Method for estimating variation in coke quality

Abstract

To provide a method for estimating variation in coke quality that enables estimation of within-lot variation of a coke quality characteristic value within a unit period from information obtainable during normal operation.SOLUTION: A method for estimating variation in coke quality comprises applying a local level model of a state-space model to time-series data of a coke quality characteristic value for each lot in a unit period where a target coke quality characteristic value is the same in a coke oven, and calculating an observation error of the state-space model as an index of within-lot variation of the coke quality characteristic value.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a method for estimating variation in coke quality, which estimates the variation in quality characteristic values ​​of coke produced in the operation of a coke oven. [Background technology]

[0002] In order to control the quality of coke produced by carbonizing coal in a coke oven, the produced coke is periodically sampled and the quality of the obtained samples is analyzed. For example, the quality of the coke, such as the drum strength (DI) and mean particle size (MS), affects blast furnace operation and is closely monitored in the coke production process. The quality of the coke is measured at any time, and based on the results of the sample analysis, it is confirmed whether coke that meets the required quality is being produced, and the operation of the coke oven and the blending of coal for producing coke are controlled.

[0003] Regarding coke quality control, for example, Patent Document 1 discloses a coke quality control method that separates the total variation in coke strength of multiple lots produced in a unit period into intra-lot variation and inter-lot variation, and separates the intra-lot variation into test variation and non-test variation. By isolating the causes of the total variation, it becomes possible to identify the dominant factor for the total variation from the multiple variation factors. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent No. 7028096 Summary of the Invention [Problem to be solved by the invention]

[0005] In the technology described in Patent Document 1, in order to calculate the intra-lot variation in coke strength, it is necessary to sample the coke based on the alternating sampling method. However, in general, the coke sampling usually performed in operations is not based on the alternating sampling method. Therefore, in order to implement the technology described in Patent Document 1, in addition to the work usually performed in operations, a new work of sampling the coke based on the alternating sampling method (hereinafter also referred to as "alternating sampling") must be performed, and this requires securing personnel due to the increased labor.

[0006] Therefore, the present invention has been made in consideration of the above problems, and an object of the present invention is to provide a method for estimating variation in coke quality, which is capable of estimating the within-lot variation of coke quality characteristic values ​​within a unit period from information that can be obtained during normal operation. [Means for solving the problem]

[0007] In order to solve the above problem, according to one aspect of the present invention, there is provided a method for estimating variation in coke quality, which applies a local level model of a state space model to time series data of coke quality characteristic values ​​for each lot in a unit period during which a target coke quality characteristic value is the same in a coke oven, and calculates the observation error of the state space model as an index of variation within a lot of the coke quality characteristic values.

[0008] Alternatively, an estimated value of intra-lot variation may be calculated from the observation error of the state space model. In this case, a correlation equation between an actual measurement value of intra-lot variation of coke quality characteristic values ​​obtained by alternately sampling coke and the observation error of the state space model may be calculated in advance, and the estimated value of intra-lot variation may be calculated from the observation error of the state space model using the correlation equation. Effect of the Invention

[0009] As described above, according to the present invention, it is possible to estimate the intra-lot variation of coke quality characteristic values ​​within a unit period from information that can be obtained during normal operation. [Brief description of the drawings]

[0010] [Figure 1] FIG. 2 is an explanatory diagram for explaining a change in the variation in coke strength during operation. [Diagram 2] 1 is a flowchart illustrating an example of a method for estimating intra-lot variation using a intra-lot variation index. [Diagram 3] 13 is a flowchart showing an example of a process for calculating a correlation equation between a state space model observation error σ and an actual measured value σ1A of variation within a lot. [Figure 4] FIG. 1 is an explanatory diagram for explaining an alternating sampling method. [Diagram 5] 1 is a graph showing, as an example, a correlation equation obtained from the state space model observation error (σ) and the actual measurement value of intra-lot variation (σ1A) in coke ovens 1 to 4 in Table 1. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In this specification and the drawings, components having substantially the same functional configurations are designated by the same reference numerals, and duplicated explanations will be omitted.

[0012] [1. Overview] Coke quality control is performed by sampling the coke produced in a coke oven and analyzing the quality of the sample obtained by sampling. For example, coke strength is measured as a quality characteristic value that indicates the quality of the coke, and the quality of the coke is controlled. In the following, coke strength is taken as an example of the coke quality characteristic value, but the coke quality characteristic value may be, for example, mean particle size (MS), coke reacted strength (CSR), etc.

[0013] In this embodiment, a production unit of coke produced within a time unit (hereinafter also referred to as a "lot time") during which a sample is taken at least once to measure coke strength for quality control is defined as a "lot". The lot time is usually set to about one day to four hours. The unit period is a period during which the coke strength targeted in the coke oven is the same, and includes a plurality of lots, and is set to a relatively long period of, for example, about one month or longer. For example, as shown in FIG. 1, the lot time is set to 8 hours and the unit period is set to one month. In this case, the coke strength of three lots is measured in one day, and the coke strength of three lots is measured in the unit period (the number of days in the month). In this case, the variation in coke strength in one month is the total variation in the unit period.

[0014] As shown in Figure 1, the measured coke strength varies between lots (lot-to-lot variation). Lot-to-lot variation represents the total variation minus the variation within a lot. Lot-to-lot variation includes variation caused by long-term operation, such as variation due to changes in the blending of raw coal, variation due to changes in operational control conditions such as particle size and temperature, and variation due to changes in the properties of the same brand of coal.

[0015] Furthermore, there is variation in coke strength even within the same lot (intra-lot variation). Intra-lot variation is classified into test variation caused by sample testing (e.g., variation during coke sampling, variation in quality test conditions, etc.) and non-test variation caused by factors other than test variation (e.g., variation in the soundness (combustibility) of the manufactured coke oven body (intra-kiln / inter-kiln variation), etc.).

[0016] In this way, the total variation in a unit period includes the lot-to-lot variation and the intra-lot variation of the coke strength of each lot. Therefore, the total variation in a unit period changes depending on the variation within the same lot and the variation between lots, and the variation factors that have a large effect on the total variation also change.

[0017] For example, as disclosed in Patent Document 1, if the factors of the total variation in coke strength can be separated, it is possible to identify the factor that is dominant over the total variation from among multiple variation factors, and the variation in coke strength can be appropriately suppressed by strengthening the variation reduction measures for the identified variation factor. However, as described above, in the technology described in Patent Document 1, alternating sampling must be performed in addition to the work normally performed in the operation in order to calculate the intra-lot variation in coke strength.

[0018] Therefore, in the method for estimating variation in coke quality according to the present embodiment, the variation within a lot of a coke quality characteristic value (e.g., coke strength) measured within a unit period is estimated from information that can be obtained during normal operation without alternate sampling, thereby making it possible to easily control the quality of the coke. The method for estimating variation in coke quality according to the present embodiment will be described in detail below.

[0019] [2. Method for estimating variation in coke quality] [2-1. Estimation of intra-lot variation using intra-lot variation index] First, as an example of a method for estimating variation in coke quality according to the present embodiment, a method for estimating variation in a lot using an index (intra-lot variation index) representing the magnitude of variation in a lot of a coke quality characteristic value measured within a unit period will be described with reference to Fig. 2. Fig. 2 is a flowchart showing an example of a method for estimating variation in a lot using the intra-lot variation index.

[0020] (S11: Acquire time series data of coke quality characteristic values ​​for each lot in a unit period) First, time-series data of the coke quality characteristic value for each lot in a unit period during which the target coke quality characteristic value in a coke oven is the same is obtained (S11).

[0021] Here, the target coke quality characteristic value (e.g., coke strength) is set for each coke oven, and the target coke quality characteristic value is not necessarily the same for different coke ovens. Furthermore, even for the same coke oven, the target coke quality characteristic value may be changed depending on the properties of the incoming coal and the operating status of the blast furnace. In step S11, time series data of the coke quality characteristic value for each lot is acquired during a unit period in which the target coke quality characteristic value set for each coke oven is the same.

[0022] The time series data includes a plurality of lot data in which the measured coke quality characteristic values ​​are associated with lot identification information that can identify the lot from which the coke quality characteristic values ​​were obtained. The lot identification information may be any information that can identify the chronological order of the measured coke quality characteristic values. For example, it is preferable to use the sampling time when the coke is sampled, or the time when the coke quality characteristic values ​​are measured may be used.

[0023] In addition, the time intervals of the lot data in the time series data may be uniform or non-uniform. For this reason, a code representing the sampling date and the number of samplings (indicating which number of samplings on that day) may be used as the lot identification information. For example, if the first sampling is on August 1, 2023, a code such as "2023080101" may be used as the lot identification information. Alternatively, the lot identification information may be expressed by the sampling date and a code representing the worker's working time zone. For example, if the working time zone of a day is divided into three zones, "time zone A," "time zone B," and "time zone C," if the time zone is on August 1, 2023, a code such as "20230801A" may be used as the lot identification information.

[0024] If the unit period of the time series data is too short, the accuracy of the intra-lot variation index estimated using the time series data will be low, and if it is too long, it will be difficult to capture changes in the variation of the coke quality characteristic values. The unit period of the time series data can be set appropriately taking these trends into account. For example, if the coke quality characteristic values ​​of three lots are to be obtained per day, the unit period should be about one month.

[0025] The time-series data for a unit period may be missing some lot data. For example, the coke quality characteristic value may not be measured for some reason, or the measurement result may contain a clear error. In such a case, the lot data for the coke quality characteristic value may be excluded.

[0026] Moreover, if there is an obvious outlier in the coke quality characteristic value, it is considered that it is due to a problem other than the one assumed to be the target of this evaluation. Therefore, it is desirable to exclude the lot data of the coke quality characteristic value from the time-series data of the unit period. The criterion for excluding lot data from the time-series data of the unit period may be, for example, to calculate the standard deviation of the coke quality characteristic values ​​for all lot data in the unit period and exclude those that fall outside the range of the average value of the coke quality characteristic values ​​±3 × standard deviation.

[0027] Furthermore, if there is a change in the operating conditions during a unit period, it is desirable to exclude the lot data for that period from the time-series data. For example, a period in which wet quenching is performed in contrast to a period in which dry quenching is normally performed is a period in which the operating conditions are changing, and therefore it is desirable to exclude the lot data for that period from the time-series data. Note that if the period in which lot data is excluded is long, the number of data used to calculate the intra-lot variation index of the coke quality characteristic value described below is reduced, and the estimation accuracy of the intra-lot variation index is reduced. In such a case, for example, the unit period may be extended by the period in which the operating conditions change, so that a sufficient number of data used to calculate the intra-lot variation index of the coke quality characteristic value is ensured.

[0028] (S13: Calculation of intra-lot variation index) Next, the time series data of the coke quality characteristic value for each lot in the unit period acquired in step S11 is used to calculate the intra-lot variation index of the coke quality characteristic value (S13). In this embodiment, the intra-lot variation index is calculated by applying a state space model to the time series data. In the state space model, an analyst can define a model to be applied to a phenomenon to be analyzed (e.g., transition of coke strength). In this embodiment, which analyzes the time series change of the coke quality characteristic value, for example, a local level model shown in the following formula (1-1) and formula (1-2) is applied.

[0029]

number

[0030] Here, t is the time point in the time series data, and y t is the coke quality characteristic value (actual measurement value), μ t is the coke quality characteristic value (estimated value). ε t and η t is the error term, and ε t has a mean value of 0 and a standard deviation of σ ε The normal distribution is t has a mean value of 0 and a standard deviation of σ η It follows a normal distribution with standard deviation σ ε and σ η is constant during the analysis period. That is, y t is the known data, and μ t , ε t , η t , σ ε and σ η are the unknown parameters to be estimated. In particular, σ ε is called the state space model observation error σ.

[0031] These unknown parameters may be estimated, for example, by using a maximum likelihood estimation method. That is, the case where the probability (likelihood) that the unknown parameters can take is highest for given known data (i.e., coke quality characteristic values ​​(actual measurements)) may be calculated, and the parameter value at the maximum probability may be adopted as the estimated value. The unknown parameters may be calculated using statistical analysis software. When the local level model of the state space model is applied to analyze the time series data, the time series data may be analyzed by applying a Kalman filter, which is a process mathematically equivalent to the local level model.

[0032] Here, we consider the physical meaning of the state space model applied to the coke manufacturing process. First, the coke quality characteristic value (estimated value) μ t Ideally, it can be considered that there is no lot-to-lot variation in the coke quality characteristic value (estimated value) μ t is based on the fact that it can be considered as the coke quality characteristic value when all the coke in the lot is sampled and measured. Therefore, the coke quality characteristic value (actual value) y t and coke quality characteristic value (estimated value) μ t The error term ε t The value of corresponds to the variation within a lot. And this error term ε t Standard deviation of σ ε (That is, the state space model observation error σ) can be said to express the magnitude of intra-lot variation. The state space model observation error σ is a quantitative measure of intra-lot variation, and can be used as an index (intra-lot variation index) that expresses the degree of intra-lot variation.

[0033] In this manner, in step S13, the local level model of the state space model is applied to the time series data of the coke quality characteristic values ​​for each lot in a unit period, and the observation error of the state space model is calculated as an index of intra-lot variation of the coke quality characteristic values.

[0034] The state space model observation error σ, which is an index of intra-lot variation, may be used to, for example, perform process control of a coke manufacturing process. For example, assume that the state space model observation error σ is calculated periodically (e.g., once a month) for the same furnace. In this case, if the state space model observation error σ becomes large, it can be inferred that the intra-lot variation (e.g., variation at the time of sampling of coke, variation in quality test conditions, etc.) has become large, rather than the inter-lot variation (e.g., variation due to a change in the raw coal mixture, variation due to changes in operational control conditions such as particle size and temperature, variation due to changes in the properties of the same brand of coal, etc.).

[0035] In this way, the cause of the increase in the total variability can be identified using the state space model observation error σ without performing alternate sampling, which makes it possible to quickly take action to reduce the variability.

[0036] [2-2. Calculation of estimated intra-lot variation] The state space model observation error σ calculated by the flowchart shown in Figure 2 is a quantitative measure of the variation within a lot. Therefore, the actual value of the variation within a lot σ1 obtained by performing alternating sampling is A However, the state space model observation error σ and the actual intra-lot variation σ1 A Therefore, the state space model observation error σ and the actual measurement value of the lot variation σ1 A The correlation equation is calculated in advance, and the intra-lot variation estimate σ1 is calculated from the state space model observation error σ by using the correlation equation. E This allows the intra-lot variation estimate σ1 to be calculated without the need for alternate sampling, which is a normal operation. E can be calculated.

[0037] Based on Figure 3, the state space model observation error σ and the actual measurement value of the variation within the lot σ1 A The calculation process of the correlation equation between the state space model observation error σ and the actual measurement value of the variation within the lot σ1A 13 is a flowchart showing an example of a process for calculating a correlation equation with

[0038] (S21: Actual measurement value of variation within a lot by alternating sampling σ1 A (Acquisition of First, multiple alternating sampling is performed to obtain the actual intra-lot variation value σ1 A In step S21, for example, as described in Patent Document 1, a double sample is taken based on an alternate sampling method based on JIS-M8811 (coals and cokes-sampling and sample preparation method), and the coke quality characteristic values ​​are measured. The actual variation value σ1 within the lot is calculated from the measured coke quality characteristic values. A may be calculated.

[0039] As described in Patent Document 1, in the alternate sampling method, as shown in FIG. 4, first, when sampling coke at equal intervals for each sampling time within a lot, the sampled coke is alternately divided to create two sample aggregates (hereinafter referred to as "alternate samples") Ai and Bi. Next, for each of the alternate samples Ai and Bi, the coke is separated into a plurality of particle size divisions, and the ratio of the coke of each division contained in the entire alternate sample is calculated, and then two samples (duplicate samples) are created as specimens for strength measurement according to the ratio. That is, as shown in FIG. 4, a pair of two samples Ai-1 and Ai-2 (duplicate samples) is created from the alternate sample Ai, and a pair of two samples Bi-1 and Bi-2 (duplicate samples) is created from the alternate sample Bi.

[0040] In addition, the actual variation value σ1 within multiple lots A may be obtained by alternate sampling at different coke ovens, or may be obtained by alternate sampling at the same coke oven at different time periods.

[0041] Next, the values ​​of the coke quality characteristic values ​​are measured for the duplicate samples prepared based on the alternate sampling method. As an index of coke strength, which is an example of a coke quality characteristic value, drum strength as described in JIS-K2151 may be used. The drum strength is measured by charging a predetermined amount of coke (10 kg) into a cylindrical drum with a diameter of 1500 mm and a length of 1500 mm, rotating the drum 30 or 150 times at 15 rpm, sieving the coke through sieves of 50 mm, 25 mm, 15 mm, and 6 mm, and expressing the strength for each rotation as a percentage of the sieved mass relative to the charged mass. Generally, the 15 mm sieve weight percentage index (15 mm) after 150 rotations is widely used, and the DI 150 15 It is written as follows.

[0042] Then, the actual variation value σ1 within the lot was calculated from the measured coke quality characteristic value. A Calculate the actual value of variation within a lot σ1 A is calculated from the following formula (2) based on JIS-M8811.

[0043]

number

[0044] Here, n' is the number of increments constituting the alternating samples Ai, Bi, and in the example shown in FIG. R is the range of measurements versus R i The average value of the measured values ​​is expressed by the above formula (2-1). i As expressed by the above formula (2-2), is the absolute value of the difference between the coke strength DI(Ai) of the alternate sample Ai and the coke strength DI(Bi) of the alternate sample Bi. n is in the range R i and corresponds to the number of tests based on the series of alternate sampling methods for measuring coke strength, which were performed using samples sampled at the lot times shown in Fig. 4. For example, if tests based on the series of alternate sampling methods for measuring coke strength were performed 10 times using samples sampled at one lot time in Fig. 4, n=10. d 2 is the coefficient that estimates the standard deviation from the range R. For example, in the case of two data, 1 / d 2 =0.8862.

[0045] (S23: Calculation of state space model observation error σ) Next, the local level model of the state space model is applied to each of the time series data of the coke quality characteristic values ​​for each lot in the unit period including the period during which the alternating sampling was performed in step S21, and the state space model observation error σ is calculated (S23). The process of step S23 may be performed based on the flowchart shown in FIG.

[0046] The number of lot data included in the time series data used to calculate the state space model observation error σ in step S23 does not necessarily have to match the number of times of alternating sampling performed in step S21. The number of times of alternating sampling in step S21 may be less than the number of lot data included in the time series data, but according to JIS-M8811, alternating sampling should be performed multiple times (at least 10 times). Therefore, in the unit period of the time series data used in step S23, the alternating sampling in step S21 should be performed at least 10 times.

[0047] (S25: State space model observation error σ and actual value of intra-lot variation σ1 A (Obtaining the correlation equation with Then, the actual variation value σ1 within the lot acquired in step S21 A Based on the state space model observation error σ obtained in step S23, a correlation equation expressing the correlation between them is obtained (S25). For example, A The correlation equation may be an approximation equation obtained, for example, by the least squares method from a set of and the state space model observation error σ. The correlation equation may be, for example, a monotonically increasing linear function. Furthermore, the correlation equation may be another function, such as a quadratic function, that monotonically increases within the range in which the intra-lot variation is estimated.

[0048] Above, the state space model observation error σ and the actual measurement value of the intra-lot variation σ1 A The calculation process of the correlation equation with the state space model observation error σ is actually performed in operation. E Then, during operation, the correlation equation is used to calculate the intra-lot variation estimate σ1 from the state space model observation error σ calculated based on the flowchart of FIG. E can be calculated.

[0049] The method for estimating variation in coke quality according to the present embodiment has been described above. According to the present embodiment, a local level model of a state space model is applied to time-series data of coke quality characteristic values ​​for each lot in a unit period in which the target coke quality characteristic value is the same, and the observation error of the state space model is calculated as an intra-lot variation index of the coke quality characteristic value. This makes it possible to estimate the degree of intra-lot variation without alternate sampling, and to easily perform coke quality control.

[0050] In addition, by calculating a correlation equation between the state space model measurement error and the actual measurement value of the intra-lot variation in advance, it is possible to calculate an estimated intra-lot variation value from the state space model measurement error, which makes it possible to grasp the intra-lot variation of the coke quality characteristic value more specifically. EXAMPLES

[0051] For the different coke ovens 1 to 5, alternating sampling was performed in a unit period in which the target coke strength set for each coke oven was the same, and data acquisition was performed to obtain time-series data of the coke strength for each lot in that unit period. In other words, the target strength for each coke oven is different. Then, based on the flowchart shown in Figure 2, a local level model of the state space model was applied to the time-series data of coke strength, and the observation error (σ) of the state space model was calculated. In addition, the actual intra-lot variation value of coke strength obtained by alternating sampling (σ1 A The unit period was one month. The data acquisition times for coke ovens 1 to 5 were different.

[0052] To verify the accuracy of this method, the state space model observation error (σ) and the actual intra-lot variation (σ1 A ) and the data from coke oven 5 were used to calculate the actual intra-lot variation (σ1 A ) and the estimated intra-lot variation calculated from the correlation equation (σ1 E ) was compared.

[0053] Table 1 shows the state space model observation error (σ) and the actual measured value of the variation within the lot (σ A ) is shown. In Fig. 5, the state space model observation error (σ) and the actual measured value of the variation within the lot (σ A ) and the correlation equation obtained from

[0054] [Table 1]

[0055] As shown in Figure 5, the state space model observation error (σ) and the actual measurement value of the variation within the lot (σ A) a good linear function was obtained as the correlation equation with σ. This confirmed that the state space model observation error σ can be an index for evaluating intra-lot variation, and also showed that it is possible to construct a correlation equation for a linear function.

[0056] In addition, in Table 2, the correlation equation shown in Figure 5 is used to calculate the intra-lot variation estimate (σ1 E ) are calculated and the results are shown below.

[0057] [Table 2]

[0058] As shown in Table 2, the actual variation within the lot of coke oven 5 (σ1 A ) and the estimated intra-lot variation (σ1 E ) has an error rate of 2.2%, which shows that the within-lot variation calculated from the state space model observation error (σ) can be estimated with high accuracy using the correlation equation.

[0059] Although the preferred embodiment of the present invention has been described in detail above with reference to the accompanying drawings, the present invention is not limited to such an example. It is clear that a person having ordinary knowledge in the technical field to which the present invention pertains can conceive of various modified or altered examples within the scope of the technical ideas described in the claims, and it is understood that these also naturally belong to the technical scope of the present invention.

Claims

1. A local level model of the state space model is applied to time series data of coke quality characteristic values ​​for each lot during a unit period in which a target coke quality characteristic value is the same in a coke oven; A method for estimating variation in coke quality, comprising: calculating an observation error of the state space model as an index of variation within a lot of a coke quality characteristic value.

2. The method for estimating variation in coke quality according to claim 1 , further comprising the step of calculating an estimated value of variation within a lot from an observation error of the state space model.

3. calculating a correlation equation between an actual measurement value of variation in a lot of coke quality characteristic values ​​obtained by alternately sampling coke and an observation error of the state space model; The method for estimating variation in coke quality according to claim 2 , further comprising the step of calculating the estimated value of variation within a lot from an observation error of the state space model by using the correlation equation.

Images