I. V. Gebler.

On Control of the Operation of Coal-Blending Machines

For blending various grades of coal, which occurs in various practical cases (coking, briquette production, etc.), various machines are used. If the blending of coals is accompanied by their simultaneous crushing, then in such a case disintegrators are usually employed; they give a homogeneous mixture. Satisfactory blending is also achieved in hammer crushers. In some cases disintegrators are used only for blending already crushed coals, whereby the output of these machines, all other conditions being equal, increases considerably in comparison with their operation for crushing.¹ However, owing to the impacts experienced by coal particles during the operation of the disintegrator, in this case additional crushing must occur, mainly of the larger grains, and as a result—a different granulometric composition of the mixture as compared with the individual coals fed to blending, which in some cases may prove undesirable.

Furthermore, for blending crushed coals, blade screws with ordinary or shaped (Fig. 1) blades are used, as well as mixers similar to that shown in Fig. 2, which are found at new coke plants operating with the dosing of crushed coals and with high capacity. Unlike high-speed machines (disintegrators and hammer crushers), which by the nature of their operation give uniform and sufficiently complete blending, the blending effect on screws may be very different; it depends on the shape, number, and arrangement of the blades and on the length of the screw.

Fig. 1.

The mixing action of mixers of the second type (Fig. 2) must be based on the throwing aside of particles of coal falling from above by the fingers of shafts A and B and on directing them along different trajectories and at different velocities, depending on the size of the particles.

The completeness of blending in this case, evidently, must depend on the quantity of coal fed per unit time onto the working shafts of the machine, the rotational speed of these shafts, and the size, number, and arrangement of the fingers. As practice has shown, the blending of coals on such machines at a rotational speed of 50 rpm, at which they operate, proves entirely unsatisfactory.

Thus the degree of blending of coals on some machines may be very different. For a coke charge, moreover, even with satisfactory blending on one machine or another, completeness of blending—

¹ Chechott G. O., “Beneficiation of Minerals,” Issues IV and V, 1927.

may be disturbed during further movement of the charge. Such a phenomenon, called segregation, may occur as a result of jolting on the belt that feeds the charge from the blending department to the tower, and also when the charge falls into the tower bunker. All this entails the need for appropriate control.

In the present article an attempt is made to outline the ways of carrying out this control as applied to coke charge, where the degree of blending is of great importance, affecting the quality of the coke obtained.

Two such ways may be indicated:

1) Determination of the degree of blending on the basis of comparing the analytical data for the individual coals and for the blend.

2) Modeling of blending apparatuses and, with their aid, establishment of the operating conditions necessary for complete blending of the coals.

I.

In principle, a judgment as to the completeness of blending of coals could be made by comparing the data of technical analysis (ash and volatile matter), to which the individual components and the finished charge are always subjected. To determine how reliable this is in practice in terms of accuracy, let us consider several examples.

Fig. 2.

Let us suppose that a sample of the charge is taken from the belt feeding the charge to the tower. With incomplete blending, in the portion taken the individual components will not be present in the ratio in which they should be according to the specified composition of the charge; this ratio will be preserved during the subsequent division of the portion taken for the analytical sample, in which, consequently, the ash and volatile-matter content will differ from the calculated value. Numerically this may be expressed as follows.

By ash content. According to existing standards, the ash content in coke \((A_k^c)\) must not exceed 11–12%. Taking the gross yield of coke on average as 75%, the amount of ash in the coals going to coking \((A_y^c)\) must be:

\[ A_y^c=\frac{A_k^c\cdot 75}{100}=\frac{12\cdot 75}{100}=9\% \]

Under Kuzbass conditions, without coal beneficiation, this figure may be increased to 10%. The lowest ash content in coals going to coking may be taken as 7%¹.

Let us assume that, with these limiting figures for ash, the charge consists of two components in a ratio of 1:1 (Table 1).

Table 1.

The accuracy of ash determination for low-ash coals \((A^c<12\%)\) is taken as equal to 0.2% for one and the same laboratory and 0.5% for different laboratories².

Let us assume that for the above-mentioned charge, in two parallel determinations, ash \((A^c)\) was found to be 8.85% and 9.00%. With an accuracy of 0.2%, this may be accepted. The average of the two figures gives \(A^c=8.92\%\). If the amounts of coals I and II in the mixture with this amount of ash are \(x\%\) and \(y\%\), then, according to Table 1, we have:

\[ 10x+7y=8.92\times 100; \qquad x=64\%; \qquad y=36\%. \]
\[ x+y=100. \]

On the graph (Fig. 3, scale: \(1\ \mathrm{mm}=1\%\)), the normal charge of 50% coal I and 50% coal II corresponds to point \(A\); when the composition of the mixture changes, this point moves along the straight line \(BC\) in one direction or the other. For the example taken, at the ratio I coal : II coal \(=64:36\), the composition of the mixture will correspond to point \(A^1\).

The discrepancy is expressed by the segment \(AF\), which we shall call \(P\). The magnitude of this segment in the given case may vary from 0 to 50; if \(P=0\), then mixing is complete and, introducing the concept of the mixing coefficient \((k_{\mathrm{cm}})\), we may say that in this case, i.e. at \(P=0\), \(k_{\mathrm{cm}}=100\). If \(P=50\), then there is no mixing, and \(k_{\mathrm{cm}}=0\).

Fig. 3.

Fig. 4.

Thus, the value \(\kappa_{\text{mix}}\) varies along the straight line \(mn\) (Fig. 4), so that for every value of \(P\) (in mm), determined from the graph in Fig. 3, we find the mixing coefficient from the graph in Fig. 4.

For the example taken, \(P = 14\) mm, whence \(\kappa_{\text{mix}} = 72\).

If the normal charge were to consist of coals I and II in the ratio \(64 : 36\), and the mixing corresponded to the ratio \(1 : 1\), then the discrepancy would be expressed by the same value of the segment \(A = P\), and the quality of mixing by the same coefficient, equal to 72.

The limits of variation of the discrepancy and the maximum of its value, at which \(\kappa_{\text{mix}} = 0\), differ depending on the ratio of the component. Thus, for example, if the normal mixture is to consist of 70% coal I and 30% coal II, while the mixture obtained contains coal I: 80% and coal II: 20% (Fig. 3, dashed line), then the discrepancy will vary from 0 to 30. Correspondingly, the graph in Fig. 4 is constructed.

The sensitivity of determining the degree of mixing by the quantity of ash, with an accuracy of 0.2%, will be

\[ 8.50 + 0.2 = 8.70. \]

Whence, according to the preceding, we have.

Krugloe:

\[ \begin{aligned} \text{coal I: }&57\%;\\ \text{coal II: }&43\%;\\ \kappa_{\text{mix}}:&83. \end{aligned} \]

With an accuracy of 0.5%:

\[ 8.50 + 0.50 = 9.00\% \]

\[ \begin{aligned} \text{coal I: }&67\%;\\ \text{coal II: }&33\%;\\ \kappa_{\text{mix}}:&67. \end{aligned} \]

In the latter case the sensitivity is very low, i.e., with an accuracy of 0.5% one can detect only very poor mixing with a mixing coefficient no higher than 67.

As another factor for determining the degree of mixing of coals, one may take the content of volatile matter.

For coals going into coke charges, the average amount of volatile matter may be taken as the following figures¹.

Suppose that we have a charge of Osinovskii PZh (I) and Prokop’evskii PS (II) coals in the ratio \(1 : 1\) (Table II).

Table II

The permissible discrepancies in determining volatiles according to the standard are, for coking coals, for one and the same laboratory: 0.5%, and for different laboratories: 1%. Correspondingly, we have:

\[ 23.75 + 0.5 = 24.25. \]

and further:

\[ \begin{aligned} 29x + 18.5y &= 2425; \\ x + y &= 100. \end{aligned} \qquad \begin{aligned} x &= 54.7 \\ y &= 45.3 \end{aligned} \]

Thus, the mixture contains, in round numbers, coal I: 55% and coal II: 45%; for this mixture, from the preceding, we find \(k_{\mathrm{mix}} = 88\).

Allowing for a discrepancy in determining volatiles of 1%, we shall have to take for the calculation \(V^{r} = 24.75\%\).

Proceeding as before, we obtain the content in the corresponding mixture: coal I: 60% and coal II: 40%. In this case \(k_{\mathrm{mix}} = 75\).

Consequently, when volatile substances are taken into account, we have greater sensitivity in determining the quality of the mixing.

In the case of a three-component charge, let us assume that it has the following composition:

Table III

The amount of ash and volatile substances with three components will have the indicated value not only at the ratio in which they have been taken, but also at many other ratios. Let \(x\), \(y\), and \(z\) be the contents, in percent, of coals I, II, and III in the mixture; then, for the given amount of volatiles, we have:

\[ \begin{aligned} 30x + 18.5y + 22z &= 2305. \\ x + y + z &= 100. \end{aligned} \]

Whence:

\[ 8x - 3.5y = 105, \]

i.e., we obtain an indeterminate equation giving a multitude of values for the unknowns.

It is therefore necessary to take two factors into account, i.e., the amount of ash and volatile substances jointly.

Suppose that, in analyzing the mixture, the following has been obtained:

\[ A^{c} = 8.50\% \quad \text{and} \quad V^{r} = 23.55\%. \]

We have:

\[ \begin{aligned} 10x + 8y + 7z &= 850; \\ 29x + 18.5y + 22z &= 2355; \\ x + y + z &= 100. \end{aligned} \qquad \begin{aligned} x &= 38.8; \\ y &= 33.3; \\ z &= 27.9. \end{aligned} \]

Thus, the composition of the mixture will be, in round numbers: coal I = 39%; coal II = 33%; coal III = 28%.

Let us combine the coals into two groups, selecting the one whose content has changed the most compared with what should be in the normal mixture. In the present case this will be coal III; accordingly, we make the following grouping for both mixtures:

For these groups, as for individual coals, from the preceding (Figs. 3 and 4) we find \(k_{\mathrm{sm}} = 70\).

It may happen, of course, that the coals participating in the mixture will differ too little from one another, for example in ash content, so that it will be impossible to take it into account.

In such a case it may prove possible to use another index, for example the thickness of the plastic layer, which in coal mixtures, as is assumed, changes additively; in this case the layer thickness must, of course, be measured with particular care, and this parameter in a three-component charge can be taken into account together with the quantity of volatile matter.

It is also possible to apply the softness numbers proposed by the author, the logarithms of which in coal mixtures change additively¹. These numbers have the advantage that for different coals they fluctuate over a very wide range, which should give great distinctness in characterizing coals; moreover, finding these numbers is rapid and not difficult, which permits several parallel determinations to be made and thereby the final result to be refined.

The thickness of the plastic layer and the softness numbers, expressing one and the same property of coal, cannot be taken into account jointly as different indices.

In multicomponent charges consisting of more than three different coals, a larger number of indices would have to be taken into account. Thus, for example, with four coals three indices are necessary, which is possible only in individual cases, in view of the difference in the values of some indices, especially in ash content, that may prove insignificant.

With a larger number of coals, establishing the composition of the mixture by the above-indicated method becomes impossible because of the lack of indices. In such a case it is only possible, after selecting the most sensitive index, to compare its values for a coal mixture carefully prepared in the laboratory, corresponding to the given charge, and for the charge itself, using a sample taken from the mixing belt.

At the same time, however, it must be kept in mind that identical values of indices may also occur with other combinations that do not correspond to the composition of the normal mixture. Therefore, it is impossible to judge the quality of mixing from a single sample taken from the belt; it is necessary to take several, for example 10 samples, and to take into account that index which differs most from the index for the normal mixture.

Suppose that the charge is to consist of 7 coals in the following ratio:

Let us take for the coals the most sensitive indices: the quantity of volatile matter \((V)\) and the softness numbers \((M)\), which, being adopted in cir-

as numbers indicated in the two lower rows under the corresponding coals.

For such a mixture we shall have:

\[ V_c=\frac{1}{100}(22\times13+21\times20+20\times15+27\times27+18\times10+20\times10+ \]

\[ +17\times5)=22\%. \]

\[ \lg M_c=\frac{1}{100}(\lg 80\times13+\lg 70\times20+\lg 60\times15+\lg 180\times27+\lg 20\times \]

\[ \times10+\lg 65\times10+\lg 15\times5)=1.8628. \]

\[ M_c=73. \]

Let there be, in two charge samples taken from the belt:

in the first sample: coal I: \(7\%\); coal IV: \(33\%\)

in the second sample: coal IV: \(15\%\); coal V: \(22\%\).

In this case the quantity of volatiles and the value \(M\), as well as the difference in the values of these indices in comparison with the normal mixture, \(\Delta V\) and \(\Delta M\), will be:

\[ \text{for the first sample } \begin{cases} V' = 22.30\%; & \Delta V = 0.3\%,\\ M' = 69; & \Delta M = 4. \end{cases} \]

\[ \text{for the second sample } \begin{cases} V'' = 20.92\%; & \Delta V = 1.08,\\ M'' = 50; & \Delta M = 23. \end{cases} \]

Assuming that the accuracy in determining \(M\) is \(5\%\), we see that for the first mixture it is impossible to establish a difference in composition from the normal one by either of the two indices. For the second mixture this difference can be established; moreover, for volatiles the obtained difference \(\Delta V\) exceeds the accuracy of determination \((0.5\%)\) by a factor of 2, while for \(M\), in relation to the mean value

\[ \frac{73+50}{2}=61.5, \]

\(\Delta M\) is 11.5, or \(18.7\%\), i.e., it exceeds the accuracy of determination \((5\%)\) by a factor of 3.7.

Thus the softness numbers prove to be the most sensitive index for establishing the quality of mixing.

In general, however, analytical control of mixing, as is evident, is difficult.

II.

A judgment concerning the operation of a coal-blending machine with respect to the quality of mixing can in certain cases be obtained on the basis of the operation of an analogous model.

As is known, the method of modeling, based on the principle of similarity, is currently applied in some branches of engineering to study the true character of occurring phenomena, for example, in studying the motion of gases in industrial furnaces.

Having built a model of a blending machine of the desired design, with which it would be possible to work in the laboratory, one can carry out the process of blending coals or other materials, bearing in mind the greatest convenience and accuracy of subsequent control of the quality of mixing with respect to the mixture obtained. In this case the coals may be taken to differ greatly in ash and volatile-matter content, or else such free-flowing materials may be used that can easily be separated quantitatively from the mixture.

In such an investigation, all factors are taken into account which affect the quality of mixing, namely: the relative dimensions of the apparatus (which also determine productivity), the speeds of motion (rotation) of the working members, the procedure for dosing and feeding, the relation between productivity and the mixing effect under otherwise identical conditions, and, finally, the state of the material being mixed, i.e., the granulometric composition, the degree of moisture, and, in cold weather, freezing. In establishing the optimum conditions for mixing on a model, generally speaking, it may be assumed that a machine constructed analogously and operating under the same conditions will give satisfactory mixing.

However, it must be supposed that modeling can by no means be carried out with equal ease and simplicity, in the sense of achieving similarity of operating conditions and of the results obtained, for mixing machines of one type or another. It must specifically be assumed that machines in which, during operation, particles of the material being mixed move through the air space along complex trajectories, being induced to do so by the working members of the machine—as, for example, disintegrators or mixers similar to that shown in Fig. 2—would present known difficulties in establishing similarity in the operation of the machine and the model, since the character of the motion of the particles and of their mutual displacement in this case must depend directly on the dimensions of the machine and, consequently, on the space in which the motion of the particles takes place.

In a small apparatus, i.e., in a model, we may have other conditions, and therefore a detailed study of the processes taking place will be necessary in order to find correcting coefficients.

Conversely, for machines in which there is no “throwing” of particles of the material and the latter is displaced in its entire mass inside the machine, while its particles move along paths whose length is in complete dependence on the dimensions of the apparatus, and the number of moving particles corresponds to the total mass of the material, also being dependent on the size of the apparatus—for these machines the conditions of similarity must be established directly. Such machines may include, for example, mixing screws, trommels, and also coal-mixing machines similar to that shown in Fig. 5.

Fig. 5.

For such machines it is quite possible, by means of a suitably constructed model, to determine the quality of mixing, characterizing it by means of the “mixing coefficient,” and to establish the necessary dimensions of the apparatus that influence the mixing effect (for example, the length of the screw).

The corresponding data, it must be supposed, should also be applicable to a production apparatus of the corresponding design. The latter, of course, must be verified in practice.

The author carried out work in the laboratory on the investigation of the mixing process on a model of a blade screw. A screw was constructed consisting of a wooden trough and a square shaft with blades fastened to it (Fig. 6); the dimensions were as follows: length between the centers of the charging and discharge openings: 305 mm; useful volume of the trough 200 cm³; angle formed by the plane of the blade with the shaft axis 80°; the axes of neighboring blades are turned by 90°. Pitch of the screw: 20 mm; diameter of the screw: 39 mm.

For the mixing experiment, quartz sand, marble, and table salt were taken—all ground to a fineness corresponding to passage through a sieve with 1 mm openings. These substances were chosen for mixing because they could be very easily separated from the mixture by means of water and hydrochloric acid.

The weighed materials were each placed in one of the three compartments of the tin trough and from it were poured uniformly into the feed hopper of the screw. About 1 kg of the total quantity of materials was passed through the apparatus, which required about one minute.

In the middle of this period, about 15 g of the mixture was taken from the discharge opening. In the case of a ternary mixture, the selected sample was treated with hot water until the sodium chloride was completely dissolved, after which filtration was carried out through a weighed filter, which, after washing and drying, was weighed; by subtracting the weight of the dry residue on the filter

Fig. 6.

from the weight of the sample, the content of table salt in the mixture was determined. Next, the same residue was treated with dilute hydrochloric acid to dissolve the marble, and its quantity was determined in an analogous manner.

The binary mixture consisted of sand and marble and was treated only with hydrochloric acid.

Thus, in the mixture obtained, the quantities of the materials entering into it were determined. The results obtained in these experiments are as follows:

Binary mixture: sand + marble in the ratio 1:1.

Table IV

Thus, with a triple pass through the screw, very good mixing is achieved.

Triple mixture: sand + marble + table salt, in the ratio 1:1:1.

Table V

In this case, as can be seen, good mixing is achieved with a double pass, and subsequent passes of the mixture through the screw do not increase the mixing coefficient, the degree reached by it for the given screw being, apparently, the limiting one. It is quite natural that, for a mixture of three components, good mixing is achieved with a smaller number of passes, i.e., with a shorter working length of the screw as compared with a binary mixture, since with a larger number of varieties of heterogeneous particles their coming together occurs more quickly.

On the basis of the experiments performed, we can determine the screw length necessary for satisfactory mixing.

The working length of the screw was 305 mm with a screw diameter \(D = 39\) mm.

With a double pass, we therefore have 610 mm, or \(\frac{610}{39} = 15D\); with a triple pass: 915 mm, or \(\frac{915}{39} = 23D\).

Thus, the working length of the mixing screw is from 15 to 23 screw diameters. A length equal to \(23D\) evidently guarantees satisfactory mixing in all cases.

In an analogous manner, mixing apparatuses of another type can be investigated, with the establishment of the conditions for their operation.

1. “Coke and Chemistry,” 1935, No. 10; 1936, No. 11. ↩↩↩

2. \(\dfrac{\mathrm{OST}}{\mathrm{VKS}}-7151\) (s I) I 1935, p. 6. ↩