JP7044730B2 - Production equipment operation support system and production equipment operation support method - Google Patents

Description

The present invention relates to a production equipment operation support system and a production equipment operation support method for supporting the operation of production equipment and the like.

In the operation of production equipment, it is desirable at the same time to optimize the operation such as ensuring the quality of products, reducing costs, and improving production efficiency, and to enable the same operation even by unskilled persons from the operation dependent on skilled workers. ing.

For example, in Patent Document 1, a model for predicting an objective variable from an explanatory variable based on plant measurement data is created, converted into a component of coordinates having no correlation between the explanatory variables, and an explanatory variable for optimizing the objective variable. The operation condition is optimized with the explanatory variable value as the optimum operation condition.

In Patent Document 2, the input of a model simulating the characteristics of a plant is divided into a plurality of groups, and the operation method for each group is learned by a plurality of learning means to speed up the learning.

Japanese Unexamined Patent Publication No. 2012-74007 Japanese Unexamined Patent Publication No. 2009-11256

As mentioned above, in the operation of production equipment, optimization of operation such as quality assurance of products, cost reduction, improvement of production efficiency, and operation dependent on skilled workers to enable equivalent operation even by unskilled workers. Is desired at the same time.

In Patent Document 1, an explanatory variable that optimizes the objective variable by converting it into a component of coordinates that are uncorrelated between the explanatory variables is obtained. However, if the operation on the upstream side of the production process affects the downstream side, or For example, when the pressure is increased on the downstream side, the upstream side is also affected, and there is a problem that the accuracy of explaining the objective variable is lowered when the components are converted into the components of the coordinates having no correlation between the explanatory variables.

In Patent Document 2, learning is speeded up by dividing the model for each measurement signal and the model for each operation signal into a plurality of groups. For example, the number of operation items is N and the operation value is M discrete values. In this case, there is a problem that it takes time to learn and calculate operation signals (operation items and operation values) for N × M solution spaces. In addition, the accuracy of operation signals (operation items and operation values) based on the learning results depends on the model accuracy of the plant that generates rewards for learning, but it is difficult to build a highly accurate model of complex plant behavior itself. There is a problem. In addition, since there is an error in the learning itself and an error in the plant model itself, there is a problem that the error may be amplified when learning the output of the model of the plant with the error as a reward. be.

The present invention has been made to solve the above-mentioned problems, and it is intended to prevent a decrease in model accuracy by reducing the complexity of operation methods of complicated production equipment and plants, and to support an operation operation support plan. The purpose is to provide a production equipment operation support system and a production equipment operation support method that can be used.

In order to achieve the above object, the production equipment operation support system of the present invention is a production equipment operation support system that supports the operation of the production equipment, and the production equipment operation support system is a data collection means for collecting data from the production equipment. The driving operation learning means includes a driving operation learning means for learning the driving operation based on the data, a driving operation support means for supporting the driving operation based on the learning result of the data and the driving operation learning means, and a display means. , Create driving operation record information (for example, driving operation information 72) for learning past operation methods , including the value of each measurement item, the value of each operation item, and the change value of each operation item based on the data. Based on the driving operation record information creating means and the driving operation record information, the operation item and the operation value selection method are separately learned, and each learning model (for example, the operation item learning model M1 and the operation value learning model M2) is used. Information for presenting the results of the items selected based on the two-step learning method to be created, the objective value prediction method learning method for learning the objective value prediction method in production, and the learning model and the objective value prediction method. It has a learning result display means that is generated and displayed on the display means, and the driving operation support means includes a current driving information creating means for creating current driving information based on data, and current driving information and two-step learning. Operation item plan and operation value plan created based on the learning model created by the means Operation plan creation means and target value prediction method When the operation item plan and operation value plan are applied based on the prediction method of the learning means A driving operation plan evaluation means that predicts and evaluates a target value, and a driving operation support information display means that creates information for presenting an operation item plan and an operation value plan and displays a recommended operation plan on the display means. The two-step learning means has, as the first step, operation item learning information including the value of each measurement item, the value of each operation item, and the operation presence / absence of the target operation item based on the driving operation record information. Create, learn the operation item selection method, create the operation item learning model, and as the second stage, based on the operation operation record information, the value of each measurement item, the value of each operation item, the target operation item It is characterized by creating operation item learning information including the change value of, learning the method of selecting the operation value of the target operation item, and creating the operation value learning model . Other aspects of the invention will be described in embodiments described below.

According to the present invention, it is possible to prevent a decrease in model accuracy by reducing the complexity of an operation method of a complicated production facility and a plant, and to present an operation operation support plan.

It is a block diagram which shows the production equipment operation support system which concerns on embodiment. It is a flowchart which shows the process of a data collection means. It is a figure which shows the example of the production equipment information collected and stored by the data collecting means. It is a figure which shows the example of the operation operation information. It is a figure which shows the example of operation item learning information. It is a figure which shows the example of the operation value learning information. It is a flowchart which shows the process of the operation performance record information creation means. It is a flowchart which shows the details of the operation operation information creation process. It is a flowchart which shows the details of operation item and operation value learning information creation processing. It is a flowchart which shows the process of a two-step learning means. It is a flowchart which shows the detail of the learning process of an operation item. It is a flowchart which shows the detail of the learning process of an operation value. It is a flowchart which shows the process of the objective value prediction method learning means. It is a flowchart which shows the process of the learning result display means. It is a figure which shows the example of the learning result display screen. It is a flowchart which shows the process of the current operation information creation means. It is a figure which shows the example of the current operation information. It is a flowchart which shows the process of the operation plan creation means. It is a flowchart which shows the detail of the creation process of a plurality of operation plan. It is a flowchart which shows the details of the operation item draft creation process. It is a flowchart which shows the details of the operation value draft creation process. It is a figure which shows the example of a driving operation plan. It is a flowchart which shows the process of the operation plan evaluation means. It is a figure which shows the example of the data for evaluation of a driving operation plan. It is a flowchart which shows the process of the driving operation support information display means. It is a figure which shows the example of the driving operation support information display screen. It is a figure which shows the specific example of the production equipment information storage data. It is a figure which shows the specific example of the operation operation information. It is a figure which shows the specific example of operation item learning information. It is a figure which shows the specific example of the operation value learning information. It is a figure which shows the specific example of the learning result display screen. It is a figure which shows the specific example of the current operation information. It is a figure which shows the specific example of a driving operation plan. It is a figure which shows the specific example of the data for evaluation of a driving operation plan. It is a figure which shows the specific example of the driving operation support information display screen.

Embodiments for carrying out the present invention will be described in detail with reference to the drawings as appropriate.
<< Embodiment 1 >>
FIG. 1 is a block diagram showing a production equipment operation support system 100 according to an embodiment. The production equipment operation support system 100 is a production equipment operation support system that supports the operation of production equipment. The production equipment operation support system 100 includes a control means 2 for controlling the production equipment operation support means 3 and a display means 4. The production equipment operation support means 3 is a data collection means 6 that collects data from the production equipment, a driving operation learning means 5 that learns driving operations based on the data, and a driving operation based on the learning results of the data and the driving operation learning means 5. It has a driving operation support means 8 for supporting the above and a database 7. The driving operation learning means 5 includes a driving operation result information creating means 10, a two-step learning means 11, a target value prediction method learning means 12, and a learning result display means 13. The driving operation support means 8 includes a current driving information creating means 20, a driving operation plan creating means 21, a driving operation plan evaluation means 22, and a driving operation support information display means 23. Hereinafter, each means will be described in detail.

The production equipment 1 is a production equipment, a plant, an equipment, or the like, which is the subject of the present invention. The control means 2 controls the operation of the production equipment operation support means 3. The data collecting means 6 obtains the values of measurement items and the values of operation items when operated by a skilled person from the production equipment 1, and stores them in the database 7.

The driving operation learning means 5 uses the data stored in the database 7 to create driving operation result information for learning the driving operation method of a past expert. The driving operation result information creating means 10, operation items and operations. Two-step learning means 11 that learns the value selection method separately, objective value prediction method learning means 12 that learns the objective value prediction method such as quality assurance of products, cost reduction, and production efficiency improvement, learning model And by the learning result display means 13 that generates information for presenting the result when the learning model is applied to the actual data, it is possible to learn the operation method of the past expert and present the learning result to the display means 4. ..

The driving operation support means 8 uses the data and the learning model accumulated in the database 7 to create the current driving information creating means 20 for creating the current driving information, and the driving operation plan for creating the operation item plan and the operation value plan. The operation item plan and the operation value that are the optimum prediction in the current operation information by the creation means 21, the operation plan evaluation means 22 that predicts and evaluates the target value when the operation item plan and the operation value plan are applied. It is possible to generate a plan, generate display information such as an operation item plan, an operation value plan, and a predicted value of a target value by the operation operation support information display means 23, and present the operation item plan and the operation value plan to the display means 4. do.

Here, the measurement items are, for example, the current, voltage, temperature, and flow rate of a certain facility, and the operation items are the input amount of raw materials, pressure, temperature, passing speed, processing time, and the like.

Hereinafter, the processing of each means will be described.
FIG. 2 is a flowchart showing the process (step S100) of the data collection means 6. The data collection means 6 starts the process according to the instruction of the control means 2 (step S101), performs the data collection process (step S102) from the production equipment 1, and performs the storage process (step S103) in the database 7. It is determined in step S104 whether the data collection process is completed, and if the process is not completed (step S104, No), the process returns to step S102, and if the process is completed (step S104, Yes), the data collection process is terminated (step S104, No). S105).

FIG. 3 is a diagram showing an example of production equipment information 71 collected and stored by the data collecting means 6. The production equipment information 71 includes a date and time, a value of measurement item 1, ..., a value of measurement item N, a value of operation item 1, ..., a value of operation item M, and a target value.

The process of the driving operation learning means 5 will be described with reference to FIGS. 4 to 15.
The driving operation record information creating means 10 creates driving operation information 72 (see FIG. 4), operation item learning information 73 (see FIG. 5), and operation value learning information 74 (see FIG. 6) as driving operation record information (see FIG. 4). See FIG. 7).

FIG. 4 is a diagram showing an example of operation operation information 72. The operation operation information 72 includes No, the value of the measurement item 1, ..., the value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, the change value of the operation item 1, ... -, The change value of the operation item M is included.

FIG. 5 is a diagram showing an example of operation item learning information 73,731. The operation item learning information 731 is No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, and the presence / absence of the operation of the operation item 1. Similarly, the operation item learning information 73M is No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, and the presence / absence of the operation of the operation item M are included.

FIG. 6 is a diagram showing an example of the operation value learning information 74. The operation item learning information 741 is No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, and the change value of the operation item 1. Similarly, the operation item learning information 74M is No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, and the change value of the operation item M are included.

FIG. 7 is a flowchart showing the process (step S200) of the operation record information creating means 10. The process is started (step S201), the operation operation information creation process (step S202, see FIG. 8) is performed, the operation item and operation value learning information creation process (step S203, see FIG. 9) is performed, and the process ends (step S203, see FIG. 9). Step S204).

FIG. 8 is a flowchart showing the details of the operation operation information creation process (step S202). Processing is started (step S301), the production equipment information 71 shown in FIG. 3 is input (step S302), i indicating the number of the production equipment information 71 is set to 1, and L is the total number of predetermined production equipment information. (Step S303), it is determined whether or not i is L or more (step S304), and if not (step S304, No), the production equipment information excluding the i + 1th target value and the i-th and i + 1th. The difference between the values of the operation items of is added to the operation operation information (step S305), i + 1 is substituted for i (step S306), and the process returns to step S304. If i is L or more in step S304 (step S304, Yes), the process ends (step S307). An example of the driving operation information 72 created by the driving operation information creating process (step S202) is shown in FIG. 4 described above.

FIG. 9 is a flowchart showing the details of the operation item and the operation value learning information creation process (step S203). Processing is started (step S401), operation operation information 72 is input (step S402), j indicating the number of the operation item is set to 1, and M is set to the total number of predetermined operation items (step S403). ), It is determined whether or not j is larger than M (step S404), and if not (step S404, No), i indicating the number of the operation operation information 72 is set to 1, and K is set to a predetermined operation. Set to the total number of operation information 72 (step S405), determine whether i is larger than K (step S406), and if not (step S406, No), the value of the i-th measurement item and the operation item. If the value of and the change value of the operation item j is 0, no operation is added, and if the change value is not 0, the operation is yes to the learning information of the operation item j (step S407). It is determined whether or not the operation of the j-th operation item is performed (step S408), and if it is present (step S408, Yes), the value of the i-th measurement item, the value of the operation item, and the operation value of the operation item j. The change value is added to the operation value learning information (step S409), i + 1 is substituted for i (step S410), and the process returns to step S406.

If there is nothing in step S408 (step S408, No), the process proceeds to step S410. If i is K or more in step S406 (step S406, Yes), j + 1 is substituted for j (step S411), and the process returns to step S404. If j is larger than M in step S404 (step S404, Yes), the process ends (step S412). Examples of the operation item learning information 73 and the operation value learning information 74 created by the operation item and operation value learning information creation process (step S203) are shown in FIGS. 5 and 6 described above.

FIG. 10 is a flowchart showing the process (step S500) of the two-step learning means 11. The process is started (step S501), the operation item learning process (step S502, see FIG. 11) is performed, the operation value learning process (step S503, see FIG. 12) is performed, and the process is terminated (step S504).

FIG. 11 is a flowchart showing the details of the operation item learning process (step S502). The process is started (step S601), the operation item learning information 73 is input (step S602), the operation item number is set to 1 (step S603), and the operation item number is based on the presence or absence of the operation of the operation item learning information 73. Machine learning is performed (step S604).

In the machine learning, the value of the measurement item N from the measurement item 1 and the value of the operation item M from the operation item 1 are used as explanatory variables, and the multiple regression equation is obtained with the presence or absence of the operation of the operation item number as the objective variable. It is not limited to the formula, and other machine learning such as a known gradient boosting method or a neural net may be used with the value of the explanatory variable as an input value and the objective variable value as an output value. The result of the machine learning is called "operation item learning model M1".

In step S605, 1 is added to the operation item numbers, it is determined whether or not the learning of all the operation item numbers is completed (step S606), and if not (steps S606, No), the process returns to step S604. If the learning of all the operation item numbers is completed in step S606 (steps S606, Yes), the process is terminated (step S607).

FIG. 12 is a flowchart showing the details of the operation value learning process (step S503). The process is started (step S701), the operation value learning information is input (step S702), the operation item number is set to 1 (step S703), and the machine is based on the operation value of the operation value learning information 74 of the operation item number. Learning is carried out (step S704). Here, the operation value may be a continuous value, but the processing can be speeded up by treating each operation item as a plurality of discrete values.

Similar to the operation item learning process (step S502), the machine learning is performed by a machine learning method such as a multiple regression equation, a gradient boosting method, or a neural network. The result of the machine learning is called "operation value learning model M2".

In step S705, 1 is added to the operation item number, and it is determined whether or not the learning of all the operation item numbers is completed (step S706). If not (step S706, No), the process returns to step S704. If the learning of all the operation item numbers is completed in step S706 (steps S706, Yes), the process is terminated (step S707).

FIG. 13 is a flowchart showing the process (step S800) of the target value prediction method learning means 12. Processing is started (step S801), production equipment information is input (step S802), machine learning (step S803) is performed with the objective variable as the objective value and the explanatory variable as the measurement item value and the operation item value, and the processing is performed. It ends (step S804).

Similarly, the machine learning is carried out by a machine learning method such as a multiple regression equation, a gradient boosting method, or a neural network. The result of the machine learning is called "target value prediction learning model M3".

FIG. 14 is a flowchart showing the process (step S900) of the learning result display means 13. The process is started (step S901), it is determined whether or not the user of the production equipment operation support system has selected the display of the learning result (step S902), and if selected (step S902, Yes), the operation items match. Degree, operation value agreement, target value prediction formula and its correlation coefficient are displayed (step S903), and it is determined whether or not one of the production equipment information and one of the operation items is selected (step S904). When selected (step S904, Yes), the operation item name, achievement and learning result of the operation item in the production equipment information are displayed (step S905), and the achievement and learning result of the operation value of the operation item are displayed (step S904). S906), the process is terminated (step S907).

If it is not selected in step S902 (step S902, No), the process ends (step S907). If it is not selected in step S904 (steps S904, No), the process ends (step S907).

FIG. 15 is a diagram showing an example of a learning result display screen. FIG. 15 shows an example of a screen displayed on the display means 4 by the learning result display means 13. On the screen 15 of FIG. 15, there are buttons 151 and 152 for selecting learning or operation, which are selected by the user of the production equipment operation support system 100. Further, the user selects one of the production equipment information with the pull-down button 153 for selecting the production equipment information, and similarly one of the operation items with the pull-down button 154 for selecting the operation item. The screen 15 includes a matching degree 155 of the selected operation item, learning result information 156 for the operation item, a matching degree 157 of the operation value of the operation item, learning result information 158 of the operation value, and information 159 related to the target value prediction formula. .. The objective value prediction formula shows an example when a multiple regression formula is used as a machine learning method.

Next, the processing of the driving operation support means 8 will be described with reference to FIGS. 16 to 26. The processing of the driving operation support means 8 is started according to the instruction of the control means 2.

FIG. 16 is a flowchart showing the process (step S1000) of the current operation information creating means 20. The process is started (step S1001), the current production equipment information 71 is input (step S1002), the value data of the measurement item and the operation item is set as the current operation information 76 (step S1003), and the process is terminated (step S1004). ).

FIG. 17 is a diagram showing an example of the current operation information 76. The current operation information 76 includes a value of the measurement item 1, ..., a value of the measurement item N, a value of the operation item 1, ..., and a value of the operation item M.

FIG. 18 is a flowchart showing the process (step S1100) of the operation plan creating means 21. The process is started (step S1101), the current operation information is input (step S1102), the current operation information is input to the "operation item learning model M1", and the selection probability of the operation item is calculated (step S1103). The selection probability is a value of 0 to 1 which is a result obtained by inputting the current operation information into the "operation item learning model M1" created by the two-step learning means 11.

Next, the current operation information is input to the "operation value learning model M2" created by the two-step learning means 11, and the selection probability of the operation value of the operation item is calculated (step S1104). The selection probability is 0 to 1, which is the result obtained by inputting the current operation information into the "operation value learning model M2", where 1 is selected when each operation value as a discrete value is selected and 0 is not selected. Let the value be the selection probability. Next, a process of creating a plurality of operation plans (step S1105, see FIG. 19) is performed, and the process is completed (step S1106).

FIG. 19 is a flowchart showing the details of the process of creating a plurality of operation plans (step S1105). The process is started (step S1201), i indicating the number of the operation plan is set to 1, the total number of predetermined operation plans is set to P (step S1202), and it is determined whether or not i is larger than P. If (step S1203) or not (step S1203, No), the operation item draft creation process (step S1204, see FIG. 20) is performed, and the operation value draft creation process (step S1205, see FIG. 21) is performed. , The i-th operation plan is additionally stored (step S1206), i + 1 is substituted for i (step S1207), and the process returns to step S1203. If i is larger than P in step S1203 (steps S1203, Yes), the process ends (step S1208).

FIG. 20 is a flowchart showing the details of the operation item draft creation process (step S1204). Processing is started (step S1301), j indicating the number of the operation item is set to 1, M is set to the total number of operation items (step S1302), and it is determined whether or not j is larger than M (step S1303). ), If not (step S1303, No), a random number between 0 and 1 is generated, the value of the random number is assigned to r (step S1304), and the selection probability of the jth operation item is r or more. It is determined whether or not (step S1305), if it is r or more (step S1305, Yes), the jth operation item is selected (step S1306), j + 1 is substituted for j (step S1307), and the step. Return to S1303. If the selection probability of the jth operation item is less than r in step S1305 (steps S1305 and No), the process proceeds to step S1307. If j is larger than M in step S1303 (steps S1303, Yes), the process ends (step S1308).

FIG. 21 is a flowchart showing the details of the operation value draft creation process (step S1205). Processing is started (step S1401), j indicating the number of the operation item is set to 1, M is set to the total number of operation items (step S1402), and it is determined whether or not j is larger than M (step S1403). , If not (step S1403, No), it is determined whether or not the jth operation item is selected (step S1404), and if it is selected (step S1404, Yes), the jth operation item is selected. One operation value is selected by a known roulette selection from the total number of operation value options and the selection probability of the operation value (step S1405), j + 1 is substituted for j (step S1406), and the process returns to step S1403. If it is not selected in step S1404 ((steps S1404, No), the process proceeds to step S1406. If j is larger than M in step S1403 (steps S1403, Yes), the process ends (step S1407).

FIG. 22 is a diagram showing an example of the operation plan 77. FIG. 22 shows P operation plans 77 created in the process of creating a plurality of operation plans (step S1105) shown in FIG. The operation plan 77 includes No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., The value of the operation item M is included.

FIG. 23 is a flowchart showing the process (step S1500) of the operation plan evaluation means 22. The process is started (step S1501), 1 is set in i indicating the number of the operation plan, the total number of operation plans is set in P (step S1502), and it is determined whether or not i is larger than P (step). S1503), if not (steps S1503, No), the information of the i-th driving operation plan is input to the "target value prediction learning model M3" to calculate the prediction target value (step S1504), and the i-th prediction. The target value is added to the operation plan evaluation data (step S1505), i + 1 is substituted for i (step S1506), and the process returns to step S1503. If i is larger than P in step S1503 (step S1503, Yes), a plan (recommended operation plan) most suitable for the driving purpose is selected from the driving operation plans (step S1507), and the process is terminated (step S1508). ).

Here, among the driving operation plans, the plan that best suits the driving purpose is, for example, if the driving purpose is cost reduction, the target value becomes a cost, and the operation plan having the smallest value among the predicted target values is the driving purpose. Select as the best plan for you.

FIG. 24 is a diagram showing an example of data 78 for evaluating an operation plan. The operation plan evaluation data 78 is the operation plan evaluation data added in step S1505. The data 78 for evaluation of the driving operation plan includes No. , The value of the measurement item 1, ..., The value of the measurement item N, the value of the operation item 1, ..., the value of the operation item M, and the prediction target value are included.

FIG. 25 is a flowchart showing the process (step S1600) of the driving operation support information display means 23. When the process is started (step S1601), it is determined whether or not the user of the production equipment operation support system 100 has selected the display of the operation plan (step S1602), and the display of the operation plan is selected (step S1602). Step S1602, Yes), the recommended operation plan is displayed (step S1603), the actual value and the predicted value of the target value are displayed (step S1604), and the process is terminated (step S1605). If the display of the operation plan is not selected in step S1602 (step S1602, No), the process ends (step S1605).

FIG. 26 is a diagram showing an example of a driving operation support information display screen. FIG. 26 shows an example of a screen 26 displayed on the display means 4 by the driving operation support information display means 23. The screen 26 has buttons 261,262 for selecting learning or operation, which the user of the production equipment operation support system 100 selects. Further, the user selects one of the operation items with the pull-down button 264 for selecting the operation item. The screen 26 includes an operation recommendation plan 265, a predicted value of the target value 266, and a time-dependent change diagram 167 leading to the predicted value of the target value. As the recommended operation plan 265 in step S1603, there are a selection probability of each operation item, an operation presence / absence and an operation value, and a selection probability of each operation value of the selected operation item. The predicted value 266 of the target value is a predicted value of the target value when the operation recommendation plan is applied. Changes over time Figure 167 shows changes in the actual results of the target values over time, but it is not necessary to limit them. In the present embodiment, the display is displayed on the monitor of the display means 4, but it is also preferable to inform the operator or the person in charge of the operation recommendation plan by voice or e-mail.

<< Embodiment 2 >>
The process in the production equipment operation support system 100 of the first embodiment will be described in more detail using the specific data shown in FIG. 27. The description of the part overlapping with the first embodiment will be omitted.

<Driving operation learning means 5>
FIG. 27 is a diagram showing a specific example of the stored data of the production equipment information 71A. FIG. 27 shows an example of the production equipment information 71A collected and stored by the data collecting means 6. The production equipment information 71A has the date and time, the current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, the temperature value of the operation item 2, and the compressive strength of the target value. included.

The process of the operation record information creating means 10 shown in FIG. 7 is started (step S201), the operation information creation process (step S202, see FIG. 8) is performed, and the operation item and operation value learning information creation process (step S203) is performed. , FIG. 9), and the process is terminated (step S204).

In the operation operation information creation process (step S202) of FIG. 8, the production equipment information 71A shown in FIG. 27 is input (step S302), and the operation operation information 72A (see FIG. 28) is created.

FIG. 28 is a diagram showing a specific example of the operation operation information 72A. FIG. 28 shows an example of the driving operation information 27A created by the driving operation information creating process (step S202). For example, the date of row 271 in FIG. 27 is 9:30 on January 20, 2019 (2019.1.20.09.30) and the date of row 272 is 9:40 on January 20, 2019. When the difference between the pressure and temperature values, which are the operation items of the production equipment information, is calculated with reference to the minute (2019.1.20.09.40), the difference between them is 0. Therefore, No. 28 in FIG. 28. Line 1 (January 20, 2019 9:40 (2019.1.20.09.40)) has a current value of 51, a ripple voltage value of 2, a pressure value of 75, a temperature value of 60, and a pressure. The change value of is 0, and the change value of temperature is 0.

Next, when the difference between the pressure and temperature values, which are the operation items of the production equipment information in rows 272 and 273 in FIG. 27, is obtained, the difference between the pressure values is −5 and the difference between the temperature values is 0. Therefore, No. 28 in FIG. 28. In row 2, the current value is 50, the ripple voltage value is 2, the pressure value is 70, the temperature value is 60, the pressure change value is -5, and the temperature change value is 0. The same applies thereafter, and when the difference between the pressure and temperature values, which are the operation items of the production equipment information in rows 276 and 277 in FIG. 27, is obtained, the difference between the pressure values is 0 and the difference between the temperature values is 5. be. Therefore, No. 28 in FIG. 28. In row 6, the current value is 49, the ripple voltage value is 3, the pressure value is 65, the temperature value is 65, the pressure change value is 0, and the temperature change value is 5. As described above, the driving operation information 72A shown in FIG. 28 is created by the driving operation information creating process (step S202) of the driving operation information creating means 10.

In the operation item and operation value learning information creation process (step S203) of FIG. 9, the operation operation information 72A shown in FIG. 28 is input (step S402), the operation item learning information 73A (see FIG. 29), and the operation value learning information 74A. (See FIG. 30) is created.

FIG. 29 is a diagram showing a specific example of the operation item learning information 73A. FIG. 29 shows an example of the operation item learning information 73A created by the operation item and operation value learning information creation process (step S203). Refer to FIG. 9 as appropriate.

In step S407, for example, when i is 1 and j is 1, No. 28 in FIG. 28. The value of the measurement item 1 is 51, the value of the ripple voltage is 2, the value of the operation item is 75, the value of the temperature is 60, and the change value of the pressure of the operation item 1 is 0. Therefore, No. 29 (1) of FIG. The presence or absence of the pressure operation of the operation item 1 in the line 1 is "None".

In step S407, when i is 2 and j is 1, No. 28 in FIG. 28. The value of the measurement item 2 is 50, the value of the ripple voltage is 2, the value of the operation item is 70, the value of the temperature is 60, and the change value of the pressure of the operation item 1 is -5. .. Therefore, No. 29 (1) of FIG. The presence or absence of the pressure operation in the operation item 1 in the second row is "Yes".

In step S407, when i is 1 and j is 2, No. 28 in FIG. 28. The value of the measurement item 1 is 51, the value of the ripple voltage is 2, the value of the operation item is 75, the value of the temperature is 60, and the change value of the temperature of the operation item 2 is 0. Therefore, No. 29 (2) of FIG. The presence or absence of the pressure operation in the row operation item 2 of 1 is "None".
Is the value of.

In step S407, when i is 6 and j is 2, No. 28 in FIG. 28. The value of the measurement item 6 is 49, the value of the ripple voltage is 3, the value of the operation item is 65, the value of the temperature is 60, and the value of the temperature change of the operation item 2 is 5. Therefore, No. 29 (2) of FIG. The presence or absence of pressure operation in operation item 2 in line 6 is "Yes".

FIG. 30 is a diagram showing a specific example of the operation value learning information 74A. FIG. 30 shows an example of the operation value learning information 74A created by the operation item and the operation value learning information creation process (step S203). Refer to FIG. 9 as appropriate.

In step S408, when i is 2 and j is 1, the presence or absence of the pressure operation, which is the first operation item, is determined by No. 29 (1). Judging from the row of No. 1, since it is present, the process proceeds to step S409, and in step S409, FIG. 29 (1) No. The current value of 2 is 50, the ripple voltage value is 2, the pressure value is 70, and the temperature value is 60. The pressure change value of -5 in No. 2 is set to No. 5 in FIG. 30 (1). Add as the value in line 1.

In step S408, when i is 6 and j is 2, the presence or absence of the temperature operation, which is the second operation item, is determined by No. 29 (2). Judging from the row of No. 6, since it is present, the process proceeds to step S409, and in step S409, FIG. 29 (2) No. The current value of No. 6 is 49, the ripple voltage value is 3, the pressure value is 65, and the temperature value is 65. The temperature change value 5 of No. 6 is set to No. 5 in FIG. 30 (2). Add as the value in line 1.

As described above, FIGS. 29 and 30 are specific examples of the operation item learning information and the operation value learning information created by the operation item and operation value learning information creation process (step S203).

Next, the process proceeds to the process of the two-step learning means 11 shown in FIG. The learning process (step S502, see FIG. 11) of the operation item of the first step of the two-step learning means 11 will be described.

In the example of the learning information of the operation item 1 shown in FIG. 29 (1), the current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, and the temperature value of the operation item 2 are set. Learning is performed using the explanatory variables and the presence or absence of the pressure operation of the operation item 1 as the objective variable. At the time of learning, if there is an operation of the objective variable, the learning is performed with a numerical value of 1, and if there is no operation, the learning is performed with a numerical value of 0, but the learning is not limited to this.

Similarly, in the example of the learning information of the operation item 2 shown in FIG. 29 (2), the current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, and the temperature of the operation item 2 Learning is performed using the value as an explanatory variable and the presence or absence of the pressure temperature of the operation item 2 as the objective variable.

As the "operation item learning model M1" which is the learning result, an "operation item pressure learning model" whose operation item is a pressure learning model and an "operation item temperature learning model" whose operation item is a temperature learning model can be created. ..

The learning process of the operation value of the second step of the two-step learning means 11 (step S503, see FIG. 12) will be described.

In the example of the learning information of the operation value of the operation item 1 shown in FIG. 30 (1), the current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, and the temperature of the operation item 2 The value of is used as an explanatory variable, and the change value of the pressure of the operation item 1 is used as the objective variable for learning.

In the learning information of the operation value of the operation item 2 shown in FIG. 30 (2), the current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, and the temperature value of the operation item 2 are used. Learning is performed using the temperature change value of the operation item 2 as the objective variable as an explanatory variable.

As the "operation value learning model M2" which is the learning result, the operation item is the "operation item pressure operation value learning model" which is the learning model of the pressure operation value, and the operation item is the learning model of the temperature operation value. Operation item Temperature operation value learning model "can be created.

Next, the process proceeds to the process of the target value prediction method learning means 12 shown in FIG.
In FIG. 13, for example, step S802 inputs the production equipment information shown in FIG. 27, and in step S803, the objective variable is the compression strength which is the objective value, the explanatory variable is the current value of the measurement item 1, and the ripple voltage of the measurement item 2. Machine learning is performed as a value, a pressure value of operation item 1, and a temperature value of operation item 2. When the multiple regression equation is obtained using the production equipment information shown in FIG. 27, for example, the “objective value prediction learning model M3” as in equation (1) can be obtained with Y as the compressive strength of the objective variable.

Next, the process proceeds to the process of the learning result display means 13 shown in FIG. Refer to FIG. 31 as appropriate.
The operation item matching degree 155 is, for example, in the i-th production equipment information out of the total number of production equipment information L, when there are two operation items from operation item 1 to operation item 2, the presence / absence of operation of the operation item is determined. If Mi out of the two match, the operation item matching degree 155 is calculated as Σ (Mi / 2) / L for i = 1 to L, but is not limited to this.

The operation value matching degree 157 is the operation value when the number of operation items is 2 and the number of operation values that completely match is Qi in the i-th production equipment information out of the total number of production equipment information L. The degree of coincidence 157 is calculated as Σ (Qi / 2) / L for i = 1 to L, but is not limited to this, for example, it is not an exact match and is within a predetermined allowable amount. It may be the degree of agreement.

FIG. 31 is a diagram showing a specific example of the learning result display screen. FIG. 31 shows an example of the screen 15A displayed on the display means 4 by the learning result display means 13. FIG. 31 is an example in which one of the production equipment information is selected by the pull-down button 153 for selecting the production equipment information, and the pressure, which is one of the operation items, is similarly selected by the pull-down button 154 for selecting the operation item. The objective value prediction formula shows an example when a multiple regression formula is used as a machine learning method.

In FIG. 31, the operation item is the case of pressure, and the operation item agreement degree 155 is 85%.
Further, the operation value matching degree 157 is 80%. The equation (1) is displayed as the target value prediction equation-related information 159, and the correlation coefficient is 0.87.

<Driving operation support means 8>
FIG. 32 is a diagram showing a specific example of the current operation information 76A. The current operation information 76A includes the value of the current of the measurement item 1, the value of the ripple voltage of the measurement item 2, the value of the pressure of the operation item 1, and the value of the temperature of the operation item 2.

The process of the current operation information creating means 20 shown in FIG. 16 is started (step S1001), the current production equipment information 71A is input (step S1002), and the value data of the measurement item and the operation item is set as the current operation information 76A. (Step S1003), the process is terminated (step S1004).

During the process of the operation plan creating means 21 shown in FIG. 18, in step S1103, the “operation item learning model M1” uses the above-mentioned “operation item pressure learning model” and “operation item temperature learning model” as specific examples. .. Further, in step S1104, the "operation value learning model M2" uses the above-mentioned "operation item pressure operation value learning model" and "operation item temperature operation value learning model" as specific examples. In the process of creating a plurality of operation plans shown in FIG. 18 (step S1105), the operation plan 77A shown in FIG. 33 is created.

FIG. 33 is a diagram showing a specific example of the operation plan 77A. FIG. 33 shows P operation plan 77A created in the process of creating a plurality of operation plans (step S1105) shown in FIG. The operation plan 77 includes No. , The value of the current of the measurement item 1, the value of the ripple voltage of the measurement item 2, the value of the pressure of the operation item 1, and the value of the temperature of the operation item 2.

During the process of the driving operation plan evaluation means 22 shown in FIG. 23, in step S1504, the information of the i-th driving operation plan is input to the target value prediction learning model of the above equation (1) to calculate the predicted target value. ..

Further, in step S1507, among the operation plans, the plan most suitable for the driving purpose is, for example, if the driving purpose is minimized when the compressive strength is 300 or more, the target value becomes the compressive strength and the predicted target value becomes. The operation plan having the smallest value among 300 or more is selected as the plan most suitable for the driving purpose. FIG. 34 shows the operation plan evaluation data 78A added in step S1505.

FIG. 34 is a diagram showing a specific example of the operation plan evaluation data 78A. The operation plan evaluation data 78A is the operation plan evaluation data added in step S1505. The data 78A for evaluating the operation plan includes No. , The current value of the measurement item 1, the ripple voltage value of the measurement item 2, the pressure value of the operation item 1, the temperature value of the operation item 2, and the compressive strength of the prediction target value are included. Specifically, No. The second row shows the operation plan of 302 having the compressive strength closest to 300.

FIG. 35 is a diagram showing a specific example of the driving operation support information display screen. FIG. 35 shows a screen 26A displayed on the display means 4 by the driving operation support information display means 23. As the operation recommendation plan 265, an operation plan for pressure and temperature, which are operation items, is shown. Further, it is shown that the compressed strength becomes 302 as the predicted value 266 of the target value.

In this embodiment, a method of selecting an operation item and a method of selecting an operation value performed by an expert from past actual data such as production equipment, and an object for the purpose of ensuring the quality of a product, reducing costs, improving production efficiency, and the like. Learn how to predict values, select operation item proposals and operation value proposals when operating production equipment, etc., and present operation item proposals and operation value proposal operation support proposals according to the evaluation that optimizes the target value. I tried to do it.

In this embodiment, by targeting all the measurement items and the operation items as explanatory variables, it is possible to prevent a decrease in the accuracy of explaining the objective variable, and the learning of the operation items and the operation values is separated as the learning of the operation method of the expert. By performing two-step learning and creating an operation plan that separates operation items and operation values, learning and operation plan creation for, for example, N solution spaces for operation items, and solutions with, for example, M operation values. By learning and creating operation values for the space, learning, operation items and operation values are calculated for N + M solution spaces, which speeds up and improves real-time performance.

Furthermore, by separating the operation items and operation values, the complexity of the operation method of a complicated plant is reduced to prevent the deterioration of the model accuracy, and by operating based on the operation operation support plan, the driver is an unskilled person. Even if there is, it is possible to drive at the same level as a skilled person.

In the production equipment operation support system 100 of the first aspect, the control means 2 controls the production equipment operation support means 3, and the data collection means collects the actual values of the measurement items and the operation items performed by the expert from the production equipment 1 and the like. Obtain it and store it in database 7. Using the data accumulated in the database 7 as the driving operation learning means 5, the driving operation result information creating means 10 creates the driving operation result information for learning the driving operation method of a past expert, and learns in two stages. The means 11 learns the operation item and the operation value selection method, and the target value prediction method learning means 12 learns and learns the target value prediction method such as quality assurance of the product, cost reduction, and production efficiency improvement. The result display means 14 stores the learning model which is the learning result in the database 7, and makes it possible to present the result when the learning model and the learning model are applied to the actual data by the learning result display means. As the driving operation support means 8, the current driving information creating means 20 creates the current driving information, and the driving operation plan creating means 21 uses a plurality of operation item plans and operations using the data and the learning model accumulated in the database 7. A value plan is created, and the driving operation plan evaluation means 22 creates a driving operation support information which is an operation item plan and an operation value plan for which the target value is the optimum prediction in the current driving information, and the driving operation support information display means. It is characterized in that the driving operation support information at the time of driving is presented by 23.

The invention of the second aspect is characterized in that, in the production equipment operation support system 100 of the first aspect, the operation operation record information calculates the change value of the operation item. The calculation of the change value is performed from the operation value of the operation item of the operation result information immediately before the operation result information currently being focused on to the operation value of the operation item of the operation result information currently being focused on. Calculate the amount of change. The change in the operation value is important information indicating the operation method selected as the operation of the production equipment for optimizing the target value in the operation operation record information that the expert is currently paying attention to.

According to the invention of the third aspect, in the production equipment operation support system 100 of the first aspect or the second aspect, the two-step learning means 11 learns whether or not an operation item is operated and the operation value of the operation item. It is characterized by. Learning whether or not there is an operation of an operation item is learned by learning whether or not there is an operation for each operation item, and learning of an operation value is skilled by learning operation performance information in which the change value of the operation value is not zero for each operation item. It is possible to learn the driving operation method that a person has as tacit knowledge. In addition, by not learning all of the enormous amount of driving operation record information that the expert was in charge of driving operation, by learning the driving operation record information that the tacit knowledge of the expert is condensed and the change value is not zero. It is possible to obtain learning results that are condensed by speeding up learning and tacit knowledge of experts.

In the invention of the fourth aspect, in the production equipment operation support system of any one of the first aspect to the third aspect, the learning result display means 13 has an operation item matching degree, an operation presence / absence of the operation item, and an operation value. It is characterized in that display information of a matching degree, an operation value, and a target value prediction formula is generated, and the information is presented by a display means. As for the operation item matching degree, for example, in the i-th production equipment information out of the total number of production equipment information L, there are N operation items from the operation item 1 to the operation item N, and the presence / absence of the operation of the operation item is N. If Mi out of the items match, the operation item matching degree is calculated as Σ (Mi / N) / L for i = 1 to L, but is not limited to this. The degree of operation value matching is the operation value match when the number of operation items is N and the number of operation values that completely match is Qi in the i-th production equipment information out of the total number of production equipment information L. The degree is calculated as Σ (Qi / N) / L for i = 1 to L, but is not limited to this, for example, not an exact match, but a degree of match within a predetermined allowable amount. May be.

In the invention of the fifth aspect, in the production equipment operation support system 100 of any one of the first aspect to the fourth aspect, the operation operation support information display means 23 uses the selection probability of each operation item as an operation recommendation plan. It is characterized by generating display information of operation presence / absence and operation value, selection probability of each operation value of the selected operation item, predicted value of target value and actual value so far, and presenting the information to the display means 4. do. By selecting a specific operation item from the operation item selection pull-down menu, the selection probability of each operation value of the operation item can be displayed.

The invention of the sixth aspect is the driving operation learning means 5 for learning the driving operation based on the data collected from the production equipment, and the driving operation support means 8 for supporting the driving operation based on the learning result of the data and the driving operation learning means. , A production equipment operation support method for supporting the operation of production equipment related to a system including the display means 4, wherein the operation operation learning means 5 is an operation for creating operation operation record information for learning past operation methods. Operation record information creation step (step S200), two-step learning step (step S500) to learn operation items and operation value selection methods separately and create each learning model based on operation record information, and production. In the target value prediction method learning step (step S800) for learning the target value prediction method in the above, information for presenting the result of the item selected based on the learning model and the target value prediction method is generated and displayed. The learning result display step (step S900) to be displayed in the above is executed, and the driving operation support means 8 includes the current driving information creation step (step S1000) for creating the current driving information based on the data, and the current driving information. The operation item is based on the operation plan creation step (step S1100) for creating the operation item plan and the operation value plan based on the operation information and the learning model created in the two-step learning step, and the prediction method of the target value prediction method creation step. The operation plan evaluation step (step S1500) that predicts and evaluates the target value when the plan and the operation value plan are applied, and the information for presenting the operation item plan and the operation value plan is created, and the recommended operation plan is created. The driving operation support information display means step (step S1600) for displaying the above is included in the display means. According to the present production equipment operation support method, it is possible to prevent a decrease in model accuracy and present an operation operation support plan by reducing the complexity of complicated production equipment and plant operation methods.

1 Production equipment 2 Control means 3 Production equipment Operation support means 4 Display means 5 Operation operation learning means 6 Data collection means 7 Database 8 Operation operation support means 10 Driving operation performance information creation means 11 Two-step learning means 12 Objective value prediction method Learning means 13 Learning result display means 20 Current operation information creation means 21 Operation operation plan creation means 22 Operation operation plan evaluation means 23 Operation support information display means 71 Production equipment information 72 Operation operation information 73 Operation item learning information 74 Operation value learning information 76 Current Operation information 77 Operation plan 78 Operation plan evaluation data 100 Production equipment operation support system M1 Operation item learning model M2 Operation value learning model M3 Objective value prediction learning model

Claims (6)

It is a production equipment operation support system that supports the operation of production equipment.
The production equipment operation support system is
Data collection means for collecting data from production equipment,
A driving operation learning means for learning a driving operation based on the above data,
A driving operation support means that supports a driving operation based on the data and the learning result of the driving operation learning means,
With display means,
The driving operation learning means is
Based on the above data, a driving operation performance information creating means for creating driving operation performance information for learning past operation methods, including the value of each measurement item, the value of each operation item, and the change value of each operation item. ,
Based on the driving operation record information, a two-step learning means that learns operation items and operation value selection methods separately and creates each learning model, and
Learning the method of predicting the target value in production.
It has a learning result display means that generates information for presenting the result of an item selected based on the learning model and the method of predicting the target value and displays it on the display means.
The driving operation support means is
Current operation information creation means for creating current operation information based on the above data,
A driving operation plan creating means for creating an operation item plan and an operation value plan based on the current driving information and the learning model created by the two-step learning means.
The operation plan evaluation means that predicts and evaluates the target value when the operation item plan and the operation value plan are applied based on the prediction method of the target value prediction method learning means, and the operation plan evaluation means.
It has a driving operation support information display means for creating information for presenting the operation item proposal and the operation value proposal and displaying the recommended operation proposal on the display means.
The two-step learning means has, as the first stage, operation item learning information including the value of each measurement item, the value of each operation item, and the presence / absence of operation of the target operation item based on the operation operation record information. , The operation item selection method is learned, and the operation item learning model is created. As the second step, the value of each measurement item, the value of each operation item, and the above Create an operation item learning information that includes the change value of the target operation item, learn how to select the operation value of the target operation item, and create an operation value learning model.
A production equipment operation support system characterized by this. In the production equipment operation support system according to claim 1,
The operation operation result information creating means is a production equipment operation support system characterized in that the change value of an operation item is calculated as the operation operation result information. In the production equipment operation support system according to claim 1 or 2.
The two-step learning means is a production equipment operation support system characterized by learning whether or not an operation item is operated and the operation value of the operation item. In the production equipment operation support system according to any one of claims 1 to 3.
The learning result display means is characterized in that it generates display information of a matching degree of an operation item, an operation presence / absence of an operation item, a matching degree of an operation value, an operation value, and a target value prediction formula, and displays the display information on the display means. Production equipment operation support system. In the production equipment operation support system according to any one of claims 1 to 4.
The driving operation support information display means has the selection probability of each operation item, the presence / absence of operation, the operation value, the selection probability of each operation value of the selected operation item, the predicted value of the target value, and the actual value so far as an operation recommendation plan. A production equipment operation support system characterized by generating display information of the above and displaying it on the display means. A system including a driving operation learning means for learning a driving operation based on data collected from a production facility, a driving operation supporting means for supporting a driving operation based on the learning result of the data and the driving operation learning means, and a display means. It is a production equipment operation support method that supports the operation of the production equipment.
The driving operation learning means is
Based on the above data, the operation record information creation step for creating the operation record information for learning the past operation method, which includes the value of each measurement item, the value of each operation item, and the change value of each operation item. ,
Based on the driving operation record information, a two-step learning step in which operation items and operation value selection methods are separately learned to create each learning model, and
Learning the method of predicting the target value in production The target value prediction method learning step and
The learning result display step of generating information for presenting the result of the item selected based on the learning model and the method of predicting the target value and displaying it on the display means is included and executed.
The driving operation support means is
The current operation information creation step for creating the current operation information based on the above data, and
A driving operation plan creation step for creating an operation item plan and an operation value plan based on the current driving information and the learning model created in the two-step learning step, and
The operation plan evaluation step for predicting and evaluating the target value when the operation item plan and the operation value plan are applied based on the prediction method of the target value prediction method learning step, and the operation plan evaluation step.
The operation item proposal and the information for presenting the operation value proposal are created, and the operation operation support information display means step for displaying the recommended operation proposal on the display means is included and executed .
As the first step, the two-step learning step is operation item learning information including the value of each measurement item, the value of each operation item, and the presence / absence of operation of the target operation item based on the operation performance record information. , The operation item selection method is learned, and the operation item learning model is created. As the second step, the value of each measurement item, the value of each operation item, and the above Create an operation item learning information that includes the change value of the target operation item, learn how to select the operation value of the target operation item, and create an operation value learning model.
A production equipment operation support method characterized by this.

Images