IMPROVEMENT OF THE METHODOLOGY FOR TESTING COKE STRENGTH IN THE HOT STATE

N. M. Smol’yaninova, A. M. Stukalov, A. V. Vikharev

(Presented by the Scientific-Methodological Seminar of the Chemical Technology Faculty)

The intensification of the blast-furnace process and the construction of large-volume blast furnaces are leading to higher requirements for the quality of coke and, above all, for its physicomechanical properties as a loosening agent of the blast-furnace burden.

On the other hand, the expansion of the raw-material base for coking, the use of low-deficit coals with reduced caking capacity for producing coke, and the creation of fundamentally new continuous coking methods lead to the production of coke with a structure different from that of conventional coke and, consequently, with different properties, including physicomechanical ones.

All this requires new, more in-depth methods for investigating and evaluating the mechanical properties of coke that would make it possible to predict the behavior of coke in the blast-furnace process.

Existing methods for evaluating the mechanical strength of metallurgical coke (drum tests, the drop-shatter method) yield results that are not fully satisfactory for production. This is explained by the discrepancy between the conditions under which mechanical strength is tested and the actual conditions occurring in the blast furnace.

In drums, coke is destroyed only under the action of mechanical forces (abrasive and crushing), whereas in the blast furnace the destruction of pieces occurs under the simultaneous action of mechanical forces, high temperatures, and reacting gases. Therefore, for a more correct evaluation of blast-furnace coke quality, the study of the degree and character of destruction of coke pieces under conditions approximating those in the blast furnace is of greater practical interest.

Most studies devoted to this question have shown that, under the simultaneous action of mechanical forces and high temperatures, coke destruction is always greater than in the cold state.

At the Department of Chemical Technology of Fuels of TPI, a method was proposed for testing coke strength in the hot state, and a study was carried out of the combined influence of thermal and mechanical forces, as well as of the gas medium, on structural strength over a wide temperature interval¹. This method, intended for laboratory studies,

revealed a number of interesting regularities in the behavior of coke during secondary heating.

However, when various types of coke were studied, both industrial and laboratory samples differing in structure, it proved difficult to obtain good agreement between parallel determinations. Sometimes the difference reached 10 percent or even more. In our opinion, the reason for this is an insufficiently substantiated testing procedure. This applies both to the choice of the initial size of the coke pieces (15–20 mm) and to the choice of the number and diameter of the balls. In creating the method and developing the procedure, we took as a basis the method of the All-Union Institute of Mineral Raw Materials²), intended for determining the mechanical strength of coal by abrasion.

In order to increase the reliability of the indicated method, we carried out work to improve it. The influence on the structural strength of coke and on the reproducibility of the results was studied for the following factors: 1) sample weight; 2) size of the pieces; 3) number and size of the balls; 4) procedure for preparing the sample for analysis and evaluating the test results.

Kemerovo coke was used for the experiments; it was very nonuniform in quality and heterogeneous in structure. To obtain a more homogeneous material, the parts of the coke pieces adjacent to the oven-wall seam were separated before the sample was prepared. Some of the samples were prepared from whole pieces, i.e., without removing the sponge.

The coke charge for testing mechanical strength was obtained by crushing the coke in a jaw crusher, thoroughly mixing the resulting fraction, and reducing it (quartering) to the required size.

The experiments were carried out in series, mainly with five parallel tests in each series. For the series, the arithmetic mean values of abrasion in percent (I) were calculated. Then the deviations of the results of each experiment from the mean value for the given series were determined.

The arithmetic mean of these deviations is termed the “mean deviation from the mean abrasion value in absolute percent” (A). This value, expressed as a percentage relative to the mean for the series, is called the “mean deviation from the mean abrasion value in relative percent” (B).

Table 1 presents the dependence of the accuracy of the results for determining abrasion on the weight of the coke sample taken for analysis. The experiments

Table 1

²) B. E. Koifman. “A Micromethod for Testing the Mechanical Resistance of Coals,” Zavodskaya laboratoriya, No. 6, pp. 741–745, 1947.

were conducted with a lump size of 15–20 mm and a total ball weight of 200 g with a ball diameter of 15 mm. The sample weight, equal to 15 g, corresponds to the old procedure.

It follows from the table that increasing the sample weight improves the agreement of the results. Apparently, for this lump size the sample becomes more representative, if only because the number of pieces in it increases.

The influence of ball size, at the same lump size and ball weight, but with a sample weight of 25 g, corresponding to sufficiently high agreement of the results, is presented in Table 2.

Table 2

As the ball size decreases, the accuracy of the determinations increases. This fact can apparently be explained by the greater abrasive action of the larger balls, i.e., for the given size of the coke specimens tested (15–20 mm), the heterogeneity in the size and shape of the lumps in the sample is manifested more strongly when large balls are used. However, it is inadvisable to reduce the ball diameter while retaining the previous size of the coke lumps to be tested, since this will lead to a decrease in the sensitivity of the method.

The same experiments, carried out with a longer drum rotation time (1 hour instead of the 20 min adopted in the procedure), showed that the accuracy of the method is independent of this factor.

Doubling the weight of the balls, for a given ball size, increases the accuracy of the analysis.

A study carried out with coke pieces of 15–20 mm showed that this class does not make it possible to obtain high agreement of the results—the discrepancy is from 3 to 10% (relative). With this size, only 6–8 pieces enter the sample, and these can hardly represent the average properties of a certain batch of coke, which is always heterogeneous in structure. Therefore it was decided to increase the number of pieces in the charge by reducing their size, and also to make them more uniform in coarseness, i.e., to narrow the class. In addition, to improve the reproducibility of the test results it would have been possible to increase the charge to 20 and even 30 g, as against the 15 g stipulated by the procedure. However, we left the charge unchanged, since in testing laboratory coke samples one usually has to deal with small quantities of it, which makes it difficult to obtain the number of samples necessary for parallel determinations.

Experimental data on determining the influence of lump size at different ball diameters and weights are given in Table 3.

It can be seen that for the 10–12 mm class the best agreement of the results (B—1.93%) was obtained with a ball size of 9.5 mm and a ball weight of 400 g. For samples composed of pieces 5–7 mm in size, the best results are shown by 8- and 5.5-millimeter balls weighing 400 g—2.81 and 2.40%, respectively, in experiments (Nos. 12 and 13).

In general, the reproducibility of the results when testing the 5–7 mm class is significantly higher, in comparison with the other classes, at the same charge size.

Table 3

The data in Table 4 show that, as the ball diameter decreases, the discrepancy in the results of parallel tests also decreases, especially for the fine size classes.

Apparently, larger balls (13, 15 mm) exert too coarse an action on the small grains, rounding off sharp corners and crushing thin, elongated pieces. The intensified crushing action of large balls introduces an additional error into the results, since the deviation from the mean value of abrasion depends not only on the difference in abrasion of parallel coke charges, but also on the difference in their crushability—owing to the irregular and varied shape of the pieces.

Small balls exert mainly an abrading action; therefore, the results of parallel tests prove to be closer.

But excessively small balls (5 mm and less) should not be used, since their abrading action on the coke will be insufficient. Balls 8 mm in diameter already have a quite sufficient abrading effect, giving quite acceptable results when testing various coke samples, including those differing greatly in strength. This is confirmed by the results of experiments for various coke samples of the 5—7 mm class, with a ball diameter of 8 mm and with a ball weight of 400 g, presented in Table 5. Here the abrasion of coke was evaluated by the yield of the class below 1 mm.

Table 4

Returning to the data of Table 3, it may be noted that for the 10–12 mm class no regular dependence of the values of A and B on the size of the balls was observed (experiments 5–8).

It should be emphasized that most of the experiments were carried out by us with balls taken in an amount of 400 g, since, as indicated above, with an increase in their weight from 200 to 400 g the accuracy of the analysis increases (Table 3). It is possible that, as the weight of the balls increases, their number also increases and, consequently, the number of impacts falling on one piece of the charge. Each grain receives approximately the same number of mechanical actions, which increases the uniformity of abrasion of the charge as a whole.

In addition, in the course of the work the abrasion index adopted in the initial version of the procedure was subjected to critical evaluation. For this purpose, experiments were carried out with determination of the amount of coke breeze passing not only through a 3 mm sieve, as in the old procedure, but also through 2 and 1 mm sieves (see Table 3). It turned out that for coke of 5–7 mm size more reproducible results are obtained when determining the yield of the 1 mm class, which characterizes abrasion. For the 10–12 mm class no such regularity was observed.

The data on the value of abrasion (I), as well as the indices of analysis accuracy (A and B, %), presented in Table 5, were obtained under the optimum conditions found in the preceding experiments, but not for a batch of coke from whose pieces the sponge was separated, but for individual pieces with sponge, i.e., for material more heterogeneous than in the preceding experiments. It can be seen that the agreement between all experiments is quite sufficient—the error is not more than 3% relative.

Table 5

Experiments a and b were carried out for two batches obtained (after thorough mixing and division of the fraction into two equal parts) from one and the same piece of coke.

It can be seen from Table 5 that if, for one half, good agreement was observed between the experiments, then for the other, obtained from the same piece of coke, the agreement was somewhat worse.

Despite thorough mixing and quartering of the primary sample, the error is apparently introduced precisely during these operations, as a result of which grains that differ significantly in structural strength get into the different halves of the fraction.

It follows from this that, when preparing a sample for testing mechanical strength, the most thorough mixing is necessary before quartering. Mixing must always be carried out by a definite method, with the same number of operations in all cases and the same method of crushing.

Thus, on the basis of the experiments carried out, it may be asserted that, with proper taking of the charge from the primary sample, discrepancies between parallel tests will not exceed 3% for the 5–7 mm fraction, ball diameter 8 mm and ball weight 400 g, with a charge size of 15 g.

Recommended Method for Testing the Structural Strength of Coke on the Laboratory Installation of the Department of Chemical Technology of Fuels

The starting material (coke, semicoke, etc.) is crushed in a jaw crusher to 5–7 mm. After separation of the 5–7 mm class, the coke grains constituting the sample are mixed 5 times “into a cone” with a scoop, after which 15 g portions are obtained by quartering (a divider may be used). Weighing is carried out with an accuracy of 0.1 g.

Before the experiment, 400 g of balls with a diameter of 8 mm are loaded into the drum. The subsequent course of the tests is carried out according to the previously adopted procedure.

The abrasion index is taken as the amount of coke fines that, after testing, pass through a 1 mm sieve, expressed as a percentage of the initial weight of the sample. (It is better to determine this value from the difference between the initial weight of the coke and the weight remaining on the 1 mm sieve, since fine dust becomes lodged in the slots and is difficult to account for.)

Testing the strength of coke in the hot state by this method over a broad temperature range confirmed the regularities previously observed in the behavior of coke during secondary heating (Fig. 1). It should merely be noted that the minimum abrasion shifts into the region of lower temperatures.

Fig. 1. Abrasion of coke as a function of the temperature of secondary heating (coke sample from the Kemerovo Coke and Chemical Plant).

The indicated phenomenon is, in all probability, associated with more rapid heating of the fine coke grains and, consequently, with earlier relief of internal stresses, in comparison with larger pieces under the previous method.

1. N. M. Smol’yaninova, A. A. Kaplin, L. M. Vasil’eva. “Investigation of Coke Strength in the Hot State,” Coke and Chemistry, No. 5, pp. 25–28, 1961. ↩