Method and apparatus for diagnosing abnormalities

Abstract

The abnormality diagnosis method includes: an abnormality determination step of determining whether or not there is an abnormality in the manufacturing equipment using a set of first prediction models constructed based on multiple variables that indicate the normal state of the manufacturing equipment, and in which the variables are used complementaryly as dependent and independent variables; an operational influence coefficient calculation step of constructing a set of second prediction models based on multiple variables in which the variables are used complementaryly as dependent and independent variables; an abnormality cause estimation matrix calculation step of calculating an abnormality cause estimation matrix; and an abnormality cause variable identification step of identifying the variables that cause the abnormality using the abnormality cause estimation matrix.

Description

Technical Field

[0001] The present invention relates to an abnormality diagnosis method and an abnormality diagnosis device.

Background Art

[0002] As methods for diagnosing the manufacturing state of a manufacturing process, particularly an abnormal state, there are a model-based approach and a database approach. The model-based approach is an approach in which a model expressing physical or chemical phenomena in a manufacturing process by a mathematical formula is constructed, and the manufacturing state of the manufacturing process is diagnosed using the constructed model. On the other hand, the database approach is an approach in which a statistical model is constructed from operation data obtained in a manufacturing process, and the manufacturing state of the manufacturing process is diagnosed using the constructed model.

[0003] In a manufacturing process such as a steel process, since products of various types and sizes are manufactured on a single production line, there are a large number of operation patterns. Also, in a manufacturing process such as a blast furnace, since natural materials such as iron ore and coke are used as raw materials, the variation in the manufacturing process is large. Therefore, when diagnosing the manufacturing state of a manufacturing process such as a steel process, there are limitations to an approach that relies only on the model-based approach.

[0004] As a database approach, there is a diagnosis method in which the characteristics of past normal operation data are extracted and the difference from current operation data is determined. One of them is diagnosis based on the degree of deviation from the normal state using a multiple regression model.

[0005] An example of this is described in Patent Document 1. Specifically, Patent Document 1 describes a method for estimating the cause of an abnormal state using the influence coefficients of a multiple regression model created using operational data from normal conditions. The input variables used are state variables that indicate the state of equipment and quality, and causal variables that are candidates for the cause of the abnormality. The method compares the influence coefficients of a multiple regression model, in which the state variables are the dependent variables and the causal variables are the independent variables, between normal and abnormal conditions, and estimates one of the causal variables with the largest change as the cause of the abnormality. [Prior art documents] [Patent Documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 7-159289 [Overview of the Initiative] [Problems that the invention aims to solve]

[0007] In steel manufacturing processes and other applications, while numerous sensor measurements can be used, most of these measurements are taken for operational management and equipment control purposes. Therefore, as described in Patent Document 1, it is not always possible to obtain sufficient measurements that directly indicate equipment status or the cause of anomalies. Furthermore, even if there are measurements that correspond one-to-one with the cause of such anomalies, it is impossible to cover all anomalies with those measurements alone.

[0008] On the other hand, recent advancements in data collection and analysis technologies have created an environment capable of handling massive amounts of data, often referred to as big data. Considering the aforementioned situation, it can be said that comprehensively and accurately detecting abnormal situations from large amounts of data, presenting data related to those abnormalities, and leading to swift maintenance actions is necessary to maintain stable operations. In this case, large amounts of data do not necessarily correspond one-to-one with the causes of abnormalities as described in Patent Document 1, and therefore, the content of the abnormalities presented may not be limited to just one.

[0009] The present invention has been made in view of the above, and aims to provide an abnormality diagnosis method and an abnormality diagnosis device that can detect abnormal conditions from a large amount of measurement data acquired in a manufacturing process and quickly and clearly present measurement values ​​related to the abnormality. [Means for solving the problem]

[0010] To solve the above-mentioned problems and achieve the objective, the abnormality diagnosis method according to the present invention includes: an operational variable acquisition step of acquiring a plurality of variables indicating the operating state of the manufacturing equipment; an abnormality determination step of determining whether or not there is an abnormality in the manufacturing equipment using the plurality of variables acquired in the operational variable acquisition step and a set of first prediction models constructed based on the plurality of variables indicating the normal state of the manufacturing equipment, and in which the variables are used complementaryly as dependent and independent variables; an operational influence coefficient calculation step of constructing a set of second prediction models based on the plurality of variables acquired in the operational variable acquisition step, in which the variables are used complementaryly as dependent and independent variables, if an abnormality is determined to exist; an abnormality cause estimation matrix calculation step of calculating an abnormality cause estimation matrix by comparing a first influence coefficient, which is the influence coefficient of the independent variables on the dependent variable for the set of first prediction models, and a second influence coefficient, which is the influence coefficient of the independent variables on the dependent variable for the set of second prediction models; and an abnormality cause variable identification step of identifying the variables that cause the abnormality using the abnormality cause estimation matrix.

[0011] Furthermore, in the abnormality diagnosis method according to the present invention, the abnormality cause estimation matrix has a tabular structure with the objective variable and the explanatory variable as key items, and the comparison result of the first influence coefficient and the second influence coefficient is used as an element.

[0012] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality cause variable identification step, the abnormality cause estimation matrix is ​​used to focus on the explanatory variable or objective variable of the variable related to the abnormality, and the variable that causes the abnormality is identified based on the change in the influence coefficient related to the variable of focus.

[0013] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality cause variable identification step, variables whose change in the influence coefficient is greater than a predetermined threshold in both the explanatory variable column and the dependent variable row of the abnormality cause estimation matrix are identified as variables that cause the abnormality.

[0014] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality determination step, the degree of abnormality is calculated based on the prediction error of each variable using the first prediction model, and the presence or absence of an abnormality in the manufacturing equipment is determined by referring to the maximum, average, or sum of the degree of abnormality of all or some of the variables.

[0015] Furthermore, the abnormality diagnosis method according to the present invention further includes an abnormality cause estimation step in which the cause of the abnormality is estimated based on the variables identified in the above invention.

[0016] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality cause estimation step, the cause of the abnormality is estimated by referring to a table of candidate causes of the abnormality corresponding to combinations of the objective variable and explanatory variables that are candidates for the abnormality.

[0017] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality cause estimation step, the cause of the abnormality is estimated by referring to a table of candidate causes of abnormality and corresponding actions corresponding to combinations of the objective variable and explanatory variables that are candidates for abnormality, and the action corresponding to the estimated cause of the abnormality is determined.

[0018] Furthermore, in the abnormality diagnosis method according to the present invention, in the abnormality cause estimation step, a command to the control device corresponding to the determined action, or an alert, is output.

[0019] In order to solve the above problems and achieve the object, the abnormality diagnosis apparatus according to the present invention uses a set of first prediction models constructed based on a plurality of variables indicating the state during the operation of manufacturing equipment and a plurality of variables indicating the normal state of the manufacturing equipment, and in which the variables are used complementarily as target variables and explanatory variables, and an abnormality determination unit that determines the presence or absence of an abnormality in the manufacturing equipment; when it is determined that there is an abnormality, an operation-time influence coefficient calculation unit that constructs a set of second prediction models in which the variables are used complementarily as target variables and explanatory variables based on a plurality of variables indicating the state during operation; and an abnormality cause estimation matrix calculation unit that calculates an abnormality cause estimation matrix by comparing a first influence coefficient, which is an influence coefficient of an explanatory variable with respect to a target variable for the set of first prediction models, and a second influence coefficient, which is an influence coefficient of an explanatory variable with respect to a target variable for the set of second prediction models; and an abnormality cause variable identification unit that identifies a variable that causes an abnormality using the abnormality cause estimation matrix.

[0020] Further, the abnormality diagnosis apparatus according to the present invention further includes an abnormality cause estimation unit that estimates the cause of the abnormality based on the specified variables in the above invention.

Effects of the Invention

[0021] According to the abnormality diagnosis method and apparatus according to the present invention, it is possible to detect an abnormal state from a large amount of measurement data acquired in a manufacturing process and quickly and easily present measurement values related to the abnormality.

Brief Description of the Drawings

[0022] [Figure 1] FIG. 1 is a block diagram showing a schematic configuration of an abnormality diagnosis apparatus according to an embodiment of the present invention. [Figure 2] FIG. 2 is a flowchart showing the flow of an abnormality diagnosis method executed by the abnormality diagnosis apparatus according to an embodiment of the present invention. [Figure 3]FIG. 3 is a diagram showing an example of an abnormal cause estimation matrix calculated with respect to changes in influence coefficients of a multiple regression model derived by the least squares method in the abnormal cause estimation matrix calculation step of the abnormal diagnosis method according to an embodiment of the present invention. [Figure 4] FIG. 4 is a diagram showing an example of an abnormal cause estimation matrix calculated with respect to changes in influence coefficients of a multiple regression model derived by a sparse structure learning method such as Lasso regression analysis in the abnormal cause estimation matrix calculation step of the abnormal diagnosis method according to an embodiment of the present invention. [Figure 5] FIG. 5 is a diagram showing an example of a table in which an objective variable, explanatory variables that are candidates for abnormalities, and candidates for causes of abnormalities are preliminarily arranged in the abnormal cause estimation step of the abnormal diagnosis method according to an embodiment of the present invention. [[ID=**9**]]**[Mode for Carrying Out the Invention]**

[0023] An abnormal diagnosis method and an abnormal diagnosis apparatus according to an embodiment of the present invention will be described with reference to the drawings.

[0024] (Abnormal Diagnosis Apparatus) The configuration of the abnormal diagnosis apparatus according to an embodiment of the present invention will be described with reference to FIG. 1. The abnormal diagnosis apparatus 1 is for diagnosing the presence or absence of an abnormality in manufacturing equipment in a manufacturing process such as a steel process. The abnormal diagnosis apparatus 1 is realized by, for example, a general-purpose computer such as a workstation or a personal computer, or an information processing apparatus such as a server arranged on the cloud. Further, the abnormal diagnosis apparatus 1 includes an input unit 10, an output unit 20, a storage unit 30, and a calculation unit 40.

[0025] The input unit 10 inputs various sensor data indicating the state during operation of manufacturing equipment and products, various set values, etc. to the calculation unit 40 regularly or irregularly.

[0026] The output unit 20 displays various calculation results from the calculation unit 40. Examples of calculation results output by the output unit 20 include the abnormality determination result from the abnormality determination unit 42, the abnormality cause estimation matrix calculated by the abnormality cause estimation matrix calculation unit 44, and the variables causing the abnormality identified by the abnormality cause variable identification unit 45. Furthermore, the output unit 20 also outputs the cause of the abnormality estimated by the abnormality cause estimation unit 46. Finally, the output unit 20 outputs commands and alert information for the control device corresponding to the cause of the abnormality determined by the abnormality cause estimation unit 46.

[0027] The storage unit 30 is comprised of storage devices such as an EPROM (Erasable Programmable ROM), a Hard Disk Drive (HDD), or a Solid State Drive (SSD). The storage unit 30 may also include a temporary storage device such as RAM (Random Access Memory) used by the abnormality diagnosis device 1 during the abnormality diagnosis process.

[0028] The memory unit 30 stores an operational data database 31 and a normal operation data database 32. The operational data database 31 stores data such as various sensor data indicating the operational status of manufacturing equipment and products, various setting values, other prediction results, and influence coefficients described later. The normal operation data database 32 stores data such as various sensor data indicating the normal (past normal operation) status of manufacturing equipment and products, various setting values, and influence coefficients described later.

[0029] The calculation unit 40 functions as an anomaly diagnosis device 1 by having a computer process each function of anomaly diagnosis. The calculation unit 40 is composed of a processor (arithmetic processing unit) such as a CPU (Central Processing Unit). The calculation unit 40 also functions as a normal influence coefficient calculation unit 41, an anomaly determination unit 42, an operational influence coefficient calculation unit 43, an anomaly cause estimation matrix calculation unit 44, and an anomaly cause variable identification unit 45. In addition to the above, the calculation unit 40 also functions as an anomaly cause estimation unit 46 as needed.

[0030] The normal operation influence coefficient calculation unit 41 constructs multiple first-stage prediction models (multiple regression models) based on multiple variables (e.g., sensor data) obtained from manufacturing equipment and products during normal operation, and calculates the influence coefficient of each explanatory variable. Details of the processing of the normal operation influence coefficient calculation unit 41 will be described later (see Figure 2). Note that the explanatory variables used when constructing the first-stage prediction models may be not only sensor data, but also operational settings (e.g., set values ​​for operating conditions).

[0031] The abnormality detection unit 42 uses a first prediction model constructed based on multiple variables acquired during normal operation to calculate a prediction error, which is the difference between the actual value and the predicted value, and determines whether or not there is an abnormality based on this prediction error. Details of the processing of the abnormality detection unit 42 will be described later (see Figure 2).

[0032] The operational influence coefficient calculation unit 43 constructs multiple second prediction models (multiple regression models) based on multiple variables (e.g., sensor data) obtained from manufacturing equipment and products during operation, and calculates the influence coefficient of each explanatory variable. Details of the processing of the operational influence coefficient calculation unit 43 will be described later (see Figure 2). Note that the explanatory variables used when constructing the second prediction models may be not only sensor data, but also operational settings (e.g., set values ​​for operating conditions).

[0033] The abnormal cause estimation matrix calculation unit 44 calculates an abnormal cause estimation matrix (abnormal cause estimation matrix) that allows comparison between the influence coefficients during normal operation calculated by the normal operation influence coefficient calculation unit 41 and the influence coefficients during operation calculated by the operation influence coefficient calculation unit 43, and saves it in the operation data DB 31. Details of the processing of the abnormal cause estimation matrix calculation unit 44 will be described later (see Figures 2 to 4).

[0034] The abnormal cause variable identification unit 45 identifies the variable causing the abnormality based on the abnormal cause estimation matrix created by the abnormal cause estimation matrix calculation unit 44. Details of the processing of the abnormal cause variable identification unit 45 will be described later (see Figures 2 to 4).

[0035] The abnormal cause estimation unit 46 estimates the cause of the abnormality based on the variables identified by the abnormal cause variable identification unit 45 and a pre-created table, etc. Details of the processing of the abnormal cause estimation unit 46 will be described later (see Figures 2 and 5).

[0036] (Methods for diagnosing abnormalities) Details of the abnormality diagnosis method performed by the abnormality diagnosis device according to the embodiment of the present invention will be described with reference to Figures 2 to 5. The abnormality diagnosis method includes a step of calculating the influence coefficient under normal conditions (step S1), a step of acquiring operating variables (step S2), an abnormality determination step (steps S3 to S5), and a step of calculating the influence coefficient under normal conditions (step S6). The abnormality diagnosis method also includes a step of calculating the abnormality cause estimation matrix (step S7) and a step of identifying the abnormality cause variable (step S8). In addition to the above steps, the abnormality diagnosis method further includes an abnormality cause estimation step (step S9) as needed.

[0037] <Steps for calculating the normal effect coefficient> In the normal operation influence coefficient calculation step, influence coefficients are derived for a set of multiple regression models based on sensor data (variables) measured from manufacturing equipment and products (hereinafter referred to as "manufacturing equipment, etc.") under normal conditions (Step S1).

[0038] In the normal operation influence coefficient calculation step, the normal operation influence coefficient calculation unit 41 first obtains multiple variables (sensor data) indicating the normal state of the manufacturing equipment, etc., from the normal operation data DB 32. Next, based on the multiple variables obtained, the normal operation influence coefficient calculation unit 41 constructs a set of multiple regression models in which these variables are used complementaryly as dependent and independent variables, and derives the influence coefficient (first influence coefficient) of each independent variable in each multiple regression model. In the following, the multiple regression model constructed in the normal operation influence coefficient calculation step will also be referred to as the "first prediction model." In the following, the sensor data under normal conditions will also be referred to as "normal operation data."

[0039] In the step of calculating the normal influence coefficient, the least squares method, which is commonly used, can be used as a multiple regression analysis technique. Furthermore, when dealing with a large number of variables, sparse structure learning methods such as Lasso regression analysis may be used as the multiple regression analysis technique. In models created using sparse structure learning methods such as Lasso regression analysis, variables with small influence coefficients (importance) are excluded from the explanatory variables beforehand. This limits the variables that need to be examined when identifying the variables causing the anomaly, thus making it easier to identify the cause of the anomaly. Sparse structure learning methods such as Lasso regression analysis are particularly effective when it is necessary to estimate the cause of anomalies from a large amount of sensor data.

[0040] In the normal operation impact coefficient calculation step, multiple multiple regression models are constructed depending on the number of variables obtained from the normal operation data DB32, and by swapping the dependent and independent variables. For example, if 10 variables are obtained from the normal operation data DB32, a total of 10 multiple regression models are constructed in the normal operation impact coefficient calculation step by swapping the dependent and independent variables.

[0041] <Steps for acquiring operational variables> In the operational variable acquisition step, sensor data (variables) from the manufacturing equipment, etc., at the time of operation (current) are acquired (step S2). In the operational variable acquisition step, the abnormality determination unit 42 acquires multiple variables (sensor data) indicating the operating state of the manufacturing equipment, etc., from the operational data DB 31. In addition, in the operational variable acquisition step, preprocessing such as shifting the sensor data in the time series direction and sampling may be performed as needed. Furthermore, in the following, the sensor data at the time of operation will also be referred to as "operational data".

[0042] <Anomaly detection step> In the abnormality detection step, the predicted value is first calculated using a multiple regression model (first prediction model) constructed from the variables under normal conditions (step S3). The abnormality detection unit 42 uses the operational data acquired in step S2 as explanatory variables and calculates the predicted value (dependent variable) using the influence coefficients under normal conditions (first influence coefficients) derived in step S1. The calculation of the predicted value is performed for each multiple regression model constructed in step S1.

[0043] Next, the anomaly detection unit 42 calculates a deviation index (degree of anomaly) based on the difference between the predicted value calculated in step S3 and the measured value of the sensor data acquired in step S2 (step S4). As the deviation index, for example, a value obtained by dividing the square of the difference between the predicted value and the measured value by the standard deviation of the distribution of the difference between the predicted value and the actual value calculated from normal data can be used. In this case, it becomes possible to compare the statistical deviation index regardless of the unit (physical interpretation) of the variable being predicted.

[0044] Next, the abnormality determination unit 42 determines whether or not there is an abnormality in the manufacturing equipment, etc., based on the deviation index calculated in step S4 (step S5). The abnormality determination unit 42 determines whether or not there is an abnormality by comparing the deviation index calculated in step S4 with a predetermined judgment criterion, etc. If an abnormality is determined in step S5, the process proceeds to step S6; if no abnormality is determined, the process returns to step S2.

[0045] Thus, in the abnormality detection step (steps S3 to S5), the presence or absence of abnormalities in the manufacturing equipment, etc., is determined using multiple variables obtained in the operational variable acquisition step and a set of multiple regression models (first prediction models) constructed based on normal data. Alternatively, in the abnormality detection step, the degree of abnormality (deviation index) may be calculated based on the prediction error of each variable using the multiple regression model (first prediction model), and the presence or absence of abnormalities in the manufacturing equipment, etc., may be determined by referring to the maximum, mean, or sum of the degree of abnormality of all or some of the variables.

[0046] <Steps for calculating operational impact coefficients> In the step of calculating the operational impact coefficient, the impact coefficient is derived for a set of multiple regression models based on sensor data (variables) measured from manufacturing equipment, etc., during operation (step S6).

[0047] In the operational impact coefficient calculation step, the operational impact coefficient calculation unit 43 constructs a set of multiple regression models in which multiple variables obtained in step S2 are used complementaryly as dependent and independent variables, and derives the impact coefficient (second impact coefficient) of each independent variable in each multiple regression model. In the following, the multiple regression model constructed in the operational impact coefficient calculation step will also be referred to as the "second prediction model".

[0048] In the step of calculating the operational impact coefficient, a commonly used method such as the least squares method can be used as the multiple regression analysis technique. Furthermore, when dealing with a large number of variables, a sparse structure learning method such as Lasso regression analysis may be used as the multiple regression analysis technique.

[0049] In the operational impact coefficient calculation step, multiple multiple regression models are constructed according to the number of variables obtained from the operational data DB31, while swapping the dependent and independent variables. For example, if 10 variables are obtained from the operational data DB31, a total of 10 multiple regression models are constructed in the operational impact coefficient calculation step, while swapping the dependent and independent variables. In the operational impact coefficient calculation step, multiple multiple regression models are constructed using the same number and types of variables as those used in the normal impact coefficient calculation step described above.

[0050] <Steps for calculating the matrix for estimating the cause of an anomaly> In the abnormal cause estimation matrix calculation step, the abnormal cause estimation matrix is ​​calculated by comparing the influence coefficients under normal conditions (first influence coefficients) derived in step S1 with the influence coefficients under operational conditions (second influence coefficients) derived in step S6 (step S7).

[0051] In the anomaly cause estimation matrix calculation step, the anomaly cause estimation matrix calculation unit 44 compares the influence coefficients of the explanatory variables on the dependent variable for the set of first prediction models (first influence coefficients) with the influence coefficients of the explanatory variables on the dependent variable for the set of second prediction models (second influence coefficients). Methods of comparison include, for example, calculating the change in the influence coefficients (difference) or calculating the rate of change in the influence coefficients.

[0052] More specifically, the anomaly cause estimation matrix calculation unit 44 creates a matrix whose elements are the influence coefficients (first influence coefficients) of the set of first prediction models, and a matrix whose elements are the influence coefficients (second influence coefficients) of the set of second prediction models. Then, by comparing the elements (influence coefficients) within each matrix of both matrices (calculating the amount of change or rate of change), it calculates the anomaly cause estimation matrix.

[0053] Figure 3 shows an example of an anomaly cause estimation matrix calculated from the multiple regression model constructed using the least squares method in steps S1 and S6. As shown in Figure 3, the anomaly cause estimation matrix consists of a matrix with the dependent variable arranged vertically and the independent variables arranged horizontally. The anomaly cause estimation matrix also has a tabular structure with the dependent variable and independent variables as key items, and its elements are the comparison results of the influence coefficients (first influence coefficient and second influence coefficient). The comparison results of the influence coefficients shown in Figure 3 represent the absolute value of the difference between the influence coefficients, but the mean or the sum of the absolute values ​​of the influence coefficients may also be used.

[0054] Figure 4 shows an example of an anomaly cause estimation matrix calculated from a multiple regression model constructed using a sparse structure learning method such as Lasso regression analysis in steps S1 and S6. The comparison results of the influence coefficients shown in Figure 4 represent the absolute value of the difference in influence coefficients, but the mean or sum of the absolute values ​​of the influence coefficients may also be used.

[0055] Figures 3 and 4 show that the larger the numerical value of an element in the matrix, the greater the change in the influence coefficient during operation compared to normal conditions. Furthermore, in Figures 3 and 4, the change in the influence coefficient is large in the columns of the explanatory variables Xi and Xj, and in the rows of the dependent variables Xi and Xj. The abnormal cause estimation matrix calculation unit 44 may also color-code the matrix within the abnormal cause estimation matrix according to the magnitude of the change in the influence coefficient and output it from the output unit 20. By color-coding the matrix within the abnormal cause estimation matrix and presenting it to operators, the variables causing the abnormality can be easily identified.

[0056] <Steps to identify abnormal causative variables> In the abnormality cause variable identification step, the abnormality cause estimation matrix calculated in step S7 is used to identify the variables that cause the abnormality (step S8). In the abnormality cause variable identification step, the abnormality cause variable identification unit 45 uses the abnormality cause estimation matrix to focus on the explanatory variables or dependent variables of the variables related to the abnormality, and identifies the variables that cause the abnormality based on the changes in the influence coefficients related to the variables of focus.

[0057] The abnormality-causing variable identification unit 45 identifies, for example, an explanatory variable corresponding to the dependent variable determined to be abnormal in step S5 as a variable causing the abnormality. At that time, the influence coefficient for that variable in the second prediction model, which uses the variable causing the abnormality as an explanatory variable, is considered to have changed significantly from the influence coefficient for that variable in the first prediction model constructed with the normal variable. For this reason, if the influence coefficient (second influence coefficient) of the second prediction model that predicts other dependent variables (for a dependent variable determined to be abnormal in step S5) has changed significantly from the influence coefficient (first influence coefficient) of the first prediction model constructed with the normal variable, the explanatory variable corresponding to that dependent variable is also identified as a variable causing the abnormality.

[0058] On the other hand, prediction errors can also deviate due to anomalies in other explanatory variables. Therefore, we focus on the influence coefficients of each explanatory variable, extract the influence coefficients of explanatory variables that have changed significantly from the normal state, and identify the corresponding variables as the cause of the anomalies. Then, we compare the influence coefficients of the above anomaly candidates in the other second prediction models with the normal state, and if the amount of change in the influence coefficient is judged to be greater than a predetermined range, it is confirmed as the variable causing the anomaly.

[0059] In the anomaly cause estimation matrix shown in Figures 3 and 4, the anomaly cause variable identification unit 45 identifies a variable as the cause of an anomaly if, for example, the change in the influence coefficient is greater than a predetermined threshold in both the explanatory variable column and the dependent variable row. For example, if the threshold is "0.1", the change in the influence coefficient is greater than the threshold in both the explanatory variable column Xi and Xj and the dependent variable row Xi and Xj. In this case, the anomaly cause variable identification unit 45 identifies the explanatory variable Xi and Xj as variables that cause the anomaly.

[0060] Thus, the anomaly diagnosis method according to this embodiment can detect abnormal conditions from a large amount of measurement data acquired in the manufacturing process and quickly and clearly present the measurement values ​​related to the anomaly. Furthermore, by identifying the explanatory variable causing the anomaly, it is possible to identify the abnormality of the sensor corresponding to the explanatory variable. Moreover, if setting values ​​of operating conditions other than sensor data are used as explanatory variables, it becomes possible to identify the setting values ​​of operating conditions that are candidates for anomalies.

[0061] <Steps for estimating the cause of the anomaly> In the abnormal cause estimation step, the cause of the abnormality is estimated based on the variables identified in the abnormal cause variable identification step.

[0062] Estimating the cause of an anomaly largely depends on knowledge gained from previous anomalies in specific processes or equipment. Therefore, the following explanation will use the steelmaking process as an example to describe the approach to estimating the cause of an anomaly. The causes of anomalies in manufacturing processes can be broadly classified into equipment-related anomalies such as mechanical failures and electrical failures, process-related anomalies such as operational factors and manufacturing process instability, systemic factors such as abnormal sensor data and control software malfunctions, and human factors such as errors in operating procedures and settings.

[0063] For example, in the steel rolling process, differential load, which is the difference in load generated at both ends of the mill, is monitored. If an abnormality occurs in the process system, such as the rolled material meandering and passing through a position off-center from the center of the line, an imbalance occurs in the load distribution on the rolling rolls, which is ultimately detected as a load difference. In addition, differential load may occur as a result of a discrepancy between the target leveling value and the actual value due to equipment system abnormalities such as malfunctions in the reduction mechanism or mechanical play in the mill components. Furthermore, differential load can also occur due to systemic factors such as an abnormal output from one of the load cells (sensors) that detect the loads at each end of the mill, or due to operator errors in leveling operations.

[0064] In the abnormality diagnosis method according to this embodiment, for example, in a multiple regression model with differential load as the dependent variable, items such as off-center amount and leveling amount can become explanatory variables with significant influence coefficients. Therefore, in step S8, if off-center amount is identified as the variable causing the abnormality based on the abnormality cause estimation matrix, it is considered to be the differential load of the operational factor caused by the increase in off-center amount due to plate meandering. In this case, the cause of plate meandering is assumed to be plate meandering due to excessive operation by the operator, etc.

[0065] On the other hand, if the leveling amount is identified as the variable causing the abnormality in step S8, it is possible that there is a malfunction in the rolling mill's reduction mechanism, such as the cylinder of the reduction device. Also, if the differential load itself is identified as the variable causing the abnormality in step S8, there is concern that the differential load may be caused by the shape of the rolled material (excessive wedge shape) or temperature (temperature non-uniformity).

[0066] Therefore, in the abnormality cause estimation step, as shown in Figure 5 for example, candidate causes of abnormalities are pre-organized in a table according to the combination of the dependent variable and the explanatory variables that are candidates for abnormalities, and stored in the storage unit 30 or the like. Then, the abnormality cause estimation unit 46 estimates the cause of the abnormality by referring to the above table in accordance with the observed abnormality, i.e., the dependent variable and the candidate abnormalities.

[0067] Furthermore, in the abnormality cause estimation step, it is desirable to organize the causes of the abnormality and the corresponding action candidates in a table in advance and store them in the memory unit 30 or the like, according to the combination of the dependent variable and the explanatory variables that are candidates for the abnormality. The abnormality cause estimation unit 46 may then estimate the cause of the abnormality and determine the corresponding action by referring to the above table in accordance with the observed abnormality, i.e., the dependent variable and the candidate for the abnormality.

[0068] Furthermore, in the abnormal cause estimation step, the output unit 20 may output action commands to the control devices of the processes and equipment corresponding to the determined action. In addition, in the abnormal cause estimation step, the output unit 20 may output alert information to the operator's monitoring monitor.

[0069] For example, if a differential load occurs in the steel rolling process, and the cause is a malfunction in the reduction mechanism or mechanical play in the mill components, the output unit 20 will output an emergency stop command for the rolling process to prevent serious trouble. Also, if the rolled material meanders and passes off-center, the output unit 20 will output a leveling correction command corresponding to the amount of off-center movement to the mill's control device. Furthermore, in the case of systemic factors such as the output of a load cell (sensor), the output unit 20 will output a command to switch to an alternative sensor to the mill's control device. In addition, if the leveling operation is mishandled by the operator, the output unit 20 will output alert information to the operator's monitoring monitor.

[0070] In the abnormality diagnosis method and abnormality diagnosis device according to the embodiments described above, a large number of variables (sensor data) obtained during operation are collected, influence coefficients are calculated in a set of multiple regression models in which these variables are complementary as dependent and independent variables, and an abnormality cause estimation matrix is ​​calculated by comparing the calculated influence coefficients with the influence coefficients under normal conditions. Then, the variables causing the abnormality are identified using the abnormality cause estimation matrix.

[0071] In the anomaly cause estimation matrix, as shown in Figures 3 and 4, for example, the vertical direction (rows) represents the dependent variable and the horizontal direction (columns) represents the independent variables. When the change in the influence coefficient is input, the rows and columns containing variables related to the anomaly will change significantly compared to variables not related to the anomaly. Therefore, variables related to the anomaly can be clearly presented by displaying them, for example, in a color chart.

[0072] Thus, the anomaly diagnosis method and anomaly diagnosis device according to the embodiment can improve estimation accuracy compared to a single-model-based cause diagnosis as proposed in Patent Document 1, because the anomaly causative variable is estimated from a set of numerous multiple regression models. Furthermore, the anomaly diagnosis method and anomaly diagnosis device according to the embodiment can clearly and easily present the variable causing the anomaly by calculating an anomaly cause estimation matrix that comprehensively compares the influence coefficients for normal and abnormal conditions for the set of multiple regression models.

[0073] Furthermore, the anomaly diagnosis method and anomaly diagnosis device according to the embodiment can use an anomaly cause estimation matrix to identify the variables that are causing the anomaly in the multiple regression model, thereby estimating the cause of the anomaly and accelerating the decision-making of specific actions. As described above, the anomaly diagnosis method and anomaly diagnosis device according to the embodiment can detect an abnormal state from a large amount of measurement data acquired in the manufacturing process and quickly and clearly present the measurement values ​​related to the anomaly.

[0074] The abnormality diagnosis method and abnormality diagnosis apparatus according to the present invention have been specifically described above with reference to embodiments and examples for carrying out the invention. However, the spirit of the present invention is not limited to these descriptions and must be interpreted broadly based on the claims. Furthermore, it goes without saying that various modifications and alterations based on these descriptions are also included in the spirit of the present invention. [Explanation of Symbols]

[0075] 1. Anomaly Diagnosis Device 10 Input section 20 Output section 30 Storage section 31 Operational Data Database 32 Normal Operation Data Database 40 Arithmetic section 41. Calculation unit for normal operation effect coefficient 42 Abnormality determination section 43. Operational Impact Factor Calculation Unit 44 Abnormal cause estimation matrix calculation unit 45 Abnormal Causative Variable Identification Unit 46 Abnormal cause estimation section

Claims

1. An operational variable acquisition step is performed to acquire multiple variables that indicate the operating state of the manufacturing equipment, An abnormality determination step that determines whether or not there is an abnormality in the manufacturing equipment, using a set of first prediction models constructed based on a plurality of variables obtained in the aforementioned operating variable acquisition step and a plurality of variables indicating the normal state of the manufacturing equipment, wherein the variables are used complementaryly as dependent and independent variables; If an abnormality is determined, the operation influence coefficient calculation step involves constructing a set of second prediction models in which multiple variables obtained in the operation variable acquisition step are used complementaryly as the dependent and independent variables, based on the multiple variables obtained in the operation variable acquisition step, An anomaly cause estimation matrix calculation step, which calculates an anomaly cause estimation matrix by comparing the first influence coefficient, which is the influence coefficient of the explanatory variables on the dependent variable for the set of first prediction models, with the second influence coefficient, which is the influence coefficient of the explanatory variables on the dependent variable for the set of second prediction models. Using the aforementioned abnormality cause estimation matrix, the abnormality cause variable identification step identifies the variable causing the abnormality, An abnormality diagnosis method that includes this.

2. The abnormality diagnosis method according to claim 1, wherein the abnormality cause estimation matrix has a tabular structure with the objective variable and the explanatory variable as key items, and the comparison result of the first influence coefficient and the second influence coefficient is an element.

3. The abnormality diagnosis method according to claim 1 or 2, wherein in the abnormality cause variable identification step, the abnormality cause estimation matrix is ​​used to focus on the explanatory variable or dependent variable of the variable related to the abnormality, and the variable causing the abnormality is identified based on the change in the influence coefficient related to the variable of focus.

4. The abnormality diagnosis method according to claim 3, wherein in the abnormality cause variable identification step, variables whose change in the influence coefficient is greater than a predetermined threshold in both the explanatory variable column and the dependent variable row of the abnormality cause estimation matrix are identified as variables that cause the abnormality.

5. The abnormality diagnosis method according to claim 1, wherein in the abnormality determination step, the degree of abnormality is calculated based on the prediction error of each variable using the first prediction model, and the presence or absence of an abnormality in the manufacturing equipment is determined by referring to the maximum, average, or sum of the degrees of abnormality of all or some of the variables.

6. The abnormality diagnosis method according to claim 1, further comprising an abnormality cause estimation step of estimating the cause of the abnormality based on the identified variables.

7. The abnormality diagnosis method according to claim 6, wherein in the abnormality cause estimation step, the cause of the abnormality is estimated by referring to a table of candidate causes of the abnormality corresponding to combinations of the dependent variable and explanatory variables that are candidates for the abnormality.

8. An anomaly diagnosis method according to claim 6, wherein in the anomaly cause estimation step, the cause of the anomaly is estimated by referring to a table of candidate causes of the anomaly and corresponding actions according to a combination of the dependent variable and candidate explanatory variables for the anomaly, and the action corresponding to the estimated cause of the anomaly is determined.

9. The abnormality diagnosis method according to claim 8, wherein the abnormality cause estimation step outputs a command to a control device corresponding to the determined action, or an alert.

10. An abnormality determination unit determines whether or not there is an abnormality in the manufacturing equipment, using a set of first prediction models constructed based on a plurality of variables indicating the operating state of the manufacturing equipment and a plurality of variables indicating the normal state of the manufacturing equipment, wherein the variables are used complementaryly as dependent and independent variables. If an abnormality is detected, the system includes an operational influence coefficient calculation unit that constructs a set of second prediction models in which multiple variables indicating the operating conditions are used complementaryly as dependent and independent variables, and An anomaly cause estimation matrix calculation unit calculates an anomaly cause estimation matrix by comparing a first influence coefficient, which is the influence coefficient of the explanatory variables on the dependent variable for the set of first prediction models, with a second influence coefficient, which is the influence coefficient of the explanatory variables on the dependent variable for the set of second prediction models. An abnormality cause variable identification unit identifies variables that cause abnormalities using the aforementioned abnormality cause estimation matrix, An abnormality diagnosis device equipped with the following features.

11. The abnormality diagnosis device according to claim 10, further comprising an abnormality cause estimation unit that estimates the cause of the abnormality based on identified variables.

Images