Quality abnormality factor analysis assistance system

Abstract

In the present invention, a data storage unit stores a manufacturing condition for a product, a setting condition for facility equipment, and, as actual results data acquired from measurement equipment, time-series data for the time of manufacturing. A quality abnormality factor estimation calculation unit acquires a quality abnormality site of an abnormal product of which a quality evaluation value calculated from quality data exceeds an allowable range. The quality abnormality factor estimation calculation unit constructs a quality abnormality factor estimation model on the basis of the time series data as relates to a normal product group or on the basis of the manufacturing condition and setting condition, estimates a factor candidate for a quality abnormality on the basis of the constructed quality abnormality factor estimation model, and calculates the degree of relevance of the estimated factor candidate to the quality abnormality. A display information generation unit generates information for displaying, on a display unit, the quality data of the abnormal product, the quality abnormality site, the time series data or the manufacturing condition and setting condition, and the factor candidate and degree of relevance estimated and calculated by the quality abnormality factor estimation calculation unit.

Description

Quality abnormality cause analysis support system

[0001] The present disclosure relates to a quality abnormality factor analysis support system, and more particularly to a quality abnormality factor analysis support system that supports an analysis of the factor of a quality abnormality when a product such as a steel plate manufactured in a rolling plant is determined to have a quality abnormality.

[0002] If a product manufactured in a rolling plant does not satisfy a quality control standard value determined based on customer requirements, etc., it is determined to have a quality abnormality (including poor quality). Patent Document 1 listed below discloses a manufacturing process abnormality determination device.

[0003] Products that are judged to have quality defects are sold as low-grade material or disposed of, resulting in a decline in production efficiency.When a quality defect occurs, it is necessary to identify its characteristics, analyze and investigate the causes, and determine and implement measures, such as changing operating conditions or control settings, to prevent recurrence.

[0004] The series of tasks from determining the characteristics of quality abnormalities to implementing countermeasures requires experienced experts to make comprehensive judgments based on a large amount of data and information, which requires a great deal of time and effort. Patent Document 2 listed below discloses a system for estimating the cause of quality abnormalities.

[0005] Japanese Patent No. 6116445 Japanese Patent No. 4365536

[0006] In Patent Document 2, a group of products with similar input information conditions at the time of manufacturing is found, and quality abnormalities can be detected in advance based on whether the output results of the group of products match the trends.In addition, the accuracy of prediction calculations for a target product can be improved based on the difference between the representative value of the output results of a group of products with similar input information conditions and the prediction result calculated based on the input information of the target product.

[0007] However, Patent Document 2 focuses on detecting quality abnormalities in relation to the manufacturing conditions and setting conditions of individual products and analyzing the causes of the abnormalities. As a result, Patent Document 2 is unable to determine the longitudinal position of the product where the quality abnormality occurred. This is because products are long in the longitudinal direction, and the amount of information in the time-series data (performance data) is too large, so it is common to compress the time-series data to detect quality abnormalities. Therefore, when a quality abnormality occurs, it is necessary to identify the location where the quality abnormality occurred and then analyze the cause of the identified location.

[0008] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a quality abnormality cause analysis support system that, when a quality abnormality is determined in a product, can display the position in the longitudinal direction of the product where the quality abnormality occurred and the cause of the occurrence.

[0009] The first aspect relates to a quality abnormality factor analysis support system. The quality abnormality factor analysis support system includes a data storage unit, a quality abnormality factor estimation calculation unit, and a display information generation unit. The data storage unit stores manufacturing conditions for products manufactured in a rolling plant, setting conditions for equipment installed in the rolling plant, and time-series data during manufacturing as performance data acquired from measuring instruments installed in the rolling plant. The quality abnormality factor estimation calculation unit calculates a quality evaluation value from quality data related to product quality selected from the time-series data during manufacturing, identifies abnormal products or product groups whose calculated quality evaluation value exceeds an allowable range as abnormal products, and acquires quality abnormality locations that exceed the allowable range in the abnormal products. The quality abnormality factor estimation calculation unit constructs a quality abnormality factor estimation model based on time-series data for normal product groups whose quality evaluation value is within an allowable range or on manufacturing conditions and setting conditions, and estimates candidate factors of the quality abnormality from the time-series data or the manufacturing conditions and setting conditions based on the constructed quality abnormality factor estimation model, and calculates the relevance of the estimated candidate factors to the quality abnormality. The display information generation unit generates information for displaying on the display unit the quality data of the abnormal product, the location of the quality abnormality, time series data, or the manufacturing conditions and setting conditions, and the candidate factors and correlations estimated and calculated by the quality abnormality factor estimation calculation unit.

[0010] The second aspect has the following characteristics in addition to the first aspect: The display information generating unit generates information for displaying a list of the manufacturing conditions, setting conditions, quality evaluation values, and quality abnormality locations of the defective product.

[0011] The third aspect has the following feature in addition to the first aspect: the display information generation unit generates information for displaying a trend chart of the quality data of the abnormal product, a trend chart of time-series data corresponding to the most relevant factor candidate among the factor candidates estimated by the quality abnormality factor estimation calculation unit, and quality abnormality locations in each trend chart.

[0012] The fourth aspect has the following feature in addition to the first aspect: the display information generation unit generates information for displaying a trend chart of quality data of the abnormal product, a plurality of candidate factors estimated by the quality abnormality factor estimation calculation unit, the degree of association of each candidate factor, a trend chart of time-series data corresponding to each candidate factor, and quality abnormality locations in each trend chart.

[0013] The fifth aspect has the following characteristics in addition to the first aspect: the display information generation unit generates information for displaying setting conditions as cause candidates and the relevance of each setting condition to the quality abnormality for abnormal products having the same quality data and quality abnormality location.

[0014] The sixth aspect has the following characteristics in addition to the first aspect: The quality abnormality factor estimation calculation unit uses machine learning or a statistical method as a quality abnormality factor estimation model.

[0015] The seventh aspect has the following feature in addition to the first aspect: the quality abnormality factor estimation calculation unit reconstructs the quality abnormality factor estimation model every time a product is manufactured, and updates the quality abnormality factor estimation model when the prediction error of the reconstructed quality abnormality factor estimation model is smaller than the prediction error of the quality abnormality factor estimation model before the reconstruction.

[0016] According to the present disclosure, by acquiring the quality anomaly location where the quality evaluation value of the abnormal product exceeds the allowable range, it is possible to display the longitudinal position of the product where the quality anomaly has occurred. Furthermore, by constructing a quality anomaly factor estimation model using normal data, inferring candidate factors of the quality anomaly using the constructed quality anomaly factor estimation model, and calculating the relevance of the inferred candidate factors to the quality anomaly, it is possible to display highly relevant time-series data as the cause of the quality anomaly. Therefore, it is possible to provide a quality anomaly factor analysis support system that, when a product is determined to have a quality anomaly, can display the longitudinal position of the product where the quality anomaly has occurred and the cause of the quality anomaly.

[0017] 1 is a schematic diagram showing an example of a rolling plant to which a quality abnormality factor analysis support system according to an embodiment 1 is applied. FIG. 2 is a diagram showing the configuration of data collected and registered by a data storage unit. FIG. 3 is a diagram showing the processing flow of the quality abnormality factor analysis support system according to an embodiment 1. FIG. 4 is a diagram showing an example of input / output variables of a quality abnormality factor estimation model. FIG. 5 is a diagram showing various screens displayed on an HMI device based on information generated by a quality information display function. FIG. 6 is a diagram showing typically the display content of a steel plate information list screen. FIG. 7 is a diagram showing typically the display content of a screen that is popped up on the steel plate information list screen. FIG. 8 is a diagram showing typically the display content of a quality abnormality factor estimation result display screen. FIG. 9 is a diagram showing typically the display content of a similar quality abnormality factor estimation analysis screen. FIG. 10 is a diagram showing typically the display content of a quality abnormality factor / setting analysis display screen. FIG. 11 is a diagram showing the processing flow of a quality abnormality factor analysis support system according to an embodiment 2. FIG. 12 is a diagram showing an example of the hardware configuration of a quality abnormality factor analysis support system.

[0018] Hereinafter, a quality abnormality factor analysis support system according to an embodiment of the present disclosure will be described with reference to the drawings. Note that common elements in the drawings will be assigned the same reference numerals and redundant description will be omitted.

[0019] First Embodiment Fig. 1 is a schematic diagram showing an example of a rolling plant 1 to which a quality abnormality cause analysis support system 20 according to a first embodiment is applied.

[0020] The rolling plant 1 is, for example, a hot rolling plant having a hot rolling line, but may also be a cold rolling plant. The rolling plant 1 is equipped with a heating furnace 2, a roughing mill 3, a crop shear 4, a finishing mill 5, a cooling device 6, and a winder 7 as main equipment constituting the hot rolling line. The rolling plant 1 is equipped with a conveying table (not shown) for conveying the steel sheet Pr between the rolling facilities. These rolling facilities are driven by an electrical system of electric motors and actuators.

[0021] The heating furnace 2 is configured to heat the steel plate (slab) Pr to a predetermined temperature (e.g., 1200°C) before rolling. The roughing mill 3 has at least one rolling stand (usually one to three stands) and rolls the steel plate (slab) Pr heated in the heating furnace 2 in multiple passes in the forward direction (from upstream to downstream of the rolling line) and the reverse direction (from downstream to upstream of the rolling line). The crop shear 4 uses upper and lower blades to cut off any shape defects present at the leading or trailing end of the steel plate Pr based on the shape measured by a shape detector 81 (described later). The finishing mill 5 is a tandem rolling mill equipped with, for example, seven rolling stands F1 to F7 arranged side by side in the rolling direction of the steel plate Pr. Each rolling stand F1 to F7 has two upper and lower work rolls 51, two upper and lower backup rolls 52, and an electric motor 53 for roll rotation. The backup rolls 52 are provided with a screw down device 54, which is configured to be able to adjust the gap between the upper and lower work rolls 51. The rolling load of each rolling stand F1 to F7 is measured by a rolling load sensor 55. The cooling device 6 cools the steel sheet Pr by injecting water onto the steel sheet Pr using a cooling bank. The cooled steel sheet Pr is wound into a coil by a winder 7.

[0022] Various sensors are installed as measuring instruments at key points in the rolling plant 1. Key points in the rolling plant 1 include, for example, the outlet side of the heating furnace 2, the outlet side of the roughing mill 3, the outlet side of the finishing mill 5, and the inlet side of the winder 7. Various sensors may also be installed between rolling stands F1 to F7 of the finishing mill 5. The various sensors include a shape detector 81 capable of measuring the shape of the steel sheet Pr at the outlet side of the roughing mill 3, a thermometer 82 that measures the surface temperature of the steel sheet Pr at the inlet side of the finishing mill 5, a speed detector 83 that measures the speed Va of the steel sheet Pr at the outlet side of the finishing mill 5, a thickness / width meter 84 that measures the thickness and width of the steel sheet Pr at the outlet side of the finishing mill 5, a thermometer 85 that measures the surface temperature of the steel sheet Pr at the inlet side of the winder 7, and the rolling load sensor 55. The various sensors sequentially measure the state of the steel plate Pr and the state of each facility device. The performance data measured by the various sensors is transmitted from moment to moment to the control computer 11. Therefore, the performance data is time-series data.

[0023] The rolling plant 1 is operated by a control system using computers with a hierarchical structure. The computers include a process control computer (hereinafter referred to as "control computer") 11 and a host computer 12, which are connected to each other via a network. The control computer 11 is connected to a quality abnormality factor analysis support system 20 (described later) via the network. The control computer 11 has a control controller such as a PLC (programmable logic controller). An HMI (human machine interface) device 13 is connected to the control computer 11 via the network. Data on the monitored object (rolling equipment) is presented to the HMI 13, allowing users (including managers) to monitor or operate (control) the monitored equipment. When a rolling plan designer inputs hot rolling command information, which is a rolling plan, into the host computer 12, the hot rolling command information is sent from the host computer 12 to the control computer 11. The hot rolling command information includes a target plate thickness, a target plate width, a target temperature, etc. The target temperatures include a target temperature on the delivery side of the finishing rolling mill 5 (hereinafter referred to as "finishing delivery temperature") and a target temperature on the entry side of the winder 7 (hereinafter referred to as "coiling temperature"). The process control computer 11 receives hot rolling command information (rolling plan), which is a manufacturing condition, from the host computer 12, calculates design data including setting values ​​of each rolling equipment to be controlled, and transmits the calculated design data to the rolling plant 1, thereby controlling the various rolling equipment.

[0024] The quality abnormality factor analysis support system 20 collects data (rolling data) exchanged between each rolling facility in the rolling plant line 1 and the control computer 12, calculates information to support quality abnormality factor analysis using the collected rolling data, and provides the information to users (including administrators).

[0025] The quality abnormality factor analysis support system 20 includes a data storage unit 21, a quality abnormality factor estimation calculation unit 22, and a display information generation unit 23. The functions of the units 21 to 23 of the quality abnormality factor analysis support system 20 can be realized, for example, by a processor 20b shown in Fig. 12 (described later) reading and executing a program stored in a memory 20c.

[0026] The data storage unit 21 has the function of collecting rolling data and storing it in a database DB. The rolling data includes the aforementioned hot rolling command information, setting data, and performance data. The quality abnormality factor estimation calculation unit 22 uses the rolling data stored in the data storage unit 21 to evaluate product quality and calculate (calculate) information to support analysis of quality abnormality factors based on the evaluation results. The display information generation unit 23 displays the calculation results of the quality abnormality factor estimation calculation unit 22 on the HMI device 13, which serves as a display unit. The operator checks and operates the information (calculation results) displayed on the HMI device 13, and performs analysis work.

[0027] 2 is a diagram showing the configuration of data collected and registered by the data storage unit 21. As shown in Fig. 2, the information collected by the data storage unit 21 includes, for example, information assigned to each steel sheet Pr, such as hot rolling command information and setting data, and time-series data obtained moment by moment for the rolling of the steel sheet Pr, such as performance data, and these pieces of information and data are stored as a database (DB) linked to the manufacturing number of the steel sheet Pr (hereinafter referred to as a "coil ID") and the time when the steel sheet Pr was extracted from the heating furnace 2. Note that, in the example shown in Fig. 1, the database (DB) is provided inside the quality abnormality factor analysis support system 20, but the database (DB) may also be provided outside the quality abnormality factor analysis support system 20, and information and data may be exchanged via a network connection.

[0028] Next, the procedure for calculating information for supporting the evaluation of product quality and the analysis of quality abnormality factors based on the evaluation results, performed by the quality abnormality factor estimation calculation unit 22, will be described with reference to Fig. 3. Fig. 3 is a diagram showing the processing flow of the quality abnormality factor analysis support system 20 according to the first embodiment.

[0029] The quality abnormality factor estimation calculation unit 22 first acquires quality data xi used to evaluate the quality of the steel sheet Pr manufactured in the rolling plant 1, such as the finish exit thickness, finish exit width, shape, finish exit temperature, and coiler entry temperature, selected from the performance data, which is the time-series data described above, from the data storage unit 21 (database DB) (step S1), and calculates a quality evaluation value (step S2). Each of these quality data xi has a defined tolerance range, which is a reference value, and the quality of the steel sheet Pr is evaluated based on this tolerance range. The quality evaluation value is calculated, for example, using the following formulas (1) and (2):

[0030] Here, performance Ａ indicates any one of the quality evaluation values ​​of the steel sheet Pr, such as the finish exit thickness, finish exit width, shape, finish exit temperature, and coiler entry temperature. ｓ The quality evaluation value calculation start point for the steel plate longitudinal position is I Ｌ and xi respectively indicate the end points for calculating the quality evaluation value for the longitudinal position of the steel plate. The longitudinal direction of the steel plate Pr corresponds to the rolling direction and the conveying direction. ｓ From I Ｌ The tolerance indicates a quality standard value of any of the finish outlet plate thickness, finish outlet plate width, shape, finish outlet temperature, and coiler inlet temperature. The start and end points for calculating the quality evaluation value in the longitudinal direction of the steel sheet Pr are often taken as the length excluding the non-steady deformed portions at the leading and trailing ends in the longitudinal direction of the steel sheet Pr (so-called leading and trailing end cuts), but for example, the length excluding the non-steady deformed portions at the leading and trailing ends in the longitudinal direction of the steel sheet Pr may be further divided into several sections, and the quality evaluation value may be calculated for each of the sections.

[0031] The quality evaluation value calculated in step S2 is stored by the data storage unit 21 in association with the coil ID of the steel sheet Pr (step S3). Next, it is determined whether the calculated quality evaluation value satisfies an allowable range as a reference value (step S4). If the quality evaluation value satisfies the allowable range, it is treated as normal quality data. If it does not, it is treated as abnormal quality data. These data are then linked to the coil ID of the steel sheet Pr and labeled (step S5), and stored by the data storage unit 21. The label for the abnormal quality data does not necessarily need to be assigned to distinguish it from normal data, but may be registered as a label corresponding to the characteristics of the abnormality. For example, if the finish delivery thickness exceeds the allowable range, it may be registered as information indicating whether the thickness is thicker than the upper limit or thinner than the lower limit, and further, the extent of the deviation. In addition, by similarly registering the measurement point where the allowable range is exceeded, i.e., the longitudinal position and range of the steel sheet Pr, it is possible to identify which portion (position) of the steel sheet Pr in the longitudinal direction exceeds the allowable range. In this case, for example, labels may be registered as indices (indicators) based on predetermined numerical ranges for the degree of deviation from the upper or lower limit of the tolerance range, or for the position or range of the measurement point that exceeds the tolerance range.

[0032] The number of steel plates Pr that have been labeled as normal data for the target quality is an arbitrary number N. Ｌ Whether or not a certain number N of normal data is collected Ｌ It is determined whether or not the number N has been collected (step S6). ＬIf the quality evaluation value exceeds the allowable range, a quality abnormality factor estimation model (hereinafter also referred to as "model") for the target quality is constructed (step S7), the constructed model is stored in the data storage unit 21 (step S8), and this routine is temporarily terminated. If the quality evaluation value described above exceeds the allowable range after the model construction, the result of step S4 is determined as NO, and the result of step S9 is determined as YES, and the process proceeds to step S10. In step S10, the abnormality factors for the target quality are estimated using the quality abnormality factor estimation model. That is, the quality abnormality factor estimation model estimates candidate factors for the quality abnormality, and the degree of association between each estimated candidate factor and the quality abnormality is calculated. The candidate factors related to the target quality abnormality and the estimated performance data are quantified as "association degrees," and the calculated degree of association between each item of performance data and the coil ID of the steel sheet Pr is stored in the data storage unit 21 in association with the coil ID of the steel sheet Pr, etc.

[0033] In this embodiment, the construction of a quality abnormality factor estimation model is performed only when the above-mentioned conditions are met, and thereafter, as will be explained in the second embodiment described later, the construction can be performed at any timing determined by the user (including the administrator) of the quality abnormality factor analysis support system 20.

[0034] The quality anomaly factor estimation model described above may be, for example, an autoencoder (AE), a random forest (RF), a support vector regression (SVR), or a method utilizing a statistical method, which fall under the category of machine learning. Here, the NN, RF, and SVR algorithms are widely known, and will be briefly described below.

[0035] AE is a type of neural network model (hereinafter referred to as "NN model") that detects anomalies from the discrepancy between input and output variables using an NN model trained only on a group of data labeled as normal data (hereinafter referred to as "normal data") so as to output a variable similar to the input variable. In its simplest configuration, an NN model has a three-layer structure consisting of an input layer, an intermediate layer, and an output layer, but it is also possible to add more intermediate layers. Adding multiple intermediate layers also enables deep learning. Each layer consists of one or more neurons, each with a weight coefficient and a bias value, and they are connected to each other. Typically, an activation function is defined for each neuron, and the output state is designed to change depending on the level of the input value.

[0036] RF is a method that connects multiple decision trees, which are tree-structured weak learners whose output is determined by the conditions of input variables, in parallel, and predicts the final output by taking a majority vote or averaging the outputs of each.SVR is a method that applies the support vector machine algorithm, which determines the boundary line and hyperplane that divides data into two groups, to regression problems.

[0037] In either method, a model for predicting and calculating output variables from input variables must be constructed using a group of data prepared for learning. In this embodiment, a quality abnormality factor estimation model is constructed using the time-series data of steel plates labeled as normal data or the setting data of steel plates labeled as normal data.

[0038] FIG. 4 shows an example of the configuration of input variables and output variables for a model that predicts and calculates output variables from input variables. The input variables include cases where setting data and quality evaluation values ​​related to the target quality data are used (Cases C and D), and cases where actual data related to the target quality data and the quality data are used (Cases A and B). A model that predicts and calculates output variables from input variables may be constructed using either one of these methods, or both. When quality data and actual data related to the quality data are used (Case A), the output variables are the same as the input variables. Here, in RF and SVR, the input variables may be only actual data related to the quality data (Case B), and the quality data may be used as the output variable. Furthermore, if the length of the steel plate Pr is divided into several sections excluding the non-steady deformation portions at the leading and trailing ends in the longitudinal direction, and an abnormality label is assigned to each section, the target section can be extracted and used as an input variable.

[0039] A model that predicts and calculates output variables from input variables is constructed so that the input variables and output variables match, that is, so that the prediction error of the model that predicts and calculates output variables from input variables is minimized. When only actual data related to the target quality data is used and the target quality data is used as the output variable, the model is constructed so that the actual target quality data matches the predicted output variable. In either case, the error is defined as a mathematical formula (error function), and the model is constructed (learned) so that this is minimized. In the case of AE, this error function may be defined, for example, as follows:

[0040] where L is the loss function, h is the hidden layer of the NN, g is the activation function, and W １ is the weight coefficient of the NN in the first hidden layer, RMSE is the root mean square error, and λ is the regularization parameter. １is defined as a matrix of the number of input variables j = {1, 2, ..., j} and the number of neurons in the first hidden layer k = {1, 2, ..., k}. As mentioned above, AE constructs a model that predicts and calculates output variables from input variables so as to accurately predict normal data. In other words, if there is a discrepancy between the input variables and the output variables for the data to be subsequently detected, it can be determined to be an anomaly. Using the loss function described above and the score indicating the relevance defined by the following equations (5) and (6), it is possible to define anomaly determination criteria and detect anomalies for the input variables of the data to be subsequently detected.

[0041] where γ ｊ is the score indicating the relevance to input variable j, γ ｊ ' is the standardized relevance score. The relevance score γ calculated from the normal data used for learning ｊ The abnormality determination criterion can be defined using the above. The abnormality determination criterion can be any numerical value, but for example, it may be the top eighth of the scores indicating the relevance calculated from the normal data used for learning sorted in descending order, but is not limited to this. Data related to quality abnormalities is detected by deviation from this abnormality determination criterion. In the case of AE, the degree of deviation Abnormality ｊ is taken as the "degree of association" in quality abnormality factor estimation, as shown in the following equation (7).

[0042] In the case of RF or SVR, for example, the root mean square error between the input variable and the output variable, or the root mean square error between the actual target quality data and the predicted output variable, may be defined as a loss function to build a model with improved prediction performance for normal data, and the output of subsequently input data may be evaluated to detect data related to quality anomalies. In this case, a method for detecting data related to quality anomalies may be, for example, SHAP (Shapley Additive Explanation), a method utilizing the Shapley value. Since SHAP is a well-known method, it will be briefly described here without going into detail. SHAP can express the degree to which each input variable contributed to the prediction result using the average marginal contribution (Shapley value) of players in a cooperative game in game theory. As in this case, when abnormal data is input into a model built using only normal data, the contribution of the abnormal data item becomes large, and it can be estimated as a quality anomaly factor.

[0043] When a statistical method is used for the anomaly factor estimation model, for example, a representative waveform is constructed for each data item of performance data and quality data related to the target quality data in the normal data described above, and the quality anomaly factors can be estimated by checking the similarity between the representative waveform and the data to be detected later. The representative waveform may be, for example, the average value of each point, as shown in the following equation (8):

[0044] Here, rep ｊ，ｉ is the data of the i-th point of the representative waveform in data item j. The similarity to the representative waveform may be expressed as, for example, the root mean square error. In other words, the larger the value, the lower the similarity can be evaluated. By treating the similarity in the same way as the score indicating the relevance in AE, the relevance can be calculated and the cause of the quality abnormality can be estimated.

[0045] In either method, it is desirable to standardize the input variables for each data item. Standardization may be calculated using the maximum and minimum values ​​of normal data, or may be calculated using the average value and standard deviation. Furthermore, even for a single steel plate Pr, the number of measurement points, i.e., the length of the steel plate at the measurement location, often differs between the performance data related to the quality data and each data item of the quality data. Therefore, the number of measurement points may be interpolated by approximation, and the length of the steel plate may also be standardized.

[0046] The display information generating unit 23 supports the analysis of quality abnormality factors by using the information in which the above-mentioned relevance is registered and performance data associated with the steel sheet Pr. Various screens generated by the display information generating unit 23 are shown in FIG. 5. FIG. 5 is a diagram showing various screens displayed on the HMI device 13 based on the information generated by the display information generating unit 23. The display information generating unit 23 generates information for displaying a steel sheet information list screen 231, a quality abnormality factor estimation result display screen 232, a similar quality abnormality factor estimation analysis screen 233, and a quality abnormality factor / setting analysis display screen 234, and displays these screens 231 to 234 on the HMI device 13. The steps involved in supporting quality abnormality factor analysis will be explained in order.

[0047] First, the steel plate information list screen 231 displays key product information, such as the coil ID and target thickness of the steel plate, as well as quality evaluation values. FIG. 6 is a diagram schematically illustrating the display contents of the steel plate information list screen 231. Key product information, such as the coil ID and target thickness of the steel plate Pr, as described above, and the quality evaluation values ​​are displayed in a table format, with one line per steel plate Pr (one coil ID). For example, a line for a steel plate Pr whose quality evaluation value exceeds the allowable range, such as coil ID "xxx2," may be highlighted by changing the background color or other means. Furthermore, highlighting may be performed by changing the color of the cell or text for a quality item (e.g., quality A "A2") whose quality evaluation value exceeds the allowable range. This screen 231 allows users to check whether there are any steel plates Pr whose quality evaluation value exceeds the allowable range and to confirm their quality data. By pressing the cell for the target quality (quality A "A2") in the row for the steel sheet Pr with the target coil ID "xxx2," a screen 231a such as that shown in FIG. 7 pops up, allowing the user to see an overview of the quality abnormality factor estimation results for the steel sheet Pr. FIG. 7 is a diagram schematically illustrating the display contents of screen 231a that pops up on the steel sheet information list screen 231. Here, the target quality data "quality A" and the actual data "data b" related to the target quality that has the highest correlation among the quality abnormality factor estimations are displayed as a trend chart. The areas where the target quality exceeds the allowable range are highlighted in the diagram with shading, allowing the user to see where (at what position) the allowable range is exceeded on the steel sheet Pr. In this pop-up screen 231a, only the data b with the highest relevance is displayed as the target quality abnormality factor. However, by pressing the detailed display button at the bottom of this pop-up screen 231a, the screen transitions to the quality abnormality factor estimation result display screen 232 shown in Figure 8, where the details of the quality abnormality factor estimation result for the target quality of the steel plate Pr can be confirmed.

[0048] FIG. 8 is a diagram schematically illustrating the display contents of the quality abnormality factor estimation result display screen 232. The quality abnormality factor estimation result display screen 232 displays details of the quality abnormality factor estimation results for the target quality A of the steel sheet Pr (coil ID: xxx2) selected on the steel sheet information list screen 231 described above. Key product information, such as the coil ID and target thickness of the steel sheet Pr, a list of quality evaluation values, and a trend chart of the target quality data are displayed. Among the quality abnormality factor estimation results for the target quality data, performance data related to the target quality A with the highest relevance are displayed as a trend chart. Performance data related to the target quality A with the highest relevance are displayed in ascending order of relevance, allowing users to quickly identify data that may be considered as a quality abnormality factor. Additionally, users can scroll the screen to view data with lower relevance levels. Similar to the pop-up screen 231a of the steel sheet information list screen 231 described above, areas of the target quality A that exceed the allowable range are highlighted in the figure, allowing users to identify where the allowable range is exceeded on the steel sheet. Highly relevant performance data and areas that exceed tolerances can be checked at a glance, so if the user is an experienced person who has been working on hot rolling lines for a long time, they can easily arrive at measures to improve quality. Even if the user is not an experienced person, highly relevant data is displayed at the top, so they can quickly take next actions such as reporting problems and exchanging measures.

[0049] Furthermore, by pressing the Similarity Analysis button 232a on the quality abnormality factor estimation result display screen 232, the user can transition to the similar quality abnormality factor estimation analysis screen 233 shown in FIG. 9 . FIG. 9 is a diagram schematically illustrating the display contents of the similar quality abnormality factor estimation analysis screen 233. The "Actual Data" tab button is pressed on the similar quality abnormality factor estimation analysis screen 233. The similar quality abnormality factor estimation analysis screen 233 allows the user to compare and analyze the quality abnormality factor estimation results with other steel sheets Pr registered with similar quality abnormality characteristics as the steel sheet Pr selected on the steel sheet information list screen 231. The target quality data for steel sheets registered with similar quality abnormality characteristics are displayed together as trend charts in the same graph. The coil IDs (eight coil IDs in the figure) of each steel sheet are also displayed simultaneously. Pressing these IDs may provide a more specialized visualization function, such as highlighting only the trend chart for the steel sheet Pr. At the same time, actual data related to the target quality abnormality is displayed as a list of trend charts. As with the quality abnormality factor estimation result screen 232 described above, in the trend chart of the target quality data and related performance data, points that exceed the target quality tolerance range (points where quality abnormalities have occurred) are highlighted by shading in the diagram, but the range of the relevant points may also be displayed in color, for example, by minimum, maximum, or average value. Similarly, the distribution of the relevance for each of the data b, d, and a is also displayed in a list. In the diagram, the distribution of the relevance for each data is displayed in table format as the number of occurrences of steel plates, but the highest values ​​may be highlighted by using a darker background color, for example. Alternatively, it may be displayed as a histogram chart.

[0050] By pressing the "Setting Data" tab button on the similar quality abnormality factor estimation analysis screen 233, it is possible to transition to a quality abnormality factor / settings analysis display screen 234 shown in FIG. 10. FIG. 10 is a diagram schematically showing the display contents of the quality abnormality factor / settings analysis display screen 234. The quality abnormality factor / settings analysis display screen 234 allows analysis of not only performance data related to the target quality abnormality but also setting data. For example, it is possible to support analysis of the target quality abnormality and setting data by displaying a box-and-whisker chart showing the degree of association with setting data such as design values ​​4, 7, 2, 8, and 5 as factor candidates.

[0051] By checking similar quality anomalies and their estimated causes for a large number of steel plates, analysis can be performed to determine whether the quality anomalies occur consistently, whether they are always caused by the same cause, or whether they occur suddenly under specific conditions, such as the steel type, target thickness, or related setting data. While both charts are shown as trend charts spanning the entire length of the steel plate, if the quality evaluation values ​​are calculated for each section of the length of the steel plate excluding the non-steady deformation portions at the leading and trailing ends in the longitudinal direction of the steel plate, a chart corresponding to the section may be used. Furthermore, since the set values ​​are set before rolling, analyzing the set values ​​estimated as possible causes before rolling based on the correlation can prevent the occurrence of quality anomalies. Combining this with analysis of quality anomalies based on time-series data (actual data) after rolling can improve user usability.

[0052] As described above, according to this embodiment, by acquiring the quality anomaly location where the quality evaluation value of the abnormal product exceeds the allowable range, it is possible to display the longitudinal position of the product where the quality anomaly has occurred. Furthermore, by constructing a quality anomaly factor estimation model using normal data, inferring candidate factors of the quality anomaly using the constructed quality anomaly factor estimation model, and calculating the relevance of the inferred candidate factors to the quality anomaly, it is possible to display highly relevant time-series data as the cause of the quality anomaly. Therefore, it is possible to provide a quality anomaly factor analysis support system that, when a product is determined to have a quality anomaly, can display the longitudinal position of the product where the quality anomaly has occurred and the cause of the quality anomaly.

[0053] Second Embodiment Next, a second embodiment of the present disclosure will be described, focusing on differences from the first embodiment. In the first embodiment, the construction of a quality abnormality factor estimation model was limited to timings that met the above-described conditions or arbitrary timings determined by a user (including a manager) of the quality abnormality factor analysis support system 20. It is difficult to cover all quality abnormality patterns in the rolling plant 1 simply by constructing a quality abnormality factor estimation model at such timings, that is, with only data sets at those timings.

[0054] This embodiment is characterized in that the above-mentioned quality abnormality factor estimation model is constructed every time the rolling of the steel sheet Pr is completed after the above timing. Fig. 11 is a diagram showing a processing flow of the quality abnormality factor analysis support system according to the second embodiment.

[0055] The routine shown in Fig. 11 is started each time rolling is completed. In this routine, similar to the routine shown in Fig. 3, quality data xi is acquired (step S1), and a quality evaluation value is calculated (step S2). Thereafter, it is determined whether or not the quality evaluation value falls outside the allowable range, which is a reference value (step S4). The processing up to this point is the same as in the first embodiment.

[0056] In this embodiment, if the quality evaluation value of the steel sheet Pr is determined to be within the allowable range in step S5, the steel sheet Pr is labeled as normal data (step S5), and new data for the steel sheet Pr is added to the normal data used in constructing the previous model, thereby reconstructing a quality abnormality factor estimation model (step S12). Next, the error calculated from the loss function of the quality abnormality factor estimation model reconstructed in step S12 is compared with the error calculated from the loss function of the previously constructed model (step S13), and if the error of the reconstructed model is small, the reconstructed quality abnormality factor estimation model is saved (step S14).

[0057] According to this embodiment, after constructing a quality abnormality factor estimation model in the above-described first embodiment, the quality abnormality factor estimation model can be updated by reconstructing the model each time rolling is completed. This improves the reproducibility of input variables or the prediction performance of the target quality data, and also makes it possible to estimate quality abnormality factors by using a quality abnormality factor estimation model that is always suited to the latest state of the rolling plant 1. Furthermore, by updating (saving in the data saving unit 21) the quality abnormality factor estimation model when the prediction error of the reconstructed quality abnormality factor estimation model is smaller than the prediction error of the quality abnormality factor estimation model currently in use before reconstruction, it is possible to prevent a decrease in accuracy due to model reconstruction.

[0058] In this embodiment, the model is reconstructed every time rolling is completed, but this is not limiting, and the user can set the timing of reconstruction as appropriate. For example, the model can be reconstructed if the user feels something is wrong when checking the quality abnormality factor estimation result display screen 232. This can improve accuracy.

[0059] FIG. 12 is a diagram illustrating an example of the hardware configuration of the quality abnormality factor analysis support system 20. The above-described functions of the quality abnormality factor analysis support system 20 can be realized by the processing circuit illustrated in FIG. 12. The processing circuit 20 may be dedicated hardware 20a. The processing circuit may include a processor 20b and a memory 20c. The processing circuit may be partially formed as dedicated hardware 20a and further include a processor 20b and a memory 20c. In the example illustrated in FIG. 12, the processing circuit 20 is partially formed as dedicated hardware 20a, and also includes a processor 20b and a memory 20c. The processing circuit 20 may be at least one dedicated hardware 20a. In this case, the processing circuit 20 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof. The processing circuit 20 may include at least one processor 20b and at least one memory 20c. In this case, each function of the quality abnormality factor analysis support system 20 is realized by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the memory 20c. The processor 20b realizes each function of the quality abnormality factor analysis support system 20 by reading and executing the programs stored in the memory 20c. The processor 20b is also called a CPU (Central Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, or DSP. The memory 20c corresponds to a storage device such as a non-volatile or volatile semiconductor memory, such as RAM, ROM, flash memory, EPROM, or EEPROM. The memory 20c can also serve as a database (DB). In this way, the processing circuit 20 can realize each function of the quality abnormality factor analysis support system 20 by hardware, software, firmware, or a combination thereof.

[0060] While the embodiment of the present invention has been described above, the present invention is not limited to the above embodiment and can be implemented in various modifications without departing from the spirit of the present invention. In the above embodiment, the information generated by the display information generation unit 23 is transmitted to the HMI device 13, and the various screens 231, 232, 233, and 234 are displayed on the HMI device 13. However, a display unit may be provided inside the quality abnormality cause analysis support system 20, and the various screens may be displayed on this display unit.

[0061] Furthermore, when the number, quantity, amount, range, etc. of each element is mentioned in the above-mentioned embodiments, the present invention is not limited to the mentioned numbers unless otherwise specified or clearly specified in principle. Furthermore, the structures, etc. described in the above-mentioned embodiments are not necessarily essential to the present invention unless otherwise specified or clearly specified in principle.

[0062] 1...Rolling plant, 13...HMI device (display unit), 20...Quality abnormality factor analysis support system, processing circuit, 20a...dedicated hardware, 20b...processor, 20c...memory, 21...data storage unit, 22...quality abnormality factor estimation calculation unit, 23...display information generation unit, 231...steel plate information list screen, 232...quality abnormality factor estimation result display screen, 233...similar quality abnormality factor estimation analysis screen, 234...quality abnormality factor / setting analysis display screen

Claims

1. A system comprising: a data storage unit; a quality abnormality factor estimation calculation unit; and a display information generation unit; the data storage unit is configured to store manufacturing conditions of products manufactured in a rolling plant, setting conditions of equipment installed in the rolling plant, and time-series data at the time of manufacturing as performance data acquired from measuring equipment installed in the rolling plant; The quality abnormality factor estimation calculation unit calculating a quality evaluation value from quality data related to product quality of the product selected from the time-series data at the time of production; An abnormal product or a group of products whose calculated quality evaluation value exceeds an allowable range is determined as an abnormal product, and a quality abnormality portion of the abnormal product whose quality evaluation value exceeds the allowable range is obtained; registering the time-series data, the manufacturing conditions, and the setting conditions relating to a group of normal products whose quality evaluation values ​​are within the tolerance range as normal data; constructing a quality anomaly factor estimation model using machine learning or a statistical method based on the time-series data registered as the normal data or the manufacturing conditions and the setting conditions registered as the normal data; a factor candidate of the quality abnormality is estimated from the time-series data or the manufacturing conditions and the setting conditions based on the constructed quality abnormality factor estimation model, and a degree of association of the estimated factor candidate with the quality abnormality is calculated; the display information generation unit is configured to generate information for displaying on a display unit the quality data of the abnormal product, the quality abnormality location, the time-series data, or the manufacturing conditions and the setting conditions, and the factor candidate and the degree of association estimated and calculated by the quality abnormality factor estimation calculation unit.

2. 2. The quality abnormality cause analysis support system according to claim 1, wherein the display information generating unit generates information for displaying a list of the manufacturing conditions, the setting conditions, the quality evaluation value, and the quality abnormality location of the abnormal product.

3. 2. The quality anomaly cause analysis support system according to claim 1, wherein the display information generation unit generates information for displaying a trend chart of the quality data of the abnormal product, a trend chart of the time-series data corresponding to the candidate factor having the highest degree of association among the candidate factor estimated by the quality anomaly cause estimation calculation unit, and the quality anomaly location in each trend chart.

4. 2. The quality anomaly cause analysis support system according to claim 1, wherein the display information generation unit generates information for displaying a trend chart of the quality data of the abnormal product, the plurality of candidate cause factors estimated by the quality anomaly cause estimation calculation unit, the relevance of each candidate cause factor, a trend chart of the time-series data corresponding to each candidate cause factor, and the quality anomaly location in each trend chart.

5. 2. The quality anomaly cause analysis support system according to claim 1, wherein the display information generation unit generates information for displaying, for the abnormal products having the same quality data and the same quality anomaly location, the setting conditions as the cause candidates and the relevance of each setting condition to the quality anomaly.

6. A quality anomaly factor analysis support system as described in any one of claims 1 to 5, wherein the quality anomaly factor estimation calculation unit reconstructs the quality anomaly factor estimation model each time the product is manufactured, and updates the quality anomaly factor estimation model if the prediction error of the reconstructed quality anomaly factor estimation model is smaller than the prediction error of the quality anomaly factor estimation model before reconstruction.

Images