Quality defect diagnostic device and quality defect diagnostic method

Abstract

This quality defect diagnostic device comprises: a data division unit for dividing product data to be diagnosed in a two-dimensional manner into prescribed blocks; a data aggregation unit for extracting operation data and facility maintenance data for each pre-established classification of quality defects on the basis of the tendency of occurrence of quality defects in the post-division blocks, and aggregating the extracted operation data and facility maintenance data for each region including two or more blocks in a specific direction of the product; and a quality defect determination unit for determining whether a quality defect has occurred by using a plurality of prediction models that are created for each classification of quality defects and are trained on the relationship between quality data and the post-aggregation operation data and facility maintenance data.

Description

Quality defect diagnosis device and quality defect diagnosis method

[0001] The present invention relates to a quality defect diagnosis device and a quality defect diagnosis method.

[0002] Patent Document 1 discloses a method for predicting and diagnosing product quality, in which operational condition data during product manufacturing and quality judgment data in the final process are linked together in consideration of the quality judgment position on the product, and quality is predicted using a prediction model that has learned from the data.

[0003] Patent No. 6953990

[0004] However, if quality data is used uniformly regardless of product characteristics when predicting product quality, information on multiple types of quality defects caused by different mechanisms will be missing. Therefore, even if a prediction model is created based on such quality data, the prediction performance may not be sufficient.

[0005] The present invention has been made in consideration of the above, and aims to provide a quality defect diagnosis device and a quality defect diagnosis method that can predict product quality defects with high accuracy and estimate the causes of the quality defects.

[0006] In order to solve the above-mentioned problems and achieve the object, the quality defect diagnosis device of the present invention comprises a data dividing unit that divides product data to be diagnosed, which consists of operation data, equipment maintenance data, and quality data, into predetermined blocks in two dimensions; a data aggregating unit that extracts the operation data and the equipment maintenance data for each predetermined quality defect classification based on the tendency of quality defects to occur in the divided blocks, and aggregates the extracted operation data and the equipment maintenance data for each area including two or more blocks in a specific direction of the product; and a quality defect determination unit that determines whether or not a quality defect has occurred using a plurality of prediction models that are created for each quality defect classification and that have learned the relationship between the aggregated operation data and the equipment maintenance data and the quality data.

[0007] Furthermore, in the above-described quality defect diagnosis device according to the present invention, the device further comprises a factor estimation unit that estimates the cause of the quality defect based on the result of the quality defect determination.

[0008] Furthermore, in the above-described quality defect diagnosis device according to the present invention, the device further comprises a result display unit that displays at least one of the determination result from the quality defect determination unit and the factor estimation result from the factor estimation unit on an output unit.

[0009] Furthermore, in the quality defect diagnosis device of the present invention, in the above invention, the predictive model is created by two-dimensionally dividing past performance product data consisting of the operation data, the equipment maintenance data, and the quality data into predetermined blocks, classifying quality defects contained in the quality data based on the tendency of quality defects to occur in the divided blocks, aggregating the operation data and the equipment maintenance data extracted for each quality defect classification into areas including two or more blocks in a specific direction of the product, and learning by machine learning the relationship between the aggregated operation data and equipment maintenance data and the quality data for each quality defect classification.

[0010] In the quality defect diagnosis device according to the present invention, in the above invention, the operational data is data that varies over the length of the product in a specific direction.

[0011] In the quality defect diagnosis device according to the present invention, in the above invention, the facility maintenance data is data that changes while the product is being processed and treated in the facility.

[0012] In the quality defect diagnostic device according to the present invention, in the above invention, the quality data is data that varies over the length of the product in a specific direction.

[0013] In addition, in the quality defect diagnosis device of the present invention, in the above invention, the equipment maintenance data is data indicating the equipment state and its maintenance state that affect a specific position of the product by directly or indirectly acting on the product.

[0014] In addition, in the quality defect diagnosis device of the present invention, in the above invention, the data aggregation unit extracts the operation data and the equipment maintenance data for each predetermined quality defect classification based on a quality defect occurrence trend having multiple quality defect occurrence locations along the longitudinal direction of the product or a quality defect occurrence trend having multiple quality defect occurrence locations along the width direction of the product.

[0015] In addition, in the quality defect diagnosis device of the present invention, in the above invention, the quality defect judgment unit judges whether or not a quality defect has occurred using a plurality of prediction models that are created for each classification of quality defects and that have learned the relationship between the aggregated equipment maintenance data and the quality data, and the factor estimation unit estimates the factor of the quality defect based on the quality defect judgment result.

[0016] In the quality defect diagnosis device according to the present invention, in the above invention, the result display unit causes the output unit to display the data group that has been subjected to the alignment process by the data aggregation unit.

[0017] In order to solve the above-mentioned problems and achieve the object, the quality defect diagnosis method of the present invention includes: a data division step in which a data division means provided in a computer divides product data to be diagnosed, which includes operation data, equipment maintenance data, and quality data, into two-dimensional blocks; a data aggregation step in which a data aggregation means provided in the computer extracts the operation data and the equipment maintenance data for each predetermined quality defect classification based on the tendency of quality defects to occur in the divided blocks, and aggregates the extracted operation data and the equipment maintenance data for each area including two or more blocks in a specific direction of the product; and a quality defect judgment step in which a quality defect judgment means provided in the computer judges whether or not a quality defect has occurred using a plurality of prediction models that are created for each quality defect classification and that have learned the relationship between the aggregated operation data and equipment maintenance data and the quality data.

[0018] Moreover, in the quality defect diagnosis method according to the present invention, in the above invention, the cause estimation means included in the computer includes a cause estimation step of estimating a cause of the quality defect based on the result of the quality defect determination.

[0019] According to the quality defect diagnosis device and quality defect diagnosis method of the present invention, by creating and using a prediction model that represents the relationship between operational data and equipment maintenance data and quality defects for each classification according to the tendency of quality defects to occur, it is possible to predict product quality defects with high accuracy and estimate the causes of the quality defects.

[0020] FIG. 1 is a diagram illustrating a schematic configuration of an information processing device that realizes a quality defect diagnosis device according to an embodiment of the present invention. FIG. 2 is a diagram illustrating details of a data division step performed by a data division unit of the quality defect diagnosis device according to an embodiment of the present invention. FIG. 3 is a diagram illustrating details of a quality defect classification step performed by a quality defect classification unit of the quality defect diagnosis device according to an embodiment of the present invention. FIG. 4 is a diagram illustrating a first example of classifying continuous quality defects in the longitudinal direction of a product in the quality defect classification step performed by the quality defect classification unit of the quality defect diagnosis device according to an embodiment of the present invention. FIG. 5 is a diagram illustrating a second example of classifying continuous quality defects in the longitudinal direction of a product in the quality defect classification step performed by the quality defect classification unit of the quality defect diagnosis device according to an embodiment of the present invention. FIG. 6 is a diagram illustrating a first example of classifying continuous quality defects in the width direction of a product in the quality defect classification step performed by the quality defect classification unit of the quality defect diagnosis device according to an embodiment of the present invention. FIG. 7 is a diagram illustrating a second example of classifying continuous quality defects in the width direction of a product in the quality defect classification step performed by the quality defect classification unit of the quality defect diagnosis device according to an embodiment of the present invention. Fig. 8 is a diagram for explaining details of a first example of a result display step performed by the result display unit of the quality defect diagnosis device according to an embodiment of the present invention. Fig. 9 is a diagram for explaining details of a second example of a result display step performed by the result display unit of the quality defect diagnosis device according to an embodiment of the present invention. Fig. 10 is a flowchart showing the steps of a method for creating a prediction model used in the quality defect diagnosis method according to an embodiment of the present invention. Fig. 11 is a flowchart showing the steps of the quality defect diagnosis method according to an embodiment of the present invention.

[0021] A quality defect diagnosis device and a quality defect diagnosis method according to an embodiment of the present invention will be described with reference to the drawings. The quality defect diagnosis device and the quality defect diagnosis method are used to diagnose quality defects occurring in products.

[0022] The "product" in this embodiment is, for example, a steel product manufactured through multiple processes, such as a semi-finished product such as a slab, or a product such as a steel plate manufactured by rolling the slab. The "quality defect" in this embodiment includes, for example, a surface defect, an internal defect, a characteristic value of the material (e.g., a result of determining mechanical properties), etc.

[0023] 1 shows the configuration of an information processing device 1 that realizes a quality defect diagnosis device according to an embodiment. As shown in FIG. 1, the information processing device 1 includes an input unit 10, a storage unit 20, a calculation unit 30, and an output unit 40.

[0024] The quality defect diagnosis device according to the embodiment can be realized by the components of the information processing device 1, excluding the quality defect classification unit 32 and the model creation unit 34 of the calculation unit 30. Furthermore, the prediction model creation device used in the quality defect diagnosis device according to the embodiment can be realized by the components of the information processing device 1, excluding the quality defect determination unit 35, the factor estimation unit 36, and the result display unit 37 of the calculation unit 30. Each component of the information processing device 1 will be described below.

[0025] The input unit 10 is an input means for the calculation unit 30, and is realized by an input device such as a keyboard, a mouse pointer, a numeric keypad, etc. The input unit 10 inputs information required for various processes in the calculation unit 30.

[0026] The storage unit 20 is composed of recording media such as an EPROM (Erasable Programmable ROM), a hard disk drive (HDD), and removable media. Examples of removable media include a USB (Universal Serial Bus) memory, a CD (Compact Disc), a DVD (Digital Versatile Disc), and a Blu-ray (registered trademark) Disc (BD). The storage unit 20 can store an operating system (OS), various programs, various tables, various databases, and the like. The storage unit 20 also stores a product database (DB) 21.

[0027] Data on products manufactured in the past is stored in the product DB 21. This product data includes, for example, operation data, equipment maintenance data, and quality data.

[0028] Operational data is data obtained by measuring the manufacturing state during product manufacturing or by estimating it by some other means. This operational data is preferably data that varies over the length of the product in a specific direction. "Data that varies over the length of the product in a specific direction" refers to data that varies continuously, for example, in the longitudinal or transverse direction of the product.

[0029] Examples of operational data include data on manufacturing conditions in the steelmaking process, hot rolling process, cold rolling process, and annealing process in each manufacturing process of steel products. Specific examples of operational data include steel composition, steel sheet temperature during hot rolling, hot rolling speed, cold rolling speed, steel sheet temperature during annealing, and annealing time. The operational data is measured, for example, at predetermined intervals along the longitudinal direction of the steel sheet, depending on the travel distance in the conveyance direction in each process.

[0030] Equipment maintenance data is data obtained by measuring or estimating by some means the state of equipment and its maintenance state during product manufacturing. This equipment maintenance data is preferably data that changes while the product is being processed and treated in the equipment. Furthermore, the equipment maintenance data is preferably data that indicates the state of equipment and its maintenance state that affects a specific position on the product by directly or indirectly acting on the product, such as data that directly or indirectly indicates the contact state of equipment that comes into contact with the product during the manufacturing process.

[0031] Equipment maintenance data includes, for example, data on the contact state of transport rolls that rotate in contact with the product to transport the product in each manufacturing process of steel products. Equipment maintenance data also includes, for example, data on the contact state of walking beams that contact and lift the product in a heating furnace to transport the product in the heating process before rolling. Specifically, equipment maintenance data includes data on the shape of the transport rolls across their width, data on the contact position of the walking beam with the product, etc.

[0032] Furthermore, among the equipment maintenance data, data on the shape of the transport roll across its width is, for example, measured as a roll profile, which is the wear depth of the roll at predetermined intervals across the width of the transport roll. Furthermore, among the equipment maintenance data, data on the contact position of the walking beam with the product is, for example, measured by measuring the leading and trailing ends of the product in its longitudinal direction within a heating furnace, and calculated from the total length of the product within the heating furnace and the position of the walking beam within the heating furnace. In this way, by using equipment maintenance data when creating a prediction model and diagnosing quality defects, it is possible to accurately predict consecutive quality defects in a specific direction (longitudinal or transverse) of the product, as described below.

[0033] The quality data is data that indicates whether the quality of a product is good or bad, based on the results of measuring the quality of the product or estimating it by some other means. This quality data is preferably data that varies over the length of the product in a specific direction. Examples of quality data include data on surface defects (surface flaws) of the product. Specifically, the quality data may include data on the position and shape of surface flaws on the product.

[0034] The above product data includes "past performance" product data used when creating a prediction model (see FIG. 10) and "diagnosis target" product data used when diagnosing quality defects (see FIG. 11).

[0035] In addition to the product DB 21, the storage unit 20 also stores, for example, product data associated with each block by the data dividing unit 31, information on quality defect classification by the quality defect classification unit 32, etc. In addition to the product DB 21, the storage unit 20 also stores, for example, operation data and equipment maintenance data aggregated by block unit by the data aggregation unit 33. The storage unit 20 may also store, as necessary, the results of quality defect determination by the quality defect determination unit 35, the results of quality defect cause estimation by the cause estimation unit 36, etc.

[0036] The calculation unit 30 is realized by a processor such as a CPU (Central Processing Unit) and a memory (main storage unit) such as a RAM (Random Access Memory) or a ROM (Read Only Memory).

[0037] The calculation unit 30 loads a program into the working area of ​​the main memory and executes it, and controls each component part through the execution of the program to realize functions that meet a predetermined purpose. Through the execution of the program, the calculation unit 30 functions as a data division unit 31, a quality defect classification unit 32, a data aggregation unit 33, a model creation unit 34, a quality defect determination unit 35, a factor estimation unit 36, and a result display unit 37. Note that while FIG. 1 shows an example in which the functions of each unit are realized by, for example, one computer (calculation unit), the means for realizing the functions of each unit is not particularly limited, and the functions of each unit may be realized by, for example, multiple computers.

[0038] The data dividing unit 31 divides the product data into two-dimensional blocks. First, the data dividing unit 31 collects operation data, equipment maintenance data, and quality data, for example, from the product DB 21. Next, the data dividing unit 31 divides the entire length of the product data in the product longitudinal direction into p blocks and the entire length of the product data in the product width direction into q blocks, as shown in FIG. 2 . In the figure, the leftmost block in the product longitudinal direction is numbered 1, and the rightmost block is numbered q. Also in the figure, the uppermost block in the product width direction is numbered 1, and the lowermost block is numbered p. The data dividing unit 31 stores the product data, consisting of operation data and quality data, in the storage unit 20, in association with each block into which the product has been divided, as shown in FIG. 2 .

[0039] The data division unit 31 divides the past performance product data when creating a prediction model (see FIG. 10). The data division unit 31 also divides the product data to be diagnosed when performing a quality defect diagnosis (see FIG. 11). It is desirable to use the same data division method (number of divisions, area of ​​each block, etc.) when creating a prediction model and when performing a quality defect diagnosis.

[0040] The data division unit 31 divides each product data item's total length in the product longitudinal direction into p blocks and its total length in the product width direction into q blocks. The number of divisions (p, q values) for the product data items is not particularly limited, but it is desirable to divide the product data items into at least three divisions (q, p values ​​> 3). That is, p is an integer greater than or equal to 3, and q is an integer greater than or equal to 3, and the values ​​of p and q can be determined arbitrarily in advance. Furthermore, the area of ​​each divided block may be the same for each product data item, or may be different. For example, in the classification of quality defects described below (see FIG. 3), blocks classified as continuous quality defects in the product longitudinal direction or in the product width direction may have different areas (regions) for each product data item.

[0041] The quality defect classification unit 32 classifies the quality defects contained in the quality data into two or more categories based on the tendency of quality defects occurring in the divided blocks. Here, the "proneness of quality defects" refers to, for example, the location and distribution of quality defects occurring in the divided blocks (see FIG. 2).

[0042] Specifically, the quality defect classification unit 32 classifies quality defects in the quality data into those that occur continuously along the longitudinal direction of the product (continuous in the product longitudinal direction) and those that occur continuously across the width direction of the product (continuous in the product width direction). The quality defect classification unit 32 classifies quality defects in the quality data into either continuous in the product longitudinal direction or continuous in the product width direction, as shown in Figure 3, for example. The quality defect classification unit 32 also classifies quality defects into continuous in the product longitudinal direction and continuous in the product width direction, for example, based on the following criteria (1) to (4).

[0043] (1) When multiple quality-defective blocks exist in the longitudinal direction of the product at each position along the product width direction, and the number of quality-defective blocks is s, if s / p≧x, the quality defect is classified as "continuous in the longitudinal direction of the product." In this case, x is an arbitrary real number in the range of 0<x≦1, and is set individually depending on the type of quality defect. (2) When multiple quality-defective blocks exist in the longitudinal direction of the product at each position along the product width direction, and m or more quality-defective blocks are adjacent to each other in a row, the quality defect is classified as "continuous in the longitudinal direction of the product." In this case, m is an arbitrary integer of 2 or greater, and is set individually depending on the type of quality defect. (3) When multiple quality-defective blocks exist in the longitudinal direction of the product at each position along the product width direction, and the number of quality-defective blocks is t, if t / q≧y, the quality defect is classified as "continuous in the longitudinal direction of the product." In this case, y is an arbitrary real number in the range of 0<y≦1, and is set individually depending on the type of quality defect. (4) When there are multiple quality-defective blocks in the product width direction at each position in the product longitudinal direction, if n or more quality-defective blocks are adjacent to each other in a row, the quality defect is classified as "continuous in the product width direction." In this case, n is an arbitrary integer of 2 or more and is set individually depending on the type of quality defect.

[0044] FIG. 4 shows a case where the quality defect classification unit 32 classifies a quality defect as continuous in the product longitudinal direction based on the above-mentioned criterion (1). For example, if "x = 0.7" in the above-mentioned criterion (1) is set, "s = 5" is obtained, and "s / p = 0.71 ≧ x", so the quality defect in the figure is classified as "continuous in the product longitudinal direction." Note that while the figure shows an example where the quality defect blocks are continuous in the product longitudinal direction, even if the quality defect blocks are not continuous in the product longitudinal direction, the above-mentioned criterion (1) can be met. For example, even if the quality defect blocks are scattered in the product longitudinal direction, they can be classified as "continuous in the product longitudinal direction" if the above-mentioned criterion (1) is met.

[0045] 5 shows a case where a quality defect is classified as a continuous quality defect in the longitudinal direction of the product based on the above criterion (2) by the quality defect classification unit 32. For example, when "m=3" in the above criterion (2), the number of blocks adjacent to each other in the longitudinal direction of the product is "3", so the quality defect in the figure is classified as "continuous in the longitudinal direction of the product".

[0046] FIG. 6 shows a case where the quality defect classification unit 32 classifies a quality defect as continuous in the product longitudinal direction based on the above-mentioned criterion (3). For example, if "y = 0.7" in the above-mentioned criterion (3) is set, "t = 6" and "t / q = 0.86 ≧ y", so the quality defect in the figure is classified as "continuous in the product width direction." Note that while the figure shows an example where the quality defect blocks are continuous in the product width direction, even if the quality defect blocks are not continuous in the product width direction, the above-mentioned criterion (1) can be met. For example, even if the quality defect blocks are scattered in the product width direction, they can be classified as "continuous in the product width direction" if the above-mentioned criterion (3) is met.

[0047] 7 shows a case where a quality defect is classified as a continuous quality defect in the longitudinal direction of the product based on the above criterion (4) by the quality defect classification unit 32. For example, when "n=4" in the above criterion (4) is set, the number of blocks adjacent to each other in the width direction of the product is "5", so the quality defect in the figure is classified as "continuous in the width direction of the product".

[0048] The quality defect classification unit 32 stores information related to the quality defect classification processed as described above in the storage unit 20. Furthermore, when creating a prediction model (see FIG. 10 ), the quality defect classification unit 32 classifies past performance quality data. When performing quality defect diagnosis (see FIG. 11 ), the quality defect classification unit 32 aggregates the operation data and equipment maintenance data of the diagnosis target using the quality defect classification used when creating the prediction model.

[0049] The data aggregator 33 aggregates operational data and equipment maintenance data for each block. The data aggregator 33 first extracts operational data and equipment maintenance data for each quality defect classification (see FIG. 3 ) based on the tendency of quality defects to occur in the divided blocks. The quality defect classification used here refers to the quality defect classification (product longitudinal direction continuous, product width direction continuous) classified by the quality defect classification unit 32 when creating the prediction model.

[0050] In this case, the data aggregating unit 33 extracts operation data and equipment maintenance data for each quality defect classification based on a quality defect occurrence trend (see FIGS. 4 and 5) having multiple quality defect occurrence locations (quality defect occurrence blocks) along the product longitudinal direction. The data aggregating unit 33 also extracts operation data and equipment maintenance data for each quality defect classification based on a quality defect occurrence trend (see FIGS. 6 and 7) having multiple quality defect occurrence locations along the product width direction.

[0051] Next, the data aggregating unit 33 aggregates the extracted operation data and equipment maintenance data for each region including two or more blocks in a specific direction of the product (continuous in the product longitudinal direction, continuous in the product width direction). The data aggregating unit 33 aggregates the operation data and equipment maintenance data, for example, as follows.

[0052] 4 and 5, for quality defects classified as "continuous in the product longitudinal direction," the data aggregating unit 33 extracts only the operation data and equipment maintenance data of p blocks in the product longitudinal direction at each position along the product width direction. The data aggregating unit 33 then aggregates the extracted operation data and equipment maintenance data by block.

[0053] 6 and 7, for quality defects classified as "continuous in the product width direction," the data aggregating unit 33 extracts only the operation data and equipment maintenance data of q blocks in the product width direction at each position along the product longitudinal direction. The data aggregating unit 33 then aggregates the extracted operation data and equipment maintenance data by block.

[0054] For example, a basic statistic representing the value at each position within a block can be used as an aggregation method in the data aggregation unit 33. Specifically, the basic statistic is a value such as the average value, maximum value, minimum value, standard deviation, variance, or a range represented by the difference between the maximum and minimum values.

[0055] In this way, the data aggregator 33 aggregates the operation data and equipment maintenance data by calculating one or more of the average value, maximum value, minimum value, standard deviation, variance, and difference between the maximum and minimum values ​​of the operation data and equipment maintenance data for each block. Note that, when aggregating the operation data and equipment maintenance data, at least one type of basic statistical quantity is used, but multiple types of basic statistical quantities may also be used. The data aggregator 33 stores the operation data and equipment maintenance data aggregated down to the block unit in the memory unit 20.

[0056] The data aggregating unit 33 aggregates operational data and equipment maintenance data of past results when creating a prediction model (see FIG. 10 ). Also, the data aggregating unit 33 aggregates operational data and equipment maintenance data of the diagnosis target when performing quality defect diagnosis (see FIG. 11 ).

[0057] The model creation unit 34 creates a plurality of prediction models according to the classification of quality defects. For each classification of quality defects, the model creation unit 34 creates a plurality of prediction models according to the classification of quality defects by learning, through machine learning, the relationship between the aggregated operation data and equipment maintenance data and the quality data.

[0058] That is, the model creation unit 34 constructs a prediction model that shows the relationship between the operation data and equipment maintenance data aggregated by block for each quality defect classification and the quality defects that occur in the corresponding block or product. The model creation unit 34 creates, for example, two types of prediction models that predict continuous quality defects in the product's longitudinal direction and continuous quality defects in the product's width direction, respectively. These prediction models can be created using various machine learning algorithms, such as deep neural networks, decision trees, and support vector machines, as well as other statistical algorithms.

[0059] The quality defect determination unit 35 determines whether or not a quality defect has occurred using multiple prediction models created by the model creation unit 34. The quality defect determination unit 35 determines whether or not a quality defect has occurred for each category by inputting the operation data and equipment maintenance data of the diagnosis target into each classification of prediction model. Here, the operation data and equipment maintenance data input into the prediction model are data aggregated in block units after being data divided by the data division unit 31 and data aggregated by the data aggregation unit 33 (see FIG. 11 ).

[0060] The quality defect determination unit 35 determines whether a product is defective, for example, before or during product manufacturing. When the determination is performed before product manufacturing, the set values ​​of pre-processing manufacturing conditions and equipment conditions are input into the prediction model as the operation data and equipment maintenance data of the product to be predicted, and calculations are performed to predict quality defects in the manufactured product. When the determination is performed during product manufacturing, the set values ​​of processed manufacturing conditions and equipment conditions are input into the prediction model as the operation data and equipment maintenance data of the product to be predicted, and calculations are performed to predict quality defects in the manufactured product.

[0061] The factor estimation unit 36 ​​estimates the factor of the quality defect based on the result of the quality defect determination by the quality defect determination unit 35. The "factor of the quality defect" refers to, for example, operation data and equipment maintenance data that are important factors in the quality defect of the product.

[0062] For example, when it is determined that a quality defect will occur in a specific product being diagnosed, the cause estimation unit 36 ​​can estimate the operational data and equipment maintenance data necessary to prevent the occurrence of the quality defect. In other words, when a quality defect is predicted to occur in a product that is not yet manufactured or is in the middle of being manufactured, the cause estimation unit 36 ​​estimates the operational data and equipment maintenance data that need to be changed or adjusted to prevent the occurrence of the quality defect. This makes it possible to consider countermeasures to prevent the occurrence of the quality defect.

[0063] Furthermore, when a quality defect occurs in a specific product, the cause estimation unit 36 ​​can estimate the operation data and equipment maintenance data that caused the quality defect. In other words, when a quality defect occurs in a manufactured product, the cause estimation unit 36 ​​estimates the operation data and equipment maintenance data that caused the quality defect. This makes it possible to identify the cause of the quality defect.

[0064] Furthermore, when quality defects occur in a large number of products, the cause estimation unit 36 ​​can estimate the operational data and equipment maintenance data that caused the quality defects. In other words, when quality defects occur in a large number of manufactured products, the cause estimation unit 36 ​​can estimate the operational data and equipment maintenance data that caused the quality defects. This makes it possible to identify the cause of the quality defects.

[0065] When the factor estimation unit 36 ​​determines that a specific product to be diagnosed will suffer from a quality defect and estimates the operational data and equipment maintenance data necessary to prevent the occurrence of the quality defect, it calculates an index corresponding to the contribution of each factor in the above-mentioned determination for the product data.The factor estimation unit 36 ​​then estimates factors with large values ​​for this index as important factors.In this case, it uses a general-purpose prediction important factor calculation method that does not depend on a prediction model creation algorithm, such as a method called "SHAP (Shapley Additive exPlanations)."

[0066] Furthermore, when a quality defect occurs in a specific product and the operation data and equipment maintenance data that caused the quality defect are to be estimated, the factor estimation unit 36 ​​calculates an index corresponding to the contribution of each factor in the above-mentioned judgment for the data of the product.The factor estimation unit 36 ​​then estimates the factor for which the index indicates a large value as an important factor.In this case, a general-purpose method for calculating important prediction factors, such as the above-mentioned SHAP, that does not depend on the algorithm for creating a prediction model, is used.

[0067] Furthermore, when quality defects occur in a large number of products and the operation data and equipment maintenance data that caused the quality defects are estimated, the factor estimation unit 36 ​​calculates an index corresponding to the influence of each factor in the entire prediction model, and estimates factors with large values ​​for the index as important factors.

[0068] The factor estimation unit 36 ​​may also use a method for calculating prediction importance factors specific to the algorithm that created the prediction model, such as importance based on Gini impurity in a decision tree algorithm. Alternatively, the factor estimation unit 36 ​​may use a general-purpose method for calculating prediction importance factors that is independent of the algorithm that created the prediction model, such as "Permutation Importance."

[0069] The result display unit 37 displays (outputs) at least one of the determination result from the quality defect determination unit 35 and the factor estimation result from the factor estimation unit 36 ​​on the output unit 40. Furthermore, it is preferable that the result display unit 37 displays the data group that has been subjected to the alignment process by the data aggregation unit 33 on the output unit 40 in an integrated manner.

[0070] When the quality defect judgment results by the quality defect judgment unit 35 are output to the output unit 40, for example, a product plan view divided into blocks by quality defect classification, with the product longitudinal direction as the horizontal direction and the product width direction as the vertical direction, is displayed on a screen on which information about the product to be diagnosed can be viewed. Then, the quality prediction results for each position on the product are displayed on the product plan view.

[0071] In this case, for example, as shown in Fig. 8, on the product plan view, parts (blocks) with quality defects may be marked with an X. Alternatively, on the product plan view, parts without quality defects may be marked with a ○, and parts with quality defects may be marked with an X. Alternatively, different colors may be used depending on the prediction result of whether or not there is a quality defect. Alternatively, only when the prediction result shows that there is a quality defect, the output unit 40 may output a statement such as "The product to be diagnosed (or a specific block of the product to be diagnosed) is predicted to have a quality defect."

[0072] 9, one or more pieces of operational data and equipment maintenance data that vary in the product longitudinal direction may be displayed in chart form below (or above) the product plan view. Also, one or more pieces of operational data and equipment maintenance data that vary in the product width direction may be displayed in chart form on the left (or right) side of the product plan view.

[0073] When the result of the factor estimation of quality defects by the factor estimation unit 36 ​​is output to the output unit 40, for example, a product plan view divided into blocks by quality defect classification, with the product longitudinal direction as the horizontal direction and the product width direction as the vertical direction, is displayed on a screen on which information about the product to be diagnosed can be viewed. Then, quality data and quality prediction results for each product position are displayed on the product plan view. Alternatively, two product plan views may be displayed, with quality data for each product position displayed on one of the product plan views and quality prediction results for each product position displayed on the other plan view.

[0074] 9, the factor estimation results of the quality defect by the factor estimation unit 36 ​​may be displayed next to the product plan view. In this case, the factor estimation results may be displayed in a list of all items of each input data together with the calculated importance index values, or may be displayed up to an arbitrary certain number of items of each input data in descending order of the calculated importance index values. The method of displaying the factor estimation results is not limited, and other methods may also be used.

[0075] Furthermore, when displaying the results of the factor estimation of quality defects by the factor estimation unit 36, one or more pieces of operational data and equipment maintenance data that vary in the longitudinal direction of the product may be displayed in chart form above or below the product plan view. In this case, the data to be displayed may be selected up to a certain number of pieces in descending order of the value of the calculated importance index.

[0076] Furthermore, when displaying the results of the factor estimation of quality defects by the factor estimation unit 36, one or more pieces of operational data and equipment maintenance data that vary in the product width direction may be displayed in the form of a chart on the left or right side of the product plan view. In this case, the data to be displayed may be selected up to a certain number of pieces in descending order of the value of the calculated importance index.

[0077] The output unit 40 is realized by a display device such as an LCD display, a CRT display, etc. Based on the display signal input from the result display unit 37, the output unit 40 displays, for example, the result of the quality defect determination by the quality defect determination unit 35, the result of the quality defect cause estimation by the cause estimation unit 36, etc. in the form of characters, figures, etc.

[0078] (Method for Creating a Prediction Model) A method for creating a prediction model used in the quality defect diagnosis method according to the embodiment will be described with reference to Fig. 10. The method for creating a prediction model includes a block division step (step S1), a quality defect classification step (step S2), a data aggregation step (step S3), and a model creation step (step S4).

[0079] In the block division step, the data division unit 31 divides the past performance product data into predetermined blocks two-dimensionally (step S1). Subsequently, in the quality defect classification step, the quality defect classification unit 32 classifies the quality defects included in the quality data according to the tendency of quality defects occurring in the divided blocks (step S2).

[0080] Next, in a data aggregation step, the data aggregation unit 33 aggregates the operation data and equipment maintenance data extracted for each quality defect classification into regions including two or more blocks in a specific direction of the product (product longitudinal direction, product width direction) (step S3). Next, in a model creation step, the model creation unit 34 creates a prediction model by learning the relationship between the aggregated operation data and equipment maintenance data and the quality data for each quality defect classification through machine learning (step S4).

[0081] (Quality Defect Diagnosis Method) A quality defect diagnosis method according to an embodiment will be described with reference to Fig. 11. The prediction model creation method includes a block division step (step S11), a data aggregation step (step S12), a quality defect determination step (step S13), a factor estimation step (step S14), and a result display step (step S15).

[0082] In the following explanation, it is assumed that a prediction model created in advance at a different timing by steps S1 to S4 in Fig. 10 is used, but a prediction model may also be created when diagnosing a quality defect. That is, steps S1 to S4 in Fig. 10 and steps S11 to S15 in Fig. 11 may be performed at different times, or steps S1 to S4 in Fig. 10 may be performed followed by steps S11 to S15 in Fig. 11.

[0083] In the block division step, the data division unit 31 divides the product data to be diagnosed into predetermined blocks two-dimensionally (step S11). Subsequently, in the data aggregation step, the data aggregation unit 33 aggregates the operation data and equipment maintenance data extracted for each predetermined quality defect classification into regions including two or more blocks in a specific direction of the product (product longitudinal direction, product width direction) (step S12). Note that in step S12, the operation data and equipment maintenance data are extracted using, for example, information on the quality defect classification created in step S2 of FIG. 10.

[0084] Next, in the quality defect determination step, the quality defect determination unit 35 determines whether or not a quality defect has occurred using multiple prediction models (step S13). Next, in the factor estimation step, the factor estimation unit 36 ​​estimates the factor of the quality defect based on the quality defect determination result (step S14). Next, in the result display step, the result display unit 37 displays the processing results of the quality defect determination step and the factor estimation step on the output unit 40 (step S15).

[0085] The quality defect diagnosis device and quality defect diagnosis method according to the above-described embodiment classify the parts of a product where quality defects will occur, and diagnose quality defects using a prediction model that incorporates the classification results. That is, the quality defect diagnosis device and quality defect diagnosis method according to the embodiment create and use a prediction model that represents the relationship between operation data and equipment maintenance data and quality defects for each classification based on the tendency of quality defects to occur. This makes it possible to predict product quality defects with high accuracy and estimate the causes of the quality defects.

[0086] The quality defect diagnosis device and quality defect diagnosis method according to the present invention have been specifically described above using the preferred embodiment and examples, but the scope of the present invention is not limited to these descriptions and should be broadly interpreted based on the claims. It goes without saying that various changes and modifications based on these descriptions are also included in the scope of the present invention.

[0087] For example, in the above-described embodiment, a diagnosis of quality defects was made using a plurality of predictive models that had been trained on the relationship between operational data and equipment maintenance data, and quality data, but a diagnosis may also be made using a plurality of predictive models that had been trained on the relationship between equipment maintenance data and quality data.

[0088] In this case, the model creation unit 34 uses machine learning to learn the relationship between the aggregated equipment maintenance data and the quality data for each quality defect classification, thereby creating multiple prediction models corresponding to the quality defect classification. The quality defect determination unit 35 determines whether or not a quality defect has occurred using multiple prediction models created for each quality defect classification and that have learned the relationship between the aggregated equipment maintenance data and the quality data. The factor estimation unit 36 ​​estimates the factor of the quality defect based on the quality defect determination result. In this way, even when only equipment maintenance data and quality data are used as input data for the prediction model, it is possible to accurately predict product quality defects and estimate the factor of the quality defect.

[0089] REFERENCE SIGNS LIST 1 Information processing device 10 Input unit 20 Storage unit 21 Product DB 30 Calculation unit 31 Data division unit 32 Quality defect classification unit 33 Data aggregation unit 34 Model creation unit 35 Quality defect determination unit 36 ​​Factor estimation unit 37 Result display unit 40 Output unit

Claims

1. A quality defect diagnosis device comprising: a data division unit that divides product data to be diagnosed, which consists of operation data, equipment maintenance data, and quality data, into a two-dimensional shape for each predetermined block; a data aggregation unit that extracts the operation data and the equipment maintenance data for each predetermined classification of quality defects based on the occurrence tendency of quality defects in the divided blocks, and aggregates the extracted operation data and equipment maintenance data for each region including two or more blocks in a specific direction of the product; and a quality defect determination unit that determines the presence or absence of the occurrence of quality defects using a plurality of prediction models created for each classification of quality defects and in which the relationship between the aggregated operation data and equipment maintenance data and the quality data has been learned.

2. The quality defect diagnosis device according to claim 1, further comprising a cause estimation unit that estimates the cause of the quality defect based on the determination result of the quality defect.

3. The quality defect diagnosis device according to claim 2, further comprising a result display unit that causes at least one of the determination result in the quality defect determination unit and the cause estimation result in the cause estimation unit to be displayed on an output unit.

4. The prediction model divides the product data of past performance, which consists of the operation data, the equipment maintenance data, and the quality data, into a two-dimensional shape for each predetermined block, classifies the quality defects included in the quality data based on the occurrence tendency of the quality defects in the divided blocks, aggregates the operation data and the equipment maintenance data extracted for each classification of the quality defects for each region including two or more blocks in a specific direction of the product, and is created by learning the relationship between the aggregated operation data and equipment maintenance data and the quality data for each classification of the quality defects by machine learning. The quality defect diagnosis device according to claim 1.

5. The quality defect diagnosis device according to claim 1, wherein the operation data is data that varies over the length in a specific direction of the product.

6. The quality defect diagnosis device according to claim 1, wherein the equipment maintenance data is data that varies during the period when the product is processed in the equipment.

7. The quality defect diagnosis device according to claim 1, wherein the quality data is data that varies over the length in a specific direction of the product.

8. The quality defect diagnosis apparatus according to claim 2, wherein the equipment maintenance data is data indicating the state of the equipment and its maintenance state that affects a specific position of the product by a direct or indirect action on the product.

9. The quality defect diagnosis apparatus according to claim 1, wherein the data aggregation unit extracts the operation data and the equipment maintenance data for each predetermined classification of quality defects based on a quality defect occurrence tendency having a plurality of quality defect occurrence locations over the longitudinal direction of the product, or a quality defect occurrence tendency having a plurality of quality defect occurrence locations over the width direction of the product.

10. The quality defect diagnosis apparatus according to claim 8, wherein the quality defect determination unit determines the presence or absence of the occurrence of a quality defect using a plurality of prediction models created for each classification of the quality defect and in which the relationship between the aggregated equipment maintenance data and the quality data has been learned, and the cause estimation unit estimates the cause of the quality defect based on the determination result of the quality defect.

11. The quality defect diagnosis apparatus according to claim 3, wherein the result display unit causes the output unit to display the data group subjected to alignment processing by the data aggregation unit.

12. A quality defect diagnosis method including: a data division step in which data division means provided in a computer divides product data to be diagnosed, which consists of operation data, equipment maintenance data, and quality data, into a two-dimensional shape for each predetermined block; a data aggregation step in which data aggregation provided in the computer extracts the operation data and the equipment maintenance data for each predetermined classification of quality defects based on the quality defect occurrence tendency in the divided blocks, and aggregates the extracted operation data and equipment maintenance data for each region including two or more blocks in a specific direction of the product; and a quality defect determination step in which quality defect determination means provided in the computer determines the presence or absence of the occurrence of a quality defect using a plurality of prediction models created for each classification of the quality defect and in which the relationship between the aggregated operation data and equipment maintenance data and the quality data has been learned.

13. The quality defect diagnosis method according to claim 12, including a cause estimation step in which cause estimation means provided in the computer estimates the cause of the quality defect based on the determination result of the quality defect.

Images