JP7401260B2 - Information processing system, information processing method and computer program - Google Patents

Description

The present invention relates to an information processing system, an information processing method, and a computer program.

In order for a plant to operate stably, it is required to properly manage and operate the plant. Stable operation of plants is often achieved through know-how such as experience and tacit technical knowledge cultivated over many years of work. However, it is predicted that there will be a shortage of personnel in the future to continue managing and operating the plant. For this reason, those who manage and operate plants run the risk of having difficulty passing on know-how such as experience and tacit technical knowledge to others. In plants, as described in Non-Patent Document 1, a monitoring system is utilized that centrally aggregates and processes on-site information to efficiently monitor and control the plant. Such a monitoring system is a system that aims to efficiently monitor and control a plant based on on-site information. However, the problem with the monitoring system is that it does not take into account the practical transfer of know-how such as experience and tacit technical knowledge in a one-stop system, nor the development of human resources. For this reason, there is a need to develop human resources while passing on know-how such as experience and tacit knowledge of technology in a one-stop system through daily routine work.

Additionally, systems that utilize AI (Artificial Intelligence) are being considered to efficiently manage and operate plants. However, conventional systems that utilize AI require new data to be collected using devices such as sensors and robots, and AI needs to be customized for each plant. In addition, in systems that utilize conventional AI, consideration is being given to using existing data or data collected in the past to make the AI learn. Systems that utilize AI cannot achieve sufficient accuracy with a small amount of data. For this reason, systems that utilize AI need to collect a huge amount of new data. For this reason, it takes a long time for systems that utilize AI to reach the level of accuracy that allows them to manage and operate plants. For this reason, there is a need for a system that utilizes AI that can use existing data, past data, and small amounts of data without customization while increasing accuracy by utilizing ICT technology. Further, the system is required to realize a CPS (Cyber Physical System) in accordance with improved accuracy.

Akira Watanabe, “Research on local government management support systems when outsourcing to the private sector”, [online], [searched on July 22, 2021], Internet <URL: https://www.jiwet.or.jp/quarterly /n005/pdf/n005-006.pdf>

In view of the above circumstances, the present invention provides an information processing system and an information processing system that can practically inherit know-how such as experience and tacit knowledge of technology in a one-stop system, and that appropriately manages and operates a plant using a small amount of data. The purpose is to provide processing methods and computer programs.

One aspect of the present invention includes a learned model generation unit that generates a learned model for each machine learning algorithm based on a plurality of teacher data indicating conditions related to plant control and information used for controlling the plant; Control information generation that generates control information indicating information for controlling the plant for each of the learned models based on condition information indicating the latest conditions regarding control of the plant and the plurality of learned models. a control information selection unit that selects control information indicating that the plant can be controlled from among the plurality of pieces of control information; and a control information selection unit that selects control information indicating that the plant can be controlled from among the plurality of control information; an analysis unit that receives an instruction to select the teacher data from a user based on control information, and the trained model generation unit generates the trained model based on the teacher data selected by the selection instruction. It is an information processing system.

One aspect of the present invention is the above information processing system, which includes a communication unit that receives the teacher data generated by a mobile terminal device operated by a user who performs work related to plant management from the mobile terminal device. Furthermore, in response to a request from the mobile terminal device, the communication unit transmits work information indicating the content of work related to the management of the plant in the form of a plurality of images to the mobile terminal device.

One aspect of the present invention is the above information processing system, in which the communication unit transmits the selected control information to the plant and acquires measurement information from the plant.

One aspect of the present invention is the above information processing system, in which the analysis unit estimates prediction accuracy of the selected control information based on the selected control information and the measurement information.

One aspect of the present invention is trained model generation in which a computer generates a trained model for each machine learning algorithm based on a plurality of teacher data indicating conditions related to plant control and information used for controlling the plant. control information indicating information for controlling the plant is generated for each learned model based on the step, condition information indicating the latest conditions regarding control of the plant, and the plurality of learned models. a control information generation step; a control information selection step of selecting control information indicating that the plant can be controlled from among the plurality of control information; and measurement information measured from the plant operated based on the selected control information. and an analysis step of receiving an instruction to select the teacher data from a user based on the selected control information, and in the trained model generation step, the learning is performed based on the teacher data selected by the selection instruction. This is an information processing method that generates a completed model.

One aspect of the present invention is a computer program for causing a computer to function as the above information processing system.

According to the present invention, it is possible to practically inherit know-how such as experience and tacit knowledge of technology in a one-stop system, and it is possible to appropriately manage and operate a plant using a small amount of data.

FIG. 1 is a system configuration diagram showing the system configuration of an information processing system 1 according to an embodiment. 1 is a functional block diagram showing a functional configuration of an information processing device 100 according to an embodiment. FIG. FIG. 2 is a functional block diagram showing the functional configuration of a mobile terminal device 200 according to an embodiment. It is a functional block diagram showing a functional configuration of a terminal device 300 according to an embodiment. FIG. 3 is a diagram illustrating a specific example of the degree of coincidence between control information and measurement information according to the embodiment. FIG. 3 is a diagram illustrating a specific example of information comparing control information and measurement information according to an embodiment. FIG. 7 is a diagram illustrating a specific example of features of the intermediate layer of the learned model according to the embodiment. FIG. 3 is a diagram illustrating a specific example of a screen related to work information according to the embodiment. FIG. 3 is a diagram illustrating a specific example of the flow of processing for generating control information according to the embodiment. FIG. 3 is a diagram illustrating a specific example of the process flow for analyzing measurement information according to the embodiment.

FIG. 1 is a system configuration diagram showing the system configuration of an information processing system 1 according to an embodiment. The information processing system 1 includes an information processing device 100, a mobile terminal device 200, and a terminal device 300. In the information processing system 1, control information is generated. Control information indicates information for controlling plant 400. For example, if the plant 400 is a plant related to sewage treatment, the control information may indicate the hypochlorite injection rate or the residual chlorine concentration. If the plant 400 is, for example, a plant related to iron manufacturing, the control information may indicate a blend of raw materials or a combustion pattern. Note that the information indicated by the control information is not limited to the above. The control information may indicate different information depending on the type of plant 400. Information processing system 1 may control plant 400 based on control information. Further, the user of the information processing system 1 may control the plant 400 based on the control information. In this embodiment, the plant 400 will be described as a facility related to sewage treatment. Note that the plant 400 may be composed of a plurality of plants, such as a first plant and a second plant.

The information processing device 100, the mobile terminal device 200, the terminal device 300, and the plant 400 may all be configured to be able to communicate via the network 500. The network 500 may be, for example, a wide area network (WAN) or a wide area communication network such as the Internet. The network 500 may be a network using wireless communication or a network using wired communication. Network 500 may be configured by combining a plurality of networks. The network 500 may be a closed communication network such as a VPN (Virtual Private Network). Note that the network 500 is only a specific example of a network for realizing communication between devices, and other configurations may be adopted as a network for realizing communication between devices. For example, communication between specific devices may be achieved using a different network from the network used for communication between other devices. Specifically, the communication between the information processing device 100 and the plant 400 may be realized by a different network from the communication between the information processing device 100, the mobile terminal device 200, and the terminal device 300.

FIG. 2 is a functional block diagram showing the functional configuration of the information processing device 100 according to the embodiment. The information processing device 100 is configured using an information processing device such as a personal computer or a server. The information processing device 100 is equipped with a function for generating control information (hereinafter referred to as "control information generation processing"). The function of the control information generation process may be implemented in the information processing apparatus 100 by hardware, or by installing software. The information processing device 100 may include a communication section 101, a teacher data storage section 102, a learned model storage section 103, a work information storage section 104, a measurement information storage section 105, and a control section 106.

The communication unit 101 may be a communication device such as a network interface. The communication unit 101 may be communicably connected to the network 500 using a predetermined protocol. The communication unit 101 may perform data communication with other devices via the network 500 under the control of the control unit 106.

The teacher data storage unit 102 may be configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The teacher data storage unit 102 may store teacher data received from the mobile terminal device 200. The training data may be data regarding example questions and answers in a machine learning algorithm. Further, the teacher data storage unit 102 may store information generated based on past information and data in advance as teacher data. The teacher data storage unit 102 may be configured to accumulate information and data generated in daily work as teacher data. The teacher data may be used to generate a trained model. The teacher data may be associated with predetermined information indicating the same content as the control information output by the trained model and information related to the predetermined information. The predetermined information may be, for example, information used to control the plant 400, such as hypochlorite injection rate and residual chlorine concentration. The predetermined information may be information corresponding to data regarding the answer of the teacher data. The information related to the predetermined information may be information related to conditions for controlling the plant 400 (hereinafter referred to as "condition information"). The information related to the predetermined information may be information corresponding to data related to the example problems of the teacher data. The condition information may indicate conditions that are taken into account to properly control the plant 400. The condition information may be a numerical value of an instrument installed in the plant 400 when the predetermined information is used to control the plant 400. The numerical values of the meters are, for example, the residual chlorine concentration, the flow rate of return water, the amount of residual salt in the return water, the amount of water distributed, the amount of hypochlorite remaining, the amount of content such as the amount of ammonia, etc. by the meters installed in the plant 400. Any numerical value may be used as long as it is a numerical value to be measured. Moreover, the condition information may be weather conditions. The weather condition may be, for example, weather, humidity, temperature, day/night, season, or precipitation. Condition information may differ depending on the type of plant 400 and the content of control. The teacher data may indicate different data depending on the type of plant 400. The teacher data storage unit 102 may be configured to store a plurality of pieces of teacher data.

The learned model storage unit 103 may be configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The trained model storage unit 103 may store a trained model generated based on teacher data. The trained model may be a trained machine learning model. A trained model may be generated for each machine learning algorithm. The trained model may be used to generate control information. The trained model storage unit 103 may store a plurality of trained models.

The work information storage unit 104 may be configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The work information storage unit 104 may store work information. The work information storage unit 104 may store a plurality of pieces of work information in advance. The work information may indicate the content of work related to the management of the plant 400. The work information may be information that shows a person working in the plant 400 how to perform the work. The work information may, for example, indicate the content of the work using a plurality of images. The work information may indicate different work contents for each piece of work information.

The measurement information storage unit 105 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device. The measurement information storage unit 105 stores a plurality of pieces of measurement information. The measurement information storage unit 105 may store a plurality of pieces of control information. The control information stored in the measurement information storage section 105 may be control information generated by the control information generation section 164 in the past. The measurement information is information measured from the plant 400 that operates based on the control information. The measurement information may indicate the same content as the control information. The measurement information storage unit 105 may store the control information generated on the same day and the measurement information acquired in association with each other.

The control unit 106 controls the operation of each unit of the information processing device 100. The control unit 106 is configured using a processor such as a CPU (Central Processing Unit) and a RAM (Random Access Memory). The control unit 106 functions as a communication control unit 161, a teacher data acquisition unit 162, a learned model generation unit 163, a control information generation unit 164, a control information selection unit 165, and an analysis unit 166 by the processor executing a specific program. Function.

The communication control unit 161 may communicate with other devices by executing a predetermined communication program. For example, the communication control unit 161 may receive teacher data from the mobile terminal device 200. The communication control unit 161 may record the received teacher data in the teacher data storage unit 102. Further, the communication control unit 161 may transmit work information to the mobile terminal device 200 in response to a request from the mobile terminal device 200. In this case, the communication control unit 161 may acquire the work information indicated by the request from the work information storage unit 104. The communication control unit 161 may transmit the acquired work information to the mobile terminal device 200. Further, the communication control unit 161 may receive a selection instruction from the terminal device 300. The selection instruction may indicate teacher data selected by the user from among the teacher data used to generate the learned model. Further, the communication control unit 161 may transmit control information to the plant 400. In this case, the plant 400 may operate based on information indicated by the control information.

The teacher data acquisition unit 162 may acquire teacher data used to generate a trained model. The teacher data acquisition unit 162 may acquire teacher data from the teacher data storage unit 102 that satisfies predetermined conditions regarding generation of a trained model (hereinafter referred to as “model generation conditions”). The model generation conditions will be explained. The model generation condition may be a condition for the teacher data acquisition unit 162 to determine which teacher data is used for machine learning among the teacher data stored in the teacher data storage unit 102. A method of clustering teacher data may be used as the model generation condition. A known clustering method may be used. For example, the teacher data acquisition unit 162 may cluster the teacher data based on date and time. The date and time may be, for example, the date and time when the teacher data was stored in the teacher data storage unit 102. The teacher data acquisition unit 162 may select teacher data clustered at a specific date and time from among the clustered teacher data as a target for machine learning. Note that the model generation conditions are not limited to date and time. For example, the teacher data acquisition unit 162 may cluster the teacher data based on any information indicated by the teacher data, such as the weather, water temperature, or contained substances, as the model generation condition. Model generation conditions may differ depending on the machine learning algorithm. In this case, the teacher data acquisition unit 162 may select teacher data for each machine learning algorithm. Further, one of the model generation conditions may be that the teacher data is specified by the selection instruction received from the terminal device 300.

The trained model generation unit 163 generates a trained model based on the teacher data. The trained model generation unit 163 stores a plurality of machine learning algorithms. The learned model generation unit 163 generates a learned model for each stored machine learning algorithm. The trained model generation unit 163 may record the generated trained model in the trained model storage unit 103. The machine learning algorithm may be any known machine learning algorithm such as SGD (Stochastic Gradient Descent), random forest, linear regression, decision tree, or CNN (Convolutional Neural Network).

The control information generation unit 164 generates control information based on the learned model and the latest condition information. The latest condition information is the condition information acquired last. The control information generation unit 164 generates control information by inputting condition information into the trained model. The control information generation unit 164 generates control information for each trained model. That is, the control information generation unit 164 generates a plurality of pieces of control information. Note that the control information generation unit 164 may acquire condition information from an external device via the network 500 when generating the control information. The external device may be, for example, the terminal device 300, a device provided in the plant 400, or a public server.

The control information selection unit 165 selects control information indicating that the plant 400 can be controlled from among the plurality of generated control information. For example, the control information selection unit 165 generates control information using a trained model that has the highest degree of agreement between control information generated by the same trained model within a predetermined past period and measurement information acquired within a predetermined past period. control information may be selected. The control information selection unit 165 may select control information indicating that the plant 400 can be controlled most stably. The past predetermined period may be, for example, one month or three months. Any period may be used as long as it is a period from when the teacher data is stored in the teacher data storage section 102 until the control information generation section 164 finally generates the control information.

FIG. 5 is a diagram illustrating a specific example of the degree of coincidence between control information and measurement information according to the embodiment. FIG. 5A shows a scatter diagram of control information generated by the first learned model and measurement information measured from the plant 400 that operated based on the control information. The graph on the left side of FIG. 5(a) shows the importance of each piece of information indicated by the teacher data. The degree of importance represents the degree of influence of training data on the machine learning algorithm. Information with a larger absolute value of importance has more influence on control information output by the learned model. FIG. 5A shows a specific example in which the absolute value of importance decreases in the order of water distribution amount, residual chlorine concentration, day and night, return water flow rate, residual salt amount of return water, precipitation amount, and ammonia amount. When the absolute value of the degree of importance is small, the information indicated by the teacher data may be configured to have no influence on the control information. The degree of importance is determined for each machine learning algorithm. The importance level may be changed according to an instruction from the user of the terminal device 300.

The graph on the right side of FIG. 5(a) (hereinafter referred to as the "first scatter diagram") is a scatter diagram of the control information output by the first learned model and the measurement information of the plant 400 at that time. show. The vertical axis of the first scatter diagram indicates control information. The horizontal axis of the first scatter diagram indicates measurement information. If the control information and measurement information have the same value, the samples are plotted on the line drawn on the scatter diagram. That is, the more information is plotted on or near the line on the scatter diagram, the higher the accuracy of the control information output by the first trained model becomes.

FIG. 5(b) shows a scatter diagram of the control information generated by the second learned model and the measurement information measured from the plant 400 that operated based on the control information. The graph on the left side of FIG. 5(b) shows the importance of each piece of information indicated by the teacher data. The degree of importance represents the degree of influence of training data on the machine learning algorithm. In FIG. 5(b), the absolute value of importance decreases in the order of water distribution amount, residual chlorine concentration, day/night, return water flow rate, residual salt amount of return water, precipitation amount, and ammonia amount.

The graph on the right side of FIG. 5(b) (hereinafter referred to as the "second scatter diagram") is a scatter diagram of the control information output by the second learned model and the measurement information of the plant 400 at that time. show. The vertical axis of the second scatter diagram indicates control information. The horizontal axis of the second scatter diagram indicates measurement information. If the control information and measurement information have the same value, the samples are plotted on the line drawn on the scatter diagram. That is, the more information that is plotted on or near the line on the scatter diagram, the higher the accuracy of the control information output by the second learned model.

The analysis unit 166 may be configured to perform settings based on the selection instruction received from the terminal device 300 so that the teacher data is used to generate a trained model from the next time onwards. For example, the analysis unit 166 may acquire measurement information from the plant 400. Next, the analysis unit 166 may acquire past control information generated by the trained model that generated the control information selected by the control information selection unit 165 from the measurement information storage unit 105. Furthermore, the analysis unit 166 may acquire measurement information from the measurement information storage unit 105 for the same period as the control information acquired from the measurement information storage unit 105. The analysis unit 166 may generate information that compares the control information and the measurement information. The analysis unit 166 may transmit the generated information to the terminal device 300.

FIG. 6 is a diagram illustrating a specific example of information comparing control information and measurement information according to the embodiment. FIG. 6 shows information obtained by comparing the measurement information and control information of the first plant and the second plant of the plant 400. The horizontal axis in FIG. 6 indicates the date. The vertical axis in FIG. 6 shows the hypochlorite injection rate [mg/l]. The diagram of the first plant in FIG. 6 shows that there are days, like region 10, from August 9, 2018 to around August 15, 2018, when the measurement information greatly exceeds the control information. The diagram of the second plant in FIG. 6 shows that there is a day, like region 20, around August 24, 2018, when the control information is significantly lower than the measurement information. In this way, areas 10 and 20 indicate that training data that lowers the accuracy of the learned model is used. Based on the information sent by the analysis unit 166, the user of the terminal device 300 can make predictions about data that will reduce the accuracy of the learned model. Note that, in addition to the information shown in FIG. 6, the analysis unit 166 may transmit a scatter diagram as shown in FIG. 5 to the terminal device 300.

Returning to FIG. 2, the explanation of the analysis section 166 will be continued. The analysis unit 166 may transmit information indicating the feature amount of the intermediate layer in the process of generating the learned model to the terminal device 300. This will be explained in detail below. The analysis unit 166 may acquire the teacher data used to generate the learned model that generated the control information selected by the control information selection unit 165 from the teacher data storage unit 102. The analysis unit 166 may acquire an intermediate layer generated in the process of generating a trained model based on teacher data. The analysis unit 166 may acquire the acquired feature amount of the intermediate layer. For example, the analysis unit 166 may reduce the dimensionality of the acquired intermediate layer to two dimensions using a dimensionality reduction method such as principal component analysis. The analysis unit 166 may acquire the feature amount by mapping the dimension-reduced intermediate layer onto a two-dimensional plane. The analysis unit 166 may transmit the acquired feature amount to the terminal device 300.

FIG. 7 is a diagram illustrating a specific example of the feature amount according to the embodiment. FIG. 7 shows the feature amount of the middle layer generated on August 24, 2018 of the second plant in FIG. 6. The feature amount may be a feature amount of an intermediate layer of the generated trained model. In FIG. 7, the horizontal axis represents the first principal component subjected to principal component analysis, and the vertical axis represents the second principal component subjected to principal component analysis. In FIG. 7, the teacher data may be grouped according to the types classified by the machine learning algorithm. According to FIG. 7, a small group can be seen in region 30. The number of teacher data mapped to the area 30 is still small. Therefore, such groups become a factor that reduces the accuracy of control information output by the trained model.

Returning to FIG. 2, the explanation of the analysis section 166 will be continued. The analysis unit 166 may receive an instruction to select teacher data from the terminal device 300. The analysis unit 166 may set the teacher data selected based on the received selection instruction to be used for generation of trained models from the next time onwards. This will be explained in detail below. The terminal device 300 that has received the information from the analysis unit 166 may accept a selection of teacher data from the user. The terminal device 300 may transmit a selection instruction indicating the selected teacher data to the information processing device 100. The analysis unit 166 may acquire the teacher data indicated by the selection instruction from the teacher data storage unit 102. The analysis unit 166 may set the acquired teacher data to be used for generating trained models from next time onwards. The analysis unit 166 may set the teacher data by setting a flag or increasing the priority, for example. Furthermore, the settings made by the analysis unit 166 are not permanent and may be deleted after a predetermined number of trained models have been generated.

FIG. 3 is a functional block diagram showing the functional configuration of the mobile terminal device 200 of the embodiment. The mobile terminal device 200 may be configured using a portable information processing device such as a tablet computer, a notebook computer, or a smartphone. The mobile terminal device 200 may be operated by a user who performs work related to the management of the plant 400. The mobile terminal device 200 may be equipped with a function for transmitting teacher data to the information processing device 100. The function for transmitting teacher data to the information processing device 100 may be implemented in the mobile terminal device 200 using hardware, or may be implemented by installing software. The mobile terminal device 200 may be configured to include a communication section 201, an input section 202, a display section 203, and a control section 204. Note that the mobile terminal device 200 of the embodiment may be any device as long as it is capable of collecting teacher data.

The communication unit 201 may be a communication device such as a network interface. The communication unit 201 may be communicably connected to the network 500 using a predetermined protocol. The communication unit 201 may perform data communication with other devices via the network 500 under the control of the control unit 204.

The input unit 202 may be configured using an input device such as a keyboard, a pointing device (mouse, tablet, etc.), a button, a touch panel, or the like. The input unit 202 may be operated by the user when inputting the user's instructions into the mobile terminal device 200. The input unit 202 may be an interface for connecting an input device to the mobile terminal device 200. In this case, the input unit 202 may input to the mobile terminal device 200 an input signal generated in response to a user's input on an input device.

The display unit 203 may be an image display device such as a liquid crystal display, an organic EL (Electro Luminescence) display, an electrophoretic display, or a CRT (Cathode Ray Tube) display. The display unit 203 may display images under the control of the control unit 204. The display unit 203 may be an interface for connecting an image display device to the mobile terminal device 200. In this case, the display unit 203 may generate a video signal for displaying an image according to the control of the control unit 204, and may output the video signal to an image display device connected to itself.

The control section 204 may control the operation of each section of the mobile terminal device 200. The control unit 204 may be configured using a processor such as a CPU and a RAM. The control unit 204 may be configured to function as the communication control unit 241, the teacher data generation unit 242, and the screen generation unit 243 when the processor executes a specific program.

The communication control unit 241 may communicate with other devices by executing a predetermined communication program. For example, the communication control unit 241 may transmit teacher data to the information processing device 100. Further, the communication control unit 241 may transmit a request for work information to the information processing apparatus 100. In this case, the communication control unit 241 may receive a request for work information via the input unit 202 operated by the user of the mobile terminal device 200. The communication control unit 241 may receive work information transmitted by the information processing device 100. Further, the communication control unit 241 may receive information obtained by comparing control information and measurement information, and feature amounts from the information processing device 100.

The teacher data generation unit 242 may generate teacher data based on information input by the user of the mobile terminal device 200. For example, in the plant 400, the user of the mobile terminal device 200 inputs date measurement information to the plant 400, information used for controlling the plant 400, such as numerical values of instruments installed in the plant 400, weather information, date and time, etc. , may be input into the mobile terminal device 200 in association with each other. The teacher data generation unit 242 may generate teacher data based on the input information. The input unit 202 may be used for input to the mobile terminal device 200.

The teacher data generation unit 242 may be configured to add attribute information to the teacher data. The attribute information may be information for uniquely identifying information input by the mobile terminal device 200. The attribute information may be, for example, an identifier such as a URI (Uniform Resource Identifier). The teacher data generation unit 242 may add attribute information using a known technique such as object storage, or may add attribute information when transmitting the teacher data to the information processing apparatus 100. By adding attribute information to the teacher data, the teacher data generation unit 242 can clarify the relationship between the teacher data and the relationship between existing information and the teacher data. Therefore, the teacher data generation unit 242 can handle information input via a plurality of different applications or information managed for each application in an integrated manner. Therefore, the teacher data generation unit 242 can generate teacher data based on information managed for each of a plurality of different applications.

The screen generation unit 243 may generate a screen based on information received from the information processing device 100. The screen generation unit 243 may output the generated screen to the display unit 203. For example, upon receiving work information from the information processing device 100, the screen generation unit 243 may generate a screen as shown in FIG.

FIG. 8 is a diagram showing a specific example of a screen related to work information according to the embodiment. FIG. 8 may be a screen for work information regarding the flow of maintenance work for the plant 400 (hereinafter referred to as a "work screen"). The display unit 203 may output a work screen. The work screen may be configured to include partial screens 231-236. The user of the mobile terminal device 200 can easily perform work on the plant 400 based on the work screen including the six partial screens 231 to 236.

The partial screen 231 may show a screen related to the preparation of tools necessary for the work. The partial screen 231 may show a screen in which tools necessary for the work are arranged. The user of the mobile terminal device 200 can grasp the tools required for the work based on the partial screen 231.

Partial screen 232 may show a screen related to the location where the work is performed. The partial screen 232 may show the appearance of the equipment of the plant 400 that is the work target and the location that is the work target. The user of the mobile terminal device 200 can understand where to perform the work based on the partial screen 232.

The partial screen 233 may show a screen related to the state in which tools used for work are worn. Partial screen 233 may show a worker and a tool attached to the worker. Based on the partial screen 233, the user of the mobile terminal device 200 can understand which tool should be worn and how.

The partial screen 234 may show a screen related to a state in which tools used for work are attached to equipment of the plant 400. Based on the partial screen 234, the user of the mobile terminal device 200 can understand which tool should be used to perform the task.

The partial screen 235 may show a screen related to operations associated with work. Based on the partial screen 235, the user of the mobile terminal device 200 can understand how to use tools to perform the work.

Partial screen 236 may show the screen after the work has been performed. The partial screen 236 may show the appearance of the equipment of the plant 400 that is the work target and the location that is the work target. Based on the partial screen 236, the user of the mobile terminal device 200 can understand what state should be in order for the work to be considered completed.

Note that the configuration of the work screen is not limited to that shown in FIG. In FIG. 8, the work screen is configured to include six partial screens, but the number of partial screens is not limited to six. The work screen may be configured to include any number of partial screens. The partial screen may be a video. For example, the partial screen 235 may be shown as a moving image of a screwdriver rotating a screw.

FIG. 4 is a functional block diagram showing the functional configuration of the terminal device 300 according to the embodiment. The terminal device 300 may be configured using an information processing device such as a personal computer, a tablet computer, or a server. The terminal device 300 may have a business application installed therein, such as a word processor or spreadsheet software. The terminal device 300 may be, for example, a device used by a user in daily work. The terminal device 300 may be equipped with a function for transmitting a selection instruction to the information processing device 100. The function for transmitting a selection instruction to the information processing device 100 may be implemented in the terminal device 300 using hardware, or may be implemented by installing software. The terminal device 300 may be configured to include a communication section 301, an input section 302, a display section 303, and a control section 304.

The communication unit 301 may be a communication device such as a network interface. The communication unit 301 may be communicably connected to the network 500 using a predetermined protocol. The communication unit 301 may perform data communication with other devices via the network 500 under the control of the control unit 304.

The input unit 302 may be configured using an input device such as a keyboard, a pointing device (mouse, tablet, etc.), buttons, a touch panel, or the like. The input unit 302 may be operated by the user when inputting the user's instructions into the terminal device 300. The input unit 302 may be an interface for connecting an input device to the terminal device 300. In this case, the input unit 302 may input to the terminal device 300 an input signal generated in response to a user's input on an input device.

The display unit 303 may be an image display device such as a liquid crystal display, an organic EL display, an electrophoretic display, or a CRT display. The display unit 303 may display images under the control of the control unit 304. The display unit 303 may be an interface for connecting an image display device to the terminal device 300. In this case, the display unit 303 may generate a video signal for displaying an image according to the control of the control unit 304, and may output the video signal to an image display device connected to itself.

The control section 304 may control the operation of each section of the terminal device 300. The control unit 304 may be configured using a processor such as a CPU and a RAM. The control unit 304 may be configured to function as the communication control unit 341 by a processor executing a specific program.

The communication control unit 341 may communicate with other devices by executing a predetermined communication program. For example, the communication control unit 341 may receive information indicating the degree of coincidence between the control information and the measurement information from the information processing device 100. In this case, the communication control unit 341 may output a screen as shown in FIG. 5 to the display unit 203. Further, the communication control unit 341 may receive information comparing control information and measurement information from the information processing device 100. In this case, the communication control unit 341 may output a screen as shown in FIG. 6 to the display unit 203. Further, the communication control unit 341 may receive information indicating the feature amount of the intermediate layer of the trained model from the information processing device 100. In this case, the communication control unit 341 may output a screen as shown in FIG. 7 to the display unit 203. Further, the communication control unit 341 may transmit a selection instruction to the information processing device 100. In this case, the communication control unit 341 may receive a selection instruction from the user of the terminal device 300 via the input unit 302. The selection instruction may indicate teacher data selected via the input unit 302. The user may select teacher data based on the information displayed on the display unit 203.

FIG. 9 is a diagram illustrating a specific example of the flow of processing for generating control information according to the embodiment. The process of generating control information is executed at a predetermined timing when controlling the plant 400. The predetermined timing may be, for example, once a day, once a week, or a predetermined timing. The control information generation unit 164 of the information processing device 100 acquires the latest condition information (step S101). For example, the control information generation unit 164 acquires condition information necessary for generation of control information from an external device via the network 500.

The teacher data acquisition unit 162 of the information processing device 100 acquires teacher data (step S102). For example, the teacher data acquisition unit 162 may acquire teacher data that satisfies the model generation conditions from the teacher data storage unit 102.

The trained model generation unit 163 of the information processing device 100 generates a trained model based on the acquired teacher data (step S103). Note that the trained model generation unit 163 stores a plurality of machine learning algorithms. Therefore, the trained model generation unit 163 generates a trained model based on a machine learning algorithm that has not yet been used.

The control information generation unit 164 generates control information based on the generated trained model and the latest condition information (step S104). The control information generation unit 164 determines whether control information has been generated for all learned models (step S105). For example, the control information generation unit 164 may generate a learned model for each machine learning algorithm stored in the learned model generation unit 163. Next, when the control information generation unit 164 generates the control information based on the latest control information and the generated trained models, the control information generation unit 164 may determine that the control information has been generated using all the trained models. . If there are trained models that have not yet been generated, the control information generation unit 164 may determine that control information has not been generated for all trained models.

If all trained models have not generated control information (step S105: NO), the process transitions to step S103. In this case, the trained model generation unit 163 may generate a trained model based on a machine learning algorithm that has not yet been used. If all the trained models have generated control information (step S105: YES), the control information selection unit 165 of the information processing device 100 selects one of the plurality of generated control information that indicates that the plant 400 can be controlled. Control information is selected (step S106). For example, the control information selection unit 165 may select control information indicating that the plant 400 can be controlled most stably from among the plurality of generated control information.

The communication control unit 161 of the information processing device 100 may transmit the selected control information to the plant 400 (step S107). The analysis unit 166 of the information processing device 100 may analyze the result of transmitting the control information to the plant 400 (step S108).

FIG. 10 is a diagram illustrating a specific example of the process flow for analyzing measurement information according to the embodiment. The measurement information analysis process may be executed at the timing when the measurement information is acquired from the plant 400 when the plant 400 is controlled, or may be executed at a predetermined timing after the measurement information is acquired. good. The analysis unit 166 acquires measurement information from the plant 400 (step S201). The analysis unit 166 generates information that compares the control information and measurement information (step S202). For example, the analysis unit 166 may acquire past control information generated by the trained model that generated the control information selected by the control information selection unit 165 from the measurement information storage unit 105. Next, the analysis unit 166 may acquire measurement information for the same period as the control information acquired from the measurement information storage unit 105. The analysis unit 166 may generate information that compares the control information and the measurement information. The analysis unit 166 may transmit the generated information to the terminal device 300.

The analysis unit 166 acquires the feature amount (step S203). For example, the analysis unit 166 may acquire the teacher data used to generate the trained model that generated the control information selected by the control information selection unit 165 from the teacher data storage unit 102. The analysis unit 166 may acquire an intermediate layer generated in the process of generating a trained model based on teacher data. For example, the analysis unit 166 may acquire feature amounts for the acquired intermediate layer using a dimension reduction method such as principal component analysis. The analysis unit 166 may transmit the acquired feature amount to the terminal device 300.

The analysis unit 166 receives an instruction to select teacher data from the terminal device 300 (step S204). The analysis unit 166 sets the acquired teacher data so that it will be used to generate trained models from next time onwards (step S205).

In the information processing system 1 configured in this way, control information regarding plant control is generated for each learned model. Next, the control information selection unit 165 selects control information that can control the plant 400 from among the generated control information. The analysis unit 166 may accept selection of teacher data based on the selected control information and measurement information measured from the plant. With this configuration, the user can understand which teacher data influences the accuracy of the trained model based on the control information and measurement information. Furthermore, the user may select teacher data that influences the accuracy of the learned model based on the control information and measurement information. In this way, by operating the information processing system 1, the user can manage, operate, and control the plant 400 in a one-stop system. Therefore, the information processing system 1 can inherit know-how such as experience and tacit technical knowledge related to the management and operation of the plant 400 and develop human resources through routine operations such as daily control of the plant 400. can.

The teacher data storage unit 102 is configured to store information generated based on past information and data as teacher data in advance, or to accumulate information and data generated in daily work as teacher data. may be done. With this configuration, the information processing system 1 can always generate a learned model based on new data. Therefore, the information processing system 1 can adapt to the environment such as the plant 400 and improve its accuracy while being used for normal work. Therefore, the user can use the information processing system 1 without customizing it for each environment.

Further, in the information processing system 1, the trained model generation unit 163 may generate a trained model based on important teacher data selected by the user. Therefore, by selecting important training data, the information processing system 1 is able to control the plant 400 with higher accuracy even with a small amount of data. Further, in the information processing system 1, by continuing machine learning using the selected teacher data, it becomes possible to manage, operate, and control the plant 400 with even higher accuracy. For example, when a substance is transferred from the plant 400 to another plant, the information processing system 1 exchanges information between the plants. Users control their respective management and operations based on the exchanged information. However, the information processing system 1 performs machine learning using training data that shows changes caused by the movement of substances, and can perform everything from management and operation to control of the plant 400 without the need to exchange information between plants. becomes possible. The information processing system 1 may be configured to provide each plant 400 with a learned model that has been machine learned to a level that does not require information exchange between plants. With this configuration, the information processing system 1 can realize a CPS that can independently control the plant 400 in place of the user.

Further, in the information processing system 1, the control information generation unit 164 selects control information indicating that the plant 400 can be controlled from the plurality of learned models. Therefore, in the information processing system 1, by selecting control information indicating that the plant 400 can be controlled, the plant 400 can be controlled with high accuracy even when the amount of training data used for machine learning is small. becomes possible.

Further, the information processing system 1 may acquire teacher data from the mobile terminal device 200. The mobile terminal device 200 may generate teacher data based on information input by the user of the mobile terminal device 200 through management of the plant 400. In this way, the information processing system 1 can acquire and store tacit knowledge such as know-how possessed by the user of the mobile terminal device 200 as teacher data.

Furthermore, the information processing system 1 does not have to control the plant 400 based on the generated control information. With this configuration, it becomes possible to separate the providers of the information processing system 1 and the plant 400. Therefore, vendor lock-in between the information processing system 1 and the plant 400 can be avoided. For example, the user can control plants 400 provided by different vendors based on the control information generated by the information processing system 1.

The analysis unit 166 may be configured to estimate the prediction accuracy of the generated control information. Prediction accuracy is an index indicating the closeness between generated control information and measurement information. Specifically, the analysis unit 166 may estimate the prediction accuracy based on the scatter diagram shown in FIG. 5. The analysis unit 166 may use a known method to estimate the degree of deviation between the case where the control information and the measurement information match and the actually plotted scatter diagram. A known technique may be, for example, to estimate a correlation coefficient. For example, the prediction accuracy of the first scatter diagram shown in FIG. 5 is 63%. Further, the prediction accuracy of the second scatter diagram shown in FIG. 5 is 93%. Between the prediction accuracy of the first scatter diagram and the prediction accuracy of the second scatter diagram, the second scatter diagram has higher prediction accuracy. This is thought to be because the amount of water distributed has a large effect on the prediction accuracy. In this case, the user may instruct the information processing device 100 to increase the importance of water distribution amount in the machine learning algorithm. With this configuration, the information processing system 1 can improve prediction accuracy.

The information processing device 100 may be configured to include a teacher data generation section 242. With this configuration, the information processing device 100 can generate teacher data based on information stored in another system. Specifically, the information processing device 100 acquires information from other systems. The teacher data generation unit 242 of the information processing device 100 generates teacher data based on the acquired information. The information stored in other systems may be existing information used to manage and operate the plant 400, for example. The information stored in other systems may be information that was used in the past to manage and operate the plant 400, for example.

The information processing device 100 may be implemented using a plurality of information processing devices communicatively connected via a network. In this case, each functional unit included in the information processing device 100 may be distributed and implemented in a plurality of information processing devices. For example, the teacher data acquisition unit 162, trained model generation unit 163, control information generation unit 164, control information selection unit 165, and analysis unit 166 may be implemented in different information processing devices.

The information processing device 100, the mobile terminal device 200, or the terminal device 300 in the embodiments described above may be realized by a computer. In that case, a program for realizing this function may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read into a computer system and executed. Note that the "computer system" herein includes hardware such as an OS and peripheral devices. Furthermore, the term "computer-readable recording medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into computer systems. Furthermore, a "computer-readable recording medium" refers to a storage medium that dynamically stores a program for a short period of time, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. It may also include a device that retains a program for a certain period of time, such as a volatile memory inside a computer system that is a server or client in that case. Further, the above-mentioned program may be one for realizing a part of the above-mentioned functions, or may be one that can realize the above-mentioned functions in combination with a program already recorded in the computer system. It may be realized using a programmable logic device such as an FPGA (Field Programmable Gate Array).

Although the embodiments of the present invention have been described above in detail with reference to the drawings, the specific configuration is not limited to these embodiments, and includes designs within the scope of the gist of the present invention.

DESCRIPTION OF SYMBOLS 1... Information processing system, 100... Information processing device, 101... Communication unit, 102... Teacher data storage part, 103... Learned model storage part, 104... Work information storage part, 105... Measurement information storage part, 106... Control part , 161...Communication control section, 162...Teacher data acquisition section, 163...Learned model generation section, 164...Control information generation section, 165...Control information selection section, 166...Analysis section, 200...Mobile terminal device, 201...Communication 202...Input unit, 203...Display unit, 204...Control unit, 241...Communication control unit, 242...Teacher data generation unit, 243...Screen generation unit, 300...Terminal device, 301...Communication unit, 302...Input unit , 303...Display unit, 304...Control unit, 341...Communication control unit, 400...Plant, 500...Network

Claims (10)

a trained model generation unit that generates a plurality of trained models using a plurality of machine learning algorithms based on a plurality of teacher data indicating conditions related to plant control and information used for controlling the plant;
a control information generation unit that generates a plurality of pieces of control information indicating information for controlling the plant based on condition information indicating the latest conditions regarding control of the plant and the plurality of learned models;
a control information selection unit that selects control information indicating that the plant can be controlled from among the plurality of control information;
an analysis unit that receives an instruction to select the teacher data from a user based on measurement information measured from a plant operated based on the selected control information and the selected control information;
a communication unit that transmits information from the analysis unit to a mobile terminal device operated by the user and receives information from the mobile terminal device;
Equipped with
The analysis unit acquires features of an intermediate layer generated in the process of generating the plurality of trained models, and transmits the acquired features of the intermediate layer to the mobile terminal device,
The mobile terminal device displays the feature amount of the intermediate layer transmitted from the analysis unit,
The trained model generation unit generates at least one of the plurality of trained models based on the teacher data selected by the selection instruction.
Information processing system. The mobile terminal device maps and displays the feature amount of the intermediate layer on a two-dimensional plane.
The information processing system according to claim 1. The communication unit transmits, to the mobile terminal device, work information indicating the content of the work related to the management of the plant in a plurality of images in response to a request from the mobile terminal device.
The information processing system according to claim 1 or 2 . The communication unit transmits the selected control information to the plant and acquires measurement information from the plant.
The information processing system according to claim 3 . The analysis unit estimates prediction accuracy of the selected control information based on the selected control information and the measurement information.
The information processing system according to any one of claims 1 to 4 . an information processing device having a first control section, a first communication section, and a storage section;
The mobile terminal device is a mobile terminal device operated by a user and has a second control section, a second communication section, a display section, and an input section;
An information processing system comprising:
The first control unit includes:
a trained model generation step of generating a plurality of trained models using a plurality of machine learning algorithms based on a plurality of training data indicating conditions related to plant control and information used for controlling the plant;
a control information generation step of generating a plurality of pieces of control information indicating information for controlling the plant based on condition information indicating the latest conditions regarding control of the plant and the plurality of learned models;
a control information selection step of selecting control information indicating that the plant can be controlled from among the plurality of control information;
an analysis step of receiving an instruction to select the teacher data from a user based on measurement information measured from a plant operated based on the selected control information and the selected control information;
an importance output step of outputting the importance, which is the importance of information indicated by the teaching data and represents the degree of influence of the teaching data on the machine learning algorithm;
has
The plurality of trained models include a first trained model and a second trained model,
The second control unit displays first information indicating the degree of importance based on the first trained model and second information indicating the degree of importance based on the second trained model on the display unit. a display step for displaying the
in the trained model generation step, generating at least one of the trained models based on the teacher data selected by the selection instruction;
Information processing system. The displaying step displays the first information and the second information as a scatter diagram or a graph on the display unit.
The information processing system according to claim 6. The computer is
a trained model generation step of generating a plurality of trained models using a plurality of machine learning algorithms based on a plurality of training data indicating conditions related to plant control and information used for controlling the plant;
a control information generation step of generating a plurality of pieces of control information indicating information for controlling the plant based on condition information indicating the latest conditions regarding control of the plant and the plurality of learned models;
a control information selection step of selecting control information indicating that the plant can be controlled from among the plurality of control information;
an analysis step of receiving an instruction to select the teacher data from a user based on measurement information measured from a plant operated based on the selected control information and the selected control information;
a display step of displaying intermediate layer feature amounts generated in the process of generating the plurality of trained models on a mobile terminal device operated by the user;
has
In the trained model generation step, the trained model is generated based on the teacher data selected by the selection instruction.
Information processing method. The displaying step displays the feature amount of the intermediate layer by mapping on a two-dimensional plane.
The information processing method according to claim 8. A computer program for causing a computer to function as the information processing system according to claim 1 .

Images