JPWO2020161835A1 - Management system and machine learning device for it and management method - Google Patents

Abstract

Create an Ishikawa diagram based on expert knowledge about the relationship between managed items and multiple factors. In this characteristic factor diagram, the plurality of factors are data acquired at the manufacturing or service site where the control target item is set, and the importance of the control target item is attached to each of the plurality of factors. In machine learning, first, the link structure of the probabilistic model is set based on the relationship between the managed item in the characteristic factor diagram and multiple factors, and each of the probabilistic models is based on the importance attached to each of the multiple factors. Set the value of the node. Then, the link structure or the value of each node is learned using the data acquired at the manufacturing or service site.

Description

The present invention relates to a management system suitable for management of items that may be affected by various factors in the field of manufacturing or service, a machine learning device for that purpose, and a management method.

As a prior art document in which the prior art related to the present invention is disclosed, for example, International Publication No. 2016/170574 can be mentioned. In the system disclosed in International Publication No. 2016/170574, the operation data (big data) of a large number of distributed target bases are aggregated in the central base, and the operation data is utilized to detect an abnormality sign. The operation data is sensor data acquired by various sensors. In this system, a model for detecting signs of abnormality is constructed based on the collected sensor data for learning. Then, the predictive analysis of the abnormality is executed by utilizing the constructed model and the sensor data for the predictive analysis, and the cause analysis of the abnormality is executed based on the sensor data for the abnormality analysis.

International Publication No. 2016/170574

Today, in the field of manufacturing or service, various controlled items are set. For example, an abnormality in which a sign is detected in the above-mentioned prior art is also one management target item. When the managed item is affected by some factor, if the relationship between the managed item and the factor is known in advance, the state of the managed item and the change in the numerical value can be predicted from the change of the factor, and vice versa. In addition, it is possible to estimate the cause of the change from the state of the managed item and the change in the numerical value. In the above-mentioned prior art, the relationship between the managed item and a plurality of factors is modeled, and the sign of an abnormality which is the managed item is detected by using the learning sensor data corresponding to the factor and the model.

However, it is known that it takes a lot of time to build a model showing the relationship between the managed item and a plurality of factors. The greater the number of factors that can cause changes in the state of managed items and numerical values, the longer it takes to build a model. Today, machine learning is used to build such models, but the time required to build a model and the accuracy of the model largely depend on the scale of computer resources used for machine learning. In an environment where large-scale computer resources are not available, it takes a lot of time to establish a relationship between managed items and multiple factors. In addition, a large amount of data is required to build a highly accurate model, and a large amount of computer resources are inevitably required to process the large amount of data.

However, a large amount of cost is required to prepare a large-scale computer resource. Depending on the manufacturing or service site, it is often possible to prepare only small-scale computer resources due to cost constraints, and if the same results can be obtained, the cost of investing in computer resources should be kept low. For this reason, even if it is a small computer resource, it is possible to obtain results in a short time and with a small amount of data. Machine learning related to managed items and multiple factors in the field of manufacturing or service. Is required in.

The present invention has been made in view of the above problems, and an object of the present invention is to enable efficient machine learning of the relationship between a managed item at a manufacturing or service site and a plurality of factors.

As a means for achieving the above object, in the present invention, in machine learning for associating a managed item with a plurality of factors, expert knowledge about the relationship between the managed item and the plurality of factors is used in the initial modeling. Take advantage of. The expert knowledge is the knowledge of an expert who is familiar with the manufacturing or service site where the items to be managed are set, and is the knowledge explicit knowledge by the method described later. The present invention provides a management system, a machine learning device for the management system, and a management method. The management system provided by the present invention is a system that enables the utilization of expert knowledge for machine learning, and the machine learning device for the management system provided by the present invention utilizes expert knowledge for machine learning. The management method provided by the present invention is a method that enables the utilization of expert knowledge for machine learning.

The management system provided by the present invention includes a data acquisition means, a characteristic factor diagram creation means, a probability model initialization means, and a probability model learning means. The data acquisition means acquires data at the manufacturing or service site where the management target item is set. The Ishikawa diagram creation means creates a characteristic factor diagram that shows the relationship between a managed item and multiple factors that are presumed to affect it, based on expert knowledge about the relationship between the managed item and multiple factors. create. However, the plurality of factors in the characteristic factor diagram are at least a part of the data acquired by the data acquisition means, and each of the plurality of factors is given the importance to the managed item. The probability model initial setting means sets the link structure of the probability model based on the relationship between the managed item in the characteristic factor diagram and a plurality of factors, and the probability model is based on the importance attached to each of the plurality of factors. Set the value of each node. The probabilistic model learning means learns the link structure of the probabilistic model or the value of each node using the data acquired by the data acquisition means.

The management system may further include a characteristic factor diagram updating means for updating the characteristic factor diagram based on the link structure of the learned probability model or the value of each node. In addition, the management system further includes a monitoring means for identifying the presumed cause of the abnormality in the managed item based on the probability model and notifying the operator of the identified factor when an abnormality is detected in the managed item. You can also do it.

In another embodiment, the management system is a data acquisition device that acquires data at a manufacturing or service site in which managed items are set, and a terminal for creating the above-mentioned characteristic factor diagram based on expert knowledge. And a machine learning device that learns a probabilistic model related to the relationship between the managed item and a plurality of factors using the data acquired by the data acquisition device. However, the machine learning device initializes the link structure of the probabilistic model based on the relationship between the managed item in the characteristic factor diagram and multiple factors, and the probabilistic model is based on the importance attached to each of the multiple factors. It is configured to initialize the value of each node of.

In the machine learning device for the management system provided by the present invention, a data input unit for inputting data acquired at a manufacturing or service site in which management target items are set, and the above-mentioned characteristic factor diagram are set. It includes a characteristic factor diagram setting unit, a probability model initial setting unit for initializing a probability model, and a probability model learning unit for learning a probability model. The probability model initialization unit sets the link structure of the probability model based on the relationship between the managed item in the characteristic factor diagram and multiple factors, and the probability model is based on the importance attached to each of the multiple factors. It is configured to set the value of each node. The probability model learning unit is configured to learn the link structure or the value of each node using the data input to the data input unit.

The management method provided by the present invention includes the following first to fourth steps. The first step is a step of acquiring data at a manufacturing or service site where a controlled item is set. The second step is to create the above-mentioned characteristic factor diagram based on expert knowledge about the relationship between the managed item and the plurality of factors. The third step is to set the link structure of the probabilistic model based on the relationship between the managed item and the plurality of factors in the characteristic factor diagram, and each of the probabilistic models is based on the importance attached to each of the plurality of factors. This is the step to set the value of the node. Then, the fourth step is a step of learning the link structure of the probabilistic model or the value of each node using the acquired data.

According to the present invention, the relationship between the managed item and the plurality of factors at the site of manufacturing or service is represented by a probabilistic model, and the expert knowledge about the relationship between the managed item and the plurality of factors is explicit knowledge. By utilizing the factor diagram for the initial modeling of the probabilistic model, it is possible to efficiently machine learn the relationship between the managed item and a plurality of factors. The details of the effect will be apparent from the drawings briefly described below and the embodiments described in detail in connection therewith.

It is a figure which shows the whole structure of the management system which concerns on embodiment of this invention. It is a figure for demonstrating the structure of a fishbone diagram. It is a figure for demonstrating the method of making a fishbone diagram. It is a figure for demonstrating the method of making a fishbone diagram. It is a figure for demonstrating the method of making a fishbone diagram. It is a figure for demonstrating the outline of machine learning. It is a figure for demonstrating the search method of a latent factor. It is a figure for demonstrating the learning method of a Bayesian network. It is a figure for demonstrating the method of creating the data for machine learning. It is a figure which represented the whole flow of the machine learning executed in embodiment of this invention as an image. It is a figure for demonstrating the update method of a fishbone diagram. It is a figure for demonstrating the update example of a fishbone diagram. It is a figure for demonstrating the update example of a fishbone diagram.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, when the number, quantity, quantity, range, etc. of each element is referred to in the embodiment shown below, the reference is made unless otherwise specified or clearly specified by the number in principle. The present invention is not limited to the number of the inventions. In addition, the structures and steps described in the embodiments shown below are not necessarily essential to the present invention unless otherwise specified or clearly specified in principle.

1. 1. Management System Configuration The present invention is applicable to various manufacturing and service sites. The manufacturing site includes, for example, a food or machine manufacturing line, a painting line, a chemical plant, and the like. Service sites include, for example, distribution centers, dry cleaners, restaurants, and the like. That is, a suitable site for using the present invention is a site where at least one management target item exists and a plurality of factors that can affect the management target item can exist. In the present embodiment, among these sites, an example in which the present invention is applied to a painting line will be described.

FIG. 1 is a diagram showing an overall configuration of a management system according to the present embodiment. The management system 1 is applied to the painting line 10. The painting line 10 includes a pretreatment device 12 that pretreats the product 11, a painting device 14 that paints the product 11, a drying device 16 that dries the painted product 11, and a product from the pretreatment device 12 to the painting device 14. A transport device 13 for transporting the product 11, a transport device 15 for transporting the product 11 from the coating device 14 to the drying device 16, a utility facility 17, and a mixing chamber 18 for mixing paints are provided.

One or a plurality of sensors (not shown) are arranged in each of the equipments 11-18 constituting the painting line 10. The sensor includes various sensors such as a temperature sensor, a pressure sensor, a flow rate sensor, a weight sensor, an atmospheric pressure sensor, an outside temperature sensor, and a wind speed sensor. At each facility 11-18, sensor data is acquired by these sensors. However, the data acquisition means is not limited to the sensor. In at least some facilities 11-18, device data such as the rotation speed of the motor is acquired. Further, the manually input data may be acquired by the operator who manages the site inputting the data to the tablet terminal or the like. [Xi] in the figure represents various data (including sensor data, device data, and manually input data) acquired at various locations on the painting line 10.

The management system 1 includes an FBD creation terminal 2, a machine learning device 3, and a monitoring device 4 in addition to the above-mentioned data acquisition device for acquiring various data. The FBD creation terminal 2 is a terminal for creating a fishbone diagram (FBD) described later based on expert knowledge. The machine learning device 3 is a device that performs machine learning for finding a relationship between the data acquired by the data acquisition device in the painting line 10 and the customer problem. The data acquired in the painting line 10 is input to the machine learning device 3. At that time, the data obtained at the same time and in the same space are grouped, and the data is managed in group units. The details of machine learning will be described later.

The monitoring device 4 is a device that identifies the factor that caused the occurrence of an abnormality by utilizing the relationship between the data obtained by machine learning and the customer problem. The monitoring device 4 includes a monitor, and dynamically displays the location and content of the abnormality and the cause of the abnormality on the monitor. However, the monitoring device 4 is not indispensable, and the system including the FBD creation terminal 2 and the machine learning device 3 can also be called the management system 1.

2. 2. Structure and creation method of fishbone diagram In this embodiment, a fishbone diagram (hereinafter referred to as FBD), that is, a characteristic factor diagram is used as a tool for explicit knowledge of the knowledge possessed by an expert. The FBD is a diagram in which the relationship between a problem (characteristic) and a plurality of factors presumed to influence it is organized and systematically summarized. As shown in FIG. 2, in the FBD, a large bone is drawn diagonally with respect to the spine representing the task. Large bone means a major classification of factors. In this embodiment, five 4M1E bones, that is, MEN (person), MAC (mechanical equipment), MAT (material), MET (working method), and ENV (environment) are drawn. There is. MEN, MAC, MAT, and MET are on-site working conditions, and ENV is on-site environmental conditions.

The data acquired from the painting line 10 is classified into any of the above 4M1E. For example, in the example shown in FIG. 2, three middle bones are drawn with respect to the large bone of the MEN. These middle bones mean factors that affect MEN, and corresponding data are given for each middle bone. Experts determine which data (factors) are classified as 4M1E based on their own knowledge.

In addition, experts determine the importance of each piece of data. The importance is the magnitude of the influence of each data on the task, and the larger the influence, the larger the number is given. In the present embodiment, numbers from 0 to 5 and X are set as the rank of importance. From ranks 1 to 5, the larger the number, the greater the importance. Rank 0 means that the importance is unknown, and rank X means that the importance is judged to be zero by an expert. However, the rank X data is not displayed on the FBD. In the example shown in FIG. 2, the data (2nd_3rd_4th_5th_001) is assigned a rank 5 importance, and the data (2nd_3rd_4th_5th_002) is assigned a rank 3 importance. In the FBD creation terminal 2, an FBD in which each factor is ranked, that is, a ranked FBD (R-FBD) is created.

In the ranked FBD, the issues are largely such as coating quality, safety, equipment maintenance, paints and chemicals, productivity, etc., and specific issues subdivided under the major issues are set. For example, as shown in FIG. 3, specific issues such as film thickness, coating NV, and dust are set under the major issue of coating quality. These specific issues are the items to be managed by the management system 1. A ranked FBD is created for each task. In the example shown in FIG. 3, it is shown that a ranked FBD is created for each of the film thickness, the coated NV, and the dust. Further, when the problem is a problem that occurs independently for each facility, the ranked FBD is created for each facility and for each problem. In the example shown in FIG. 2, it is shown that a ranked FBD for which preventive maintenance is an issue is created for each of the three facilities PT11, PT31, and PT01.

4 and 5 show specific examples of the central bone of the ranked FBD. In the example shown in FIG. 4, two middle bones are drawn with respect to the large bone of MET. These cores are the factors that experts have determined to affect MET. Here, the chemical side of the solution of the preliminary degreasing device of the equipment PT11 and the free alkalinity of the solution of the preliminary degreasing device of the equipment PT11 are listed as factors for MET. On the other hand, in the example shown in FIG. 5, two middle bones are drawn with respect to the large bone of the MEN. Then, a small bone is drawn diagonally with respect to one middle bone. The data corresponding to the middle bone (PT01-1410) is clogged and the data corresponding to the small bone (PT01-1210) is pressure. As shown in the table, such a ranked FBD is created if the expert determines that clogging and pressure are related to results and factors.

As described above, in the present embodiment, the ranking FBD is created by the FBD creation terminal 2. In terms of correspondence with the claims, the FBD creating terminal 2 corresponds to the characteristic factor diagram creating means described in the claims.

3. 3. Contents of machine learning In the machine learning of the present embodiment, a method of searching for a causal relationship based on space-time is used. FIG. 6 is a diagram for explaining an outline of machine learning. Here, if the problem corresponding to the product j is expressed as Y _j , the set of data acquired from the painting line 10 is expressed as X _i, and the number of data is 100, the problem Y _j is set X _i = {X ₁ , ₁ , It is composed of X ₂ , X ₃ ... X ₃₅ , X ₃₆ ... X ₉₉ , X ₁₀₀ }. Task Y _j is quantified. The set X _i corresponding to the task Y _j is a group of data acquired at the same time as the task Y _j .

In machine learning, a parameter ΔY indicating a change in task Y and a parameter ΔX indicating a change in each data X are calculated. As a method of calculating the parameters ΔY and ΔX, there are a method of calculating the differential value shown in Equation 1 and a method of calculating the rate of change shown in Equation 2. Either calculation method may be used.

As shown in the upper part of FIG. 6, ΔY _j and ΔX _i are calculated and grouped for each product. Since the products flow on the painting line at regular intervals, calculating for each product is equivalent to performing calculations at regular time intervals. In the figure, the parameter surrounded by a circle means that the value exceeds the limit value. For example, for the 19th product, ΔX ₂ exceeds the limit value. On the other hand, for the 20th product, ΔY _20, ΔX _2, ΔX _{3, and} ΔX ₁₀₀ exceed the limit values. Since ΔY ₂₀ exceeds the limit value when ΔX ₃ and ΔX ₁₀₀ exceed the limit value, it can be inferred that X ₃ and X ₁₀₀ affect the task Y.

The physical structure of the painting line 10 is drawn in the lower part of FIG. The data acquired from the painting line 10 is also classified according to the acquired space. For example, X ₁ , X ₂ , X ₃ are the data obtained by the painting device, X ₇ , X ₈ , X ₉ are the data obtained in the mixing chamber, and X ₁₁ is the transfer device between the painting device and the drying device. The data obtained in, is classified as. Further, the arrow lines connecting the facilities in FIG. 6 indicate the overlap of spaces on the data. In the example shown in FIG. 6, the coating apparatus 14 and the mixing chamber 18 are regarded as the same space in terms of data. As described above, in the machine learning of the present embodiment, a set of data obtained at the same time and in the same space is created, and machine learning is performed in units of the set.

The logical structure of the painting line 10 is drawn in the middle of FIG. In the logical structure, {Pro ○○, Pro △△} is a set of products included in one batch, and for example, {Pro19, Pro20} represents a set of products Nos. 19 and 20. The frame drawn by the broken line represents the classification of the data X that changed when the task Y changed. In the example shown in FIG. 6, {Pro19, Pro20} in the set _{{X 1, X 2, X} 3} the set _{{X 7, X 8, X} 9} and a change occurs in the set _{{X 11}, {} Pro44, Pro45} in the set _{{X 1, X 2, X} 3} the set _{{X 7, X 8, X} 9} and a change occurs in the set _{{X 11}, {Pro102,} Pro103} in the set _{{X 7} , X ₈ , X ₉ } and the set {X ₁₁ } have changed. From the example shown in FIG. 6, each of the set {X ₁ , X ₂ , X ₃ }, the set {X ₇ , X ₈ , X ₉ } and the set {X ₁₁ } is a cause of repeated occurrence, and , Set {X ₁ , X ₂ , X ₃ } and set {X ₇ , X ₈ , X ₉ } are the causes of superposition in the same space.

By performing the above machine learning on the data obtained from the painting line 10, the relationship between the problem and a plurality of factors can be sorted out, and the factors causing the problem can be searched for. It becomes. Here, the learning method of the probability model and the search method of the latent factor in machine learning will be described in more detail with reference to FIG. 7. What is shown in FIG. 7 is the time series data of ΔY _j and the corresponding sets {ΔX ₁ , ΔX ₂ ... ΔX ₃₀ , ΔX ₃₁ , ΔX ₃₂ ... ΔX ₉₉ , ΔX ₁₀₀ }. In machine learning, the change of ΔY _j is determined for each time. When ΔY _j is Boolean data, the change from true to false is determined as the change in ΔY _j . When ΔY _j is a numerical value, it is determined as a change in ΔY _j that a difference of the numerical value (measured value) or more occurs.

When a change occurs in ΔY _j , the set {ΔX ₁ , ΔX ₂ ... ΔX ₃₀ , ΔX ₃₁ , ΔX ₃₂ ... ΔX ₉₉ , ΔX ₁₀₀ } at that time is searched for X _{i in} which the difference is greater than or equal to the limit value. And the counting result Count (X _i | Y _j ) is obtained. Then, the Count (X _i | Y _j ) is converted into discrete data, and the probability model is learned based on the data. In this embodiment, a Bayesian network model is used as a probabilistic model. The problem of finding the link structure and the value of each node in the Bayesian network model can be solved by integer programming. In machine learning, a search for latent factors is performed using a Bayesian network model. In this case, a new factor (variable) is searched by the belief propagation method based on the time lag effect based on the space.

Next, a method of learning the Bayesian network by the machine learning device 3 will be described with reference to FIG. As described in the Bayesian pattern recognition algorithm section, a Bayesian network is a computer-readable structure that makes the number of cases finite and the actual phenomenon that is physical / chemical with an infinite number of cases. It is a model represented by. Specifically, there are observable factor X _i and unobservable factor ω _i in the actual phenomenon in the painting line 10, but the causal relationship between the result and the factor for the observable factor X _{i in} the Bayesian network. , And the causal relationship between the factors is represented by a link structure.

FIG. 8 depicts a graphical statistical model of a recurrent neural network. The update structured model in this graphical statistical model is a model for updating the structure of the Bayesian network, that is, the link structure and the value of each node. The set of data obtained at the current time on the painting line 10 and the output of the updated structured model at the previous time are input to the updated structured model. The output of the updated structured model is the previous value of the above-mentioned unobservable factors. The above-mentioned belief propagation method is used for training in the updated structured model.

The machine learning device 3 performs machine learning by the above method. However, although the number of data has been described as 100 here, the actual number of data is large and there are many problems to be managed, so that it takes a long time to calculate with a small computer resource. The above-mentioned ranked FBD created by experts is utilized there. As shown in FIG. 9, the machine learning device 3 converts the ranked FBD created by the FBD creation terminal 2 into an orthogonal array. Alternatively, the ranked FBD may be converted into an orthogonal array in advance, and the orthogonal array may be input to the machine learning device 3. In the orthogonal array, in addition to the relationship between the task (result) and the factor (cause), the importance of each factor and the causal relationship between the factors are also input.

The machine learning device 3 initializes the Bayesian network based on the orthogonal array. Since the orthogonal array is converted from the ranked FBD, the Bayesian network link structure is initialized based on the relationship between the issues and multiple factors in the ranked FBD, and the importance attached to each of the multiple factors. The value of each node of the Bayesian network will be initialized based on the degree. There are many factors that can influence the task, but the factors that experts have determined to have no effect are not displayed in the ranked FBD and are not reflected in the Bayesian network. As a result, the factors are excluded from the target of machine learning, so that machine learning can be performed efficiently and even small-scale computer resources can be processed.

The learned Bayesian network is used in the monitoring device 4. When an abnormality is detected in the managed item (issue), the monitoring device 4 identifies the presumed cause of the abnormality in the managed item based on the learned Bayesian network, and notifies the operator of the identified factor. .. In correspondence with the claims, the monitoring device 4 corresponds to the monitoring means described in the claims.

4. Update of Fishbone Diagram As described above, in this embodiment, the ranked FBD in which expert knowledge is explicit knowledge is applied to machine learning using a Bayesian network. Machine learning clarifies the relationship between tasks that are managed items and various factors, and it is also possible to further improve the accuracy of ranked FBD by feeding back the results of machine learning to ranked FBD. .. FIG. 10 is an image representation of the overall flow of machine learning executed in the present embodiment.

The expert stores the relationship between the observable data (factors) obtained at the painting line 10 and the problem as knowledge (step S1). Then, a ranked FBD is created based on that knowledge. Observable data obtained from the painting line 10 is associated with the middle and small bones of the ranked FBD (step S2).

In the machine learning device 3, the link structure of the Bayesian network and the value of each node are initialized based on the ranked FBD created by the expert (step S3). Further, learning is performed based on the observable data obtained on the painting line 10 (step S4). Training is performed by a recurrent neural network, and unobservable data is used as a hidden layer. The learning result is reflected in the Bayesian network, and the value of each node changes or the link structure changes according to the learning result (step S5). Further, the learning result is reflected in the ranked FBD via the Bayesian network, and the bone structure and the importance of each factor are updated based on the learned link structure or the value of each node (step S6).

Here, a method of updating the ranked FBD will be specifically described with reference to FIG. There is a change in 100 data in total in a predetermined time a change occurs in the object Y, 4 pieces of which data X _1, the data X ₂ is 3, 10 data X _3, the data X ₄ are included three Suppose it was. Without considering the context of the problem Y, for better probability of data X ₃ is higher than the probability of the data X _2, in general, towards the importance of the data X ₃ is higher than the importance of the data X ₂ Is also judged to be high. Moreover, since the probability of data X ₄ is not high, it is generally judged that the importance of data X ₄ is low. Here, it is assumed that the ranked FBD is created by an expert according to such a judgment.

However, in machine learning using a Bayesian network, the relationship between task Y and each data is statistically calculated. According to FIG. 11, the probability P (Y | X ₂ ) that the task Y changed when the data X ₂ changed was 1/3, whereas the task Y changed when the data X ₃ changed. The probability P (Y | X ₃ ) is 2/10. Therefore, the machine learning device 3 determines that the importance of the data X ₂ is higher than the importance of the data X ₃ . According to this judgment, the values of the nodes corresponding to the data X ₂ and X ₃ are changed in the Bayesian network. Furthermore, the change in the Bayesian network is also reflected in the ranked FBD, and the importance of the factors corresponding to the data X ₂ and X ₃ is changed.

Also, the probability P that problems Y when the data X ₄ is changed is changed (Y | X ₄₎ is the statistical probability problems Y when the data X ₂ is changed is changed P (Y | X ₂ ) And the probability P (X ₂ | X ₄ ) that the data X ₂ changed when the data X ₄ changed. Therefore, the probability P (Y | X ₄ ) is 1/9 instead of zero. According to this judgment, the link structure is changed in the Bayesian network to add the node corresponding to the data X ₄ . In addition, changes in the Bayesian network will be reflected in the ranked FBD, and new small bones corresponding to data X ₄ will be added.

FIG. 12 is a diagram for more specifically explaining an example of updating the ranked FBD, specifically, an example in which the importance of factors changes. In the learning of the Bayesian network, each of the data X _i and the task Y _j are based on the history of each change of the data X _i obtained from the painting line 10 and the history of the evaluation result of the task Y _j which is the management target item. The degree of association with is checked, and the value of each node is reviewed according to the degree of association.

In the example shown in FIG. 12, the positions of the nodes of data X ₂ and the positions of the nodes of data X ₃ are exchanged before and after learning. This means that the magnitude relationship of the nodes between data X ₂ and data X ₃ changed before and after learning. When a node's value is reviewed in a Bayesian network, the ranked FBD modifies the importance assigned to the factor corresponding to that node. In the example shown in FIG. 12, before and after learning, the importance of the data corresponding to the middle bone (1st_2nd_3rd_4th_5th_003) was changed from rank 5 to rank 3, and as a result, the importance was lower than that of the data of rank 4 (1st_2nd_3rd_4th_5th_002). There is.

FIG. 13 is a diagram for more specifically explaining an example of updating the ranked FBD, specifically, an example in which the result-factor relationship changes. In the Bayesian network learning, each of the data X _i and the task Y _j are based on the history of each change of the data X _i obtained from the painting line 10 and the history of the evaluation result of the task Y _j which is the management target item. The degree of relevance to is examined. Then, if the node corresponding to the highly relevant data does not exist in the link structure of the Bayesian network, the node is newly added to the link structure.

In the example shown in FIG. 13, after learning, a node of data X ₄ is newly added below the node of data X ₂ . This means that for data X ₂ , data X _{4, which} has a relationship between results and factors, was newly found by machine learning. When a new node is added to the link structure, the factor corresponding to the node is newly added to the ranked FBD. In the example shown in FIG. 13, after learning, a small bone corresponding to the data (1st_2nd_3rd_4th_5th_004) is newly added to the middle bone corresponding to the data (1st_2nd_3rd_4th_5th_002).

The process described above is performed by the machine learning device 3. In correspondence with the claims, the function of the machine learning device 3 to acquire data from the painting line 10 corresponds to the function as the data input unit described in the claims. Further, the function of the machine learning device 3 to receive the ranked FBD created by the FBD creation terminal 2 corresponds to the function as the characteristic factor diagram setting unit described in the claim. The function of initializing the link structure of the Bayesian network and each node value based on the ranked FBD corresponds to the probability model initialization means described in the claims and corresponds to the probability model initialization unit described in the claims. To do. The function of learning the link structure of the Bayesian network and the value of each node based on the data acquired from the painting line 10 corresponds to the probability model learning means described in the claim, and the probability model learning unit described in the claim. Corresponds to. Further, the function of updating the ranked FBD based on the Bayesian network after learning corresponds to the characteristic factor diagram updating means described in the claims.

5. Other Embodiments Some of the functions provided by the machine learning device 3 in the above embodiment can be transferred to another computer. It is also possible to transfer some or all of the functions of the machine learning device 3 to a server on the cloud.

1 Management system 2 FBD creation terminal 3 Machine learning device 4 Monitoring device 10 Painting line

Claims (10)

Data acquisition means for acquiring data at the manufacturing or service site where the items to be managed are set, and
It is a characteristic factor diagram showing the relationship between the management target item and a plurality of factors presumed to affect the management target item, and the plurality of factors are at least a part of the data acquired by the data acquisition means. In addition, in order to create a characteristic factor diagram in which the importance of the management target item is attached to each of the plurality of factors based on expert knowledge about the relationship between the management target item and the plurality of factors. Ishikawa diagram creation means and
A link structure of the probability model is set based on the relationship between the management target item and the plurality of factors in the characteristic factor diagram, and each of the probability models is based on the importance attached to each of the plurality of factors. Probability model initial setting means to set the value of the node,
A management system comprising: a stochastic model learning means for learning the link structure or the value of each node using the data acquired by the data acquisition means. The management system according to claim 1, further comprising a characteristic factor diagram updating means for updating the characteristic factor diagram based on the learned link structure or the value of each node. The probability model learning means has a degree of association between each of the data and the managed item based on the history of each change of the data acquired by the data acquisition means and the history of the evaluation result of the managed item. Is configured to examine and review the value of each node according to the degree of association.
When the value of each of the nodes is reviewed, the characteristic factor diagram updating means is configured to correct the importance assigned to each of the plurality of factors based on the value of each of the reviewed nodes. The management system according to claim 2, wherein the management system is characterized by the above. The probability model learning means has a degree of association between each of the data and the managed item based on the history of each change of the data acquired by the data acquisition means and the history of the evaluation result of the managed item. If the node corresponding to the highly relevant data does not exist in the link structure, the node is newly added to the link structure.
A claim characterized in that the characteristic factor diagram updating means is configured to newly add a factor corresponding to the node to the characteristic factor diagram when a new node is added to the link structure. The management system according to 2 or 3. When an abnormality is detected in the management target item, a monitoring means for identifying a factor presumed to be the cause of the abnormality in the management target item based on the probability model and notifying the operator of the identified factor is further provided. The management system according to any one of claims 1 to 4. A claim characterized in that the data acquired by the data acquisition means includes data relating to evaluation of the management target item, data relating to working conditions at the site, and data relating to environmental conditions at the site. The management system according to any one of 1 to 5. The management system according to any one of claims 1 to 6, wherein the probability model is a Bayesian network. A data acquisition device that acquires data at the manufacturing or service site where managed items are set, and
It is a characteristic factor diagram showing the relationship between the management target item and a plurality of factors presumed to affect the management target item, and the plurality of factors are at least a part of the data acquired by the data acquisition device. In addition, in order to create a characteristic factor diagram in which the importance of the management target item is attached to each of the plurality of factors, based on expert knowledge about the relationship between the management target item and the plurality of factors. Terminal and
A machine learning device that learns a probabilistic model regarding the relationship between the managed item and the plurality of factors by using the data acquired by the data acquisition device is provided.
The machine learning device initially sets the link structure of the probability model based on the relationship between the management target item and the plurality of factors in the characteristic factor diagram, and the importance assigned to each of the plurality of factors. A management system characterized in that it is configured to initialize the value of each node of the probability model based on the above. A data input unit where data acquired at the manufacturing or service site where managed items are set is input, and
It is a characteristic factor diagram created based on expert knowledge about the relationship between the managed item and a plurality of factors presumed to affect it, and the plurality of factors are input to the data input unit. A characteristic factor diagram setting unit in which a characteristic factor diagram that is at least a part of the data and has an importance to the managed item is set for each of the plurality of factors.
A link structure of the probability model is set based on the relationship between the management target item and the plurality of factors in the characteristic factor diagram, and each of the probability models is based on the importance attached to each of the plurality of factors. Probability model initial setting part that sets the value of the node,
A machine learning device for a management system, comprising: a probabilistic model learning unit that learns the link structure or the value of each node using the data input to the data input unit. Steps to acquire data at the manufacturing or service site where managed items are set, and
It is a characteristic factor diagram showing the relationship between the management target item and a plurality of factors presumed to affect the management target item, and the plurality of factors are at least a part of the acquired data and the plurality of factors. A step of creating a characteristic factor diagram in which the importance of each of the factors is assigned to the managed item based on expert knowledge about the relationship between the managed item and the plurality of factors, and
A link structure of the probability model is set based on the relationship between the management target item and the plurality of factors in the characteristic factor diagram, and each of the probability models is based on the importance attached to each of the plurality of factors. Steps to set the value of the node and
A management method comprising: learning the link structure or the value of each node using the acquired data.

Images