Patent Application
20230244210
This application is based upon and claims the benefit of priority from the Japanese Patent Application No. 2022-015113, filed Feb. 2, 2022, the entire contents of which are incorporated herein by reference.
Embodiments described herein relate generally to a data processing apparatus, a method, and a storage medium.
In the manufacturing industry, it is important to identify the cause of a specific state of a product at an early stage. For example, when a product is in an abnormal state different from a normal state, identifying the cause of the abnormal state at an early stage leads to maintenance and improvement of yield. In many manufacturing industries, various kinds of data acquired during the manufacturing process are monitored, and this helps detect abnormality and identify the cause.
The contents of the data are various. For example, data on conditions for manufacturing a product includes contents such as the name of a material used for manufacturing the product and the name of an apparatus used for manufacturing the product. In addition, data on the state of the product includes contents such as the size, physical characteristics, and appearance quality of the manufactured product. Usually, the data is often associated with an ID, a serial number, or the like, which is information that can identify each individual of the product.
By monitoring each item of individual data of the product, it may be possible to detect an abnormality of the product or the apparatus. For example, when a value of individual data of a product included in a product group manufactured in a certain period is different from a normal value, there is a possibility that an abnormality has occurred in the product. In this case, manufacturing data including data of conditions for the manufacturing is reviewed in order to search for the cause of the abnormality of the product. For example, when the fact that the abnormal product was manufactured only by a specific apparatus is identified from the manufacturing data, the apparatus may be the cause of the abnormality.
The more steps required to complete the product and the more apparatuses used to complete the product, the more data required to be monitored. Furthermore, with the recent development of Internet of Things (IoT) technologies, various data related to manufacturing can be easily acquired. Therefore, the number of items of manufacturing data has significantly increased. As a result, it is difficult to monitor manufacturing data manually. Under such circumstances, there is a demand for an apparatus that supports monitoring of manufacturing data by a user.
In general, according to one embodiment, a data processing apparatus includes a processor including hardware. The processor acquires state data related to a state of each product and order data related to order related to manufacturing of each product. The processor calculates a first score by using the acquired state data and the acquired order data. The first score regards a change in the acquired state data and is based on a first change model. The first change model represents the change in the state data using the order data. The processor outputs the first score.
A first embodiment will be described. A data processing apparatus according to the first embodiment estimates a cause of a certain state of a product by using information of order related to the product, and visualizes the cause based on a result of the estimation. As a result, the data processing apparatus assists the user in finding the cause of the certain state of the product.
The acquisition unit 101 acquires manufacturing data from the manufacturing database 104. The manufacturing data includes key data for identifying each product, state data related to a state of each product, and order data related to order related to the manufacturing of each product. The acquisition unit 101 inputs the manufacturing data to the first calculation unit 102.
The first calculation unit 102 calculates a first score regarding a change in the state data using the state data and the order data. The first score is a score indicating an extent to which a first change model fits the actual state data and the order data. The first change model is a model representing a change in the state data using the order data.
The output unit 103 generates visualized data from the first score or information describing the first score. Then, the output unit 103 outputs the visualized data to the display unit 105. The visualized data is data for visually presenting the user with a possible cause of a change in the state data.
The manufacturing database 104 stores the manufacturing data. The manufacturing database 104 may be configured as a general relational database management system (RDBMS). The manufacturing database 104 may be, for example, a Not only SQL (NoSQL) database. In addition, the manufacturing data stored in the manufacturing database 104 may be configured as a file in a predetermined format such as CSV (Comma Separated Value).
The display unit 105 displays various types of information based on the visualized data. The display unit 105 can be constituted by any of various displays such as a liquid crystal display and an organic EL display.
Next, an operation of the data processing apparatus 1 according to the first embodiment will be described. First, an operation of the acquisition unit 101 will be described. The acquisition unit 101 acquires manufacturing data to be used for calculating the first score from the manufacturing database 104. The manufacturing data will be described below in detail.
In the example of
In the embodiment, the acquisition unit 101 acquires at least two or more pieces of manufacturing data satisfying conditions, that is, two or more rows of manufacturing data. The acquisition unit 101 inputs the manufacturing data to the first calculation unit 102. The condition for acquiring the manufacturing data is, for example, whether or not an ID included in the manufacturing data is included in an ID list of the product as a target for identifying the cause. Under such an acquisition condition, the acquisition unit 101 may acquire manufacturing data having the ID included in the list. The number of pieces of manufacturing data acquired by the acquisition unit 101, that is, the number of rows is represented as D.
The condition for acquiring the manufacturing data may be designated by the user. For example, the condition may be designated to acquire manufacturing data for a product group manufactured in any period of one hour, one day, or the like. The state data and the order data which are elements of the manufacturing data will be further described below.
First, the order data will be described. For the following description, the order data is represented by a set {Ok: k=1 . . . L}. Here, L represents the number of items of the acquired order data. The product to be analyzed in the present embodiment is manufactured according to specific L types of order. The order data is data related to the order related to the manufacturing of the product. The order data may include, for example, data of the order in which processes are started in manufacturing of the product, that is, the order in which tasks for the product are started. As described above, when the order data is data indicating the order in which the processes are started, the value of L in the order data is equal to the number of processes.
The order data may not be data in units of processes. For example, the order data may be data in units of manufacturing lines, data in units of manufacturing facilities, data in units of manufacturing apparatuses, data in units of lots of products, data in units of batches, and data in units of manufacturing. In addition, the order data may be data in units of manufacturing bases or the like. For example, the order data in a case where the product reaches an inspection site from a factory A through a factory B and a factory C may be recorded in a time zone or the like in which the product is at each factory and the inspection site.
Furthermore, the order in the order data may be represented by, for example, serial numbers from 1 or may be represented by time. For example, the order data may be data of the start time of each of the processes in manufacturing the product or the time when a task for the product is started in each of the processes. In
In addition, the order data may be, for example, instead of the start time, time of passage, time of completion, a number representing a certain period or a certain number of product units, a character string, a serial number, a product serial number, an ID, the number of times of processing of an apparatus, and other various data from which the order related to the manufacturing of the product can be grasped.
In addition, in the embodiment, for example, a situation is assumed in which the product is automatically or manually moved by a belt conveyor and a carriage, or the like, and processed or inspected by predetermined equipment or the like. In addition, for example, even in a situation where the product is fixed and is processed or inspected by equipment or the like different for each process, the order can be specified, and thus the techniques described in the first embodiment can be applied.
Further, the product with which the order data is associated may not be a product that undergoes multiple processes. For example, even if the processes are not clearly defined, in a case where the product undergoes processes by a plurality of apparatuses or in a case where a plurality of tasks are performed on the product, the order data may indicate the order in which the product reaches the apparatuses or the order in which the product reaches work places.
Furthermore, in a case where the order data is recorded based on time, the order data may include data of products in the same order due to the effect of the accuracy of recording the time or the like. However, it is desirable that the order data be recorded so that the order is unique.
Next, the state data will be described. For the following description, the state data is represented by a set {Yi: i=1 . . . N}. Here, N represents the number of items of the acquired state data.
The state data Yi represents, for example, the state of each product. The state of each product is represented, for example, as a measured value obtained by performing an inspection on the product. The state data Yi includes, for example, data of measured values related to inspection items, such as a dimension and weight of the product. In addition, depending on the type of product, the state data Yi may include data of measured values such as electrical characteristics and physical characteristics of the product. As described above, the state data Yi can be, for example, a measured value related to an inspection item at the time of shipment of the product.
The state data Yi may be any data that is not the measured value related to the inspection item and from which the state of the product can be grasped. For example, the state data Yi may be data of a result of some determination. For example, the state data Yi may be an integer value representing a 5-point scale on the quality of the product, a binary flag representing a result of determining whether or not the product is good, or the like. The acquisition unit 101 may acquire any state data Yi determined by the user to be useful for analysis.
As illustrated in
Next, an operation of the first calculation unit 102 will be described. When receiving the state data Yi and order data Ok from the acquisition unit 101, the first calculation unit 102 calculates a first score S1(Yi, Ok).
First, the first calculation unit 102 initializes k and i to initial values, for example, 1. Then, in step S1, the first calculation unit 102 applies a first change model ƒ(Yi)=aOk to one set of the state data Yi and the order data Ok. The first change model is a model representing a change in the state data Yi using the order data Ok. In the embodiment, the first change model is, for example, a regression model using the order data Ok as a step basis.
In step S2, the first calculation unit 102 calculates a first score S1 (Yi, Ok) regarding a change in the state data Yi. The first score S1 (Yi, Ok) is the goodness of fit of the first change model to the state data Yi and the order data Ok. For example, as the first score, an existing evaluation index such as a residual sum of squares (RSS) or a log likelihood can be used. As other examples of the first score, a correlation coefficient, a regression coefficient, a mean square error (MSE), a root mean square error (RMSE), a mean absolute error (MAE), a determination coefficient (R2), indexes derived from these evaluation indexes, and the like can be used. Furthermore, the first score may be an evaluation index calculated from a plurality of indexes among these indexes.
In the embodiment, as a more preferable example, a score indicating a sum of the goodness of fit to the data and a penalty for complexity of the model may be used as the first score. The change in the state data Yi includes various types of noise caused by various events different from original causes, such as a measurement error and replacement of processes in addition to a change caused by the original cause of a specific state of the product. In a case where the effect of noise is large, if only the goodness of fit is the score, the score of the model that excessively fits the change including the noise becomes high. In this case, it is difficult to grasp the change in the state data Yi due to the cause. In the embodiment, the Bayesian Information Criterion (BIC), known as a score that represents the sum of the goodness of fit to the data and the penalty for the complexity of the model, is used as the first score. The BIC is calculated according to Equation (1). In Equation (1), In(L) is the maximum log likelihood, n is the number of samples, and k is the degree of freedom of data.
BIC=−2·ln(L)+k ln(n) (1)
In addition, under the Gaussian error model, Equation (1) can be rewritten into Equation (2). In Equation (2), n is the number of samples, Yitrue is a measured value of the state data Yi, and Yipred is a predicted value of the state data Yi calculated from the first change model.
In Equation (1) or (2), in a case where the value of the BIC is small, the value of the BIC means that the goodness of fit of the model to the data is high. In a case where the value of the BIC is small, it can be considered that the order data Ok is highly likely to be related to the cause of the change in the state data Yi. By scoring the relationship between the order data Ok and the state data Yi in this manner, it is possible to evaluate, as a numerical value, an extent to which the first change model ƒ(Yi)=aOk can explain the change in the state data Yi.
Here, the description returns to
In step S4, the first calculation unit 102 determines whether i is larger than N. In a case where it is determined in step S4 that i is not larger than N, the first calculation unit 102 adds 1 to i and then returns the processing to step S1. In a case where it is determined in step S4 that i is larger than N, that is, in a case where it is determined that the processing for all the order data Ok for all the state data Yi has been completed, the first calculation unit 102 ends the processing illustrated in
In this manner, the first calculation unit 102 applies the first change model to the state data Yi to calculate the first score S1(Yi, Ok). As a result, the change in the state data Yi can be grasped as a numerical value, and the cause of the change in the state data Yi can be identified.
In the embodiment, it is assumed that a change that is the cause occurs once or several times in a certain period, so that an index with a penalty for complexity, such as the BIC, is suitable. On the other hand, the first score is not necessarily limited to the BIC. For example, as the first score, the Akaike information criterion (AIC), the deviation degree information criterion (DIC), the generalized information criterion (GIC), the widely applicable Akaike information criterion (WAIC), and the widely applicable Bayesian information criterion (WBIC) may be used, or derived indexes thereof may be used.
In the embodiment, the step basis is generated from the order data. On the other hand, a basis other than the step basis can be used as long as a change in the state data Yi can be captured. For example, a sigmoid basis, a ramp basis, or the like may be used, or a Fourier basis, a wavelet basis, or the like may be used.
Furthermore, in the above-described example, the first change model is a regression model. On the other hand, other models can be used as long as a change in the state data Yi can be captured. For example, in a case where the moving average model is used as the first change model, noise that is estimated to have a low relationship with the cause of the change in the state data Yi can be removed. Such a first change model can handle a change that is more likely to be the cause. In a case where the moving average model is used as the first change model, not the BIC or the like but a threshold for detecting a change, the number of changes, and the like can be used as the first score. Furthermore, the first score may be directly calculated by a machine learning method such as a support vector machine, deep learning, or the like.
Next, an operation of the output unit 103 will be described. After the first score is calculated by the first calculation unit 102, the output unit 103 generates visualized data from the first score or information describing the first score.
First, an operation of the output unit 103 in a case where the output unit 103 receives the first score from the first calculation unit 102 will be described. The same number of first scores S1(Yi, Ok) as the number of combinations of the state data Yi and the order data Ok may be present. The output unit 103 may sort the first scores S1(Yi, Ok), for example, in the order in which the order data Ok has been acquired. Then, the output unit 103 may generate visualized data based on the sorted data, and output the generated visualized data as an analysis result to the display unit 105. In addition, the output unit 103 may sort the order data Ok in the order of the first scores S1(Yi, Ok). Then, the output unit 103 may generate visualized data based on the sorted data, and output the generated visualized data as an analysis result to the display unit 105. In this case, the visualized data can be generated such that the first scores S1(Yi, Ok) and the order data Ok are displayed in a ranking format. Therefore, the user can check the results in order of importance by checking the order data Ok in the ranking order.
Next, an operation of the output unit 103 in a case where the output unit 103 receives, from the first calculation unit 102, a result of applying the first change model as information describing the first change model will be described. When receiving the result of applying the first change model, the output unit 103 generates, from the result of applying the first change model, visualized data in a format that is, for example, a diagram and enables the user to visually recognize the data. By using the result of applying the first change model, the user can check a change point in the distribution of the state data Yi as visual information, instead of a numerical value such as the first score. This allows the user to more intuitively understand the result.
When receiving both the first score and the result of applying the first change model, the output unit 103 may combine the first score and the result of applying the first change model to generate visualized data that can be displayed. As a result, the user can search for the cause of the change in the state data Yi from numerical values and the scatter diagram.
Here, the format of the visualized data will be described. The visualized data can be generated as, for example, image data or diagram data. In addition, the visualized data may be data in a format displayable on the display unit 105, for example, the Hypertext Markup Language (html), the Extensible Markup Language (xml), or JavaScript (registered trademark) Object Notation (JSON).
The creation date display field 301 is a display field for displaying, for example, a creation date of manufacturing data used for analysis as information to be analyzed.
The state data name display field 302 is a display field for displaying the name of the state data Yi to be analyzed.
The second display regions 303, 304, and 305 are display regions for displaying information on a combination of the state data Yi and the order data Ok displayed in the corresponding first display region 300. A maximum of L second display regions can be provided for one first display region.
The second display regions 303, 304, and 305 include order data name display fields 3031, 3041, and 3051, respectively. In addition, each of the second display regions 303, 304, and 305 includes at least one of a display region for a first score and a display region for a diagram representing a result of applying the first change model. For example, the second display region 303 includes both a first score display region 3032 and a display region 3033 for a diagram representing the result of applying the first change model. On the other hand, the second display regions 304 and 305 include only first score display regions 3042 and 3052, respectively.
The amount of information displayed in each of the second display regions 303, 304, and 305 may be determined based on, for example, the value of the first score. For example, when the value of the first score is smaller than a predetermined threshold Th1, both the first score and the diagram representing the result of applying the first change model are displayed. In addition, when the value of the first score is larger than the predetermined threshold Th1 and smaller than a threshold Th2 larger than the threshold Th1, one of the first score and the diagram representing the result of applying the first change model, for example, the value of the first score is displayed. In addition, when the value of the first score is larger than the predetermined threshold Th2, for example, as indicated in the first score display region 3052 of
In addition, display priorities of the second display regions 303, 304, and 305 may also be determined according to the values of the first scores. For example, the priorities may be determined such that the second display regions are arranged from the top of the display screen in ascending order of the first scores.
In addition, when the value of the first score is smaller than the threshold Th1, the corresponding second display region may be highlighted. The highlighting can be performed by, for example, coloring the corresponding second display region, making bold the name of the order data displayed in the corresponding second display region, or adding a warning mark or the like to the corresponding second display region. On the other hand, when the value of the first score is larger than the threshold Th2, the corresponding second display region may be made unnoticeable. The second display region may be made unnoticeable by, for example, lightening a color of the second display region.
As described above, according to the first embodiment, the order data Ok related to the change in the state data Yi is estimated based on the first score calculated using the state data Yi related to the state of the product and the order data Ok related to the order related to the manufacturing of the product. That is, in the first embodiment, a process or the like related to the change in the state data Yi can be estimated without using information directly indicating the cause of the change in the state data Yi, such as information indicating how manufacturing conditions have been changed. Therefore, since the user does not need to monitor the state data one by one, it is expected that the burden on the user is reduced.
In the first embodiment, the priorities and amounts of information when the analysis result is displayed are determined based on the values of the first scores. Therefore, the user can preferentially check the order related to the manufacturing and expected to have a high relationship with the change in the state data Yi, for example, a specific passage place during the manufacturing. Furthermore, since the number of places required to be checked is reduced, it is expected that the burden of checking on the user is reduced. Furthermore, it is expected that the user's oversight is reduced.
In addition, each of the first scores in the first embodiment indicates the likelihood of the cause of the change in the state data Yi based on the presence or absence of the change in the state data Yi caused by the order data Ok and the number of changes in the state data Yi caused by the order data Ok. For example, in a case where the product is in a specific state such as an abnormal state due to an event such as different conditions before and after maintenance, different lots of members, or a failure of an apparatus that has been normally operated, the state data Yi indicating the state of the product often changes when the specific event occurs in specific order. That is, in such a case, the state data Yi can greatly change at a certain point of time as a boundary. When the specific event is the cause of the change in the state data Yi, the number of changes in the state data Yi in a process in which the specific event has occurred or the like is considered to be small. On the other hand, when a measurement error is included in the state data Yi or the order of products is changed between processes, the state data Yi may change complicatedly. Therefore, it is considered that the number of changes in the state data Yi is large. Since the first score is a score that is based on the presence or absence of the change in the state data Yi caused by the order data Ok and the number of changes in the state data Yi caused by the order data Ok, and indicates the likelihood of the cause of the change in the state data Yi, the first change model that fits the state data Yi considered to be in a specific state such as an abnormal state can be determined. It can be estimated that the order or time that is indicated in the order data Ok from which the first change model is generated and that causes a change in the state data Yi caused the state of the product to be a specific state such as an abnormal state. In this way, according to the first embodiment, the order data Ok related to the change in the state data Yi can be estimated without using information of a change in the manufacturing conditions.
Next, a second embodiment will be described. Similarly to the first embodiment, a data processing apparatus according to the second embodiment also estimates a cause of a certain state of a product by using order information, and visualizes the cause based on a result of the estimation. As a result, the data processing apparatus assists the user in finding the cause of the certain state of the product. In the first embodiment, the first calculation unit 102 calculates a first score regarding a change in the state data Yi based on the order data Ok. On the other hand, when manufacturing condition data, which is information indicating conditions for manufacturing the product, can be acquired from a manufacturing database 104, a second score regarding a change in state data can be calculated based on the manufacturing condition data in a similar manner to the first score. If the first score and the second score are present, the relationship between the order causing the change in the state data and the manufacturing conditions can be presented to the user by comparing the first score with the second score. As described above, the second embodiment is an example applied to a case where the manufacturing condition data can be acquired. Description of parts similar to those described in the first embodiment will be omitted or simplified, and only parts different from those described in the first embodiment will be described.
In the second embodiment, the acquisition unit 101 acquires manufacturing data. In the second embodiment, the manufacturing data is data to be used for calculating the first score and the second score, and includes key data for identifying each product, state data related to a state of each product, order data related to order related to the manufacturing of each product, and manufacturing condition data related to conditions for manufacturing of each product. The acquisition unit 101 inputs the key data, the state data, and the order data among the manufacturing data to the first calculation unit 102. In addition, the acquisition unit 101 inputs the key data, the state data, and the manufacturing condition data among the manufacturing data to the second calculation unit 106.
The second calculation unit 106 calculates a second score regarding a change in the state data based on a second change model using the state data and the manufacturing condition data. The second score is a score indicating an extent to which the second change model fits the actual state data and the manufacturing condition data. The second change model represents a change in the state data using the manufacturing condition data.
The output unit 103 generates visualized data from the first score or information describing the first score, and the second score or information describing the second score. Then, the output unit 103 outputs the visualized data to the display unit 105. As in the first embodiment, the visualized data is data for visually presenting a user with a possible cause of a large change in the manufacturing data.
Next, an operation of the data processing apparatus 1 according to the second embodiment will be described. First, an operation of the acquisition unit 101 will be described. In the second embodiment, the acquisition unit 101 acquires the manufacturing data including the manufacturing condition data from the manufacturing database 104. The manufacturing condition data will be described below in detail. For the following description, the manufacturing condition data is represented by a set {Cj: j=1 . . . M}. Here, M represents the number of items of the acquired manufacturing condition data.
In addition, as the manufacturing conditions recorded as the manufacturing condition data, information such as a name of a material used for the product and a name of an apparatus used for processing or assembling the product can be used. More generally, information on 5M1E can be used as the manufacturing conditions. 5M1E is a term in which acronyms of man, machine, material, method, measurement, and environment are arranged, and is widely known as six factors for managing a manufacturing process. Information of man includes information such as a name of a processing person. Information of machine includes information such as a name of an apparatus used for manufacturing the product, a name of a manufacturing line, and the state of the apparatus at the time of processing, such as a temperature and pressure in the apparatus. Information of material includes information such as an ID or a name of a material used for manufacturing the product and an ID or a name of a component constituting the product. Information of method includes information such as a method for processing the product and a type of a processing program. Information of measurement includes information such as a name of an apparatus where measurement is performed and a measurement position of the product where the measurement is performed. Information of environment includes information such as a building name of a factory where the measurement is performed and the temperature and humidity when the measurement is performed. The acquisition unit 101 acquires the manufacturing condition data from the manufacturing database 104 so as to include manufacturing conditions necessary for analysis. At this time, the acquisition unit 101 may acquire the manufacturing condition data so as to include manufacturing conditions determined by the user to be useful for analysis and visualization.
Next, an operation of the first calculation unit 102 will be described. When receiving state data Yi and order data Ok from the acquisition unit 101, the first calculation unit 102 calculates a first score S1(Yi, Ok). The first score is calculated in the same manner as in the first embodiment.
Next, an operation of the second calculation unit 106 will be described. When receiving the state data Yi and manufacturing condition data Cj from the acquisition unit 101, the second calculation unit 106 calculates a second score S2(Yi, Cj).
First, the second calculation unit 106 initializes i and j to initial values, for example, 1. Then, in step S101, the second calculation unit 106 applies a second change model ƒ(Yi)=aCj to one set of the state data Yi and the manufacturing condition data Cj. The second change model represents a change in the state data Yi using the manufacturing condition data Cj. The second change model may be a regression model similar to the first change model. The second change model may be, for example, a regression model using the manufacturing condition data Cj as a step basis.
Here, in a case where the manufacturing condition data Cj is data of a nominal scale such as an ID of a material or of a component and the name of the material or of the component, the second change model is applied after the manufacturing condition data Cj is converted into a dummy variable by a method such as one-hot encoding. The method for converting the manufacturing condition data Cj into the dummy variable is not limited to one-hot encoding. For example, as the method for converting the manufacturing condition data Cj into the dummy variable, a method such as label encoding, count encoding, target encoding, or leave one out encoding may be used. In addition, in a case where the manufacturing condition data Cj is data to which the regression model can be applied without being converted into a dummy variable such as a ratio scale, it is not necessary to use these conversion methods.
In step S102, the second calculation unit 106 calculates a second score S2(Yi, Cj) regarding a change in the state data Yi. The second score S2 (Yi, Cj) is the goodness of fit of the second change model to the state data Yi and the manufacturing condition data Cj. The second score is desirably a score that can be compared with the first score. For example, a BIC similar to the first score can be used as the second score. Of course, each score that can be used as the first score may be used as the second score. With the second score, an extent to which the second change model can explain the change in the state data Yi can be evaluated as a numerical value. For example, by displaying the ranking of first scores and the ranking of second scores in parallel, the user can evaluate which item of the order data Ok and the manufacturing condition data Cj is likely to be the cause of the change in the state data Yi.
In step S103, the second calculation unit 106 determines whether j is larger than M. In a case where it is determined in step S103 that j is not larger than M, the second calculation unit 106 adds 1 to j and then returns the processing to step S101. In a case where it is determined in step S103 that j is larger than M, that is, in a case where it is determined that the processing for all the manufacturing condition data Cj for the current state data Yi has been completed, the second calculation unit 106 causes the processing to proceed to step S104.
In step S104, the second calculation unit 106 determines whether i is larger than N. In a case where it is determined in step S104 that i is not larger than N, the second calculation unit 106 adds 1 to i and then returns the processing to step S101. In a case where it is determined in step S104 that i is larger than N, that is, in a case where it is determined that the processing for all the manufacturing condition data Cj for all the state data Yi has been completed, the second calculation unit 106 ends the processing illustrated in
Next, an operation of the output unit 103 will be described. After the first score is calculated by the first calculation unit 102 and the second score is calculated by the second calculation unit 106, the output unit 103 generates visualized data from the first score or the information describing the first score and the second score or the information describing the second score. The output unit 103 may further perform processing of comparing the first score with the second score. When the cause of the change in the state data Y±occurs at a certain timing and the order of products is changed, the first score is considered to be a score equivalent to the score of the true cause. Therefore, by comparing the first score with the second score, it can be determined to what extent the corresponding manufacturing condition data Cj is likely to be the cause of the change. For example, when the first score and the second score are almost the same, it is determined that the change in the corresponding manufacturing condition data Cj is likely to be the cause of the change in the state data Yi. On the other hand, when the first score and the second score are different, it cannot be said that the change in the corresponding manufacturing condition data Cj is likely to be the cause of the change in the state data Yi. In this case, it is suggested that there is another cause. Therefore, in such a case, the output unit 103 may generate the visualized data so as to notify the user of the possibility that there is another cause of the change.
The first score to be compared with the second score will be described. First, the degree of freedom in calculating the second score is determined based on the number −1 of conditions of the manufacturing condition data Cj. The number of conditions of the manufacturing condition data Cj is the number of elements included in the manufacturing condition data Cj. For example, when one of two processing conditions can be selected as a processing condition in a certain process, the number of conditions of the manufacturing condition data Cj is two, and the degree of freedom in calculating the second score is one. On the other hand, the degree of freedom in calculating the first score can take all values from 0 to L−1. Therefore, the degree of freedom of the first score when the first score and the second score are compared may be the degree of freedom at the time of the maximum likelihood score. For example, in a case where the BIC is used, the degree of freedom of the first score when the first score and the second score are compared may be the degree of freedom when the BIC is minimized. As another method, the degree of freedom of the order data Ok may be limited. For example, by adjusting the degree of freedom of the order data Ok to the number of elements of the manufacturing condition data. Cj, the comparison can be performed while suppressing excessive fit to the order data Ok. In the case of performing the calculation after designating the degree of freedom as described above, a greedy algorithm, an orthogonal matching pursuit (OMP) algorithm, lasso regression, or a derivative method thereof may be used. In these methods, it is possible to obtain a score by designating not only the degree of freedom of the maximum likelihood but also an arbitrary degree of freedom and performing regression calculation. Therefore, these techniques are suitable techniques.
Further, to compare the first score with the second score, a Δ score that is a difference value between the two scores may be used as a reference value of the comparison. For example, when the Δ score is 0, it indicates that the first score and the second score are exactly the same value. In addition, the larger the value of the Δ score, the larger the difference between the first score and the second score. In a case where the first score and the second score are compared on the assumption that the first score is equivalent to the true cause as in the present embodiment, the smaller the A score, the higher the possibility that the manufacturing condition data Cj corresponding to the second score at that time is likely to be the true cause, and the larger the Δ score, the higher the possibility that there is a different cause from the manufacturing condition data Cj corresponding to the second score at that time. That is, by using the Δ score, the reliability for the true cause of the manufacturing condition data Cj can be presented. The Δ score in a case where each of the first score and the second score is the BIC has a relationship shown in Formula (3) with the Bayesian factor BF01 shown in Jeffreys, H. 1961. Theory of probability. Oxford University Press.
Δ score˜21nBF01 (3)
Based on Formula (3), a value obtained by taking a logarithm of the reference value and doubling the logarithm may be treated as a reference value for comparison based on the Δ score. Of course, the user may set an appropriate reference value according to the situation. As a result, the comparison result can be evaluated according to the user's needs.
Furthermore, the diagram related to the manufacturing condition data Cj may be, for example, a scatter diagram corresponding to
An operation of the output unit 103 will be described.
The creation date display field 301 and the state data name display field 302 in the first display region 300 are similar to those illustrated in
The second display regions 303, 304, and 305 are display regions for displaying information on a combination of the state data Yi and the order data Ok displayed in the corresponding first display region 300. On the other hand, the second display regions 306 and 307 are display regions for displaying information on the combination of the state data Yi and the manufacturing condition data Cj displayed in the corresponding first display region 300. A maximum of L x M second display regions can be provided for one first display region.
The second display regions 303, 304, and 305 include order data name display fields 3031, 3041, and 3051, respectively. In addition, each of the second display regions 303, 304, and 305 includes at least one of a display region for a first score and a display region for a diagram representing a result of applying the first change model. For example, the second display region 303 includes both a first score display region 3032 and a display region 3033 for a diagram representing the result of applying the first change model. On the other hand, the second display regions 304 and 305 do not include the display region for the diagram representing the result of applying the first change model, but include first score display regions 3042 and 3052, respectively.
The second display regions 306 and 307 include manufacturing condition data name display fields 3061 and 3071, respectively. Each of the second display regions 306 and 307 includes at least one of a display region for a second score and a display region for a scatter diagram representing a relationship between the manufacturing condition data and the state data. For example, the second display region 306 includes both a second score display region 3062 and a display region 3063 for a scatter diagram representing the relationship between the manufacturing condition data and the state data. On the other hand, the second display region 307 does not include the display region for the scatter diagram representing the relationship between the manufacturing condition data and the state data, but includes a second score display region 3072. The scatter diagram representing the relationship between the manufacturing condition data and the state data can be generated, for example, by plotting the state data Yi corresponding to each piece of the manufacturing condition data Cj. In
The amount of information displayed in each of the second display regions may be determined based on, for example, the value of the first score and the value of the second score. For example, when the value of the first score is smaller than a predetermined threshold Th1, both the first score and the diagram representing the result of applying the first change model are displayed. In addition, when the value of the first score is larger than the predetermined threshold Th1 and smaller than a threshold Th2 larger than the threshold Th1, one of the first score and the diagram representing the result of applying the first change model, for example, the value of the first score is displayed. For example, when the value of the second score is smaller than a predetermined threshold Th3, both the second score and the diagram representing the result of applying the second change model are displayed. In addition, when the value of the second score is larger than the predetermined threshold Th3 and smaller than a threshold Th4 larger than the threshold Th3, one of the second score and the diagram representing the result of applying the second change model, for example, the value of the second score is displayed. In addition, when the value of the second score is larger than the predetermined threshold Th4, for example, it may be displayed that the value of the score is larger than the threshold, similarly to the first score. In this case, the corresponding second display region itself may be hidden. Note that the threshold Th1 and the threshold Th3 may be the same value. Note that the threshold Th2 and the threshold Th4 may be the same value.
When manufacturing condition data Cj indicating the second score equivalent to the first score is present, the second display region related to the manufacturing condition data Cj may be highlighted. The second display region can be highlighted by, for example, changing a color of a frame, making bold the name of a manufacturing condition, displaying a specific mark, or the like.
In addition, each of display priorities of the second display regions 303, 304, 305, 306, and 307 may also be determined according to the values of the first scores and the values of the second scores. In this case, in a case where the first scores and the second scores can be compared, the first scores and the second scores are compared together, and the priorities may be determined such that the second display regions are arranged from the top of the display screen in ascending order of the score values. On the other hand, when the first scores and the second scores cannot be compared, the priorities may be determined such that the first scores and the second scores are separately compared, the second display regions based on the order data are arranged in ascending order of the values of the first scores, and the second display regions based on the manufacturing condition data are arranged in ascending order of the values of the second scores.
Instead of the scatter diagram representing the relationship between the manufacturing condition data and the state data illustrated in
With the scatter diagram in which the order data Ok, the manufacturing condition data Cj, and the state data Yi are combined, a change in the state data Yi accompanying a change in the manufacturing condition data Cj with reference to the order data Ok can be visually distinguished. Therefore, the user can easily evaluate the manufacturing condition data Cj based on the order data Ok.
In a case where the degree of freedom in calculating the first score matches the number of elements of the manufacturing condition data Cj when the second score is calculated, the display screen including the second display regions 303, 304, and 305 as illustrated in
The second display regions 303, 304, and 305 include manufacturing condition data name display fields 3037, 3043, and 3053, respectively. In addition, each of the second display regions 303, 304, and 305 includes at least one of a display region for the Δ score and the display region for the scatter diagram in which the order data Ok, the manufacturing condition data Cj, and the state data Yi are combined. For example, the second display region 303 includes a display region 3038 for the score and a display region 3039 for the scatter diagram in which the order data Ok, the manufacturing condition data Cj, and the state data Yi are combined. On the other hand, the second display regions 304 and 305 include only the display regions 3044 and 3054 for Δ scores, respectively. Each of the second display regions may include an order data name display field instead of the manufacturing condition data name display fields 3037, 3043, and 3053.
The amount of information displayed in each of the second display regions 303, 304, and 305 may be determined based on, for example, the value of the Δ score. For example, when the value of the Δ score is within a threshold range R1 that is a range close to 0, both the Δ score and the scatter diagram in which the order data Ok, the manufacturing condition data Ci, and the state data Yi are combined are displayed. When the value of the Δ score is out of the threshold range R1 and is within a threshold range R2 wider than the threshold range R1, any one of the Δ score and the scatter diagram in which the order data Ok, the manufacturing condition data Cj, and the state data Yi are combined, for example, the value of the Δ score is displayed. Similarly, when the value of the Δ score is out of the threshold range R2, it may be displayed that the value of the Δ score is out of the threshold range. In this case, the corresponding second display region itself may be hidden.
In addition, display priorities of the second display regions 303, 304, and 305 may also be determined according to the values of the Δ scores. For example, the priorities may be determined such that the second display regions are arranged from the top of the display screen in ascending order of the Δ scores.
When the value of the Δ score is within the threshold range R1, the corresponding second display region may be highlighted. The highlighting can be performed by, for example, coloring the corresponding second display region, making bold the name of the manufacturing condition data or the order data displayed in the corresponding second display region, or adding a warning mark or the like to the corresponding second display region. Conversely, when the value of the Δ score is out of the threshold range R2, the corresponding second display region may be made unnoticeable. The second display region may be made unnoticeable by, for example, lightening a color of the second display region.
As described above, according to the second embodiment, in addition to the first embodiment, the manufacturing condition data Cj related to the change in the state data Yi is estimated based on the second score calculated using the state data Yi related to the state of the product and the manufacturing condition data Cj related to the conditions for the manufacturing of the product. The user can evaluate whether or not a change in the corresponding manufacturing condition data Cj is likely to be a cause of a change in the state data Yi by comparing the first score with the second score.
In addition, in the second embodiment, the priority and the amount of information when the analysis result is displayed are determined based on both the first score and the second score. Therefore, the user can preferentially check the order related to the manufacturing and the manufacturing conditions that are expected to have a high relationship with the change in the state data Yi.
Next, a third embodiment will be described. Similarly to the first embodiment and the second embodiment, a data processing apparatus according to the third embodiment also estimates a cause of a certain state of a product by using order information, and visualizes the cause based on a result of the estimation. As a result, the data processing apparatus assists the user in finding the cause of the certain state of the product.
In
For example, the second display region 303 illustrated in
On the other hand, the second display region 304 illustrated in
The information amount change buttons 401, 402, 403, and 404 desirably include information related to changes in amounts of information of the second display regions. The information related to the changes in the amounts of the information of the second display regions is, for example, “+” and “−”. Specifically, “+” may be displayed on the information amount change buttons 401, 402, 403, and 404 when the amounts of information of the second display regions are increasing, and “−” may be displayed on the information amount change buttons 401, 402, 403, and 404 when the amounts of information of the corresponding second display regions are decreasing.
In addition, the first embodiment and the second embodiment describe the case where the second display region in which the first score or the second score is larger than the corresponding threshold is hidden. In the third embodiment, the first display region 300 includes a switch button 405 for switching between display and non-display of a second display region in which a first score or a second score is larger than a threshold. For example, while the display of the switch button 405 is “show more”, the second display region in which the first score or the second score is larger than the threshold is hidden in the folded state. When the switch button 405 is selected in this state, the second display region in the folded state is displayed. Then, the display of the switch button 405 is switched to “hide”. When the switch button 405 is selected in this state, the second display region in which the first score or the second score is larger than the threshold is hidden again in the folded state. All second display regions in the folded state may be displayed, or a predetermined number of second display regions in the folded state may be displayed. The order in which the second display regions are displayed may be determined by the first scores and/or the second scores. In a case where the predetermined number of second display regions are displayed, only a display region with the first score and/or the second score higher than those in the other display regions may be displayed.
As described above, the same number of first display regions as the number of pieces of the state data Yi to be analyzed may be present. In a case where a plurality of pieces of state data Yi to be analyzed are present, an information amount change button and a switch button may be provided for each first display region.
In
In
As illustrated in
In addition, the first display region 300 may include an interface for changing the thresholds for the first scores and the second scores.
As described above, in the third embodiment, for the state data Yi, the order data Ok, and the manufacturing condition data Cj related to the state of the product, the amount of information and the priority to be displayed in the second display region are determined based on the first score or the second score regarding the change in the state data Yi. In addition, the user can change the amount of information and the display priority of the second display region via the operation unit 107. In addition, the user can change the amount of information and the display priority of the first display region via the operation unit 107. As a result, the user can not only preferentially monitor information expected to have a high relationship with an abnormality, but also check the analysis result in a more diversified manner. For example, in a daily monitoring task, the user can properly use the display region to check an item expected to have a high relationship with an abnormality displayed first, and monitor an item having a relatively low relationship with an abnormality in a scene where detailed monitoring is required.
Next, a fourth embodiment will be described. Similarly to the first to third embodiments, a data processing apparatus according to the fourth embodiment also estimates a cause of a certain state of a product by using order information, and visualizes the cause based on a result of the estimation. As a result, the data processing apparatus assists the user in finding the cause of the certain state of the product.
The manufacturing process database 108 records the manufacturing process data indicating a process of manufacturing the product. Usually, a subdivided task until the final commercialization is completed is called a manufacturing process. Manufacturing data is usually recorded for each process. The manufacturing process data in the present embodiment is data indicating which manufacturing process each item of the manufacturing data acquired by the acquisition unit 101 belongs to.
In
The first display regions are rearranged according to priorities designated by the priority change button 702. For example, in
In
The first display region 300 may include an interface for changing the thresholds for the first score and the second score for each process. Such an interface is effective when a process in which an abnormality easily occurs and a process in which an abnormality hardly occurs are known.
The user database 109 records user data of a user who performs analysis. As the user in the embodiment, for example, a worker of each manufacturing process, a manager of each manufacturing process, a manager of the entire manufacturing, or the like is assumed. The user data is data indicating a priority at the time of display for each attribute of the user.
The user data may include numerical data indicating the level of the user. For example, an analysis result related to some manufacturing data may be displayed for a user with a low numerical value, and an analysis result related to a large number of manufacturing data may be displayed for a user with a high numerical value. For example, the numerical value of the level of an inexperienced user may be set low, and the numerical value of the level of an expert or an administrator may be set high. In
As described above, according to the fourth embodiment, the visualized data is generated based on not only the score but also the manufacturing process data. As a result, it is possible to enable display that is easy for the user to monitor, such as displaying analysis results in the order of the manufacturing processes. Furthermore, by using not only the score but also the user data, it is possible to enable display according to a target to be monitored by the user who uses the user data and/or the skill level of the user. As a result, the burden on the user can be reduced.
In the fourth embodiment, the manufacturing process database 108 and the user database 109 are included in the data processing apparatus 1. However, the manufacturing process database 108 and the user database 109 may be provided separately from the data processing apparatus 1.
Next, a fifth embodiment will be described. Similarly to the first to fourth embodiments, a data processing apparatus according to the fifth embodiment also estimates a cause of a certain state of a product by using order information, and visualizes the cause based on a result of the estimation. As a result, the data processing apparatus assists the user in finding the cause of the certain state of the product.
In the fifth embodiment, the first calculation unit 102 records a first score or information describing the first score in the analysis database 110. Similarly, the second calculation unit 106 records a second score or information describing the second score in the analysis database 110.
In addition, in the fifth embodiment, the acquisition unit 101 records data related to acquired manufacturing data in the analysis database 110. For example, the acquisition unit 101 records data of conditions for acquiring a manufacturing condition such as a time period for acquiring the manufacturing data.
In addition, in the fifth embodiment, the output unit 103 generates visualized data using the data recorded in the analysis database 110. For example, the output unit 103 may generate a graph representing the transition of the first score and the second score in addition to the diagram representing the result of applying the first change model described above and the diagram representing the result of applying the second change model described above. In addition, the output unit 103 may generate visualized data that enables a previous analysis result to be compared with a current analysis result.
As described above, in the fifth embodiment, the past first score or information describing the first score, and the second score or information describing the past second score are recorded in the analysis database 110. As a result, it is possible to provide the visualized data without applying the change models again or the like. This is effective for monitoring daily manufacturing data and the like.
Next, a hardware configuration of each of the data processing apparatuses according to the above-described embodiments will be described.
The CPU 801 is a processor that executes arithmetic processing, control processing, and the like according to a program. The CPU 801 uses a predetermined area of the RAM 802 as a work area, and executes various processes as the acquisition unit 101, the first calculation unit 102, the second calculation unit 106, and the output unit 103 described above in cooperation with programs stored in the ROM 803, the storage 804, and the like.
The RAM 802 is a memory such as a synchronous dynamic random access memory (SDRAM). The RAM 802 functions as a work area of the CPU 801. The ROM 803 is a memory that stores programs and various types of information in a non-rewritable manner.
The storage 804 is an apparatus that writes and reads data to and from a semiconductor storage medium such as a flash memory, a magnetically recordable storage medium such as a hard disk drive (HDD), an optically recordable storage medium, or the like. The storage 804 writes and reads data to and from the storage medium under the control of the CPU 801. The storage 804 may operate as the manufacturing database 104, the manufacturing process database 108, the user database 109, and the analysis database 110. The manufacturing database 104, the manufacturing process database 108, the user database 109, and the analysis database 110 may be stored in a storage medium different from the storage 804.
The display 805 is a liquid crystal display (LCD) or the like. The display 805 operates as the display unit 105, and displays various display screens such as the display screen illustrated in
The input device 806 is an input device that includes a mouse and a keyboard. The input device 806 operates as the operation unit 107, receives information input based on an operation by the user as an instruction signal, and outputs the instruction signal to the CPU 801.
The communication module 807 communicates with an external device via a network in accordance with control from the CPU 801.
The instructions indicated in the processing procedures described in the above-described embodiments can be executed based on a program that is software. By storing this program in advance and reading this program, a general-purpose computer system can obtain effects similar to the effects of the data processing apparatuses described above. The instructions described in the above-described embodiments are recorded in a magnetic disk (flexible disk, hard disk, or the like), an optical disc (CD-ROM, CD-R, CD-RW, DVD-ROM, DVD±R, DVD±RW, Blu-ray (registered trademark) disc, or the like), a semiconductor memory, or a recording medium similar thereto as the program that can be executed by a computer. The storage format may be any form as long as the recording medium is readable by a computer or an embedded system. When the computer reads the program from the recording medium and causes a CPU to execute an instruction described in the program based on the program, it is possible to implement an operation similar to those of the data processing apparatuses according to the above-described embodiments. Of course, in a case where the computer acquires or reads the program, the computer may acquire or read the program via a network.
In addition, an operating system (OS) running on the computer, database management software, middleware (MW) such as a network, or the like may execute a part of each of the processes for implementing the present embodiment based on an instruction of the program installed in the computer or the embedded system from the recording medium.
Furthermore, the recording medium in the present embodiment is not limited to the medium independent of the computer or the embedded system, and includes a recording medium that downloads and stores or temporarily stores the program transmitted via a LAN, the Internet, or the like.
Furthermore, the number of recording media is not limited to one. Also in a case where the processing in the present embodiment is executed from a plurality of media, the media may be included in the recording medium in the present embodiment, and each of the configurations of the media may be any configuration.
Note that the computer or the embedded system in the present embodiment is for executing each of the processes in the present embodiment based on the program stored in the recording medium, and may have any configuration such as an apparatus including one of a personal computer, a microcomputer, and the like, a system in which a plurality of apparatuses are connected to a network, or the like.
In addition, the computer in the present embodiment is not limited to a personal computer, and includes an arithmetic processing apparatus, a microcomputer, and the like included in an information processing apparatus, and collectively refers to a device and an apparatus that are capable of implementing the functions in the present embodiment by the program.
While certain embodiments have been described, these embodiments have been presented by way of example only, and are not intended to limit the scope of the inventions. Indeed, the novel embodiments described herein may be embodied in a variety of other forms; furthermore, various omissions, substitutions and changes in the form of the embodiments described herein may be made without departing from the spirit of the inventions. The accompanying claims and their equivalents are intended to cover such forms or modifications as would fall within the scope and spirit of the inventions.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2022-015113 | Feb 2022 | JP | national |
