, and of an ammeter showing the current going to the motor armature, as well as of a second one showing the current going to the motor field coils. It was, of course, impossible to use for these reports the readings of the instruments on the distribution board, since the voltage beyond the regulating resistance was much lower—only from 45 to 128 volts—whereas the voltmeter on the board constantly showed 226 volts. As for the current strength, a part of it, and a rather considerable part, went into the field coils; for example, at the smallest number of revolutions about 1.0 amp. went into the field coils, with a current going to the armature of 6.9 amp., i.e. about 16% of the total current, and at the greatest number of revolutions respectively 0.7 and 15.0 amp., i.e. still 4.5%.

When the readings of all the instruments at the smallest number of revolutions had become quite steady, for which 20–25 min. were required, the number of revolutions and the readings of the voltmeter and ammeters were recorded; the resistance handle was moved to the next contact, reducing the magnitude of the resistance to the current going into the armature and thereby increasing the number of revolutions; again, having waited for a completely steady state, all the required readings were taken; then the resistance handle was moved to the next contact, and so on up to the highest number of revolutions. After this, the same procedure was followed, reducing the number of revolutions.

Multiplying the number of amperes going to the armature at any given number of revolutions of the electric motor by the number of volts at the terminals, we obtain the energy consumption in watts at the corresponding number of revolutions. However, from this energy it is still necessary to subtract the unavoidable losses in the electric—

the electric motor, equal to $i^2 r$, where $i$ is the corresponding current, in amperes, flowing into the armature, and $r$ is the resistance of the armature, found by us for our electric motor to be about 0.6 ohm. In addition, the work of rotating the armature of the electric motor itself must be subtracted; we found this by measuring the current and voltage when the electric motor was running idle.

The remainder of the energy, in watts, divided by 736 and multiplied by the efficiency coefficient of the belt transmission $\eta_p$, which with sufficient accuracy may be taken as $\eta_p = 0.95$, gives the expenditure of work for rotating the flywheels and shaft, in h.p., at the corresponding number of revolutions.

The results of the experiments are represented graphically in Drawing 36: crosses denote experiments with an increasing number of revolutions, circles—with a decreasing one. Connecting the separate points by a smooth curve,

Drawing 36.

Drawing 36.

we see at once that the figures for increasing $n$ lie above the curves connecting the figures for decreasing $n$. The first explanation that naturally suggests itself for the smaller expenditure of work with decreasing $n$ is the influence of the live force of the flywheels; however, this explanation falls away if one remembers that the readings in both cases, with increasing and decreasing $n$, were taken only after a fully established state had set in. The phenomenon can rather be explained by a decrease in friction owing to the warming of the oil. Unfortunately, in one evening it was not possible to repeat the entire series of experiments for lack of time, and, on the other hand, it did not occur in time to carry out the test in the reverse order, beginning with the largest number of revolutions. Several days after these experiments had been performed, the opportunity was lost

Study of a Two-Stroke Machine

to repeat them: the storage battery of the Institute’s electric station, from which current was taken for the electric motor, had been dismantled. An attempt to repeat the experiments while receiving energy from a dynamo machine gave completely unsatisfactory results: changes in the current consumption in the network were so reflected in the electric motor that it was impossible to obtain a steady state, and the results obtained in this way gave deviations both above and below the curve of Fig. 36.

In the end we decided, without resorting to new experiments, to take as the work of the resistances in Fig. 36 the mean, continuous line, as if discarding the experiments with increasing $n$. In favor of such a decision is the circumstance that, in general, the magnitude of all the resistances taken together is far from constant: greater or lesser tightening of the collars of the eccentrics, better or worse lubrication of the moving parts—all this may have a considerable influence; thus, for example, we observed that under the influence merely of moving the platform $m$ of the regulator, Figs. 1 and 2, from one extreme position to the other, corresponding to the smallest and largest number of revolutions, during the preceding tests with the electric motor the number of revolutions of the machine, with the readings of the voltmeter and ammeters almost unchanged, varied by approximately 5–8%, falling when the platform was moved toward the crankshaft, which in regulation corresponds to an increase in the number of revolutions, and conversely, increasing when it was moved away. With a mean number of revolutions, about 300, the usual change by $\pm 2\%$, i.e. within 12 revolutions, as is seen from the diagram of Fig. 36, corresponds to a change of the resistances by $\pm 0.08$ hp, i.e. considerably greater than the possible error with which our mean curve was found; this error at $n = 300$ corresponds to $\pm 0.04$ hp. If, however, the accuracy of determining the mean curve is referred to the indicated work of the machine at the corresponding $n$ and with the normal useful load of 24 kg on the brake scales, then we obtain as limits: at $n = 200$ an accuracy of 0.02 hp for $N_i = 8.6$ hp, i.e. $\pm 0.2\%$; at $n = 300$ an accuracy of $\pm 0.4$ hp for $N_i = 15.5$ hp, i.e. about $\pm 0.3\%$.

The error from not taking into account the position of the platform $m$, reaching, as stated, up to $\pm 0.04 n$, corresponds at $n = 200$ to about 0.034 hp, or, referring it to the corresponding $N_i$, $\pm 0.4\%$, and at $n = 300$ already to about 0.14 hp, or correspondingly $\pm 0.9\%$.

Thus the total possible error of the quantity $R_v$ amounts in the worst case, at $n = 300$, to only $\pm 0.54$ hp, i.e. 1.1%.

Thus the accuracy with which the mean curve of Fig. 36, according to our experiments, expresses the work of the resistances to rotation of the shaft with all parts and instruments connected with it may be considered quite satisfactory.

In the calculations in the subsequent treatment of the experimental material, the work of the resistances was taken directly as the ordinates of the mean curve of Fig. 36 for the corresponding number of revolutions.

17. Leakage of air through the blower.—In the very first experiments with the supply of air under pressure by means of the Root-system blower, we noticed the formation of bubbles from the oil at the places where the blower shaft bearings emerged. These bubbles revealed a leakage of air, drawn in by the blower through the air meter. This leakage, judging by the size and number of the bubbles, seemed at first sight not especially significant. However, in compiling the heat balance for these experiments, unusually large figures began to be obtained for the heat loss carried away by the products of combustion.

This figure is obtained as the product of the quantity of products of combustion by their heat capacity and temperature. There could not have been any error, at least a significant one, in calculating the heat capacity, nor in the temperature readings, which were always checked, as was indicated above; and in general, at high temperatures instruments are more likely to indicate a temperature below the true one, and by no means above it. Thus the only possible source of error proved to be the quantity of products of combustion, or, more precisely, the quantity of air, since the weight of kerosene was measured in such a way that there could be no doubt as to its correctness, and moreover it constitutes only a small part of the total weight of the products of combustion.

Thus the heat balance confirmed our preliminary doubts concerning the leakage of air through the blower bearings and revealed its magnitude. In view of the complete impossibility of eliminating this leakage, owing to the design of the given blower, it was necessary to reconcile ourselves to the fact of the leakage and merely try to determine the necessary correction. This correction could either be calculated, relying on other data, as will be indicated below, or found by direct experiments. In view of the importance of knowing the exact air consumption, we did not confine ourselves to calculating the correction, but tried to verify it also experimentally.

First of all it was necessary to decide on what the correction could depend. Assuming the constancy of the leakage resistances, i.e. the size of the clearances and the viscosity and quantity of the oil filling them—which in fact was observed in our experiments—it may be said with confidence that the leakage depends exclusively on the excess pressure of the air in the blower. Neither the quantity of air supplied, nor the temperature in the room, nor the barometric pressure, at least within the small limits within which these two quantities fluctuated in our experiments, nor, finally, the number of revolutions of the blower as such, could have any appreciable influence on the leakage.

Taking this into account, we immediately obtained a very simple course for the forthcoming experiments: it was necessary to close off the useful outlet of the air and, by setting, through rotation of the blower, a known pressure in the air vessel c, drawing 8, to measure by the air meter the corresponding air consumption, which would directly give the desired leakage at the given pressure.

The experiments were conducted in the following manner: having disconnected, by means of a plug, the branch of the air pipe leading to our two-stroke machine and having left the rest of the air pipe in the same condition as during the experiments, the blower was set in rotation, forcing air into the vessel up to the required pressure. As soon as the latter had been reached, readings were begun on the air meters, while at the same time an effort was made to keep the established pressure as constant as possible; this pressure was read from a water manometer or, correspondingly, from a mercury manometer.

The setting of any given pressure height was achieved simply by changing the number of revolutions of the blower.

Although, owing to the absence of electric current, the blower had to be driven by hand, this method gave quite satisfactory results, represented graphically in Fig. 37: repeated experiments gave very close results, and all the points lie sufficiently close to the mean curve.

Fig. 37

Fig. 37.

As can be seen, the curve representing the leakage of air in m³/hour as a function of pressure is very close to an inclined straight line and intersects the abscissa axis at $p = 0$ mm water column, which is readily explained: the resistance to seepage is in reality small; therefore there is no leakage at all only in the absence of pressure.

Using the diagram in Fig. 37, we obtain directly, for any pressure up to 1300 mm, the air leakage in m³/hour, which must be subtracted from the corresponding readings of the air meters.

The accuracy of the correction introduced for leakage according to Fig. 37 ranges from $\pm 0.3\ \mathrm{m}^3$ approximately to $\pm 0.7\ \mathrm{m}^3$; relating it to the corresponding total air flow, which varied from 90 to 120 m³/hour, we obtain $\pm 0.33$ to $\pm 0.60\%$.

V. L. MALYSHEV.

Addendum.

18. Difficulties during the experiments.—Although the difficulties described below, which were encountered while carrying out the experiments, have no direct bearing on the tasks of our experiments, we nevertheless consider it useful to acquaint the reader with them, since, first, this will provide material for judging whether, in our experiments, we really achieved the extreme limit of accuracy possible when working with the machine we had, and at the same time will present a picture that will give a measure for evaluating our experiments; and, secondly, we think that the indication of the means by which we dealt with the difficulties, most of which may be assigned to the category of defects of the machine, may be useful to others in carrying out similar experiments, and in part also generally to persons who have to deal with kerosene engines.

Seizing of the piston. The greatest difficulty we experienced was the circumstance that, after a certain time of operation, 1–2 hours, the machine would stop by itself as a result of an excessive increase in the friction of the piston, which became evident when turning the flywheels in an attempt to start it again. After waiting a little, 20–40 minutes, in most cases the cause could again be checked, and it was just as easy as before the experiment. Obviously, such an increase in the friction of the piston, which soon disappears when the machine cools, can be explained only by expansion of the piston as a result of its heating.

This expansion could be of two kinds: either uniform, which causes seizing simply as a result of insufficient clearance between the piston and the cylinder, or one-sided, which may cause seizing even with an entirely sufficient clearance measured when the piston and cylinder are cold, and occurs because the cross-section of the piston changes from circular to elliptical.

In view of the construction of the piston—very simple, without any ribs—and the care in workmanship for which the L. Nobel works is generally known, we initially assumed only an insufficiency of clearance between the piston and the cylinder, which should soon have disappeared of itself as a result of the inevitable wear of the piston. The correctness of this assumption seemed also to be supported by the following circumstance: we began to work with the machine in summer; during the summer no increase in the seizing of the piston was observed, partly, perhaps, because of another difficulty that appeared, which will be discussed in more detail separately, namely the lubrication of the piston.

From autumn onward, the seizing of the piston began to increase noticeably, in the sense of a decrease in the interval of time from the moment the machine was started until its spontaneous stopping as a result of an excessive increase in the friction of the pis—

Study of a Two-Stroke Machine

piston. Since at the same time the temperature of the cooling water, which was admitted into the machine directly from the municipal water supply, began to fall, we regarded this only as confirmation of our supposition concerning the influence of the temperature difference between the cooled cylinder wall and the uncooled piston. We also regarded as confirmation of our supposition the experimentally established fact that the machine stopped sooner when a larger quantity of water was admitted and, correspondingly, when its temperature at the outlet was lower.

Continuing to hope for a natural decrease in the diameter of the piston and an increase in the diameter of the cylinder as a result of unavoidable wear, we directed our efforts toward creating conditions for increasing the duration of the machine’s operation from the moment of starting to the forced stop. With these aims, we first paid attention to the extremely uneven distribution of temperature around the circumference in the outer wall of the cylinder at low temperature; as a result, with the existing admission of water in our machine from the side at the bottom, this place of the cylinder had, to the touch of the hand, a very low temperature; the latter increased from there in all directions, reaching its greatest value at the top and partly on the side opposite the water inlet. We made an additional water inlet on the other side; to the touch of the hand the temperature of the walls became more uniform; however, the sticking of the piston not only did not diminish, but even seemed to increase. Then we resorted to another, completely unusual method and one hardly ever used by anyone else—artificial heating of the cooling water.

The heating of the water was carried out directly as it moved along the water-supply pipe, by means of a special gas burner constructed by us, Drawing 38: the illuminating gas enters through the copper nozzle $a$, having a diameter of 4 mm and screwed into the cap $b$, which connects the nozzle with the burner itself—a 1-inch iron pipe $d$—by means of the reducing coupling $c$; in the cap $b$ there are two slots for drawing in air, and in the pipe $d$ there are up to 50 openings, about 4 mm in diameter, arranged in a zigzag in two rows, beginning at the end opposite the nozzle, closed by means of the cap $f$, and ending approximately 200 mm from the nozzle; at a closer distance of these openings from the nozzle, the flame jumps into the mouth of the nozzle. The distance between the two rows is about 15 mm, and the distance between the openings in each row is about 25 mm. Each opening gives a separate tongue of flame, em-

Drawing 38

Drawing 38.

V. L. Maleev.

a height of 10 to 30 mm, depending on the position of the cock $e$, by which the inflow of gas could be regulated. The burner was connected to the gas pipe by means of a rubber tube 15 mm in diameter and was suspended beneath the water pipe, at a distance of about 20 mm below it, on wire hooks.

The burner arranged in this way gave excellent results: the flame of the jets was blue, almost colorless, and the heat efficiency $\eta_t$ was sufficiently high. Namely, with the aid of two such burners, suspended under the $1 \tfrac{1}{2}''$ water pipe one after the other, the temperature of the water, at a consumption of about 400 kg/hour, rose from $+3^\circ$ to $+16^\circ$ C.; the gas consumption for both burners was about 4 m$^3$/hour; its calorific value was determined by means of a Junkers calorimeter at about 5000 heat units, which gives:

\[ \eta_t=\frac{400(16-3)}{5000.4}=0.26, \]

i.e. 26%.

However, when we started the engine with the water thus heated, no improvement occurred. Thus, after a number of fruitless attempts, we were compelled to admit that our supposition as to the simple inadequacy of the size of the clearance was wrong. There remained only the second supposition: that, when heated during operation, either the piston or the cylinder itself loses its circular form, and this causes the seizure. The idea of such an irregular deformation of the cylinder had to be rejected at once: on the one hand, because its temperature under the action of the cooling water is very low; on the other, in view of the fact that the attempt mentioned above to eliminate the seizure by admitting additional cooling water, owing to which a more uniform distribution of temperature in the walls was indeed obtained, had no favorable effect whatever on the occurrence of piston seizure.

Thus, following the method of gradual elimination, we were finally forced to acknowledge the correctness of the supposition that had arisen earlier: that the piston, having acquired internal stresses during casting, deforms irregularly when heated, most likely assuming a somewhat elliptical cross-section; moreover, this irregular deformation gradually increased with time, increasing the seizure of the piston, while we mistakenly attributed this latter circumstance to the accidental, simultaneous lowering of the temperature of the water in the municipal network.

The only radical means of correcting the piston would be to anneal it and check it on a lathe or, still better, to make a new piston. Not venturing to resort in Tomsk to the first measure, and not having the possibility of immediately resorting to the latter, we

they decided to continue the experiments with the old piston, slightly reducing its diameter and trying to remove more material in those places which entered the cylinder especially tightly, and which could be detected from the traces of wear. This work was carried out by hand, with the aid of a personal file and emery paper, by the laboratory mechanic, and very successfully. After this the machine began to run considerably better, so that it was possible to complete with this piston the entire series of experiments necessary for our purposes.

Unsatisfactory lubrication of the piston. The determination of the true cause of the piston’s seizing was also greatly hindered by difficulties with the lubrication of the piston. The very first experiments with the machine showed that it was impossible to lubricate the cylinder and piston with ordinary machine oil, with a flash point $t_{\text{vs}} = 190^\circ$: removal of the piston revealed on the walls of the piston and cylinder a heavy deposit, which could be cleaned off only with difficulty by soaking in kerosene and scraping with an iron scraper.

They tried pouring into the lubricator so-called cylinder oil with a higher flash point, $t_{\text{vs}} = 240^\circ$; the results obtained were essentially just as unsatisfactory. It was then decided to turn, for lubrication, to so-called viscosine, a product with a very high $t_{\text{vs}} = 300^\circ$, manufactured by the Nobel Brothers firm especially for lubricating the cylinders of Diesel engines. At room temperature this viscosine is so thick that it cannot flow from a drip lubricator, but must be supplied by means of a pump. In any case, however, we had to resort to a pump for supplying lubricant under pressure in those experiments in which we worked with air scavenging, since a drip lubricator supplies lubricant under the action of the vacuum that occurs in the rear cavity during the suction stroke. When working with air scavenging this vacuum either disappears or, at the very least, decreases so much that even liquid oil does not go into the cylinder.

The hopes placed on viscosine, however, were not justified: it turned out that the grade obtainable in Tomsk contains many resinous substances, which after strong heating form a kind of enamel; the action of this enamel was especially evident on the piston rings, which very quickly stuck, ceased to spring, as a result of which the preliminary compression decreased owing to leakage of air, and in general the regular operation of the machine was disturbed.

One of the circumstances that considerably interfered with the proper action of the piston lubrication proved to be a phenomenon detected only after a number of experiments—namely, the entry of combustion products into the upper channel through which the air for scavenging is admitted, and through which passes the tube supplying oil to the piston. Entry into this cha-

the products of combustion, having a high velocity and a high temperature, at once caused two undesirable phenomena: on the one hand, the splashing of the drops of lubricant falling freely from the tube, i.e., a reduction in its useful delivery; on the other hand, evaporation, under the action of the high temperature of the products of combustion, of the lightest and most fluid constituent parts of the lubricant, i.e., a reduction in its lubricating capacity and the facilitation of the formation of the deposit or enamel mentioned above.

Only after all these circumstances had gradually become clear was it finally possible to arrange the lubrication in the proper manner: splashing was eliminated by doing away with the fall of an open drop, for which purpose the tube from the pump was extended almost to the very surface of the piston, and the large opening, about 30 mm in diameter, in the cylinder wall, through which the drop of oil falls onto the piston, was covered with a tin cover, allowing only the tube of the pump to pass through it; the evaporation of the oil, in turn, had already been partly eliminated by eliminating the open drop, and then further by wrapping the end of the tube from the pump, heated by the products of combustion entering the scavenging channel, with asbestos cord and thereby insulating it from the action of the high temperature; and finally, the viscosity of Nobel Brothers’ vaseline was replaced by a special American oil for internal-combustion machines of the Vacuum Oil Company, grade “Gargoil B.” Its flash point was found by us, with the aid of a Marcusson apparatus, to be about $+270^\circ$ C.; its specific gravity is $\delta = 0.89$; on ignition it leaves only an insignificant residue in the form of a black powder.

The lubrication arranged in this way began to operate quite satisfactorily.

Lubrication of the crankshaft. A slight difficulty also arose in work with air blowing in relation to the lubrication of the crankshaft of the flywheel shaft. Normally the crankshaft is lubricated by means of an ordinary drop lubricator with a check valve; the suction of oil is produced by the vacuum in the crank chamber during the suction stroke. When operating with air blowing, as already mentioned above, this vacuum disappears, and the oil ceases to flow.

By replacing the ordinary lubricator with one having a piston pump, brought into action by means of a ratchet mechanism, we also eliminated this difficulty.

Defects in the stuffing boxes. Rather much trouble was also caused by the attainment of tightness in the rear crank chamber; owing to the construction of the machine, tightness is obtained here exclusively by the fitting of the washers $y, y$, drawing 4, over a large surface; during operation these washers wear down rather quickly, and air leakage occurs. When air is sucked in from the atmosphere, such leakage, as indicated above, has no especially great effect, but when operating with

with the air blower, it can greatly alter the actual consumption of air, which will be considerably less than that found from the readings of the air meters. The most effective means of eliminating leakage would have been to install springs between the disk cheeks of the shaft and the washers $y, y$, so that the latter would always be pressed against the covers, as is done in some small two-stroke machines of the automobile type. Not wishing, however, to complicate the design, on the one hand, and bearing in mind that the machine operates exclusively for experiments, i.e. in general for a comparatively short time, during which wear of the washers occurs slowly, we confined ourselves only to fitting the washers $y, y$ tightly to the cheeks of the shaft and to the covers, and subsequently compensating for the wear of the washers by reducing the thickness of the paper gaskets under the covers $x, x$.

The results, as regards tightness, proved quite satisfactory, as could easily be ascertained by observing the leakage of air in the form of bubbles through the oil filling of the main bearings.

Kerosene supply. The greatest difficulties were encountered with the supply and atomization of the kerosene. These difficulties, despite all efforts, could not be completely eliminated, and this circumstance noticeably affected the results of the experiments and, unfortunately, forced us to narrow considerably the range of questions to which our experiments could give an answer.

These difficulties may be divided into two groups: improper action of the kerosene pump, and unsatisfactory action of the nozzle.

Improper action of the pump manifested itself in that it would deliver kerosene sometimes in the required quantity, sometimes suddenly less, sometimes even cease delivering altogether, and then again begin to operate correctly. Such disturbances in the supply occurred periodically after 5–10, sometimes 30 minutes.

As became clear after careful investigations, the causes of the incorrect supply of kerosene were the following: a faulty condition of the suction valve in the pump cylinder, heating of the kerosene on the path from the tank to the pump, and, finally, an insufficient cross-section of the supply tube on this same path.

The first cause was, of course, very easily eliminated by lapping the valve and suitably adjusting its spring.

As for the heating of the kerosene, it occurred because initially the kerosene tube was in contact, along almost its entire length, with the tube running beside it that carried away the cooling water, heated to 50–60° C. Although it was difficult to establish the influence of heating of the kerosene by direct observations, nevertheless we considered it possible that kerosene vapors were formed under the simultaneous influence of elevated temperature and reduced pressure during the suction stroke. The formation of even small vapor bubbles, with the insignificant quantities delivered by the pump in one stroke, could greatly distort the pump delivery.

1 stroke—0.26 to 0.30 cm³; naturally, the volumetric delivery of the pump had to be greatly reduced.

The heating was completely eliminated by placing wooden blocks between the water-supply pipe and the kerosene tube fastened to it, so that the distance between the tubes became about 15 mm; in addition, the kerosene tube was wrapped with asbestos cord.

When, however, all this did not help, and the pump continued to work badly, although somewhat better than before, we turned our attention to the velocity in the kerosene-supply tube. For this purpose we determined the angle of rotation of the shaft corresponding to the time of the suction stroke of the kerosene pump; this angle proved to be equal to $32^\circ$. Taking the greatest kerosene consumption as up to 6 kg/hr, as was found from our experiments, we calculated the mean suction velocity: with the diameter of the tube sent from the factory, $d = 3.8$ mm, i.e. a cross-sectional area $f = 11.34$ mm², and with the specific gravity of kerosene $\delta = 0.821$, we obtain the required velocity of kerosene motion

\[ c = \frac{6.360}{32.0.821.3600.1000.0.00001134} = 2.0 \text{ m/sec}, \]

a rather considerable velocity. When, however, we measured the flow of kerosene from the tank by gravity, disconnecting the end of the tube from the pump, we obtained a flow of only 11.2 to 13.8 kg/hr, which corresponds to a velocity of only 0.33 to 0.41 m/sec, i.e. 5–6 times less than is required. This, however, was to be expected, in view of the rather considerable length of the pipeline, the small head, and the considerable frictional resistance at so small a diameter.

After this calculation the tube was replaced by another, with a diameter $d = 9.5$ mm. Measurement of the kerosene flow by gravity with this new tube gave from 174.6 to 190.0 kg/hr, i.e. velocities from 5.2 to 5.7 m/sec, or approximately $2\frac{1}{2}$ times greater than we required.

It is interesting to note that the difference between the greatest and the smallest kerosene flow with the thin tube, $13.8 - 11.2 = 2.6$ kg, amounted to about 19% of the greatest flow, whereas with the second tube the difference $190.0 - 174.6 = 15.4$ kg amounts to only 8%; that is, accidental, imperceptible, and unavoidable changes in the conditions of kerosene motion—such as, for example, a change in the head height or the entry of a speck—affect the kerosene consumption far less with the larger tube. The operation of the pump should become considerably more reliable and more uniform. And indeed, with the new tube the pump began to work quite properly.

Nozzle. The situation with the nozzle was much worse. Two troubles arose with it: either it clogged easily, or it sprayed poorly.

Clogging was partly eliminated, though not entirely, by the fact that, in addition to the finest metal mesh, 1600 openings per 1 cm², from which a cap was made for the end of the tube that takes kerosene from the bottom of the tank, a frame with a flannel bag—a filter—was made for the tank, through which the kerosene was also passed when filling. This same filter protected the tank from the entry of litter from outside.

Especially much labor was spent on finding the most suitable design of the nozzle, which we remade several times. With a small opening through which the kerosene is atomized, too great a resistance is obtained to the motion of the little pump piston, and the delivery of kerosene is insufficient; with a large one, the kerosene atomizes worse, and, in addition, periodically a drop of kerosene is obtained, which passes through the engine and, of course, does not have time to burn.

Attempts to replace one opening in the nozzle by 2–3 smaller ones proved unsuccessful, since, although such a nozzle at first operates well, it clogs too easily and quickly, despite any filters.

The same proved true of a nozzle of special design, without openings, but with an annular slot formed by a conical recess—countersink—at the end of the nozzle tip and by a cone screwed into it.

In the end it was necessary to return to the type of nozzle supplied by the factory, and to reconcile oneself to the fact that part of the kerosene, not atomized and falling in the form of drops, passes through the engine without burning and, by uselessly increasing the consumption of kerosene, lowers the effective thermal efficiency $\eta_t$ of the engine.

In conclusion, it may be of interest to report the following phenomenon: the pressure that occurs in the kerosene pump during impact action is so great that it gradually pierces the walls of the fitting at the delivery valve; this fitting is cast, brass; the wall thickness is about 3 mm; there are, as far as can be seen, no cavities. The leakage of kerosene occurring directly through the thickness of the fitting could be eliminated only by forcing a tightly fitted brass tube onto the fitting and soldering it into one piece with the fitting.

It was likewise difficult to obtain tightness in the joints of these fittings. Leather and lead gaskets are crushed by frequent unscrewing; merely grinding the end face is insufficient; satisfactory results were obtained only by inserting rings of red copper.

Operation with air blowing. Among the difficulties encountered by us in the experiments one may also include the difficulty of starting the engine when operating with air blowing, especially when supplying air directly into

the rear cavity, i.e., operating, so to speak, with ready-compressed air, without a pump. However, this difficulty, which consisted chiefly in the unusual nature of the operating conditions, was soon overcome, after several trials, by developing special techniques, which have already been discussed above.

CHAPTER III.

Processing of the Experimental Material.

19. General remarks. — The processing of the raw numerical material obtained in the scientific testing of an internal-combustion machine may, as we think, be divided into three stages:

the derivation of mean figures from all the records made during the experiments;
the calculation of the usual quantities characterizing the given experiment, chiefly the complete heat balance with all its component parts, and their arrangement in the corresponding tables; and, finally,
the ascertainment of new, non-standard results for the sake of which the given experiments were undertaken, or which emerged in the course of their production and processing.

Accuracy and its determination. In calculating mean figures we tried to avoid an error often encountered in technical literature, even among the best authors, namely, excessive fascination with the number of decimal places. In calculating an arithmetic mean we took it to only one digit more than each of the terms. In calculating a mean quantity in the form of a product of two or several mean quantities, we took only such a number of decimal places, or, correspondingly, replaced units, sometimes even tens, by zeros, so that the rejection of superfluous digits rounded the given number precisely within the limits of the possible accuracy of its determination in our experiments. At the same time, all calculations were carried out with the aid of A. Nestler’s “Präzision” slide rule, corresponding in accuracy to an ordinary rule 50 cm in length.

As a measure of accuracy we found the mean error of the mean quantity $m$ by the known formula:

\[ m = \pm \sqrt{\frac{\Sigma[\delta_n]^2}{n(n-1)}}, \tag{8} \]

where $\Sigma[\delta_n]^2$ is the sum of the squares of the differences between the individual measurements and the mean value, and $n$ is the number of measurements made.

The accuracy proper $\nu$ we expressed in percent of the mean value $a$

\[ \nu = \pm \frac{100m}{a} \tag{9} \]

If the measurement was made only once, as, for example, the marking of the moment when the level of kerosene passed the point, or the weight of the whole bucket of kerosene, then the error was calculated from the corresponding possible deviation of the quantity $a$ by the quantity $p$, i.e.,

\[ \nu = \pm \frac{100p}{a}. \tag{10} \]

If some quantity $A$ is formed as the product or quotient of several quantities, then the mean error of the derived quantity, $\nu'$, is, of course, equal to the sum of the $\nu$’s of the component quantities.

If the quantity $A$ is obtained as the sum or difference of several quantities, then its mean error $\nu''$ was calculated by the formula

\[ \nu'' = \pm \sqrt{\Sigma[\nu]^2}, \tag{11} \]

where the mean error $\nu$, in %, of each separate quantity $a$ must be determined from the mean absolute error $m$, not relative to the corresponding $a$, but relative to the whole $A$, i.e., in this case

\[ \nu = \frac{100m}{a}. \tag{12} \]

This kind of calculation of mean errors, besides providing an objective measure for evaluating experiments and their results, is very useful, on the one hand, in choosing the number of significant figures in the result, and, on the other, in demonstrating at times the impossibility of attaining any special accuracy in determining a given quantity, or, conversely, in revealing a considerably greater accuracy of some quantity than is ordinarily assumed.

In general, such calculation of the accuracy of individual quantities and of the entire result is almost never practiced in technical measurements, and, as we hope to show below, quite without justification. The corresponding calculations are in general not difficult, and yield very interesting, sometimes quite unexpected, results.

Criticism and correction of records. As for the further derivation of mean figures, we considered it necessary not to confine ourselves to the mechanical computation of arithmetic means for the separate reports, but always to treat them critically; that is, when deriving mean values, to examine each separate report attentively and, if it appears doubtful, either correct it, if possible, using other data, or strike it out altogether, if possible, or, finally, even reject and discard the entire given experiment—but by no means to trust blindly a record once made. The point is that we have repeatedly convinced ourselves of the possibility, one might say the inevitability, of errors in reports or even of omissions in the records when carrying out such experiments as ours, when every second counts, and when all attention is absorbed by the various reports being made in succession.

Gasmot. 5, 1905, p. 17. ↩
Gasmot. 2, 1903, p. 166. ↩
Gasmot. 2, 1903, p. 153. ↩
Gasmot. 5, 1905, p. 35. ↩
Gasmot. 5, 1905, p. 107. ↩
Gasmot. 7, 1907, p. 3; Z. V. d. I. 1907, p. 1305. ↩
Z. V. d. I. 1908, p. 261. ↩
Z. V. d. I. 1908, p. 603. ↩
Z. V. d. I. 1909, p. 503. ↩
Gasmot. 10, 1910, p. 116. ↩
H. Dubbel, Grossgasmaschinen. Berlin, 1910. ↩
V. L. Mal’ev. Testing of the Indicator. St. Petersburg, 1905, p. 43. ↩
W. Hempel, Gasanalytische Methoden, 3 Aufl. Braunschweig. 1900. S. 25. ↩
H. Landolt u. R. Börnstein. Physikalisch-chemische Tabellen. 2 Aufl. Berlin. 1905 S. 599. ↩
A. Gramberg, Technische Messungen. 2. Aufl. Berlin 1910. S. 276. ↩
Z. V. d. I. 49, 1905 S. 569; 53, 1909 S. 254; 53, S. 327; 53, S. 1808. ↩
Schweiz Bauz. 18, 1872; Z. V. d. I. 45, 1901, p. 1343. ↩
Indicator tests, p. 17. ↩
For details see Exp. study, p. 30 ff. ↩
For a description of the corresponding apparatus see V. L. Mal’ev, “Measurement of Temperatures for Technical Purposes,” p. 163, fig. 147; for the apparatus for mercury and sulfur, ibid., p. 166, fig. 149. ↩
Measurement of Temperatures, p. 161. ↩
W. Hempel, Gasanaly. Methoden. 3 Aufl. 1900, S. 181. ↩
J. Gasbel. 1896, p. 84. ↩
Engler and Ieniaranskii give, for Baku kerosene obtained by distillation at a temperature above $200^\circ$ (Nobel), the following composition of the kerosene: $220$—$270^\circ$C, $\mathrm{C}=86.77\%$, $\mathrm{H}_2=12.76\%$, $\mathrm{O}=0.30\%$. Berl. Ber. 1895 S. 2503; Muspratt, Technische Chemie, 4 Aufl., 1898, B. 6, S. 2143. ↩

Submission history

[v1] 1912

	Page
Preface	I
Noted misprints	VIII
Abbreviations and designations adopted	X

	Page
1. Significance of two-stroke machines	1
Advantages of two-stroke machines	1
Disadvantages of two-stroke machines	2

	Page
Exhaust velocity	148
Its mean errors	151
25. Other characteristics of the experiments	152
Kerosene consumption per 1 h.p.	152
Air consumption per 1 kg of kerosene	153
Air delivery	155
Relative magnitude of the charging work	155
Scavenging characteristics	155

	Page
2. Questions to be resolved	4
3. Literature on the present question	5
Review of journal literature	6
Review of book literature	9
Conclusion	10

	Page
26. Operation of the air pump	161
Change in volume \(V_3\)	161
Influence of \(V_3\) on \(\lambda_0\) and \(\lambda\)	166
Influence of \(V_3\) on \(p_i'\) and on \(\xi\)	167
Influence of \(V_3\) on \(L_0/N_i\)	167
Influence of the number of revolutions \(n\)	167
Influence of the gaskets \(r\) (window opening)	168
Influence of the check valve	168
Experiments with air blowing	169
27. Charging of the working cylinder	170
Influence of volume \(V_3\)	173
Influence of pressure \(p_c'\)	174
Duration of scavenging	175
Relation between scavenging and exhaust	175
Influence of the number of revolutions \(n\)	175
Influence of \(n_2\)	176
Influence of gaskets \(r\)	176
Influence of compression \(\varepsilon\)	176
Operation with air blowing	177
28. Compression and combustion	178
Influence of compression \(\varepsilon\)	178
Influence of water injection	179
Ignition	180
Fuel nozzle	182
Influence of the number of revolutions \(n\)	183
Influence of compression \(\varepsilon\)	183
Relation between the pressures \(p_c\), \(p_3\), and \(p_i\)	183
Temperature of the exhaust gases	184
Content of \(CO_2\) and \(CO\)	185
Data of analysis of soot	187
Water injection	187
Use of heat	188
Influence of air blowing	188
Exhaust	188
Pressure after exhaust	193
Calculation of exhaust velocity	195
Duration of exhaust	195
Influence of shadow phenomena	196
30. Additional remarks on the operation of the machine	196
Relative magnitude of the charging work	196
Mechanical coefficient of efficiency	197
Operation of rotation of the flywheels	197
Piston friction	197
Kerosene consumption	197
Economic coefficient of efficiency	197
Loss from incomplete combustion, from radiation, and kinetic energy of the exhaust gases	198
31. General conclusions	198
Data on the operation of the tested machine and its improvement	198
Improvement of the construction and operation of machines of this type	199
Clarification of the principal questions of two-stroke operation	199
Directions for carrying out tests of internal-combustion machines	200

Port No.	Crank angles: opening	Crank angles: closing	Crank angles: duration of opening	% of piston travel: opening	% of piston travel: closing	duration of opening at \(n=340\), in sec.
1	\(-69^\circ\)	\(-47^\circ\)	\(22^\circ\)	36.4	18.4	0.011
2	\(-37^\circ\)	\(-16^\circ\)	\(21^\circ\)	11.7	2.3	0.010
3	\(-9^\circ\)	\(+10^\circ\)	\(19^\circ\)	0.6	0.8	0.009
4	\(+23^\circ\)	\(+48^\circ\)	\(25^\circ\)	4.6	19.1	0.012
5	\(+22^\circ\)	\(+87^\circ\)	\(25^\circ\)	30.3	52.8	0.012

Gas composition	Initial, %	After 24 hours: composition	After 24 hours: change	After 24 hours: in % of initial	After 48 hours: composition	After 48 hours: change	After 48 hours: in % of initial	After 72 hours: composition	After 72 hours: change	After 72 hours: in % of initial
\(\mathrm{CO_2}\)	8.0	6.4	−1.6	−20.0	6.1	−1.9	−23.8	6.0	−2.0	−25.0
\(\mathrm{O_2}\)	9.2	9.4	+0.2	+2.2	9.5	+0.3	+3.3	9.6	+0.4	+4.3
\(\mathrm{CO}\)	0.6	0.4	−0.2	−33.3	0.3	−0.3	−50.0	0.3	−0.3	−50.0
\(\mathrm{N_2}\)	82.2	83.8	+1.6	+1.9	84.1	+1.9	+2.3	84.1	+1.9	+2.3
Volume in cm\(^3\)	30.0	29.3	−0.7	−2.3%	29.1	−0.9	−3.0%	29.05	−0.95	−3.2%

Series of experiments	Gas	\(H_2O\)	\(H_2O + 0.35\,NaCl\)	\(H_2O + 0.74\,CaCl_2\)	\(H_2O + 0.54\,MgCl_2\)
I	\(CO_2\)	0.875	0.265	0.205	0.124
II	\(O_2\)	0.028	0.008	0.006	0.004
I	\(CO_2\)	0.565	0.135	0.100	0.072
II	\(O_2\)	0.015	0.004	0.003	0.002

1	2	3	4	5	6	7	8	9	10	11
No.	\(G\) kg	\(t_2-t_1\) °C	\(H_k\)	\(\Delta t_{\max}\)	\(\Delta t_{\min}\)	\(\dfrac{\Delta t_{\max}}{t_2-t_1}\) %	\(\dfrac{\Delta t_{\min}}{t_2-t_1}\) %	\(p\) mm Hg	burner position	lowering time
13	3,08	36,31	11180	2,27	3,83	6,2	10,5	275	high	slow
14	3,08	35,77	11020	2,32	2,53	6,5	7,1	325	”	rapid
15	3,19	34,80	11100	2,51	2,24	7,2	6,4	400	”	slow
16	3,21	33,33	10700	1,82	2,63	6,1	7,9	325	low	”
17	5,06	20,10	10170	0,69	0,28	3,4	1,4	500	”	rapid
18	5,84	18,65	10900	0,65	0,30	3,5	1,6	570	”	slow
19	5,98	18,54	11080	1,35	0,69	7,3	3,7	400	high	rapid
20	7,13	15,58	11110	1,17	1,03	7,5	6,6	300	”	slow
21	7,71	14,31	11020	0,78	0,42	5,5	2,9	475	”	rapid
22	8,83	12,58	11100	0,85	0,95	6,8	7,5	325	”	slow
23	12,35	8,78	10930	0,97	0,98	11,0	11,2	200	”	rapid

time	water, cold \(t_5\)	water, hot \(t_6\)	air 1 \(t_2\)	air 2 \(t_3\)	exhaust gases, mercury \(t_4'\)	exhaust gases, thermoelem. \(t_4''\)	exhaust gases, cold junction \(\Delta t_4''\)
	\(0^\circ\)	\(+7.0^\circ\)	\(-11^\circ\)	\(+2^\circ\)	\(+200^\circ\)	—	—
11 h. 42 m.	19,3	51	46,5	87	415	408	20,5
47	19,3	51	46,7	87	416	410	20,5
52	19,3	51	47	87	414	408	21,0
57	19,3	51,5	47	87	413	407	21,0
1 h. 2 m.	19,3	51	47	87	410	402	20,5
7	19,3	51	47	87	413	405	22,0
12	19	51	47,5	87	416	405	22,0
17	19	50,5	47,7	87	413	407	22,5
22	19	51	47,7	87	408	400	22,5
27	19	51	47,7	87	408	398	23,0
average	19,18	51,00	47,18	87,00	412,5	404,7	21,65
corrected	19,2	51,2	47,4	87,9	425	426	—

Experimental study of the operation of a two-stroke machine: with 3 tables and 47 drawings

Full Text

EXPERIMENTAL STUDY

OF THE OPERATION

OF A TWO-STROKE MACHINE.

PREFACE.

Preface.

Contents

Introduction

Chapter I

Chapter II

Arrangement of the Experiments

Contents

Preliminary Experiments

Conduct of the Main Experiments

Auxiliary Experiments

Contents

Appendix

Chapter III

Processing of the experimental material

Contents

Chapter IV.

Results of the Experiments.

Noticed Misprints

Noticed Misprints

Abbreviations and Symbols Used

General abbreviations:

Heat:

Pressure:

Machine:

Other quantities:

Literature abbreviations:

INTRODUCTION

1. The Significance of Two-Stroke Machines

Introduction.

§ 1.

CHAPTER I.

Chapter I. § 3.

3. Literature on the Present Question

Chapter I. § 3.

CHAPTER II.

Arrangement of the Experiments.

Chapter II. Arrangement of the Experiments. § 4.

Chapter II. Experimental Setup. § 4.

Preliminary Experiments.

Study of a Two-Stroke Machine

Study of a Two-Stroke Machine

Conduct of the Main Experiments

Table 8

Table 9

Table 10.

Table 11.

Table 12.

Auxiliary Experiments

Study of a Two-Stroke Machine

Addendum.

CHAPTER III.

Processing of the Experimental Material.

Submission history

Access Paper

Citation

Share

Related Papers

Feedback