JP7154468B2 - Information processing device, information processing method and information processing program - Google Patents

Description

The present disclosure relates to techniques for performing analysis to improve performance.

In a production system such as a factory, priority targets are set according to the situation from time to time, among management indicators such as quality, cost, delivery date, and production quantity. Production control is then carried out to achieve the set priority targets.
A priority goal is called KGI (Key Goal Indicator). For example, when quality is set in KGI, daily production control is performed so as to achieve the target value of quality. Production management collects information on the current operational status from the production system, compares the current value with the target value, and if the current value does not achieve the target value, makes improvements so that the current value achieves the target value. To do an activity. Improvement activities are activities such as enhancement of equipment in the production system, adjustment of parameters, education of workers, review of work procedures, review of materials and/or inventory, and the like.

Conventionally, such improvement activities have often been carried out based on the experience of production system managers, and there has been the problem that improvement activities that do not necessarily contribute to the achievement of priority goals or that have low effectiveness are carried out.
In recent years, efforts have been made to solve such problems by promoting IoT (Internet of Things) in production systems. Specifically, the issues will be resolved through the following activities. In addition to the current value of KGI in a certain production system, detailed data is collected on the status of the processes, equipment, and equipment within the equipment within the production system at any given time. Then, by analyzing the collected data, relationships of data within the production system are clarified. Then, the clarified relationship is used as a reference for production control.

For example, Patent Literature 1 defines a hierarchical structure of relationships among a plurality of KPIs (Key Performance Indicators), each of which is a management index for a plurality of processes and/or a plurality of facilities. Further, in Patent Literature 1, information for calculating KPI is collected from processes and/or equipment. In Patent Document 1, correlation analysis is performed on each KPI to efficiently select an alarm that notifies an administrator of an abnormality when an abnormality occurs.

JP 2019-117464 A

However, in the correlation analysis used in the technique of Patent Literature 1, when the collected information is an outlier, the influence is large, so complicated preprocessing is required to remove this influence. In addition, since correlation analysis analyzes 1:1 relationships for KPIs in different processes, it is not possible to analyze one-to-many information relationships, such as the relationship between a KGI and multiple KPIs. There is

As described above, in the technique of Patent Document 1, when performing improvement activities in a system composed of a large number of elements such as a production system, an element to improve KGI, that is, an element to improve performance and other elements It is difficult to analyze relationships with multiple factors. Therefore, with the technique of Patent Document 1, there is a problem that it is not easy to identify other elements that contribute to improving the performance of the element whose performance is to be improved.

One of the main purposes of the present disclosure is to solve such problems. More specifically, the main object of the present disclosure is to efficiently identify elements that contribute to improving the performance of elements whose performance should be improved.

The information processing device according to the present disclosure is
a designation unit that designates, as a designated element, an element whose performance should be improved among the three or more elements;
an extraction unit that extracts two or more elements having a significant relationship with the designated element from elements other than the designated element as related elements;
analyzing the influence of the performance of each of the two or more related elements extracted by the extracting unit on the performance of the designated element, and selecting from among the two or more related elements to improve the performance of the designated element an estimating unit for estimating an improvement target related element, which is a related element whose performance should be improved.

According to the present disclosure, it is possible to efficiently identify an element that contributes to improving the performance of an element whose performance should be improved.

FIG. 2 is a diagram showing a functional configuration example of the improvement point analysis device according to the first embodiment; FIG. FIG. 2 is a diagram showing a hardware configuration example of the improvement point analysis device according to the first embodiment; FIG. 4 is a diagram showing a configuration example of an information model according to Embodiment 1; FIG. 4 is a diagram showing an example of setting items according to the first embodiment; FIG. 4 is a flowchart showing an operation example of an evaluation unit according to Embodiment 1; 4A and 4B are diagrams showing examples of evaluation results by an evaluation unit according to the first embodiment; FIG. 4A and 4B are diagrams showing examples of evaluation results by an evaluation unit according to the first embodiment; FIG. 4 is a flow chart showing an example of the operation of an improvement point analysis unit according to Embodiment 1; 4 is a flow chart showing an example of the operation of an improvement point analysis unit according to Embodiment 1; FIG. 4 is a diagram showing an example of an analysis target according to Embodiment 1; FIG. 4 is a diagram showing an example of analysis conditions according to Embodiment 1; FIG. FIG. 10 is a diagram showing an example of an analysis result of an improvement point analysis unit according to the first embodiment; FIG. FIG. 10 is a diagram showing an example of setting items according to the second embodiment; FIG. FIG. 10 is a diagram showing an example of analysis conditions according to Embodiment 2; FIG. FIG. 10 is a diagram showing an example of effects obtained by the second embodiment; FIG. 11 is a diagram showing an example of setting items according to the third embodiment; FIG. FIG. 11 is a diagram showing a configuration example of an information model according to Embodiment 4; FIG. 11 is a diagram showing an example of analysis conditions according to Embodiment 4; 13 is a flow chart showing an operation example of an improvement part analysis unit according to the fourth embodiment; FIG. 10 is a diagram showing an example of analysis conditions according to Embodiment 5; FIG. 11 is a diagram showing an output example according to Embodiment 5; FIG. 11 is a diagram showing a functional configuration example of an improved portion analysis device according to Embodiment 6; FIG. 12 is a diagram showing an example of an improvement record according to Embodiment 6; FIG. FIG. 12 is a diagram showing an example of analysis conditions according to Embodiment 6; FIG. 12 is a diagram showing an output example according to Embodiment 6; FIG. 12 is a diagram showing an example of analysis conditions according to Embodiment 7; FIG. 13 is a diagram showing a functional configuration example of a machine learning device according to Embodiment 7; 14 is a flowchart showing an operation example of a machine learning device according to Embodiment 7; FIG. 11 is a flowchart showing a procedure for correcting an analysis result output according to Embodiment 7; FIG. FIG. 13 is a diagram showing an example of a neural network according to Embodiment 7;

Embodiments will be described below with reference to the drawings. In the following description of the embodiments and drawings, the same reference numerals denote the same or corresponding parts.

Embodiment 1.
*** Configuration description ***
FIG. 1 shows a functional configuration example of an improved portion analysis apparatus 100 according to this embodiment.
Improvement point analysis apparatus 100 according to the present embodiment is connected to analysis target 200 via network 300 . In this embodiment, the analysis target 200 is a production system.
The improvement point analysis device 100 corresponds to an information processing device. Further, the operation procedure of the improvement point analysis device 100 corresponds to an information processing method.

As shown in FIG. 1 , the improvement point analysis device 100 has an information storage unit 101 , an evaluation unit 102 , an improvement point analysis unit 103 and an information collection unit 104 .
Details of the information storage unit 101, the evaluation unit 102, the improvement point analysis unit 103, and the information collection unit 104 will be described later.

FIG. 2 shows a hardware configuration example of the improvement point analysis device 100 .
The improvement point analysis device 100 is a computer.
The improvement point analysis apparatus 100 includes a processor 901, a storage device 902, and a communication interface 903 as hardware. Processor 901 , storage device 902 and communication interface 903 are interconnected by bus 905 .
Storage device 902 stores program 904 . A program 904 is a program for realizing the functions of the evaluation unit 102, the improvement part analysis unit 103, and the information collection unit 104 shown in FIG.
Processor 901 reads program 904 from storage device 902 and executes program 904 . By executing the program 904 by the processor 901, the functions of the evaluation unit 102, the improvement part analysis unit 103, and the information collection unit 104, which will be described later, are realized.
The program 904 corresponds to an information processing program.
Although not shown, the storage device 902 stores, in addition to the program 904, various kinds of information necessary for realizing the functions of the improvement point analysis apparatus 100 according to this embodiment. The information storage unit 101 shown in FIG. 1 is implemented by a storage device 902 .
A communication interface 903 is used for communication with the production system that is the analysis target 200 .

FIG. 3 shows a configuration example of an information model used in this embodiment.
The information model shows a plurality of elements that constitute the production system that is the analysis target 200 . Also, in the information model, relationships between elements are indicated by hierarchical structure and/or logical structure. In FIG. 3, solid lines indicate relationships based on the hierarchical structure, and dashed arrows indicate relationships based on the logical structure.

In the information model of FIG. 3, the hierarchy includes production lines, processes, facilities and equipment.
"Product A production lead time" is defined as an element constituting the hierarchy: production line. FIG. 3 shows an example of setting "product A production lead time" as the KGI of the production system and managing production.
"Process #1 lead time,""Process#2 lead time," and "Process #3 lead time," which are the lead times of each process in the production system, are defined as a hierarchy below "Product A production lead time." ing. "Process #1 lead time", "Process #2 lead time" and "Process #3 lead time" are connected to "Product A production lead time" in a hierarchical structure. That is, the "process #1 lead time", the "process #2 lead time" and the "process #3 lead time" can affect the "product A production lead time". Furthermore, each process is connected with a relationship by a logical structure based on the production process. Here, an example is shown in which production is performed in order of "process #1", "process #2", and "process #3". In this specification, "lead time" means required time. In other words, "product A production lead time" is the time required to complete the production of product A.
Similarly, for the hierarchy of equipment and equipment, the relationship with the elements of the higher hierarchy is shown.
Specifically, the hierarchy: "equipment #1-1 lead time", "equipment #1-2 lead time", "equipment #2-1 lead time", "equipment #3-1 Lead Time” and “Equipment #3-2 Lead Time” are defined. "Equipment #1-1 Lead Time" and "Equipment #1-2 Lead Time" can affect "Process #1 Lead Time". Also, the "equipment #2-1 lead time" can affect the "process #2 lead time." In addition, "equipment #3-1 lead time" and "equipment #3-2 lead time" can affect "process #3 lead time".
Hierarchy: Elements that make up the device are "PLC #1-1-1 cycle time", "servo #1-1-2 motor current value", "sensor #1-1-3 target angle", "PLC #1-2-1 cycle time” and “robot #1-2-2 reach rate” are defined. "PLC #1-1-1 cycle time", "servo #1-1-2 motor current value" and "sensor #1-1-3 target angle" affect "equipment #1-1 lead time" obtain. In addition, "PLC #1-2-1 cycle time" and "robot #1-2-2 arrival rate" can affect "equipment #1-2 lead time". Furthermore, "Servo #1-1-2 Motor Current Value" can affect "PLC #1-1-1 Cycle Time" and "Sensor #1-1-3 Target Angle". Also, "sensor #1-1-3 target angle" affects "robot #1-2-2 reach rate", and "robot #1-2-2 reach rate" affects "PLC #1-2-1 can affect cycle time.

Here, the information storage unit 101, evaluation unit 102, improvement point analysis unit 103, and information collection unit 104 shown in FIG. 1 will be described.

The information storage unit 101 stores the information model illustrated in FIG.
The information storage unit 101 also stores setting items to be described later.
Also, the information storage unit 101 stores evaluation results by the evaluation unit 102 .
Further, the information storage unit 101 stores analysis results by the improvement part analysis unit 103 .
The information storage unit 101 also stores information collected by the information collection unit 104 .

The evaluation unit 102 evaluates whether or not the performance of each element included in the analysis target 200 meets the performance criteria. The evaluation unit 102 evaluates whether or not the performance of each element matches the performance criteria defined for each element.
In addition, although each element constitutes a plurality of layers as described above, the evaluation unit 102 may evaluate whether the performance of each element meets the performance standard in a time period different for each layer. good.
Furthermore, the evaluation unit 102 may output an evaluation result each time evaluation is performed.

Based on the evaluation result by the evaluation unit 102, the improvement part analysis unit 103 designates an element whose performance should be improved among the elements included in the analysis object 200 as a designated element.
Further, the improvement part analysis unit 103 may designate an estimated improvement target related element (described later) as a new designated element. Further, the improvement portion analysis unit 103 may repeat designating the estimated new improvement target related element as a new designated element each time it estimates a new improvement target related element.

Further, the improvement point analysis unit 103 extracts one or more elements having a significant relationship with the designated element from elements other than the designated element as related elements. In this embodiment, a 1:1 relationship between a specified element and a related element can also be analyzed, so an example in which the improvement point analysis unit 103 extracts one or more elements as related elements will be described. When analyzing the relationship between one designated element and each of a plurality of related elements, the improvement part analysis unit 103 extracts two or more elements as related elements. A "significant relationship" is a relationship that can affect the performance of the specified element. More specifically, in the information model of FIG. It is a positioned element.
For example, when the evaluation unit 102 designates "product A production lead time" as a designated element, the improvement point analysis unit 103 determines that the element (related element) having a significant relationship with "product A production lead time" is: "Process #1 lead time", "Process #2 lead time" and "Process #3 lead time" are extracted. Further, for example, when the evaluation unit 102 designates "process #2 lead time" as a designated element, the improvement point analysis unit 103 determines an element (related element) that has a significant relationship with "process #2 lead time". , "process #1 lead time" and "equipment #2-1 lead time" are extracted. "Performance" is a value obtained by measurement and/or calculation. For example, the performance of "product A production lead time" is the time obtained by measuring the time required for each process and adding the time required for each process. Also, for example, the performance of "servo #1-1-2 motor current value" is a motor current value obtained by actually measuring servo #1-1-2.

In addition, the improvement part analysis unit 103 analyzes the influence of the performance of each of the extracted one or more related elements on the performance of the specified element, and selects, from among the one or more related elements, Estimate the relevant element to be improved, which is the relevant element whose performance should be improved. The improvement part analysis unit 103 uses the fact that the performance of the designated element is improved (or deteriorated) in the conclusion part, and the condition is that the performance of each of the one or more related elements is improved (or deteriorated). We estimate the related elements to be improved by performing the association analysis used for the part. Association analysis will be described later.
In addition, when estimating a plurality of improvement target related elements, the improvement part analysis unit 103 analyzes the influence of the performance of each improvement target related element on the performance of other improvement target related elements, Set priorities among related elements.
Further, when a new specified element is specified by the evaluation unit 102, the improvement part analysis unit 103 converts one or more elements other than the new specified element, which have a significant relationship with the new specified element, into a new related element. Extract as Then, the improvement part analysis unit 103 analyzes the influence of the performance of each of the extracted one or more new related elements on the performance of the new specified element, and selects the new specified element from among the one or more new related elements. In order to improve the performance of the designated element, a new related element whose performance should be improved is estimated as a new improvement target related element.
In addition, each time a new specified element is specified, the improvement part analysis unit 103 repeats extracting one or more elements other than the specified new specified element as new related elements, and extracts one or more new specified elements. Estimation of a new related element to be improved may be repeated each time a related element is extracted.
Note that the improvement point analysis unit 103 corresponds to a designation unit, an extraction unit, and an estimation unit. Processing performed by the improvement point analysis unit 103 corresponds to designation processing, extraction processing, and estimation processing.

Here, association analysis will be described.
Association analysis (also called basket analysis, association rule, etc.) is known as a technique for analyzing relationships between a plurality of pieces of information.
In association analysis, statistical information is used as input, and support, confidence, and lift values are calculated for the relation between the condition part and the conclusion part. The degree of support is the ratio of data that includes both the condition part and the conclusion part among the total number of data. Reliability is the ratio of data including both the condition part and the conclusion part to the number of data including the condition part. A lift value is a value obtained by dividing the reliability by the number of data including the conclusion part. Generally, when the lift value is greater than 1, the effect of the conditional part on the conclusion part is large, and it can be quantitatively determined that there is a relationship between the conclusion part and the conditional part.
However, in association analysis, the amount of calculation is enormous because it is necessary to calculate the support, reliability, and lift value for each combination of input information. For this reason, there is a problem that association analysis cannot be simply applied to a production system with many types of information.
In the present embodiment, the improvement point analysis unit 103, when the calculation maintenance condition is not satisfied in any of the plurality of calculation items (support, reliability, lift value) included in the association analysis, Do not calculate uncalculated calculation items among multiple calculation items. By doing so, in the present embodiment, the association analysis is effectively applied even to a production system with many types of information.

The information collection unit 104 collects information about each element shown in FIG. 3 from the production system that is the analysis target 200 .
For example, the information collection unit 104 collects performance for each element shown in FIG. Specifically, the information collection unit 104 collects the measured time required for each process as the performance of the "product A production lead time", adds the time required for each process, and obtains the time required to produce product A. get The information collecting unit 104 also collects the motor current value obtained by actually measuring the servo #1-1-2 as the performance of the "servo #1-1-2 motor current value".
Information collecting unit 104 stores the collected information in information storage unit 101 .

FIG. 4 shows an example of setting items of the information collection unit 104. As shown in FIG.
The information collection unit 104 collects information of each element included in the information model according to the setting items shown in FIG.
In FIG. 4, as setting items of the information collecting unit 104, an element, a collection target, a monitoring period, and a criterion are defined.
Among the setting items in FIG. 4, the elements included in the information model are shown in the element column.
The collection target column shows the information collection method. As a collection method, there are a method of directly collecting information from any device existing in the production system, and a method of calculation based on information collected from any device. When collecting information directly from a device, the device from which the information is collected is indicated in the collection target column.
The monitoring period column shows the time width for which the criterion is applied.
The criteria used for evaluation by the evaluation unit 102, that is, performance criteria, are shown in the criteria column.
The information collected by the information collection unit 104 is held in the evaluation unit 102 as time-series data as shown in the storage content.

***Description of operation***
FIG. 5 is a flowchart showing an operation example of the evaluation unit 102. As shown in FIG.
FIG. 6 shows an example of the evaluation result of the evaluation unit 102 in graph form.
FIG. 7 shows an example of evaluation results of the evaluation unit 102 in tabular form.
FIG. 8 is a flow chart showing an operation example of the improvement point analysis unit 103 .
FIG. 9 is a flow chart showing details of step S204 in FIG.
FIG. 10 shows an example of an analysis target according to this embodiment.
FIG. 11 shows an example of analysis conditions according to this embodiment.
FIG. 12 shows an example of the analysis result of the improvement point analysis unit 103. As shown in FIG.
Hereinafter, an operation example of the improvement point analysis apparatus 100 according to the present embodiment will be described with reference to FIGS. 1 to 12. FIG.

Presetting Phase The production system manager or designer generates the information model shown in FIG. 3 and the setting items shown in FIG. 4 for management of the production system. Then, the administrator or designer of the production system stores the generated information model and setting items in the information storage unit 101 of the improvement point analysis device 100 .
The information model and setting items are desirably generated based on the design information of the production system.

Information Collecting Phase The information collecting unit 104 collects information from the production system, which is the analysis target 200, based on the setting items in FIG. Then, the collected information is stored in the information storage unit 101 .
Note that the information collecting unit 104 may acquire information collected by an external device other than the improvement point analysis device 100 from the external device, and store the information acquired from the external device in the information storage unit 101 in association with the setting items. good.

Evaluation Phase Next, as an evaluation phase, the evaluation unit 102 performs the flowchart shown in FIG.
Specifically, the evaluation unit 102 reads the setting items shown in FIG. 4 from the information storage unit 101 (step S101).
Next, the evaluation unit 102 reads the information collected by the information collection unit 104 from the information storage unit 101 (step S102).
Next, the information collecting unit 104 evaluates the performance of each element using the set items and the collected information, and stores the evaluation result in the information storage unit 101 (step S103). In other words, the information collecting unit 104 determines whether or not the information collected by the information collecting unit 104 matches the determination criteria for each element indicated by the setting item in units of monitoring periods. Note that when the "collection target" is defined as "calculation" such as "product A production lead time", the evaluation unit 102 performs calculation based on the information collected by the information collection unit 104, and the calculation Compare the obtained value with a criterion.

6 and 7 show examples of evaluation results by the evaluation unit 102. FIG. FIG. 6 shows the evaluation results in graph form, and FIG. 7 shows the evaluation results in tabular form.
For example, "product A production lead time" has threshold value X1 as the criterion. The evaluation unit 102 evaluates the performance (required time) of "product A production lead time" to be HIGH (binary value of 1) when the performance (required time) exceeds the threshold value X1 during a certain monitoring period. Moreover, the evaluation unit 102 evaluates as LOW (0 in binary value) when the performance (required time) of the "product A production lead time" is equal to or less than the threshold value X1 in another monitoring period. In this embodiment, "product A production lead time" is set to KGI, so if the performance (required time) of "product A production lead time" is higher than the criterion (=HIGH), productivity will decrease. means that
Note that the evaluation unit 102 may sequentially output evaluation results to an HMI (Human Machine Interface). With this configuration, the person in charge of improvement of the production system can understand the situation of the production system in real time by referring to the HMI, and can judge the necessity of improvement.

Improvement Point Analysis Phase The person in charge of improvement of the production system refers to the evaluation result by the evaluation unit 102 and judges whether or not improvement is necessary based on the evaluation status of the evaluation unit 102 for the KGI. For example, as shown in FIGS. 6 and 7, if the "product A production lead time" is higher than the criterion, that is, if there are multiple periods in which productivity is declining, the person in charge of improving the production system should: Decide that improvement is necessary.
When it is determined that improvement is necessary, the person in charge of improvement of the production system instructs the improvement point analysis device 100 to perform analysis by the improvement point analysis unit 103 .
The improvement point analysis unit 103 executes the flow of FIG. 8 based on instructions from the person in charge of improvement of the production system.
Here, an example is shown in which the improvement point analysis unit 103 executes the flow of FIG. 8 based on instructions from the person in charge of improvement of the production system. may Alternatively, the improvement point analysis unit 103 may execute the flow of FIG. 8 when a specified number of evaluation results by the evaluation unit 102 are accumulated. Furthermore, the improvement part analysis unit 103 may refer to the evaluation result by the evaluation unit 102 to determine whether or not improvement is necessary, and when it is determined that improvement is necessary, the flow of FIG. 8 may be executed.

As shown in FIG. 8, the improvement point analysis unit 103 reads the information model shown in FIG. 3 from the information storage unit 101 (step S201).
Further, the improvement point analysis unit 103 reads the evaluation result by the evaluation unit 102 from the information storage unit 101 (step S202).
The order of step S201 and step S202 may be changed, or step S201 and step S202 may be performed in parallel.

Next, the improvement point analysis unit 103 designates a designated element to be analyzed, and sets or reads analysis conditions (step S203).
The improvement point analysis unit 103 may specify the specified elements according to the instruction of the person in charge of improvement of the production system, or may specify the specified elements by analyzing the evaluation result by the evaluation unit 102 . For example, the improvement point analysis unit 103 designates, as the designated element, the element of the highest layer among the elements whose performance does not meet the criteria for a plurality of times.
Details of the analysis conditions will be described later.

Next, the improvement point analysis unit 103 performs analysis and stores the analysis result in the information storage unit 101 (step S204).
If the analysis is to be continued further (YES in step S205), the process returns to step S203, and the improvement point analysis unit 103 designates a new designated element. The improvement part analysis unit 103 designates, for example, the improvement target related element obtained in step S204 as a new designated element.

FIG. 9 shows details of step S204. The details of step S204 will be described below with reference to FIG.

First, the improvement part analysis unit 103 extracts a significant combination of the conclusion part and the condition part in the association analysis from the information model (step S2041).
A significant combination means that the concluding part and the conditional part are not inconsistent from the viewpoint of both the relationship based on the hierarchical structure and the relationship based on the logical structure in the information model. In other words, a significant combination means that the concluding part and the conditional part are connected in the correct direction in the relationship based on the hierarchical structure, or in the correct direction in the relationship based on the logical structure. Elements that are connected in the correct direction in a hierarchical relationship with the conclusion element and elements that are connected in the correct direction in the logical structure relationship with the conclusion element do not affect the performance of the conclusion element. element that can be given.
Specifically, the improvement point analysis unit 103 sets the specified element specified in step S203 to the conclusion part. Then, based on the information model, the improvement part analysis unit 103 identifies elements that are connected in the correct direction in the relationship between the specified element and the hierarchical structure, and elements that are connected in the correct direction in the relationship between the specified element and the logical structure. Set in the conditional part. Elements set in the condition part correspond to related elements.
For example, when the improvement point analysis unit 103 designates "product A production lead time" as a designated element in step S203, the improvement point analysis unit 103 determines that the relationship between "product A production lead time" and the hierarchical structure is in the correct direction. "process #1 lead time", "process #2 lead time" and "process #3 lead time" connected by are extracted as related elements.
In this case, as shown in FIG. 10, the improvement point analysis unit 103 sets "the value of the product A production lead time is higher than the criterion" in the conclusion part, and sets "process #1 lead time", ""process#2 lead time" and "process #3 lead time" are set in the condition part.
Further, for example, when the improvement point analysis unit 103 designates "process #1 lead time" as a designated element in step S203, the relationship between "process #2 lead time" and "process #3 lead time" is based on the logical structure. and does not result in a significant combination (even if there is a change in the lead time of process #2 or #3, the lead time of process #1 does not change). Therefore, the improvement point analysis unit 103 does not extract "process #2 lead time" and "process #3 lead time" as related elements.

Next, the improvement part analysis unit 103 performs association analysis on the combination of the extracted condition part and conclusion part (step S2042).
Here, the improvement point analysis unit 103 is assumed to calculate the support, reliability and lift value in the association analysis. The calculation order of these calculation items is not specified.
The improvement point analysis unit 103 may refer to the analysis conditions and stop the calculation of the uncalculated calculation items when the calculation maintenance condition is not satisfied for any calculation item. By configuring in this way, the amount of calculation for association analysis can be reduced.
FIG. 11 shows an example of analysis conditions.
In the analysis conditions of FIG. 11, the interval (time width) to be analyzed and the calculation maintenance conditions of each calculation item of the association analysis are defined.
In addition, in the example of FIG. 11, the section condition is a constant section regardless of the hierarchy. Elements in higher layers are generally collected at longer intervals, and there may be a certain delay until changes in elements in lower layers are propagated to higher layers. In such a case, it is preferable to set the interval (time width) to be analyzed by shifting the element of the upper layer and the element of the lower layer.
The example of FIG. 11 also shows that the lift value calculation maintenance condition is 1 or more, the support calculation maintenance condition is greater than 0.001, and the reliability calculation maintenance condition is greater than 0.001.

FIG. 12 shows an example of an analysis result by the improvement point analysis unit 103. As shown in FIG.
Note that in FIG. 12, the combinations in which the support, confidence, and lift values are all shaded out are not significant combinations and thus were not subjected to association analysis.

The upper part of FIG. 12 shows the result obtained by analyzing the influence of the lead time of each process on the product A production lead time by the improvement point analysis unit 103 . As described above, the improvement point analysis unit 103 extracts and analyzes only significant combinations. Therefore, in the upper part of FIG. 12, "product A production lead time" is set in the conclusion part, and each of "process #1 lead time", "process #2 lead time" and "process #3 lead time" is set as a condition. Shown are the results of an association analysis that was performed using a set of parts.
In the example in the upper part of FIG. 12, the association analysis is performed with the case where the "process #1 lead time" is higher than the criterion (=HIGH) as the conditional part, and the case where the "process #2 lead time" is higher than the criterion (=HIGH). is the conditional part of the association analysis, the lift value is greater than 1. Therefore, there is a high possibility that the "process #1 lead time" and the "process #2 lead time" are the factors that cause the deterioration of the "product A production lead time". Therefore, the improvement point analysis unit 103 estimates the “process #1 lead time” and the “process #2 lead time” as related elements to be improved. In other words, the improvement point analysis unit 103 extracts the “process #1 lead time” and the “process #2 lead time” as elements whose performance should be improved in order to improve the performance of the “product A production lead time”.

The lower part of FIG. 12 shows that the improvement point analysis unit 103 analyzes the influence of each of "process #1 lead time", "process #2 lead time" and "process #3 lead time" on the lead times of other processes. and the results obtained.
In the upper part of FIG. 12, since there are a plurality of combinations with lift values greater than 1, the improvement point analysis unit 103 also analyzes the influence of the lead time of each process on the lead time of other processes. That is, the improvement point analysis unit 103 designates each process as a new designated element, extracts processes other than the process designated as the new designated element as a new related element, and performs association analysis.
In the example in the lower part of FIG. 12, the association is set in the condition part when the "process #1 lead time" is higher than the criterion (=HIGH), and in the conclusion part when the "process #2 lead time" is higher. The lift value is greater than 1 in the analysis. Therefore, there is a high possibility that the "process #1 lead time" is the cause of the deterioration of the "process #2 lead time". Therefore, the improvement point analysis unit 103 extracts the "process #1 lead time" as an element whose performance should be improved in order to improve the performance of the "process #2 lead time".
As a result, the improvement point analysis unit 103 estimates the "process #1 lead time" as the top priority improvement target related element. That is, the improvement point analysis unit 103 determines that the "process #1 lead time" is the most effective improvement target related element for improving the performance of the "product A production lead time", and determines that the "process #2 lead time ” is the next effective improvement target related element.
When a plurality of improvement target related elements are obtained as in the example in the upper part of FIG. 12, the improvement part analysis unit 103 assigns priorities among the plurality of improvement target related elements as in the example in the lower part of FIG. set.

From the above analysis results, the improvement point analysis unit 103 can estimate that "the deterioration of the production lead time of the product A is caused by the deterioration of the process #1 lead time and the deterioration of the process #2 lead time". Further, the improvement point analysis unit 103 can estimate that "the worsening of the process #2 lead time is caused by the worsening of the process #1 lead time". For this reason, the improvement point analysis unit 103 can propose an improvement point to the person in charge of improvement of the production system, saying, "First, the process #1 lead time, which is the cause of the deterioration of both lead times, should be improved." can.
The improvement point analysis unit 103 outputs such an analysis result (improvement point proposal) to the HMI, and stores the analysis result in the information storage unit 101 (step S2043).

Furthermore, the improvement point analysis unit 103 can perform a deeper analysis of factors that cause deterioration of the "process #1 lead time" as necessary (step S205 in FIG. 8).
Specifically, based on the relationship of the hierarchical structure of the information model in FIG. 1 lead time” and “equipment #1-2 lead time” are set in the condition part as new related elements to perform the association analysis.
After the analysis of the equipment hierarchy is completed, the improvement point analysis unit 103 may further perform association analysis in which the element of the equipment hierarchy is set as a new related element in the conditional part as necessary. In this manner, the improvement point analysis unit 103 can perform a deeper analysis. As an analysis condition, the number of repetitions of such deep analysis may be provided.

***Description of the effect of the embodiment***
As described above, according to the present embodiment, it is possible to efficiently specify the elements that contribute to the improvement of the performance of the elements whose performance should be improved.
In other words, the improvement point analysis apparatus 100 according to the present embodiment analyzes the relationship between the KGI and a plurality of pieces of information in a stepwise and logical manner while referring to the relationship based on the hierarchical structure and logical structure defined in the information model. For this reason, according to the present embodiment, it is possible to efficiently specify an improvement location that contributes to the improvement of KGI.

In addition, according to the present embodiment, the evaluation unit 102 sequentially outputs the evaluation results to the HMI, so that the person in charge of improvement of the production system can understand the situation of the production system in real time and immediately judge whether or not improvement is necessary. be able to.

Further, according to the present embodiment, the improvement point analysis unit 103 refers to the analysis conditions, and if any of the calculation items does not meet the calculation maintenance conditions, the uncalculated calculation items are not calculated. The amount of time calculation can be reduced.

Further, according to the present embodiment, by setting the number of repetitions for deepening the analysis by the improvement point analysis unit 103, the person in charge of improvement of the production system does not have to wait for the final result of the improvement point analysis device 100. Analysis results can be obtained within the range of experience, and improvement activities can be made more efficient.

Further, according to the present embodiment, the improvement point analysis unit 103 can adjust the analysis section according to the hierarchy when the analysis target is in a different hierarchy, so that the analysis can be performed effectively.

Embodiment 2.
***Purpose***
In Embodiment 1, evaluation results obtained by evaluation based on a single criterion are binarized. In the method of Embodiment 1, especially when trying to improve the device hierarchy in the information model, there is a possibility that it will not be possible to clearly analyze which values of the device hierarchy elements will affect the performance of the specified elements. be.
The main object of the second embodiment is to solve such problems.
In this embodiment, differences from the first embodiment will be mainly described.
Matters not described below are the same as those in the first embodiment.

*** Configuration description ***
Also in the present embodiment, the configuration of the improved portion analysis apparatus 100 is as shown in FIGS. 1 and 2. FIG.
In the present embodiment, as shown in FIG. 13, additional determination criteria are provided to the setting items (FIG. 4) for information collection shown in the first embodiment. Furthermore, as shown in FIG. 14, the analysis conditions (FIG. 11) are provided with subdivision conditions.
In FIG. 13, for "sensor #1-1-3 target angle" among the setting items for information collection, threshold X6 is set as the first determination criterion, and threshold values X7 and X8 are set as additional determination criteria. is doing. Furthermore, in FIG. 14, as conditions for subdivision, the case where the lift value is greater than 2 and the case where the degree of support is greater than 0.5 are set.

***Description of operation***
In this embodiment, the same operation as in the first embodiment is performed until the improvement point analysis phase is completed.
At this time, when an analysis result as shown in FIG. 12 is obtained, and further, when the conditions for subdividing the analysis conditions in FIG. conduct.
FIG. 15 shows an example of the effects of the second embodiment.
In other words, in the present embodiment, the improvement point analysis unit 103 sets the “equipment #1-1 lead time” to the conclusion part and the “sensor #1-1-3 target angle” to the condition part to perform the association analysis. there is Then, the improvement portion analysis unit 103 estimates the “sensor #1-1-3 target angle” as the improvement target related element. Then, the result of the association analysis satisfies the subdivision conditions (FIG. 14) described above. For this reason, the evaluation unit 102 subdivides whether the performance of the improvement target related element "sensor #1-1-3 target angle" matches the additional determination criteria 2 (thresholds X7 and X8). We are evaluating whether or not FIG. 15 shows evaluation results obtained by evaluation using criterion 2. FIG.
Although not shown in FIG. 15, it is also possible to show the results (lift value, degree of support, degree of reliability) newly analyzed using criterion 2 by segmentation. In this case, when the “sensor #1-1-3 target angle” becomes a specific value, the “sensor #1-1-3 target angle” is the performance of “equipment #1-1 lead time” It can be analyzed whether it affects deterioration.

***Description of the effect of the embodiment***
As described above, in the present embodiment, a plurality of determination criteria are provided, and subdivision conditions are provided for the evaluation unit 102 to apply the plurality of determination criteria. For this reason, according to the present embodiment, the evaluation result based on a single criterion can be subdivided by a plurality of criteria, and the element that can affect the performance of the designated element can be specified with any specific value. It is possible to clearly determine whether it will affect the performance of the specified element when it becomes.

Embodiment 3.
***Purpose***
In Embodiments 1 and 2, it is assumed that the criteria for setting items in the information collecting unit 104 and the analysis conditions in the improvement point analyzing unit 103 are set in advance with reference to design information or the like.
However, if the design information cannot be obtained or if there is a difference between the design information and the actual state of the production system, the effects of the first and second embodiments may not be obtained.
The main object of the third embodiment is to solve such problems.
In this embodiment, differences from the first embodiment will be mainly described.
Matters not described below are the same as those in the first embodiment.

*** Configuration description ***
Also in the present embodiment, the configuration of the improved portion analysis apparatus 100 is as shown in FIGS. 1 and 2. FIG. However, in this embodiment, the method of setting the determination criteria of the information collecting unit 104 is as shown in FIG.
In FIG. 16 , statistical processing for collected information is defined as a determination criterion in the setting items of the information collection unit 104 . Statistical processing is set, for example, as "average value". In addition to this, an average value with standard deviation taken into consideration (average value±standard deviation), a mode value in a frequency distribution, a local minimum value, etc. are also suitable, but are not limited to these.

***Description of operation***
In this embodiment, the same operation as in the first embodiment is performed until the middle of the information collection and evaluation phases.
In this embodiment, in the flowchart of FIG. 5, the evaluation unit 102 performs statistical processing on the information collected by the information collection unit 104 and stored in the information storage unit 101 in step S103. Then, the evaluation unit 102 sets the result of the statistical processing as the criterion of FIG. 16 . The evaluation unit 102 performs evaluation by comparing each piece of information collected by the information collection unit 104 and stored in the information storage unit 101 with the result of statistical processing set as a criterion. After that, the evaluation unit 102 stores the evaluation result in the information storage unit 101 .
After that, in the improvement point analysis phase, the same operation as in the first embodiment is performed.

***Description of the effect of the embodiment***
As described above, in the present embodiment, the result of statistical processing is set as the criterion, so that the effects of the first and second embodiments can be obtained even when the design value of the analysis object cannot be obtained.
In addition, by setting the result of statistical processing as the determination criterion, it is possible to reduce the trouble of having to set the determination criterion in advance.

Embodiment 4.
***Purpose***
Embodiments 1 to 3 are premised on defining relationships based on a hierarchical structure and/or a logical structure in an information model.
However, especially in the device hierarchy, it is difficult to strictly define device information because the designer differs from the designer of the production system. Therefore, it may not be possible to accurately define hierarchical and/or logical relationships. Incorrect analysis results may be output if the relationship is defined incorrectly or if the relationship cannot be defined. Also, analysis for all possible combinations may occur. In such cases, efficiency is reduced.
The main object of the fourth embodiment is to solve such problems.
In this embodiment, differences from the first embodiment will be mainly described.
Matters not described below are the same as those in the first embodiment.

*** Configuration description ***
Also in the present embodiment, the configuration of the improved portion analysis apparatus 100 is as shown in FIGS. 1 and 2. FIG. In this embodiment, the configuration of the information model and the analysis conditions and flow chart of the improvement part analysis unit 103 are different.
FIG. 17 shows an example of an information model according to the fourth embodiment. In FIG. 17, relationships based on hierarchical structure and logical structure are not defined in the device hierarchy. In other words, in the information model shown in FIG. 17, only elements existing in the device hierarchy are defined.
FIG. 18 shows additional items in analysis conditions in the fourth embodiment. In FIG. 18, items related to relationships are added to the analysis conditions. As a condition regarding the relationship, for example, it is conceivable to define "automatically generate the relationship of the device hierarchy". It may be defined to automatically generate relationships in layers other than the device layer, or it may be defined to generate relationships by a method other than automatic generation. Furthermore, it may be defined to periodically check the relationship once created.
FIG. 19 shows a procedure additionally performed by the improvement portion analysis unit 103 in step S204 (execution and storage of analysis) in FIG.

***Description of operation***
The same operations as in the first embodiment are performed in the information gathering phase, the evaluation phase, and step S203 in FIG.
In step S204, prior to the flow of FIG. 9, the flow of FIG. 19 is performed. That is, when the relationship between any of the elements is unknown, the improvement part analysis unit 103 estimates the relationship between the elements whose relationship is unknown, and automatically generates the relationship between the elements.
Specifically, the improvement point analysis unit 103 refers to the setting items in FIG. 18 and executes the flow in FIG. 19 to automatically generate the relationship for the device hierarchy.
In FIG. 19, among the elements included in the device hierarchy, no particular relationship is defined with "PLC#1-1-1 cycle time", which defines a hierarchical structure with the equipment hierarchy, which is the upper layer. It shows the procedure for automatically generating the relationship between "servo #x-1 motor current value".
First, in step S301, the improvement point analysis unit 103 performs association analysis on "PLC #1-1-1 cycle time" and "servo #x-1 motor current value". In the example of FIG. 19, the improvement point analysis unit 103 sets in the condition part that "PLC#1-1-1 cycle time" is higher than the criterion (=HIGH), and "servo #x-1 motor current The support, reliability, and lift value of the combination (hereinafter referred to as "combination 1") whose conclusion is set that the "value" is higher (=HIGH) than the criterion is calculated. Furthermore, the improvement point analysis unit 103 sets the condition part that the “servo #x-1 motor current value” is higher than the judgment reference (=HIGH), and the “PLC #1-1-1 cycle time” is judged. The support, reliability, and lift value of a combination (hereinafter referred to as “combination 2”) whose conclusion is set to be higher than the standard (=HIGH) are calculated. If the meaning of the information is unknown, it is desirable that the improvement part analysis unit 103 performs association analysis including combinations where the value of each element is lower than the criterion (=LOW).
Next, in step S302, the improvement point analysis unit 103 evaluates the analysis result of step S301. Here, the improvement point analysis unit 103 evaluates which combination yields the highest support, reliability, and lift value of the analysis results. In other words, the improvement point analysis unit 103 evaluates which of combination 1 and combination 2 yields a higher value. In association analysis, if the evaluation result of the lift value is greater than 1, it can be evaluated that there is a relationship between the condition part and the conclusion part. For this reason, the improvement point analysis unit 103 may perform evaluation by assigning weights to the results of each item so as to emphasize the lift value. These weights may be provided in the setting items in FIG. Also, the conditions for generating relationships may be provided in the setting items in FIG. 18 as well.
Finally, in step S<b>303 , the improvement point analysis unit 103 refers to the evaluation result to construct the relationship between the elements, and stores the constructed relationship between the elements in the information storage unit 101 .
In the example of FIG. 19, it is set in the condition part that "servo #x-1 motor current value" is higher than the criterion, and the conclusion part is set that "PLC#1-1-1 cycle time" is higher than the criterion. In the combination (combination 1) set to , the support, reliability, and lift value are high. Therefore, the improvement point analysis unit 103 adds to the information model the relationship that the "servo #x-1 motor current value" affects the performance of the "PLC #1-1-1 cycle time".
In this example, the procedure for automatically generating the relationship between "PLC #1-1-1 cycle time" and "servo #x-1 motor current value" is shown. For the "sensor #x-2 target angle" and "robot #x-3 reach rate" shown in FIG. 17, the relationship is automatically generated by the same procedure. do.

***Description of the effect of the embodiment***
As described above, in the present embodiment, when the relationship between any of the elements is unknown, the improvement part analysis unit 103 estimates the relationship between the elements whose relationship is unknown, and determines the relationship between the elements. Generate automatically. For this reason, according to the present embodiment, even when the relationship based on the hierarchical structure and/or the logical structure cannot be accurately defined in the information model, it is possible to construct the relationship between the elements based on the actually collected information. can be used, and the analysis can be performed efficiently.
Also, even if the relationship between elements is defined incorrectly, the information model can be corrected by evaluating the relationship between the elements based on the actually collected information.

Embodiment 5.
***Purpose***
In Embodiments 1 to 4, analysis results are output in multiple items such as support, reliability, and lift value. Elements to be improved need to be determined by considering the output of these multiple items. However, there is a problem that it is difficult for a production system manager who is particularly unfamiliar with analysis to determine which element is an improvement point.
The main purpose of the fifth embodiment is to solve such problems.
In this embodiment, differences from the first embodiment will be mainly described.
Matters not described below are the same as those in the first embodiment.

*** Configuration description ***
Also in the present embodiment, the configuration of the improved portion analysis apparatus 100 is as shown in FIGS. 1 and 2. FIG. In this embodiment, the analysis conditions of the improvement point analysis unit 103 are different.
FIG. 20 shows analysis conditions according to the fifth embodiment. In FIG. 20, compared with FIG. 11, an analysis result output item is newly added. For example, a calculation formula “A1+A2+A3” is defined as a condition for outputting analysis results. "A1" is set to 1 when the degree of support is greater than 0.1, and 0 when the degree of support is less than 0.1. Similarly, "A2" is set to 1 when the reliability is greater than 0.1, and set to 0 when the reliability is less than 0.1. For "A3", the lift value is set as it is when the lift value is 1 or more, and 0 is set when the lift value is less than 1.
Note that the calculation formula may be adjusted so as to emphasize any one of the support, reliability, and lift value. For example, the weight α is used to adjust A1+A2+A3×α. The weight α is set by the production system designer or the person in charge of improvement.
In the present embodiment, as shown in FIG. 20, the improvement point analysis unit 103 performs calculation using a plurality of calculated values for a plurality of calculation items included in the association analysis. output the calculated value of

***Description of operation***
The same operations as in the first embodiment are performed up to step S2042 in FIG. 9 in the information collection phase, evaluation phase, and improvement point analysis phase.
In step S2043, the improvement point analysis unit 103 calculates the analysis result output according to the analysis result output calculation formula of FIG.
FIG. 21 shows an example of output according to the fifth embodiment. As shown in FIG. 21, the output according to the present embodiment includes analysis results obtained according to the analysis result output calculation formula of FIG.

***Description of the effect of the embodiment***
As described above, in the present embodiment, the analysis conditions define a calculation method for outputting analysis results, and the analysis results calculated according to the calculation method are output. Therefore, according to the present embodiment, a production system manager who is unfamiliar with analysis can easily determine which element is to be improved.

Embodiment 6.
***Purpose***
In the fifth embodiment, it is necessary for the production system designer or the person in charge of improvement to appropriately set the calculation formula for outputting the analysis result of the improvement portion. However, especially in a complicated production system, there is a possibility that the analysis result output and the actual improvement points are different.
The sixth embodiment is mainly intended to solve such problems.
In this embodiment, differences from the fifth embodiment will be mainly described.
Matters not described below are the same as in the fifth embodiment.

*** Configuration description ***
FIG. 22 shows a configuration example of an improved portion analysis device 100 according to the sixth embodiment. In the configuration of FIG. 22, an improvement result storage unit 105 is added to the configuration of FIG. The improvement record storage unit 105 stores the improvement record. The components other than the improvement performance storage unit 105 are the same as those shown in FIG.
FIG. 23 shows an example of improvement results stored in the improvement result storage unit 105. As shown in FIG. The improvement record shown in FIG. 23 includes the event that occurred and the cause of the event (improvement point). Events and factors should be described in detail, including specific values. Improvement achievements are described by the person in charge of improvement while the analysis target is in operation. Alternatively, if there is an improvement record for the same type of analysis target, the improvement record for the same type of analysis target may be used. However, in this case, it is desirable to describe it so that it can be understood that it is the improvement performance of another analysis target. Alternatively, an evaluation result of correctness or wrongness given to the output of the analysis result through the HMI by the person in charge of improvement who referred to the output of the analysis result shown in Embodiment 5 or an evaluation result accompanied by a numerical value may be used as the improvement result. At this time, it is desirable that the relationship between the contents described in the improvement result storage unit 105 is defined in the information model. If the relationship is not defined in the information model, the relationship may be defined in the information model.
24 shows an example of analysis conditions according to Embodiment 6. FIG. In FIG. 24, as a condition for outputting the analysis result, a calculation formula for reflecting the improvement performance in the output is defined. In FIG. 24, the actual number of improvements is described as the correction value "A4", and "A4" is also included in the calculation formula.

***Description of operation***
The same operations as in the first embodiment are performed up to step S2042 in FIG. 9 in the information collection phase, evaluation phase, and improvement point analysis phase.
In step S2043, the improvement point analysis unit 103 calculates the analysis result output according to the analysis result output calculation formula of FIG. That is, in step S2043, the improvement part analysis unit 103 corrects the calculated value obtained by the calculation formula "A1+A2+A3" using the improvement results for the related element corresponding to the conditional part in the association analysis, and obtains the calculated value after correction. Output.
FIG. 25 shows an example of analysis result output according to the sixth embodiment. As shown in FIG. 25, in the present embodiment, the improvement results in FIG. 23 are output as correction values, and the analysis result output after correction based on the number of improvement results is output.

***Description of the effect of the embodiment***
As described above, in the present embodiment, the improvement results are stored, and the analysis result output is corrected according to the improvement results. For this reason, according to the present embodiment, it is possible to output improvement points that are in line with the actual state of the production system.

Embodiment 7.
***Purpose***
In Embodiments 5 and 6, analysis result output is calculated using a calculation formula, and the obtained analysis result output is output, thereby obtaining an analysis result with higher accuracy.
However, in the case of a large-scale or complicated production system, the types of elements are enormous and the relationships between the elements are complicated. For this reason, it may be difficult to define a calculation formula so as to produce an analysis result output that is in line with the actual state of the object to be analyzed.
Embodiment 7 is mainly intended to solve such problems.
In this embodiment, differences from the fifth embodiment will be mainly described.
Matters not described below are the same as in the fifth embodiment.

*** Configuration description ***
FIG. 26 shows setting of analysis conditions according to the seventh embodiment. In FIG. 26, it is set that the analysis result output in FIG. 20 or FIG. 24 is calculated using a learned model learned using machine learning, which will be described later.
FIG. 27 shows a configuration example of a machine learning device 400 utilized by the improvement point analysis device 100. As shown in FIG. The machine learning device 400 includes a data acquisition unit 401 , a teacher data acquisition unit 402 , a learning unit 403 , a trained model storage unit 405 and an output unit 404 . The machine learning device 400 includes a processor, a storage device, a communication interface, and a bus as hardware configuration, as in FIG. 2 . The data acquisition unit 401, the teacher data acquisition unit 402, and the learning unit 403 are implemented by, for example, programs. The program is executed by a processor. A trained model storage unit 405 is realized by a storage device.

The data acquisition unit 401 acquires the condition part, the conclusion part, the degree of support, the degree of reliability, and the lift value shown in FIGS. 10, 11, 20, and 24 as state variables. Note that the data acquisition unit 401 may acquire the evaluation result of the evaluation unit 102 shown in FIGS. 6 and 7. FIG.
The teacher data acquisition unit 402 acquires the factors and events shown in the improvement results in FIG. 23 .
The learning unit 403 creates a data set based on the combination of the condition part, the conclusion part, the support, the reliability, and the lift value output from the data acquisition unit 401 and the factors and events output from the teacher data acquisition unit 402. Learn how to correct the output based on That is, the learning unit 403 analyzes from the condition part, the conclusion part, the degree of support, the degree of reliability, the lift value, which are the analysis results of the improvement point analysis unit 103 of the improvement point analysis device 100, and the factors and events that are the actual improvement results. Generate a trained model that infers how to correct the result output. Here, the data set is data in which state variables and teacher data are associated with each other.

The machine learning device 400 is used to learn a correction method for the output of the improvement point analysis device 100. For example, the improvement point analysis device 100 connected to the improvement point analysis device 100 via a network and the may be a separate device. Further, the machine learning device 400 may be built in the improvement point analysis device 100 . Furthermore, machine learning device 400 may reside on a cloud server.

Any learning algorithm may be used by the learning unit 403 . In this embodiment, as an example, a case where a neural network is applied will be described.
The learning unit 403 learns an output correction method by, for example, so-called supervised learning according to a neural network model. Here, supervised learning is a model in which a large amount of data sets of certain inputs and results (labels) are given to the machine learning device 400 to learn features in those data sets and estimate results from the inputs. Say.
A neural network consists of an input layer consisting of a plurality of neurons, an intermediate layer (hidden layer) consisting of a plurality of neurons, and an output layer consisting of a plurality of neurons. The intermediate layer may be one layer, or two or more layers.
For example, in a three-layer neural network as shown in FIG. 30, when a plurality of input data are input to the input layer (X1-X3), each input data value is multiplied by a weight W1 (w11-w16). Each input data multiplied by the weight W1 is input to the intermediate layer (Y1-Y2). Then, the result of the intermediate layer (Y1-Y2) is further multiplied by the weight W2 (w21-w26), and the result of the intermediate layer (Y1-Y2) multiplied by the weight W2 is output from the output layer (Z1-Z3). . The output result varies depending on the value of weight W1 and the value of weight W2.
In the present application, the neural network is created based on the combination of the condition part, the conclusion part, the support, the reliability, and the lift value obtained by the data obtaining unit 401, and the factors and events obtained by the teacher data obtaining unit 402. The output correction method is learned by so-called supervised learning according to the data set obtained.
That is, the neural network adjusts the weights W1 and W2 so that the results output from the output layer by inputting the condition part, the conclusion part, the support, the confidence, and the lift value to the input layer are closer to the factors and events. Learn by doing.
Neural networks can also learn how to correct their outputs by so-called unsupervised learning. In unsupervised learning, a large amount of only input data is given to the machine learning device 400 so that the machine learning device 400 learns how the input data are distributed. In unsupervised learning, it is possible to learn by compressing, classifying, shaping, etc. input data without giving corresponding teacher output data. In other words, unsupervised learning can cluster similar features in multiple datasets. Using the results of clustering, output prediction can be realized by setting some criteria and assigning outputs so as to optimize the clustering results. Semi-supervised learning is an intermediate problem setting between unsupervised learning and supervised learning. In semi-supervised learning, there are pairs of input data and output data only in some cases, and only input data in other cases.

The learning unit 403 generates a trained model by executing the above learning.
A trained model storage unit 405 stores the trained model generated by the learning unit 403 .
The output unit 404 outputs a method for correcting the analysis result output of the improvement point analysis device 100, which is obtained using the learned model. That is, by inputting the condition part, the conclusion part, the support, the reliability, and the lift value to the data acquisition unit 401, the condition part, the conclusion part, the support, the reliability, and the lift value are obtained from the output unit 404 based on the learned model. A method of correcting the output suitable for the value can be obtained.
In this embodiment, the output unit 404 of the machine learning device 400 uses the learned model obtained by the learning unit 403 to output the analysis result output correction method to the improvement location analysis device 100. However, the improvement point analysis apparatus 100 may acquire a learned model and acquire a correction method for the analysis result output based on this learned model.

***Description of operation***
Next, the processing that the machine learning device 400 learns will be described with reference to FIG. 28 . FIG. 28 is a flowchart relating to learning processing of machine learning device 400 .
First, in step S401, the data acquisition unit 401 acquires the condition part, conclusion part, support, reliability, and lift value as state variables.
Next, in step S402, the training data acquisition unit 402 acquires factors and events that are improvement results. In the present embodiment, the data are acquired in the order described above, but it is sufficient to input the condition part, the conclusion part, the support, the reliability, the lift value, and the factors and events in association with each other. They may be executed simultaneously or in reverse order.
Furthermore, in step S403, the learning unit 403 acquires the combination of the condition part, the conclusion part, the support, the reliability, and the lift value acquired by the data acquisition unit 401, and the factors and events acquired by the teacher data acquisition unit 402. According to the data set created based on, by so-called supervised learning, a correction method for analysis result output is learned, and a trained model is generated.
Finally, in step S<b>404 , the learned model storage unit 405 stores the learned model generated by the learning unit 403 .

Next, with reference to FIG. 29, a process for obtaining a correction method for analysis result output using the machine learning device 400 will be described.
First, in step S501, the data acquisition unit 401 acquires the condition part, the conclusion part, the support, the reliability, and the lift value.
Next, in step S502, the learning unit 403 inputs the condition part, the conclusion part, the degree of support, the degree of reliability, and the lift value to the learned model stored in the learned model storage unit 405, and selects a correction method for the analysis result output. obtain. The learning unit 403 outputs the obtained analysis result output correction method to the output unit 404 .
Furthermore, in step S503, the output unit 404 outputs a correction method for the analysis result output obtained by the learned model.
Finally, in step S504, the improvement point analysis unit of the improvement point analysis apparatus 100 corrects the analysis result using the method for correcting the outputted analysis result output, and outputs the corrected analysis result. As a result, it is possible to output improvement points that are in line with the actual state of the production system.

In this embodiment, a case where supervised learning is applied to the learning algorithm used by learning unit 403 has been described, but the present invention is not limited to this. In addition to supervised learning, it is also possible to apply reinforcement learning, unsupervised learning, semi-supervised learning, and the like as learning algorithms.
Also, the learning unit 403 may learn an output correction method according to data sets collected from a plurality of improvement point analysis devices 100 .
Note that the learning unit 403 may acquire data sets from a plurality of improvement point analysis devices 100 used in the same area. Alternatively, the learning unit 403 may learn the analysis result output correction method using data sets collected from a plurality of improvement point analysis devices 100 operating independently in different areas. Furthermore, the learning unit 403 can also add the improvement point analysis device 100 that collects data sets in the middle. Alternatively, the learning unit 403 can conversely remove any of the improvement point analysis devices 100 from the improvement point analysis devices 100 that collect data sets.
Furthermore, the machine learning device 400 that has learned the method of correcting the analysis result output for a certain improvement point analysis device 100 is applied to another improvement point analysis device 100, and the analysis result is output for the other improvement point analysis device 100. may be re-learned and updated.
In addition, as a learning algorithm used in the learning unit 403, deep learning (Deep Learning) that learns to extract the feature amount itself can be used, and other known methods such as genetic programming, functional logic programming, support Machine learning may be performed according to vector machines and the like.
***Description of the effect of the embodiment***
As described above, in the present embodiment, machine learning is utilized to obtain a correction method for analysis result output. Therefore, according to the present embodiment, even if the analysis target is a large-scale or complicated production system, it is possible to output improvement points that are in line with the actual state of the production system.

Embodiments 1 to 7 have been described above, but two or more of these embodiments may be combined for implementation.
Alternatively, one of these embodiments may be partially implemented.
Alternatively, two or more of these embodiments may be partially combined for implementation.
Also, the configurations and procedures described in these embodiments may be changed as necessary.

*** Supplementary explanation of hardware configuration ***
Finally, a supplementary description of the hardware configuration of the improvement point analysis device 100 will be given.
A processor 901 shown in FIG. 2 is an IC (Integrated Circuit) that performs processing.
The processor 901 is a CPU (Central Processing Unit), a DSP (Digital Signal Processor), or the like.
The storage device 902 shown in FIG. 2 is RAM (Random Access Memory), ROM (Read Only Memory), flash memory, HDD (Hard Disk Drive), or the like.
The communication interface 903 shown in FIG. 2 is an electronic circuit that executes data communication processing.
The communication interface 903 is, for example, a communication chip or a NIC (Network Interface Card).

The auxiliary storage device 902 also stores an OS (Operating System).
At least part of the OS is executed by the processor 901 .
The processor 901 executes a program 904 while executing at least part of the OS.
Task management, memory management, file management, communication control, and the like are performed by the processor 901 executing the OS.
In addition, at least one of information, data, signal values, and variable values indicating the processing results of the evaluation unit 102, the improvement part analysis unit 103, and the information collection unit 104 is stored in the storage device 902, the register in the processor 901, and the cache memory. stored in at least one.
The program 904 may also be stored in a portable recording medium such as a magnetic disk, flexible disk, optical disk, compact disk, Blu-ray (registered trademark) disk, DVD, or the like. Then, a portable recording medium storing the program 904 may be distributed.

Also, the “units” of the evaluation unit 102, the improvement point analysis unit 103, and the information collection unit 104 may be read as “circuit”, “process”, “procedure”, or “processing”.
Further, the improvement point analysis device 100 may be realized by a processing circuit. The processing circuits are, for example, logic ICs (Integrated Circuits), GAs (Gate Arrays), ASICs (Application Specific Integrated Circuits), and FPGAs (Field-Programmable Gate Arrays).
In this case, the evaluation unit 102, the improvement point analysis unit 103, and the information collection unit 104 are each realized as part of the processing circuit.
In this specification, the general concept of processors and processing circuits is referred to as "processing circuitry."
Thus, processors and processing circuitry are each examples of "processing circuitry."

100 improvement point analysis device 101 information storage unit 102 evaluation unit 103 improvement point analysis unit 104 information collection unit 105 improvement result storage unit 200 analysis object 300 network 400 machine learning device 401 data acquisition unit 402 Teacher Data Acquisition Unit 403 Learning Unit 404 Output Unit 405 Trained Model Storage Unit 901 Processor 902 Storage Device 903 Communication Interface 904 Program 905 Bus.

Claims (18)

a designation unit that designates, as a designated element, an element whose performance should be improved among the three or more elements;
an extraction unit that extracts two or more elements having a significant relationship with the designated element from elements other than the designated element as related elements;
analyzing the influence of the performance of each of the two or more related elements extracted by the extracting unit on the performance of the designated element, and selecting from among the two or more related elements to improve the performance of the designated element An information processing apparatus having an estimation unit that estimates an improvement target related element, which is a related element whose performance should be improved. The estimation unit
When estimating a plurality of related elements to be improved, analyze the influence of the performance of each related element to be improved on the performance of other related elements to be improved, and prioritize among the plurality of related elements to be improved. The information processing apparatus according to claim 1, wherein the setting is performed. The specifying unit is
Designating the improvement target related element estimated by the estimation unit as a new designated element;
The extractor is
Extracting two or more elements other than the new designated element that have a significant relationship with the new designated element as new related elements,
The estimation unit
analyzing the influence of the performance of each of the two or more new related elements extracted by the extracting unit on the performance of the new designated element, and selecting the new designated element from among the two or more new related elements; 2. The information processing apparatus according to claim 1, wherein a new related element whose performance should be improved is estimated as a new improvement target related element in order to improve the performance of the element. The specifying unit is
Each time a new improvement target related element is estimated by the estimation unit, repeatedly designating the estimated new improvement target related element as a new designated element,
The extractor is
Repeatedly extracting two or more elements other than the new specified element as new related elements each time a new specified element is specified by the specifying unit,
The estimation unit
4. The information processing apparatus according to claim 3, wherein each time two or more new related elements are extracted by the extraction unit, estimation of a new related element to be improved is repeated. The estimation unit
estimating the related element to be improved by performing an association analysis using in the conclusion part that the performance of the specified element is improved and using in the condition part that the performance of each of the two or more related elements is improved; The information processing apparatus according to claim 1. The estimation unit
6. The method according to claim 5, wherein when a calculation maintenance condition is not satisfied for any calculation item among the plurality of calculation items included in the association analysis, calculation of the uncalculated calculation item among the plurality of calculation items is not performed. The information processing device described. The information processing device further includes
Each element has an evaluation unit that evaluates whether the performance meets the performance criteria,
The specifying unit is
2. The information processing apparatus according to claim 1, wherein an element whose performance is evaluated by said evaluation unit as not meeting a performance standard is designated as said designated element. The evaluation unit
8. The information processing apparatus according to claim 7, wherein the performance of each element is evaluated as to whether or not it meets a performance standard defined for each element. The three or more elements constitute a plurality of hierarchies,
The evaluation unit
8. The information processing apparatus according to claim 7, wherein whether or not the performance of each element meets a performance standard is evaluated in a time period different for each layer. The evaluation unit
8. The information processing apparatus according to claim 7, wherein the evaluation result is output each time the evaluation is performed. The evaluation unit
An additional performance standard is defined for the improvement target related element estimated by the estimating unit, and if a condition for applying the additional performance standard to the improvement target related element is established, the improvement target related element 8. The information processing apparatus according to claim 7, wherein it is evaluated whether or not the performance of the related element meets the criterion of the additional performance. The evaluation unit
8. The information processing apparatus according to claim 7, wherein evaluation is performed using the performance criteria obtained by statistical processing. The estimation unit
2. The information processing apparatus according to claim 1, wherein, when the relationship between any two or more of the three or more elements is unknown, the relationship between the elements whose relationship is unknown is estimated. The estimation unit
6. The information processing apparatus according to claim 5, wherein calculation is performed using a plurality of calculated values for a plurality of calculation items included in the association analysis, and a calculated value of the calculation using the plurality of calculated values is output. The estimation unit
A calculated value obtained by calculation using a plurality of calculated values for a plurality of calculated items included in the association analysis is corrected based on the improvement performance for at least one of the two or more related factors, and after correction 6. The information processing apparatus according to claim 5, which outputs a calculated value. The estimation unit
Obtaining by machine learning a method for correcting calculated values obtained by calculation using a plurality of calculated values for a plurality of calculation items included in the association analysis, and performing calculation using the plurality of calculated values according to the obtained correction method 6. The information processing apparatus according to claim 5, wherein the obtained calculated value is corrected, and the corrected calculated value is output. the computer designates an element whose performance should be improved among the three or more elements as a designated element;
The computer extracts two or more elements having a significant relationship with the designated element from elements other than the designated element as related elements,
The computer analyzes the influence of the performance of each of the extracted two or more related elements on the performance of the designated element, and selects the designated element from among the two or more related elements to improve the performance of the designated element. An information processing method for estimating an improvement target related element, which is a related element whose performance should be improved. a designation process for designating an element whose performance should be improved among the three or more elements as a designated element;
an extraction process of extracting two or more elements having a significant relationship with the specified element from among elements other than the specified element as related elements;
analyzing the influence of the performance of each of the two or more related elements extracted by the extraction process on the performance of the designated element, and selecting among the two or more related elements to improve the performance of the designated element An information processing program for causing a computer to execute an estimation process of estimating an improvement target related element, which is a related element whose performance should be improved.

Images