Coal quality control methods

Abstract

To detect measurement anomalies that occur during coal quality measurement at an early stage. [Solution] The method includes: a measurement step in which lots unloaded at multiple unloading sites are subjected to a quality measurement process at each unloading site, measuring the quality of the same coal at each unloading site; a bias calculation step in which relational equations are formulated based on the measured values ​​obtained in the measurement step, and multiple biases associated with the measuring device or measurement conditions at each unloading site are statistically estimated using these relational equations; and a determination step in which, if the measured value measured at unloading site X is an abnormal value, and at least one of the multiple biases at unloading site X calculated in the bias calculation step deviates from a predetermined standard value, it is determined that the measurement is abnormal.

Description

[[Technical Field]]

[0001] The present invention relates to a management method for managing the quality of coal. [[Background Art]]

[0002] In order to produce high-quality coke for blast furnaces, it is important to appropriately control the quality of coal, which is the raw material.

[0003] Even for coal of the same brand, properties may vary depending on differences in the produced coal seam and coal preparation processing. Therefore, when the coal carriers are different (in other words, when lots are different), it is necessary to re-evaluate the properties of coal. In addition, the properties of coal may change during the transportation of coal from foreign coal mines to the domestic territory, so evaluation based only on the property measurement results at the coal mine is insufficient. For this reason, the quality of coal is regularly measured at the unloading site each time.

[0004] Conventionally, when the measured value measured at the unloading site greatly deviates from the range of coal of the same brand (in other words, when the measured value is an abnormal value), it has been unclear whether the cause of the abnormality is caused by a change in coal quality or by measurement abnormality. Measurement abnormality occurs due to causes related to the measurement device, such as improper installation during update of the measurement device, insufficient output caused by heater failure (insufficient temperature rise of coal), and the like. Although it is difficult to assume that such a situation actually occurs, even when an operator mistakenly continues measurement through a procedure different from the prescribed procedure (in other words, when measurement conditions are not complied with), the result will also be a measurement abnormality.

[0005] If the detection of measurement abnormality is delayed, the quality of coal will continue to be evaluated inaccurately. The methods described in Patent Documents 1 and 2 are methods for predicting coke quality, and are different from coal quality management methods. [[Prior Art Literature]] [[Patent Documents]]

[0006] [Patent Document 1] Japanese Patent Publication No. 2022-41892 [Patent Document 2] Japanese Patent Application Publication No. 9-165579 [Overview of the project] [Problems that the invention aims to solve]

[0007] The present invention aims to detect measurement anomalies in coal quality measurement at an early stage. [Means for solving the problem]

[0008] To solve the above problems, the coal quality control method of the present invention is characterized by comprising: (I) a measurement step in which a measurement process is carried out for multiple different lots in which lots unloaded at multiple unloading sites are the same coal and their quality is measured at each unloading site; a bias calculation step in which relational equations are formulated based on the respective measurement values ​​obtained in the measurement step, and multiple biases associated with the measurement device or measurement conditions of each unloading site are statistically estimated using these relational equations; and a determination step in which, if the measurement value measured at unloading site X among the multiple unloading sites is an abnormal value, and at least one of the multiple biases at unloading site X calculated in the bias calculation step deviates from a predetermined standard value, it is determined that the measurement is abnormal.

[0009] (II) The coal quality control method described in (I) above, wherein the relation follows equation (1) below. Qi_P=α_P×Qi+β_P+εj...Equation (1) However, "i" corresponds to identification information that identifies the lot, and "P" corresponds to identification information that identifies the unloading location. "j" corresponds to identification information that identifies the measurement value. "Qi_P" is the measurement value obtained in the measurement step described above. "εj" is a measurement error and is an unknown variable. "Qi" is a hypothetical measurement value unaffected by bias α_P, bias β_P, and measurement error εj, and is therefore an unknown variable. "α_P" and "β_P" refer to the measuring devices used at each unloading site in the measurement process. Alternatively, it could be a bias tied to the measurement conditions, which is an unknown variable.

[0010] (III) The quality control method according to (II) above, characterized in that the predetermined reference value consists of a first reference value corresponding to bias α_P and a second reference value corresponding to bias β_P, and in the determination step, if bias α_P deviates from the first reference value by a first predetermined amount or more, and bias β_P deviates from the second reference value by a second predetermined amount or more, then it is determined that there is a measurement abnormality.

[0011] (IV) A coal quality control method according to (III), characterized in that the preparation period is divided into multiple periods in advance, and a preparation step is performed in each period in which the same process as the measurement step and the bias calculation step is performed, the first reference value is a representative value of the bias α_P in each period of the unloading site X obtained in the preparation step, and the second reference value is a representative value of the bias β_P in each period of the unloading site X obtained in the preparation step.

[0012] (V) The coal quality control method according to (II) above, characterized in that, in the determination step, if the measured value measured at a loading site other than loading site X is an abnormal value and Qi is an abnormal value, it is determined that a quality abnormality of coal has occurred.

[0013] (VI) The coal quality control method according to (I) above, characterized in that the measured values ​​measured in the measurement step further include measured values ​​for a single port unloading, where all coal from the same lot is unloaded at one unloading point. [Effects of the Invention]

[0014] According to the present invention, measurement anomalies occurring in coal quality measurement can be detected at an early stage. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] [Figure 1] It is a flowchart showing the procedure of a coal quality control method. [Figure 2] It is a graph plotting bias α_P and bias β_P at location A (Example). [Figure 3] It is a graph plotting bias α_P and bias β_P at location B (Example). DESCRIPTION OF EMBODIMENTS

[0016] The present inventors focused on multi-port unloading of coal as a method for detecting measurement anomalies at an early stage. Multi-port unloading refers to dividing and unloading coal of the same brand loaded on the same coal vessel (in other words, coal of the same lot) at a plurality of unloading sites (hereinafter also referred to as "locations"). Since the quality of coal from the same lot is inherently the same, the coal quality measured at each unloading site will also be generally the same, and ideally, measurement anomalies can be grasped by comparing the measurement results of multi-port unloading with each other. As described above, measurement anomalies are mainly caused by the measurement device, so measured values are constantly biased. In the present specification, this bias is referred to as bias.

[0017] However, in practice, measurement results include variations caused by measurement errors, so it is difficult to perform evaluation using only data from a specific lot. Furthermore, when there are 2 or fewer unloading locations, it is impossible in principle to determine whether the data at one location deviates compared to data from other locations. Accordingly, the present inventors have conducted intensive studies on a statistical analysis method using multi-port unloading data from a plurality of lots, and have accomplished the coal quality control method of the present invention.

[0018] Hereinafter, an embodiment of the coal quality control method of the present invention will be described. Figure 1 is a flowchart showing the procedure of the coal quality control method.

[0019] (S101: Preparation step) In preparation step S101, the same processes as those in measurement step S102 and bias calculation step S103 described below are performed in advance to obtain a first reference value and a second reference value which will be described later. The content of preparation step S101 will be described later.

[0020] (S102: Measurement step) In measurement step S102, a measurement process of measuring the quality of lots unloaded at a plurality of unloading locations, wherein the lots each consist of the same coal, at each of the respective unloading locations is performed for a plurality of different lots. The expression "the lots each consist of the same coal" refers to coal of the same brand loaded onto the same coal carrier. The term "brand" refers to the place of origin. "Quality" refers to properties of coal that affect the quality of coke, and representative examples include the total dilatation (TD) of coal, caking property, and volatile matter (VM).

[0021] As described above, measurement abnormality occurs mainly due to a measurement device, so measured values are constantly biased to either the positive or negative side relative to normally measured values. On the other hand, measurement error may fluctuate to either the positive or negative side and is not constant, so fluctuations tend to cancel out over a long period. The above-described measurement process is performed for a plurality of different lots. This is because if the measurement process is limited to one lot, evaluation of measurement abnormality becomes difficult due to the inclusion of abnormal values caused by measurement error. When there are a plurality of lots, measurement errors tend to cancel out, which facilitates evaluation of measurement abnormality.

[0022] Measurement step 102 is a routine operation performed at the unloading site, and this operation includes detecting abnormal values. An abnormal value is a value that, if true, could adversely affect coke quality. Since measurement results contain errors, the determination of which values ​​to set as abnormal values ​​is made by comprehensively considering factors such as the standard degree of variation in the quality being evaluated, the measurement error of the measuring device, and the degree of loss due to adverse effects. Based on the measured values ​​collected within the predetermined aggregation period (the judgment period described later), the following bias calculation step S103 is performed. The aggregation period (judgment period) is not particularly limited, but it is necessary that multiple port unloading has been performed for at least two different lots. In the examples described later, the aggregation period (judgment period) is set to 3 months, but the present invention is not limited to this.

[0023] (S103: Bias calculation step) The bias calculation step S103 involves formulating relational equations based on each measurement value obtained in the measurement step S102, and using these relational equations to statistically estimate multiple biases associated with the measuring device or measurement conditions at each unloading site (hereinafter also referred to as statistical analysis). The relationship may also be a linear equation as shown in equation (1) below. Qi_P=α_P×Qi+β_P+εj...Equation (1) However, "i" corresponds to identification information that identifies the lot (e.g., a number), and "P" corresponds to identification information that identifies the unloading location (e.g., abbreviated name of the unloading location). Note that "P" is an abbreviation for "place". Also, "j" corresponds to identification information that identifies the measurement value. "Qi_P" is a measured value obtained in measurement step S102 and is a known value. "εj" is a measurement error and is an unknown variable. "Qi" is a hypothetical measurement value unaffected by bias α_P, bias β_P, and measurement error εj, and is therefore an unknown variable. "α_P" and "β_P" are biases associated with the measuring device or measurement conditions at each unloading site used in measurement step S102, and are unknown variables. As mentioned above, bias is a constant bias in measured values ​​mainly caused by the measuring device, so if the unloading site (location) is the same, the bias will be the same regardless of the type or lot of coal.

[0024] When bias α_P is 1 and bias β_P is 0, equation (1) becomes "Qi_P = Qi + εj". Therefore, the closer bias α_P is to 1 and bias β_P is to 0, the smaller the bias can be considered to be.

[0025] We will explain regression analysis, an example of statistical analysis, using a linear equation as an example. For example, if coal from lot 1 is unloaded at locations A and B, and coal from lot 2 is unloaded at locations A, C, and D, then in measurement step S102, five measurement values ​​consisting of Q1_A, Q1_B, Q2_A, Q2_C, and Q2_D are aggregated, and from these measurement values, the following five linear equations can be formulated. As mentioned above, the "j" in "εj" is identification information that identifies the measurement value, so "εj" is assigned to each measurement value. Q1_A=α_A×Q1+β_A+ε1...Equation (1-1) Q2_A=α_A×Q2+β_A+ε2...Equation (1-2) Q1_B=α_B×Q1+β_B+ε3...Equation (1-3) Q2_C=α_C×Q2+β_C+ε4...Equation (1-4) Q2_D=α_D×Q2+β_D+ε5...Equation (1-5)

[0026] The biases α_A and β_A for location A, α_B and β_B for location B, α_C and β_C for location C, and α_D and β_D for location D are calculated through regression analysis. As a result, the unknown variables Q1 and Q2 are also calculated.

[0027] For regression analysis, for example, the least squares method or the maximum likelihood method can be used. Needless to say, "using these linear equations" in the bias calculation step S103 also includes transforming the linear equations, as will be explained below. When performing regression analysis using the least squares method, equations (1-1) to (1-5) are rearranged to formulate the following equation (2), which calculates the sum of squares of the measurement errors εj. Σ{Qi_P-(α_P×Qi+β_P)} 2 =Σ{εj} 2 ...Equation (2) Assuming that the measurement error εj follows a normal distribution with a mean of 0, we need to find the set of bias α_P, bias β_P, and Qi (hereinafter collectively referred to as the unknown) that minimizes the left-hand side of equation (2). As mentioned above, the measurement error εj can fluctuate positively or negatively and is easily canceled out, so it can be considered to follow a normal distribution with a mean of 0.

[0028] Because there are many sets of variables to calculate and the calculation is complex, you may use Excel's Solver function or Python's scipy to calculate the unknowns.

[0029] Furthermore, the statistical analysis using equation (1) may also be performed using Bayesian estimation. That is, since there are many unknowns, the statistical analysis may be performed using Bayesian estimation, which assumes a prior distribution for each parameter. This suppresses overfitting of parameter estimation, thereby stabilizing the computational process. For example, an unknown can be estimated by writing the following equation (3) in Stan, a library known for Bayesian estimation. Equation (3) is a transformation of equation (1). Qi_P-(α_P×Qi+β_P)=εj...Equation (3)

[0030] As mentioned above, in an unbiased state, α_P approaches 1 and β_P approaches 0. Therefore, we can assume that the prior distribution of the unknown bias α_P follows a normal distribution with mean 1, and the prior distribution of the bias β_P follows a normal distribution with mean 0. Note that we assume a normal distribution here because we cannot imagine any other viable distribution shape, and therefore we have adopted the normal distribution as the most reasonable assumption. If we have prior knowledge of the distribution shapes of α_P and β_P, we may adopt a distribution other than the normal distribution. We can assume that the prior distribution of Qi follows a normal distribution with mean Qi_P. Although this is a repetition of the explanation, we can assume that the measurement error εj follows a normal distribution with mean 0. In this embodiment, the relation is defined by the linear equation (1), but the relation may also be defined by an exponential equation. For example, the exponential equation (4) below can be formulated as the relation, and the coefficient part (β_P) and the exponential part (α_P) of the exponential equation can be used as the bias, respectively. Qi_P=β_P×Qi^(α_P)+εj...Equation (4)

[0031] The measurements taken in measurement step S102 may include data from a single port unloading. Single port unloading refers to the unloading of all coal from the same lot at a single unloading location. Even if single port unloading data is included, excluding the measurement error εj, the number of unknown variables does not increase, as only one unknown variable (Qi) and one known variable (Qi_P) are added.

[0032] (S104: Judgment Step) If the measurement value taken at unloading site X among the above-mentioned multiple unloading sites is an abnormal value, and at least one of the multiple biases at unloading site X calculated in bias calculation step S103 deviates from a predetermined standard value, then it is determined that the measurement is abnormal. Here, if the relationship is equation (1), and the measurement value measured at unloading site X among the multiple unloading sites described in measurement step S102 is an abnormal value, and the bias α_P of unloading site X deviates from the first reference value by a first predetermined amount or more, and the bias β_P of unloading site X deviates from the second reference value by a second predetermined amount or more, then it can be determined that the measurement abnormality is caused by the measuring device. This prevents situations such as coal blending design being carried out based on incorrect measurement values.

[0033] Since measurement anomalies occur infrequently, the unloading site X is usually one of the multiple unloading sites described in measurement step S102. However, if there are many unloading sites (for example, 20 or more), there may be two or more unloading sites X that are determined to have measurement anomalies. I will not repeat the explanation regarding the detection of abnormal values.

[0034] When an anomaly in coal quality is detected, Qi changes rather than α_P and β_P, allowing for differentiation between measurement anomalies and quality anomalies. Coal quality anomalies may include weathering during the coal transportation process. In other words, if, in the determination step (S104), the measured value at the unloading site is an abnormal value, and Qi is also an abnormal value, it can be considered an abnormality in coal quality rather than a measurement abnormality. The explanation of abnormal values ​​will not be repeated.

[0035] The first and second reference values ​​can be determined by the preparation step S101 described above. Specifically, the steps corresponding to the measurement step S102 and the bias calculation step 103 are carried out in advance over multiple periods. That is, when the time spent on the preparation step 101 is defined as the preparation period, the preparation period is divided into multiple periods, and the measurement step S102 and the bias calculation step 103 are carried out in each of the divided periods. As a result, bias α_P and bias β_P are obtained for each unloading site, corresponding to the number of divided periods.

[0036] Next, representative values ​​of bias α_P and bias β_P are determined for each unloading site, and these representative values ​​are set as the first and second reference values, respectively. The representative values ​​may be the mean or the median. It is desirable to select a stable period with minimal changes in bias α_P and bias β_P as the preparation period. However, if the preparation period includes a period with large changes in bias α_P and bias β_P, the representative values ​​can be calculated excluding that period. Therefore, it is not essential that the period be stable.

[0037] The first predetermined amount and the second predetermined amount cannot be quantitatively defined because they vary depending on the quality of the coal being measured. For example, if the quality of the coal being measured is the total expansion rate (TD), the first predetermined amount can be 0.05 and the second predetermined amount can be 1.5. This quantitative finding was obtained from the examples described later. When measuring a coal quality different from the total expansion rate (TD), the first predetermined amount and the second predetermined amount can be quantitatively defined by performing the analysis process described in the examples described later for that quality.

[0038] Here, depending on how the first and second predetermined values ​​are set, it is possible that only one of the biases may deviate significantly from the standard value. Specifically, it is possible that bias α_P deviates from the first reference value by a first predetermined amount or more, and bias β_P does not deviate from the second reference value by a second predetermined amount or more. In this case, if bias α_P deviates from the first reference value by a third predetermined amount or more (provided that the third predetermined amount > the first predetermined amount), it may be judged as an abnormal measurement. If the quality of coal is the total expansion rate (TD), the third predetermined amount may be twice the above 0.05 (first predetermined amount), i.e., 0.10.

[0039] Conversely, it is also possible that bias β_P deviates from the second reference value by a second predetermined amount or more, and bias α_P does not deviate from the first reference value by a first predetermined amount or more. In this case, if bias β_P deviates from the second reference value by a fourth predetermined amount or more (provided that the fourth predetermined amount > the second predetermined amount), it may be judged as a measurement anomaly. If the quality of coal is the total expansion rate (TD), the fourth predetermined amount may be twice the above 1.50 (second predetermined amount), i.e., 3.00.

[0040] The first and second reference values ​​are not limited to the representative values ​​described above. Since the probability of measurement anomalies is low, bias α_P is usually close to 1.00 and bias β_P is close to 0.00. Therefore, measurement anomalies may be determined based on a comparison between 1.00 (first reference value) and bias α_P, and a comparison between 0.00 (second reference value) and bias β_P. In this case, the first and second reference values ​​will be fixed values ​​common to all unloading sites. The first predetermined quantity and the second predetermined quantity may be appropriately determined based on the data acquired during the preparation period described above. For example, one could identify the bias α_P that is furthest from 1.00 within each period (stable period), and set the first predetermined amount to 1.2 times the difference between this bias α_P and 1.00. Alternatively, one could identify the bias β_P that is furthest from 0.00 within each period, and set the second predetermined amount to 1.2 times the difference between this bias β_P and 0.00.

[0041] The measurement step S102, bias calculation step S103, and discrimination step S104 described above should preferably be performed regularly. Performing them regularly allows for early detection of measurement abnormalities. Specifically, it is desirable to perform them within four months of the previous measurement.

[0042] In actual steel mill operations, cross-checks are conducted where samples of the same coal lot are distributed to multiple locations, and quality checks are performed at each location. If cross-checks are performed frequently, measurement anomalies can be detected by comparing the measurement results from one location with those from other locations. However, it is difficult to perform cross-checks frequently in addition to regular coal quality measurements, which may lead to measurement anomalies being overlooked for extended periods. In contrast, the coal quality control method of this embodiment utilizes the results of coal quality measurements that are carried out on a regular basis, so measurement abnormalities can be detected at an early stage.

[0043] (Examples) The present invention will be specifically described below with reference to examples. The total expansion coefficient (TD) of 11 types of coal, measured at seven locations (locations A-G) from January 1, 2018 to June 30, 2020, was used as measurement data. However, from the perspective of analytical accuracy, only measurement data with a total expansion coefficient (TD) between 2 and 174 was adopted. The preparation period (27 months) was defined as January 1, 2018 to March 31, 2020. This preparation period was divided into 3-month intervals, and the bias α_P, bias β_P, and Qi for each interval were calculated using Bayesian estimation. The preparation period is considered a stable period with minimal bias fluctuations. For each location, the average value of bias α_P (first reference value) and the average value of bias β_P (second reference value) for each period were calculated.

[0044] The evaluation period (3 months) was set from April 1, 2020 to June 30, 2020, and bias α_P, bias β_P, and Qi were calculated for each location using Bayesian estimation. At location A, the total expansion rate (TD) measurement device was updated in May 2020, so it was predicted that bias α_P and bias β_P would deviate from the first and second reference values, respectively.

[0045] Figure 2 is a graph plotting bias α_P and bias β_P at location A. Figure 3 is a graph plotting bias α_P and bias β_P at location B. The first predetermined amount was set to 0.05 Pt, and the second predetermined amount was set to 1.5 Pt. If the bias α_P obtained within the judgment period deviates from the first reference value by 0.05 Pt or more, and the bias β_P deviates from the second reference value by 1.5 Pt or more, it was determined to be a measurement abnormality. Specifically, if the values ​​fall outside the upper and lower limits shown by the dotted lines in Figures 2 and 3, it was determined to be a measurement abnormality.

[0046] As predicted, the result for location A was "measurement abnormality detected." On the other hand, the result for location B was "no measurement abnormality detected."

Claims

1. A measurement step in which a measurement process is carried out for multiple different lots, in which lots unloaded at multiple unloading sites measure the quality of the same coal at each unloading site, A bias calculation step is performed in which relational equations are formulated based on the measured values ​​obtained in the measurement step, and multiple biases associated with the measuring device or measurement conditions at each unloading site are statistically estimated using these relational equations. The system includes a determination step in which, if the measured value at unloading site X among the plurality of unloading sites is an abnormal value, and at least one of the multiple biases at unloading site X calculated in the bias calculation step deviates from a predetermined standard value, the system determines that the measurement is abnormal. A method for controlling the quality of coal, characterized by the following features.

2. The coal quality control method according to claim 1, wherein the relational expression follows the following equation (1). Qi_P=α_P×Qi+β_P+εj...Formula (1) However, "i" corresponds to identification information that identifies the lot, and "P" corresponds to identification information that identifies the unloading location. "j" corresponds to identification information that identifies the measurement value. "Qi_P" is the measured value obtained in the measurement step described above. "εj" is a measurement error and is an unknown variable. "Qi" is a hypothetical measurement value unaffected by bias α_P, bias β_P, and measurement error εj, and is therefore an unknown variable. "α_P" and "β_P" are the measurement devices used at each unloading site in the measurement process. Alternatively, it could be a bias tied to the measurement conditions, which is an unknown variable.

3. The predetermined reference value consists of a first reference value corresponding to bias α_P and a second reference value corresponding to bias β_P. In the determination step described above, if bias α_P deviates from the first reference value by a first predetermined amount or more, and bias β_P deviates from the second reference value by a second predetermined amount or more, then it is determined that there is a measurement abnormality. The quality control method according to feature 2.

4. The system has a preparation step in which the same process as the measurement step and the bias calculation step is performed in each of the multiple periods into which the preparation period is divided. The first reference value is a representative value of the bias α_P for each period at the unloading site X obtained in the preparation step, The second reference value is a representative value of the bias β_P for each period at the unloading site X obtained in the preparation step. The coal quality control method according to feature 3.

5. In the aforementioned determination step, if the measured values ​​at unloading sites other than the aforementioned unloading site X are abnormal values, and Qi is also abnormal, it is determined that a quality abnormality of the coal has occurred. The coal quality control method according to feature 2.

6. The measurements taken in the aforementioned measurement step further include measurements for a single port unloading, where all coal from the same lot is unloaded at one unloading point. The coal quality control method according to feature 1.

Images