Devices and Systems

Abstract

To provide an information processing system performing communication between a system within an intranet of a specific organization and a system of the specific organization placed outside the intranet, through a server connected by the Internet outside the intranet.SOLUTION: An information processing system 1 comprises: plant systems 4a-4c within an intranet of a specific organization; a plant system 4d of the specific organization placed outside the intranet; and a cloud server connected by the plant system 4d and the Internet outside the intranet. The plant system 4d communicates with the plant systems 4a-4c through the cloud server outside the intranet. The plant systems 4a-4c perform processing based on first information obtained from the plant system 4d through the cloud server, and generate a first model that is a machine learning model or a statistical analysis model, based on the first information.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to an information processing system. [Background technology]

[0002] Patent Document 1 discloses a technique related to an information processing system. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2019-33369 Summary of the Invention

[0004] In one embodiment, the device includes a control unit that acquires a first captured image of the wastewater in the first tank and determines a first dosage of separating agent to be dispensed into the first tank based on the first captured image. [Brief explanation of the drawings]

[0005] [Figure 1] FIG. 1 illustrates an example of the configuration of an information processing system. [Figure 2] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 3] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 4] FIG. 2 illustrates an example of the configuration of a server. [Figure 5] FIG. 1 illustrates an example of the configuration of a device gateway. [Figure 6] FIG. 2 is a diagram illustrating an example of the configuration of a cloud server. [Figure 7] FIG. 2 is a diagram illustrating an example of the configuration of a business division server. [Figure 8] FIG. 1 is a diagram illustrating an example of the operation of a factory system. [Figure 9] FIG. 1 is a diagram illustrating an example of the operation of an information processing system. [Figure 10] FIG. 1 is a diagram illustrating an example of the operation of an information processing system. [Figure 11] FIG. 1 is a diagram illustrating an example of the operation of an information processing system. [Figure 12] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 13] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 14] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 15] FIG. 1 is a diagram illustrating an example of the configuration of a factory system. [Figure 16] FIG. 1 is a diagram illustrating an example of the operation of an information processing system. [Figure 17] FIG. 1 is a diagram illustrating an example of the operation of an information processing system. DETAILED DESCRIPTION OF THE INVENTION

[0006] FIG. 1 is a block diagram showing an example configuration of an information processing system 1. The information processing system 1 is, for example, an information processing system including an intranet of a specific organization. The specific organization refers to, for example, a company. The information processing system 1 is, for example, an information processing system including an intranet of a company that has multiple factories. The information processing system 1 is a type of IoT (Internet of Things) system. The information processing system 1 can be said to be a factory IoT system.

[0007] As shown in FIG. 1, an information processing system 1 includes an intranet 2 owned by a certain company (specific organization). Hereinafter, this certain company may be referred to as a specific company. The specific company may include, for example, multiple factories A to D located in different locations. Note that the number of factories owned by the specific company is not limited to this.

[0008] An information processing system 3 is provided within the intranet 2. The information processing system 3 includes a factory A system 4a, a factory B system 4b, a factory C system 4c, and an on-premise system 5. The factory A system 4a is an information processing system provided in factory A. The factory B system 4b is an information processing system provided in factory B. The factory C system 4c is an information processing system provided in factory C. The on-premise system 5 is a core system owned by a specific company, and is a type of information processing system.

[0009] The factory A system 4a, factory B system 4b, and factory C system 4c are connected to one another. The factory A system 4a, factory B system 4b, factory C system 4c, and on-premise system 5 are connected to a system I / F 20. I / F stands for interface. The system I / F 20 has a function for connecting multiple systems within the intranet 2 and a function for connecting a system within the intranet 2 with a device outside the intranet 2. Each of the factory A system 4a, factory B system 4b, and factory C system 4c can communicate with the on-premise system 5 through the system I / F 20.

[0010] In addition to the intranet 2, the information processing system 1 also includes a factory D system 4d, servers 10, 11, and 12 (cloud servers 10, 11, and 12 in this embodiment), and firewalls 14, 15, and 16. The factory D system 4d, the cloud servers 10, 11, and 12, and the firewalls 14, 15, and 16 exist outside the intranet 2 and are connected to the Internet.

[0011] Factory D system 4d is an information processing system provided in factory D of a specific company. Although factory D system 4d is a system provided by a specific company, it is located outside intranet 2. Each of firewalls 13, 14, and 15 is a type of computer device. Each of cloud servers 10, 11, and 12 is a type of computer device.

[0012] Firewall 13 is connected to intranet 2 and cloud server 10. Firewall 14 is connected to factory D system 4d and cloud server 10. Firewall 15 is connected to cloud server 10 and system I / F 20. Cloud servers 11 and 12 are connected to system I / F 20.

[0013] The factory D system 4d is able to communicate with each of the factory A system 4a, the factory B system 4b, and the factory C system 4c within the intranet 2 through the firewall 14, the cloud server 10, and the firewall 13.

[0014] Hereinafter, when there is no need to particularly distinguish between the factory A system 4a, factory B system 4b, factory C system 4c, and factory D system 4d, they may each be referred to as a factory system 4. Furthermore, a factory system 4 within the intranet 2 may be referred to as an intranet factory system 4, and a factory system 4 outside the intranet 2 may be referred to as an intranet external factory system 4. Furthermore, the factory D system 4d may be referred to as an intranet external factory system 4d.

[0015] Cloud server 10 is managed and operated, for example, by a company other than a specific company. Cloud server 10 is capable of communicating with factory D system 4d through firewall 14. Cloud server 10 is also capable of communicating with factory A system 4a, factory B system 4b, and factory C system 4c within intranet 2 through firewall 13. Cloud server 10 is also capable of communicating with on-premise system 5 within intranet 2 through firewall 15 and system I / F 20.

[0016] In the cloud server 10, a cloud P / F 100 that provides at least one type of service is constructed by the cooperation of hardware and software included in the cloud server 10. P / F means platform. The cloud P / F 100 includes, for example, a machine learning engine 101. The machine learning engine 101 can also be said to be a machine learning device. The machine learning engine 101 generates a machine learning model that outputs an inference result. In other words, the machine learning engine 101 trains the machine learning model based on input information to generate a trained machine learning model. The machine learning engine 101 then inputs information to the trained machine learning model and uses the information output from the trained machine learning model as an inference result. Hereinafter, the machine learning model may be referred to as a trained model. The trained machine learning model may also be referred to as a trained model. The trained model may also be referred to as an inference model.

[0017] The machine learning engine 101 uses, for example, deep learning. In this case, the learning model includes a neural network. The machine learning engine 101 learns the parameters of the neural network to generate a learned model. The parameters of the neural network include weighting coefficients that indicate the weights of connections between artificial neurons. The machine learning engine 101 may perform unsupervised learning or supervised learning. The machine learning engine 101 may also perform reinforcement learning. A combination of deep learning and reinforcement learning is called deep reinforcement learning. A learning model that includes a neural network is sometimes called an AI (Artificial Intelligence) model. Note that the machine learning engine 101 may use a machine learning method other than deep learning.

[0018] The cloud server 11 is managed and operated by, for example, a company other than a specific company. The cloud server 11 is capable of communicating with each factory system 4 and the on-premise system 5 within the intranet 2 via the system I / F 20. In the cloud server 11, a cloud P / F 110 that provides at least one type of service is constructed by the cooperation of hardware and software provided in the cloud server 11. The cloud P / F 110 includes, for example, a machine learning engine 111. The performance of the machine learning engine 111 is different from the performance of the machine learning engine 101, for example.

[0019] The machine learning engine 111 uses, for example, deep learning. In this case, the machine learning engine 111 learns parameters of a neural network included in a learning model to generate a learned model. The machine learning engine 111 may perform unsupervised learning or supervised learning. The machine learning engine 111 may also perform reinforcement learning. Note that the machine learning engine 111 may use a method other than deep learning.

[0020] The cloud server 12 is managed and operated, for example, by a company other than a specific company. The cloud server 12 is capable of communicating with each factory system 4 and the on-premise system 5 within the intranet 2 via the system I / F 20. In the cloud server 12, a cloud P / F 120 that provides at least one type of service is constructed by the cooperation of hardware and software included in the cloud server 12. The cloud P / F 120 includes, for example, a statistical analysis engine 121. The statistical analysis engine 121 can also be considered a statistical analysis device. The statistical analysis engine 121 can perform, for example, regression analysis. The statistical analysis engine 121 performs statistical analysis on input information to generate a statistical analysis model that outputs predetermined information. The statistical analysis model outputs predetermined information according to the input information.

[0021] At least one of the cloud servers 10, 11, and 12 may be managed and operated by a specific company.

[0022] <Factory system configuration example> Next, we will explain an example of the configuration of the factory system 4. Each factory system 4 can transmit information obtained from devices such as sensors to the cloud server 10. Furthermore, each factory system 4 can control devices based on the information from the cloud server 10.

[0023] Fig. 2 is a block diagram showing an example of the configuration of a factory system 4. The multiple factory systems 4 included in the information processing system 1 have similar configurations. In Fig. 2, to avoid complicating the drawing, the configuration of the factory A system 4a is shown in detail.

[0024] As shown in Fig. 2, each factory system 4 includes a server 40, multiple device gateways 41, and multiple devices 42. The server 40 and each device gateway 41 are a type of computer device. The computer performance of the server 40 is, for example, higher than the computer performance of the device gateway 41. Each of the multiple devices 42 is connected to one of the multiple device gateways 41. Hereinafter, the device gateway 41 may be referred to as a device GW 41.

[0025] The information processing system 1 of this example employs fog computing. Fog computing refers to a processing environment located close to a terminal in a network environment, for example. A fog P / F 400 that controls each device GW 41 is established in the server 40 by cooperation of hardware and software provided in the server 40. The fog P / F 400 manages the factory system 4 to which it belongs. The fog P / F 400 of each intra-factory system 4 can communicate with the cloud P / F 100 through a firewall 13. The servers 40 of multiple intra-factory systems 4 are connected to each other. The fog P / Fs 400 of multiple intra-factory systems 4 can communicate with each other. The server 40 of each intra-factory system 4 is connected to a system I / F 20. The fog P / F 400 of each intra-factory system 4 can communicate with the on-premise system 5, the cloud P / F 110, and the cloud P / F 120 through the system I / F 20.

[0026] The fog P / F 400 of the intra-external factory system 4d can communicate with the cloud P / F 100 through the firewall 14. The fog P / F 400 of the intra-external factory system 4d can communicate with the fog P / F 400 of each intra-external factory system 4 through the firewall 14, the cloud server 10, and the firewall 13. The fog P / F 400 of the intra-external factory system 4d can communicate with the on-premise system 5, the cloud P / F 110, and the cloud P / F 120 through the firewall 14, the cloud server 10, the firewall 13, the fog P / F 400 of the intra-external factory system, and the system I / F 20. Note that the fog P / F 400 of the intra-external factory system 4d may also be able to communicate with the on-premise system 5, the cloud P / F 110, and the cloud P / F 120 through the firewall 14, the cloud server 10, the firewall 15, and the system I / F 20.

[0027] In this way, in the information processing system 1, the fog P / Fs 400 and cloud P / Fs 100 of multiple factory systems 4 can exchange information with each other. In addition, the fog P / Fs 400 of each factory system 4 can exchange information with the on-premise system 5, the cloud P / Fs 110, and the cloud P / Fs 120. The fog P / Fs 400 may operate, for example, based on a rule base. In this case, the fog P / Fs 400 operate according to preset rules.

[0028] At least one device 42 is connected to each device GW 41. The device GW 41 transmits information 420 obtained from at least one device 42 connected to it to the fog P / F 400. The device GW 41 may also control the device 42 connected to it based on instructions from the fog P / F 400.

[0029] At least one device 42 connected to one device gateway 41 includes, for example, at least one of a sensor and a programmable logic controller (PLC). A PLC is sometimes called a sequencer. Hereinafter, the sensor serving as the device 42 may be referred to as a sensor 42a, and the PLC serving as the device 42 may be referred to as a PLC 42b (see FIG. 2).

[0030] The sensor 42a performs a detection operation, for example, once every few seconds, and outputs the detection information. The multiple devices 42 included in the factory system 4 may include at least one of a temperature sensor, a humidity sensor, an acceleration sensor, a vibration sensor, a pH sensor, an illuminance sensor, a position sensor, an air pressure sensor, a water level sensor, a flow rate sensor, a voltage sensor, a current sensor, an electrical resistance sensor, a microphone, and a camera. However, the types of the sensors 42a included in the factory system 4 are not limited to these.

[0031] The PLC 42b can control at least one control target device in the factory. The PLC 42b controls the at least one control target device based on the outputs of switches, sensors, etc. The at least one control target device controlled by the PLC 42b may include at least one of a lamp, a buzzer, a motor, a valve, a pump, a sensor, and a robot. The device GW 41 can acquire information about the devices connected to it from the PLC 42b. For example, the device GW 41 can acquire from the PLC 42b the on / off state of a switch, the lighting state of a lamp, whether a buzzer is sounding, the motor rotation speed, the valve opening, the operating status of a pump, sensor detection information, alarm information output by the sensor, the on / off state of a sensor, or the operating status of a robot. However, the types of devices controlled by the PLC 42b are not limited to these.

[0032] The number of device GWs 41 included in a factory system 4 is set for each factory. Therefore, the number of device GWs 41 included in a factory system 4 may differ between at least two factory systems 4. Furthermore, the number and types of devices 42 included in a factory system 4 are set for each factory. Therefore, the number and types of devices 42 included in a factory system 4 may differ between at least two factory systems 4.

[0033] The fog P / F 400 is aware of the status of the factory system 4 that it manages. The status of the factory system 4 that the fog P / F 400 is aware of includes, for example, what device GWs 41 are connected to the fog P / F 400 and what operations the device GWs 41 are performing. The status of the factory system 4 that the fog P / F 400 is aware of also includes what devices 42 are connected to the device GWs 41 connected to the fog P / F 400 and what operations the devices 42 are performing.

[0034] Note that at least one of the multiple factory systems 4 may be provided with a device GW 41 that is not controlled by the server 40. FIG. 3 is a diagram illustrating a configuration example of a factory system 4 including a device GW 41 that is not controlled by the server 40. In the example of FIG. 3, the fog P / F 40a of the server 40 cannot control one device GW 41a included in the multiple device GWs 41. When the device GW 41a is provided in the intranet factory system 4, the device GW 41a communicates with the cloud server 10 through the firewall 13. On the other hand, when the device GW 41a is provided in the external intranet factory system 4d, the device GW 41a communicates with the cloud server 10 through the firewall 14. The device GW 41a can transmit information 420 obtained from the device 42 to the cloud P / F 100. In the following description, it is assumed that the configuration of the factory D system 4d outside the intranet 2 is the configuration shown in FIG. 3.

[0035] When the multiple devices 42 included in the factory system 4 include multiple PLCs 42b, a master PLC to which the multiple PLCs 42b are connected may be connected to the device GW 41. In this case, the device GW 41 can acquire information 420 obtained from the PLC 42b through the master PLC.

[0036] FIG. 4 is a block diagram showing an example configuration of the server 40. FIG. 5 is a block diagram showing an example configuration of the device GW 41. As shown in FIG. 4, the server 40 includes a control unit 401 and communication units 405 and 406. The control unit 401 can comprehensively manage the operation of the server 40 by controlling the other components of the server 40. The control unit 401 can also be referred to as a control device or control circuit. The control unit 401 includes at least one processor to provide control and processing capabilities for performing various functions, as will be described in further detail below.

[0037] According to various embodiments, the at least one processor may be implemented as a single integrated circuit (IC) or as multiple communicatively connected integrated circuits ICs and / or discrete circuits. The at least one processor may be implemented according to various known techniques.

[0038] In one embodiment, a processor includes one or more circuits or units configured to perform one or more data computational procedures or processes, for example, by executing instructions stored in associated memory. In other embodiments, a processor may be firmware (e.g., discrete logic components) configured to perform one or more data computational procedures or processes.

[0039] According to various embodiments, the processor may include one or more processors, controllers, microprocessors, microcontrollers, application specific integrated circuits (ASICs), digital signal processors, programmable logic devices, field programmable gate arrays, or any combination of these devices or configurations, or other known devices and configurations, to perform the functions described below.

[0040] In this example, the control unit 401 includes a central processing unit (CPU) 402 and a storage unit 403. The storage unit 403 includes a non-transitory recording medium that can be read by the CPU 402, such as a read-only memory (ROM) and a random access memory (RAM). The storage unit 403 stores a control program 403a for controlling the server 40. The CPU 402 executes the control program 403a stored in the storage unit 403 to realize various functions of the server 40. In the server 40, the CPU 402 executes the control program 403a to configure a fog P / F 400 as a functional block.

[0041] The configuration of the control unit 401 is not limited to the above example. For example, the control unit 401 may include multiple CPUs 402. The control unit 401 may also include at least one DSP (Digital Signal Processor). All or some of the functions of the control unit 401 may be realized by a hardware circuit that does not require software to realize the function. The storage unit 403 may also include a computer-readable non-transitory recording medium other than ROM and RAM. The storage unit 403 may also include, for example, a small hard disk drive and an SSD (Solid State Drive).

[0042] The communication unit 405 is capable of communicating with the cloud server 10 through a firewall. The communication unit 405 may be connected to the firewall using at least one of a wired connection and a wireless connection. The communication unit 405 is also capable of communicating with the on-premise system 5 and the cloud servers 11 and 12 through the system I / F 20. The communication unit 406 is capable of communicating with multiple device GWs 41. The communication unit 406 may be connected to multiple device GWs 41 using at least one of a wired connection and a wireless connection. The communication units 405 and 406 can also be considered communication circuits.

[0043] The configuration of server 40 is not limited to the above example. For example, server 40 may include a display controlled by control unit 401. Server 40 may also include an operation unit that accepts user operations. The operation unit may include a mouse, a keyboard, or a touch panel.

[0044] 5, the device GW41 includes a control unit 411, a communication unit 415, and a device I / F 416. The control unit 411 is capable of overall management of the operation of the device GW41 by controlling the other components of the device GW41. The control unit 411 includes at least one processor to provide control and processing capabilities for executing various functions, as will be described in further detail below. The above description of the processor included in the control unit 401 of the server 40 can also be applied to the processor included in the control unit 411.

[0045] In this example, the control unit 411 includes a CPU 412 and a storage unit 413. The storage unit 413 includes a non-transitory recording medium such as a ROM and a RAM that can be read by the CPU 412. The storage unit 413 stores a control program 413a for controlling the device GW 41, etc. Various functions of the device GW 41 are realized by the CPU 412 executing the control program 413a in the storage unit 413.

[0046] The configuration of the control unit 411 is not limited to the above example. For example, the control unit 411 may include multiple CPUs 412. The control unit 411 may also include at least one DSP. All or some of the functions of the control unit 411 may be realized by a hardware circuit that does not require software to realize the function. The storage unit 413 may also include a computer-readable non-transitory recording medium other than ROM and RAM.

[0047] The communication unit 415 is capable of communicating with the communication unit 406 of the server 40. At least one device 42 is connected to the device I / F 460. The control unit 411 can acquire information 420 obtained by the device 42 through the device I / F 460. The control unit 411 may also control the device 42 through the device I / F 460. The device I / F 460 may also be referred to as a device I / F circuit.

[0048] The configuration of the device GW 41 is not limited to the above example. For example, the device GW 41 may include a display controlled by the control unit 411. The device GW 41 may also include an operation unit that accepts user operations.

[0049] <Cloud server configuration example> Next, a description will be given of an example of the configuration of the cloud servers 10, 11, and 12. The cloud servers 10, 11, and 12 have, for example, the same hardware configuration. FIG. 6 is a block diagram showing an example of the configuration of the cloud servers 10, 11, and 12.

[0050] As shown in Figure 6, each of cloud servers 10, 11, and 12 includes a control unit 151 and a communication unit 155. The control unit 151 is capable of overall management of the operation of the cloud server by controlling the other components of the cloud server. As will be described in more detail below, the control unit 151 includes at least one processor to provide control and processing capabilities for executing various functions. The above description of the processor included in the control unit 401 of server 40 also applies to the processor included in the control unit 151.

[0051] In this example, the control unit 151 includes a CPU 152 and a storage unit 153. The storage capacity of the storage unit 153 is larger than, for example, the storage capacity of the storage unit 403 of the server 40 of the factory system 4. The storage unit 153 includes a non-transitory recording medium readable by the CPU 152, such as a ROM and a RAM. The storage unit 153 stores a control program 153a for controlling the cloud server. Various functions of the cloud server are realized when the CPU 152 executes the control program 153a stored in the storage unit 153. In the cloud server 10, the CPU 152 executes the control program 153a to construct a cloud P / F 100 as a functional block. In the cloud server 11, the CPU 152 executes the control program 153a to construct a cloud P / F 110. In the cloud server 12, the CPU 152 executes the control program 153a to construct a cloud P / F 120.

[0052] The configuration of the control unit 151 is not limited to the above example. For example, the control unit 151 may include multiple CPUs 152. The control unit 151 may also include at least one DSP. All of the functions of the control unit 151 or some of the functions of the control unit 411 may be realized by a hardware circuit that does not require software to realize the functions. The storage unit 153 may also include a computer-readable non-transitory recording medium other than ROM and RAM.

[0053] The communication unit 155 is connected to the Internet. The communication unit 155 can acquire various information via the Internet. The communication unit 155 of the cloud server 10 can communicate with at least one of the communication unit 405 of the server 40 of each factory system 4 and the communication unit 415 of the device GW 41 via a firewall. In addition, the communication unit 155 of the cloud server 10 can communicate with the on-premise system 5 via the firewall 15 and the system I / F 20.

[0054] The communication units 155 of the cloud servers 11 and 12 are connected to the system I / F 20. The communication units 155 of the cloud servers 11 and 12 can communicate with the communication units 405 of the servers 40 of each intra-factory system 4 and the on-premise system 5 through the system I / F 20. The communication units 155 of the cloud servers 11 and 12 can also communicate with the communication unit 405 of the server 40 of the external intra-factory system 4d through the system I / F 20, the server 40 of the intra-factory system 4, the firewall 13, the cloud server 10 and the firewall 14. The communication units 155 of the cloud servers 11 and 12 may also be able to communicate with the communication unit 405 of the server 40 of the external intra-factory system 4d through the system I / F 20, the firewall 15, the cloud server 10 and the firewall 14.

[0055] The communication unit 155 may communicate with a device not included in the information processing system 1. For example, the communication unit 155 may communicate with a server that is connected to the Internet and provides weather information. In this case, the control unit 151 may store the weather information received by the communication unit 155 in the storage unit 153. The communication unit 155 may communicate with a server that is connected to the Internet and provides economic information. In this case, the control unit 151 may store the economic information received by the communication unit 155 in the storage unit 153.

[0056] The configuration of cloud servers 10, 11, and 12 is not limited to the above example. For example, at least one of cloud servers 10, 11, and 12 may include a display controlled by control unit 151. Furthermore, at least one of cloud servers 10, 11, and 12 may include an operation unit that accepts user operations. Furthermore, at least two of cloud servers 10, 11, and 12 may have different hardware configurations.

[0057] <Example of on-premise system configuration> Next, a configuration example of the on-premise system 5 will be described. As shown in FIG. 1, the on-premise system 5 includes, for example, a business division server 50 provided for each business division of a specific company. The business division server 50 is a type of computer device. The business division server 50 manages production management information, accounting information, and technical information for the corresponding business division.

[0058] FIG. 7 is a block diagram showing an example configuration of the business department server 50. As shown in FIG. 7, the business department server 50 includes, for example, a control unit 501 and a communication unit 505. The control unit 501 is capable of overall management of the operation of the business department server 50 by controlling the other components of the business department server 50. The control unit 501 includes at least one processor to provide control and processing capabilities for executing various functions, as will be described in further detail below. The above description of the processor included in the control unit 401 of the server 40 also applies to the processor included in the control unit 501.

[0059] In this example, the control unit 501 includes a CPU 502 and a memory unit 503. The memory unit 503 includes a non-transitory recording medium such as a ROM and a RAM that can be read by the CPU 502. The memory unit 503 stores a control program 503a for controlling the business department server 50. The memory unit 503 also includes a production management information DB 503b that stores production management information, an accounting information DB 503c that stores accounting information, and a technical information DB 503d that stores technical information. DB stands for database. The various functions of the business department server 50 are realized by the CPU 502 executing the control program 503a in the memory unit 503.

[0060] The configuration of the control unit 501 is not limited to the above example. For example, the control unit 501 may include multiple CPUs 502. The control unit 501 may also include at least one DSP. All or some of the functions of the control unit 501 may be realized by a hardware circuit that does not require software to realize the functions. The storage unit 503 may also include a computer-readable non-transitory recording medium other than ROM and RAM.

[0061] The communication unit 505 is connected to the system I / F 20. The communication unit 505 can communicate with the cloud P / Fs 100, 110, and 120 and the fog P / F 400 of each factory system 4 through the system I / F 20.

[0062] The configuration of the business department server 50 is not limited to the above example. For example, the business department server 50 may include a display controlled by the control unit 501. The business department server 50 may also include an operation unit that accepts user operations.

[0063] In addition to multiple business division servers 50, the on-premise system 5 also includes a computer device 51 having a machine learning engine 510 and a computer device 52 having a statistical analysis engine 520. The computers 51 and 52 have, for example, the same hardware configuration as the business division server 50. In the computer device 51, the CPU executes a control program stored in a storage unit to implement the machine learning engine 510 as a functional block. The performance of the machine learning engine 510 differs from the performance of the machine learning engines 101 and 111 outside the intranet 2. In the computer device 52, the CPU executes a control program stored in a storage unit to implement the statistical analysis engine 520 as a functional block. The statistical analysis engine 520 can perform, for example, regression analysis. The performance of the statistical analysis engine 520 differs from the performance of the statistical analysis engine 121 outside the intranet 2.

[0064] At least one of the computers 51 and 52 may have a different hardware configuration from the business division server 50. The machine learning engine 510 and the statistical analysis engine 520 may be provided in a single computer. Instead of having a server for each business division as in the example of FIG. 1, there may be a single server that manages information for multiple business divisions. In the on-premise system 5, there may be a single server that collectively manages information for all business divisions of a specific company.

[0065] Hereinafter, when there is no need to particularly distinguish between the machine learning engines 101, 111, and 510, they may each be referred to as a machine learning engine. Also, when there is no need to particularly distinguish between the statistical analysis engines 121 and 520, they may each be referred to as a statistical analysis engine.

[0066] As described above, in this example, the information processing system 3 within the intranet 2 and the factory D system 4d outside the intranet 2 can communicate with each other through the cloud server 10. This eliminates the need to incorporate the factory D system 4d into the intranet 2. Therefore, an information processing system 1 in which the information processing system 3 and the factory D system 4d can communicate with each other can be easily constructed using the cloud server 10.

[0067] <Example of information processing system operation> Various operation examples of the information processing system 1 will be described below.

[0068] <Example of factory system operation> In the factory system 4, the fog P / F 400 can control to which server the information 420 obtained by the device GW 41 from the device 42 is sent. When the fog P / F 400 receives the information 420 obtained by the device GW 41 from the device 42 from the device GW 41, the fog P / F 400 can transmit the received information 420 to, for example, the cloud P / F 100 of the cloud server 10. The fog P / F 400 can also transmit the received information 420 to the cloud P / F 110. The fog P / F 400 can also transmit the received information 420 to the cloud P / F 120. Each of the cloud P / Fs 100, 110, and 120 stores the received information 420 in the storage unit 153. The fog P / F 400 can also transmit the information 420 from the device GW 41 to the on-premises system 5. In the on-premise system 5, information 420 from the fog P / F 400 is stored in the storage unit 503 of the business division server 50. Specifically, the information 420 from the fog P / F 400 of a certain factory system 4 is stored in the technical information DB 503d in the storage unit 503 of the business division server 50 corresponding to the business division that has the certain factory system 4.

[0069] Furthermore, the fog P / F 400 can perform processing using information 420 from the device GW 41. The fog P / F 400 can perform, for example, simple calculation processing. For example, the fog P / F 400 can perform arithmetic operations using the information 420. FIG. 8 is a diagram showing an example of the operation of the factory system 4.

[0070] As shown in FIG. 8, in step s1, the device 42 transmits information 420 to the device GW 41. Next, in step s2, the device GW 41 transmits the information 420 to the fog P / F 400. Next, in step s3, the fog P / F 400 performs processing using the information 420. Then, the fog P / F 400 generates information to be transmitted to the device GW 41 according to the processing result. Next, in step s4, the fog P / F 400 transmits the generated information to the device GW 41. Next, in step s5, the device GW 41 controls the device 42 based on the received information. The device 42 transmitting the information 420 in step s1 and the device 42 controlled in step s5 may be the same or different.

[0071] For example, consider a case where the device 42 transmitting the information 420 in step s1 is a current sensor. In this case, the fog P / F 400 determines in step s3, for example, whether the current (information 420) obtained from the current sensor has been continuously increasing for a predetermined time or more. If the current obtained from the current sensor has been continuously increasing for a predetermined time or more, the fog P / F 400 transmits instruction information to the device GW 41 in step s4, instructing the device GW 41 to output an alarm. Having received the instruction information, the device GW 41 outputs the alarm in step s5, for example, by sounding a buzzer (device 42) connected to the device GW 41. At this time, the device GW 41 having a display may display the alarm information on the display.

[0072] Also, consider a case where a relational expression expressing the relationship between the information 420 and certain information is calculated in advance and stored in the fog P / F 400. In this case, the fog P / F 400 may calculate certain information based on the stored relational expression and the input information 420. For example, consider a case where the factory system 4 executes a process to control the flow rate of wastewater flowing into a treatment tank of a wastewater treatment system installed in a factory in accordance with the wastewater level in the treatment tank. The fog P / F 400 stores a relational expression expressing the relationship between the wastewater level in the treatment tank and the corresponding set value of the wastewater flow rate. In this case, in step s3, the fog P / F 400 calculates the set value of the wastewater flow rate based on the relational expression and the wastewater level in the treatment tank (information 420) transmitted by the device GW 41 in step s2. Then, in step s4, the fog P / F 400 notifies the device GW 41 of the calculated set value. In step s5, the device GW41 sets the notified set value in the PLC 42b. A flow rate sensor that detects the flow rate of the wastewater is connected to the PLC 42b. The PLC 42b can also control the opening of a valve that controls the flow rate of the wastewater. The PLC 42b controls the opening of the valve so that the flow rate detected by the flow rate sensor becomes the set value.

[0073] As another example, consider a case where an industrial inkjet printer performs printing in the factory system 4. The fog P / F 400 stores a relational expression that indicates the relationship between the amount of misalignment of the printed area and the offset amount of the inkjet head position required to correct the misalignment. Furthermore, the devices 42 each include a sensor that detects the amount of misalignment of the printed area. In this case, in step s3, the fog P / F 400 calculates an offset amount based on the relational expression and the amount of misalignment of the printed area (information 420) sent by the device gateway 41. Then, in step s4, the fog P / F 400 notifies the device gateway 41 of the calculated offset amount. In step s5, the device gateway 41 sets the notified offset amount in the PLC 42b that controls the inkjet printer. The PLC 42b controls the inkjet printer to shift the position of the inkjet head by the set offset amount.

[0074] Furthermore, when the information 420 is a captured image obtained by a camera, the fog P / F 400 may perform simple image processing on the captured image. For example, the fog P / F 400 may reduce the image size of the captured image.

[0075] In this way, the fog P / F 400 can execute a certain amount of processing using the information 420. This reduces the processing load on the cloud P / F 100. It also reduces the amount of information sent from the intranet 2 to the cloud P / F 100. In addition, because the fog P / F 400 on the edge side, which is closer to the device 42, performs the processing, the real-time nature of the processing is improved.

[0076] The fog P / F 400 may also store a control program 413a for the device GW41 to be controlled. In this case, when the device GW41 is replaced with a new device GW41 due to a failure or the like, the fog P / F 400 may write the control program 413a to the new device GW41. This allows the new device GW41 to operate appropriately. This makes it easy to replace the device GW41.

[0077] <Information sharing between multiple fog P / Fs> In this example, a P / F similar to the fog P / F 400 that controls the device GW 41a included in the factory D system 4d is established in the cloud P / F 100. Hereinafter, for convenience of explanation, the P / F similar to the fog P / F 400 established in the cloud P / F 100 may be referred to as the cloud-side fog P / F 400.

[0078] The cloud-side fog P / F 400 is connected to the device GW 41a of the factory D system 4d through the firewall 14. The cloud-side fog P / F 400 manages a system consisting of the device GW 41a of the factory D system 4d, the device 42 connected thereto, and the cloud-side fog P / F 400. The cloud-side fog P / F 400 operates in the same manner as the fog P / F 400 in the factory system 4. For example, the cloud-side fog P / F 400 can control the device 42 connected to the device GW 41a based on information 420 from the device GW 41a. The cloud-side fog P / F 400 also understands the status of the system it manages. The system status understood by the cloud-side fog P / F 400 includes, for example, which device GWs 41 are connected to the cloud-side fog P / F 400 and what operations the device GWs 41 are performing. Furthermore, the system status grasped by the cloud-side fog P / F 400 includes what devices 42 are connected to the device GWs 41 connected to the cloud-side fog P / F 400 and what operations the devices 42 are performing. In the factory D system 4d equipped with a device GW 41a, the fog P / F 400 can be said to manage a system consisting of the fog P / F 400, the device GWs 41 connected thereto, and the devices 42 connected to the device GWs 41. Hereinafter, unless otherwise specified, the term fog P / F 400 also includes the cloud-side fog P / F 400.

[0079] In the information processing system 1 of this example, container technology is used, and multiple fog P / Fs 400 virtually configure one P / F. Each fog P / F 400 is aware of not only the status of the system it manages but also the status of systems managed by other fog P / Fs 400. In other words, each fog P / F 400 stores system status information indicating the status of the system it manages, as well as system status information indicating the status of systems managed by other fog P / Fs 400.

[0080] For example, the fog P / F 400 of the factory A system 4a stores not only system status information indicating the status of the factory A system 4a, but also system status information indicating the status of the system managed by the fog P / F 400 of the factory B system 4b, i.e., the factory B system 4b. The fog P / F 400 of the factory A system 4a also stores system status information indicating the status of the factory C system 4c. The fog P / F 400 of the factory A system 4a also stores system status information indicating the status of the system managed by the fog P / F 400 of the factory D system 4d. The fog P / F 400 of the factory A system 4a also stores system status information indicating the status of the system managed by the cloud-side fog P / F 400. Fog P / Fs 400 other than the fog P / F 400 of the factory A system 4a similarly store system status information for the systems managed by the other fog P / Fs 400.

[0081] The fog P / F 400 of each factory system 4 can notify the fog P / F 400 of the other factory systems 4 of the status of the system it manages through the cloud-side fog P / F 400. Specifically, when a change occurs in the status of the system it manages, the fog P / F 400 of each factory system 4 can notify the fog P / F 400 of the other factory systems 4 of the details of the change, for example, through the cloud-side fog P / F 400. Furthermore, the cloud-side fog P / F 400 can notify the fog P / F 400 of each factory system P / F 400 of the status of the system it manages. Specifically, when a change occurs in the status of the system it manages, the cloud-side fog P / F 400 can notify the fog P / F 400 of each factory system P / F 400 of the details of the change. An example of the operation of the information processing system 1 when the status of a system managed by a fog P / F 400 is notified to another fog P / F 400 will be described in detail below.

[0082] When a change occurs in the status of the system managed by the fog P / F 400, the fog P / F 400 of each factory system 4 notifies the cloud-side fog P / F 400 of system change information indicating the details of the change. For example, when new content is added to the rules of the fog P / F 400, information identifying the content is included in the system change information. When a new sensor is added to the system, information about the new sensor is included in the system change information. When an abnormality occurs in the system managed by the fog P / F 400, information identifying the details of the abnormality is included in the system change information. For example, when the sensor 42a breaks down and stops working, information indicating that the sensor 42a is not working is included in the system change information. Hereinafter, the fog P / F 400 that sends the system change information may be referred to as the target fog P / F 400.

[0083] When the cloud-side fog P / F 400 receives system change information from a target fog P / F 400 in a certain factory system 4, it updates the system status information stored in itself for the system managed by the target fog P / F 400 based on the received system change information. This allows the cloud-side fog P / F 400 to grasp the latest status of the system managed by the target fog P / F 400. The cloud-side P / F 400 also notifies the fog P / F 400 in the other factory systems 4 of the system change information received from the target fog P / F 400 in a certain factory system 4. The fog P / F 400 in the other factory systems 4 updates the system status information for the system managed by the target fog P / F 400 based on the received system change information. Furthermore, when a change occurs in the status of the system it manages, the cloud-side fog P / F 400 notifies the fog P / F 400 in each factory system 4 of system change information indicating the content of the change. Based on the received system change information, the fog P / F 400 in each factory system 4 updates the system status information of the system managed by the cloud-side fog P / F 400. This allows the fog P / F 400 in each factory system 4 to grasp the latest status of the systems managed by the other fog P / Fs 400.

[0084] When the information processing system 1 is introduced, system status information indicating the initial status of the system managed by the other fog P / Fs 400 is written to each fog P / F 400. Each fog P / F 400 can grasp the latest status of the system managed by the other fog P / Fs 400 by appropriately updating the system status information written at the time of system introduction as described above.

[0085] In this way, in this example, the fog P / F 400 is aware of the status of the systems managed by the other fog P / F 400. This allows the fog P / F 400 to perform processing according to the status of the systems managed by the other fog P / F 400.

[0086] For example, if an abnormality occurs in a system managed by another fog P / F 400, the fog P / F 400 can execute a process to deal with the abnormality. As an example, consider a case where Factory C is an unmanned factory and Factory B is a factory where people are present. In this case, if the fog P / F 400 of the factory B system 4b determines that an abnormality has occurred in the factory C system 4c based on the system status information of the factory C system 4c, it controls the device GW 41 of the factory B system 4b to activate the lamp and buzzer connected thereto. This allows a person in Factory B to be notified of the occurrence of an abnormality in the factory C system 4c installed in the unmanned factory C. Furthermore, if the cloud-side fog P / F 400 determines that an abnormality has occurred in the factory C system 4c based on the system status information of the factory C system 4c, the cloud-side fog P / F 400 may notify a management server owned by a specific company that manages the information processing system 1 of the occurrence of the abnormality in the factory C system 4c. The management server may be present, for example, in the on-premises system 5.

[0087] In addition, since the fog P / F 400 is aware of the status of the systems managed by other fog P / F 400, if one fog P / F 400 stops due to a malfunction or the like, another fog P / F 400 can operate in place of the stopped fog P / F 400.

[0088] Furthermore, when a server 40 in a factory system 4 is replaced with a new server 40, the fog P / F 400 in another factory system 4 or the cloud-side fog P / F 400 can notify the fog P / F 400 in the new server 40 of the status of the system managed by the fog P / F 400 in the new server 40. This allows the fog P / F 400 in the new server 40 to operate appropriately. This makes it easy to replace the server 40 in the factory system 4.

[0089] Furthermore, the cloud-side fog P / F 400 can understand the rules of the fog P / F 400 of the factory system 4 from the system status information of the factory system 4. Therefore, the cloud-side fog P / F 400 can determine whether the operation of the fog P / F 400 of the factory system 4 is abnormal.

[0090] For example, suppose a rule for a fog P / F 400 specifies that the fog P / F 400 transmits information 420 to the cloud P / F 100 once every five minutes. In this case, if the fog P / F 400 does not transmit information 420 to the cloud P / F 100 for more than ten minutes, the cloud-side fog P / F 400 determines that the fog P / F 400 is abnormal. When the cloud-side fog P / F 400 determines that the fog P / F 400 in the factory system 4 is abnormal, the cloud-side fog P / F 400 notifies, for example, the management server in the on-premise system 5, that an abnormality has occurred in the fog P / F 400. The management server notifies an operator that an abnormality has occurred in the fog P / F 400 by displaying a message on a display, for example. Note that the method by which the cloud-side fog P / F 400 determines that the fog P / F 400 in the factory system 4 is abnormal is not limited to this. Furthermore, the operation of the cloud-side fog P / F 400 when the cloud-side fog P / F 400 identifies an abnormality in the fog P / F 400 of the factory system 4 is not limited to this.

[0091] <Example of machine learning engine and statistical analysis engine operation> The machine learning engine 101 of the cloud P / F 100 can generate a trained model based on, for example, information in the storage unit 153 of the cloud server 10. That is, the machine learning engine 101 can train a trained model using the information in the storage unit 153 of the cloud server 10 as training data. The machine learning engine 101 can also infer predetermined information based on the trained model and the information in the storage unit 153 of the cloud server 10. When information in the storage unit 153 is input to the trained model, the machine learning engine 101 outputs the trained model as an inference result. Hereinafter, the information in the storage unit 153 of the cloud server 10 may be referred to as first cloud-side information. The trained model generated by the machine learning engine 101 may be referred to as a first cloud-side trained model.

[0092] The machine learning engine 111 of the cloud P / F 110 is capable of generating a trained model based on, for example, information in the storage unit 153 of the cloud server 11. The machine learning engine 111 is also capable of inferring predetermined information based on the trained model and the information in the storage unit 153 of the cloud server 11. Hereinafter, the information in the storage unit 153 of the cloud server 11 may be referred to as second cloud-side information. The trained model generated by the machine learning engine 111 may be referred to as the second cloud-side trained model.

[0093] The machine learning engine 510 of the on-premise system 5 can generate a trained model based on, for example, information in the storage unit 503 of the business department server 50. That is, the machine learning engine 510 can train a trained model using the information in the storage unit 503 as training data. The machine learning engine 510 can also infer predetermined information based on the trained model and the information in the storage unit 503 of the business department server 50. When information in the storage unit 503 is input to the trained model, the machine learning engine 510 outputs the trained model as an inference result. Hereinafter, the information in the storage unit 503 of the business department server 50 may be referred to as on-premise information. The trained model generated by the machine learning engine 510 may be referred to as an on-premise trained model.

[0094] When the first cloud-side information is updated, the machine learning engine 101 may retrain the first cloud-side trained model based on the updated first cloud-side information. Similarly, when the second cloud-side information is updated, the machine learning engine 111 may retrain the second cloud-side trained model based on the updated second cloud-side information. Furthermore, when the on-premise side information is updated, the machine learning engine 510 may retrain the on-premise side trained model based on the updated on-premise side information.

[0095] The statistical analysis engine 121 of the cloud P / F 120 is capable of, for example, performing statistical analysis on information in the storage unit 153 of the cloud server 12 and generating a statistical analysis model that outputs predetermined information. Furthermore, the statistical analysis engine 121 can input information in the storage unit 153 of the cloud server 12 into the generated statistical analysis model to obtain the predetermined information. Hereinafter, the information in the storage unit 153 of the cloud server 12 may be referred to as third cloud-side information. Furthermore, the statistical analysis model generated by the statistical analysis engine 121 may be referred to as a cloud-side statistical analysis model.

[0096] The statistical analysis engine 520 of the on-premise system 5 can, for example, perform statistical analysis on on-premise information and generate a statistical analysis model that outputs predetermined information. Furthermore, the statistical analysis engine 520 can input on-premise information into the generated statistical analysis model to obtain the predetermined information. Hereinafter, the statistical analysis model generated by the statistical analysis engine 520 may be referred to as an on-premise statistical analysis model.

[0097] When the third cloud-side information is updated, the statistical analysis engine 121 may reconstruct the cloud-side statistical analysis model based on the updated third cloud-side information. Similarly, when the on-premise-side information is updated, the statistical analysis engine 520 may reconstruct the on-premise-side statistical analysis model based on the updated on-premise-side information.

[0098] In the information processing system 1 of this example, information that can be obtained by statistical analysis is generated using a statistical analysis engine, and information that is difficult to obtain by statistical analysis is generated using a machine learning engine.

[0099] In the information processing system 1 of this example, the information to be inferred by the machine learning engine is determined according to the performance of the machine learning engine, and the information to be output by the statistical analysis engine is determined according to the performance of the statistical analysis engine.

[0100] A specific example of the operation of the machine learning engine and the statistical analysis engine will be described below. In the following description, a case where the machine learning engine and the statistical analysis engine are used for a wastewater treatment system in a factory will be taken as an example.

[0101] Each of the factories A to D is provided with a wastewater treatment system that purifies wastewater generated in the factory and discharges it. The wastewater treatment system includes, for example, an adjustment tank that adjusts the pH of the wastewater, and a separation tank that performs a separation process on the wastewater. Note that the wastewater treatment system may also be provided with a single treatment tank that adjusts the pH of the wastewater and performs a separation process on the wastewater.

[0102] An adjuster for adjusting the pH is added to the adjustment tank. A separating agent is added to the separation tank. In this example, for example, the amount of adjuster added is determined using a statistical analysis engine. Also, the amount of separating agent added is estimated using a machine learning engine. Below, a method for determining the amounts of adjuster and separating agent added will be described using the wastewater treatment system of factory A as an example. The following description can also be applied to the wastewater treatment systems of factories B to D.

[0103] <An example of how to determine the amount of regulator to be added> The multiple devices 42 provided in the factory A system 4a include a pH sensor that detects the pH of the wastewater in the adjustment tank. The pH sensor repeatedly detects and outputs the pH. The multiple devices 42 provided in the factory A system 4a also include a temperature sensor that detects the temperature of the wastewater in the adjustment tank. The temperature sensor repeatedly detects and outputs the temperature. The multiple devices 42 provided in the factory A system 4a also include a PLC that controls an adjuster dosing device that automatically adds an adjuster to the adjustment tank. Hereinafter, the pH detected by the pH sensor may be referred to as the detected pH. The temperature detected by the temperature sensor may be referred to as the detected temperature.

[0104] The effectiveness of the adjuster added to wastewater varies depending on the wastewater temperature. For example, as the wastewater temperature increases, the adjuster's effectiveness improves. In this example, during the statistical model generation stage, the statistical analysis engine performs statistical analysis on the detected pH and detected temperature and the required adjuster dosage (predetermined values) to generate a statistical analysis model that outputs the appropriate adjuster dosage. The statistical analysis engine then uses the generated statistical analysis model to determine the adjuster dosage. That is, the statistical analysis engine uses the output information of the statistical analysis model when the detected pH and detected temperature are input to the statistical analysis model as the adjuster dosage. Once the adjuster dosage is determined, the fog P / F 400 of the factory A system 4a notifies the determined dosage to the device GW 41, which is connected to a PLC that controls the adjuster dosage device. The device GW 41 sets the notified dosage to the PLC 42b, which controls the adjuster dosage device. The PLC 42b controls the adjuster dosage device so that the set dosage amount of adjuster is added to the adjustment tank.

[0105] The dosage of the adjuster may be determined by the statistical analysis engine 520 in the on-premise system 5. In this case, the fog P / F 400 of the factory A system 4a transmits the detected pH and the detected temperature received from the device GW 41 to the on-premise system 5. In the on-premise system 5, the detected pH and the detected temperature are stored in the technical information DB 503d of the business division server 50 corresponding to the business division that has the factory A. In the statistical model generation stage, the statistical analysis engine 520 reads the detected pH and the detected temperature from the technical information DB 503d. Then, the statistical analysis engine 520 performs statistical analysis on the read-out detected pH and detected temperature and the dosage of the adjuster required at that time, and generates a statistical analysis model that outputs an appropriate dosage of the adjuster. In this way, a statistical analysis model for the factory A is generated.

[0106] After the statistical analysis model is generated, the statistical analysis engine 520 inputs the latest detected pH and detected temperature in the technical information DB 503d into the statistical analysis model to determine the amount of adjuster to be added. The amount of adjuster to be added determined by the statistical analysis engine 520 is notified to the fog P / F 400 of the factory A system 4a. The fog P / F 400 sets the amount of adjuster to be added to the PLC 42b via the device GW 41. As a result, the adjuster is added to the adjustment tank in the amount determined by the statistical analysis engine 520. Note that the statistical analysis engine 520 may automatically reconstruct the statistical analysis model based on the detected pH and detected temperature in the technical information DB 503a.

[0107] When the statistical analysis engine 520 determines the amount of adjusting agent input for each of factories A to D, the statistical analysis engine 520 generates a statistical analysis model for factory A, a statistical analysis model for factory B, a statistical analysis model for factory C, and a statistical analysis model for factory D.

[0108] The detected pH and temperature obtained in the factory D system 4d outside the intranet 2 are input to the on-premise system 5 via the cloud server 10. At this time, the detected pH and temperature obtained in the factory D system 4d may be input to the on-premise system 5 via the firewall 14, the cloud server 10, the firewall 15, and the system I / F 20. Alternatively, the detected pH and temperature obtained in the factory D system 4d may be input to the on-premise system 5 via the firewall 14, the cloud server 10, the firewall 13, the fog P / F 400 of the intranet factory system 4, and the system I / F 20. Furthermore, the input amount for factory D determined by the statistical analysis engine 520 of the on-premise system 5 may be input to the fog P / F 400 of the factory D system 4d via the system I / F 20, the firewall 15, the cloud server 10, and the firewall 14. Alternatively, the input amount for factory D determined by the statistical analysis engine 520 may be input to the fog P / F 400 of the factory D system 4d via the system I / F 20, the fog P / F 400 of the intranet factory system 4, the firewall 13, the cloud server 10, and the firewall 14. The same applies to other information exchanges between the factory D system 4d and the on-premise system 5.

[0109] FIG. 9 is a diagram illustrating an overview of the information flow between the factory D system 4d and the statistical analysis engine 520 in the case where the adjuster input amount for factory D is calculated by the statistical analysis engine 520 in the intranet 2. As shown in FIG. 9, in step s51, the detected pH and the detected temperature are output from the pH sensor and the temperature sensor of the factory D system 4d to the device GW 41. Next, in step s52, the detected pH and the detected temperature are output from the device GW 41 to the fog P / F 400. Next, in step s53, the detected pH and the detected temperature are output from the fog P / F 400 to the cloud P / F 100. Next, in step s54, the cloud P / F 100 outputs the detected pH and the detected temperature. The detected pH and the detected temperature output from the cloud P / F 100 are input to the on-premise system 5 in the intranet 2. In the on-premise system 5, the detected pH and the detected temperature are input to the statistical analysis engine 520. When the statistical analysis engine 520 determines the dosage amount of the adjuster using the statistical analysis model, it outputs the determined dosage amount in step s61. The dosage amount output from the statistical analysis engine 520 is input to the cloud P / F 100. In step s62, the cloud P / F 100 outputs the dosage amount to the fog P / F 400 of the factory D system 4d. Next, in step s64, the fog P / F 400 outputs the dosage amount to the device GW 41. Next, in step s54, the device GW 41 sets the dosage amount in the PLC 42b that controls the adjuster dosage device.

[0110] The dosage of the adjuster may be calculated by the statistical analysis engine 121 of the cloud server 12. In this case, the fog P / F 400 of the factory A system 4a transmits the detected pH and detected temperature received from the device GW 41 to the cloud P / F 120. The cloud P / F 120 stores the received detected pH and detected temperature in the memory unit 153 of the cloud server 12. In the statistical model generation stage, the statistical analysis engine 121 of the cloud P / F 120 performs statistical analysis on the detected pH and detected temperature stored in the memory unit 153 and the dosage of the adjuster required at that time, to generate a statistical analysis model that outputs an appropriate dosage of the adjuster. After the statistical analysis model is generated, the statistical analysis engine 121 inputs the latest detected pH and detected temperature stored in the memory unit 153 into the statistical analysis model to calculate the dosage of the adjuster. The dosage of the adjuster calculated by the statistical analysis engine 121 is notified to the fog P / F 400 of the factory A system 4a. The fog P / F 400 sets the dosage of the adjuster in the PLC 42b, which controls the adjuster dosage device, via the device GW 41. As a result, the adjuster is dispensed into the adjustment tank in the dosage amount determined by the statistical analysis engine 520. The statistical analysis engine 121 may automatically reconstruct the statistical analysis model based on the detected pH and detected temperature in the memory unit 153.

[0111] FIG. 10 is a diagram illustrating an overview of the information flow between the factory A system 4a and the cloud P / F 120 when the adjuster dosage for factory A is determined by the statistical analysis engine 121. As shown in FIG. 10, in step s11, the pH sensor and temperature sensor of the factory A system 4a output the detected pH and temperature to the device GW 41. Next, in step s12, the device GW 41 outputs the detected pH and temperature to the fog P / F 400. Next, in step s13, the fog P / F 400 outputs the detected pH and temperature to the cloud P / F 120. The cloud P / F 120 stores the detected pH and temperature. After determining the adjuster dosage using the statistical analysis model, the cloud P / F 120 outputs the determined dosage to the fog P / F 400 of the factory A system 4a in step s21. Next, in step s22, the fog P / F 400 outputs the dosage to the device GW 41. Next, in step s23, the device GW41 sets the dosage amount in the PLC 42b that controls the adjuster dosage device.

[0112] The detected pH and temperature obtained in the factory D system 4d are input to the cloud server 12 via the cloud server 10. At this time, the detected pH and temperature obtained in the factory D system 4d may be input to the cloud server 12 via the firewall 14, the cloud server 10, the firewall 15, and the system I / F 20. Alternatively, the detected pH and temperature obtained in the factory D system 4d may be input to the cloud server 12 via the firewall 14, the cloud server 10, the firewall 13, the fog P / F 400 of the intra-factory system 4, and the system I / F 20. Furthermore, the input amount for factory D determined by the statistical analysis engine 121 of the cloud server 12 may be input to the fog P / F 400 of the factory D system 4d via the system I / F 20, the firewall 15, the cloud server 10, and the firewall 14. Alternatively, the input amount for factory D determined by the statistical analysis engine 121 may be input to the fog P / F 400 of the factory D system 4d via the system I / F 20, the fog P / F 400 of the intra-factory system 4, the firewall 13, the cloud server 10, and the firewall 14. The same applies to other information exchanges between the factory D system 4d and the cloud server 12. The same applies to information exchanges between the factory D system 4d and the cloud server 11.

[0113] FIG. 11 is a diagram illustrating an overview of the information flow between the factory D system 4d and the cloud P / F 120 when the adjuster dosage for factory D is determined by the statistical analysis engine 121. As shown in FIG. 11, in step s31, the detected pH and temperature are output from the pH sensor and temperature sensor of the factory D system 4d to the device GW 41. Next, in step s32, the detected pH and temperature are output from the device GW 41 to the fog P / F 400. Next, in step s33, the detected pH and temperature are output from the fog P / F 400 to the cloud P / F 100. Next, in step s34, the cloud P / F 100 outputs the detected pH and temperature. The detected pH and temperature output from the cloud P / F 100 are input to the cloud P / F 120. The cloud P / F 100 stores the detected pH and temperature. After determining the adjuster dosage using the statistical analysis model, the cloud P / F 120 outputs the determined dosage in step s41. The input amount output from the cloud P / F 120 is input to the cloud P / F 100. In step s42, the cloud P / F 100 outputs the input amount to the fog P / F 400 of the factory D system 4d. Next, in step s43, the fog P / F 400 outputs the input amount to the device GW 41. Next, in step s44, the device GW 41 sets the input amount in the PLC 42b that controls the adjuster input device.

[0114] <An example of how to determine the amount of separating agent to be added> The multiple devices 42 provided in the factory A system 4a include a camera that takes pictures of the wastewater in the separation tank. The camera repeatedly takes pictures of the wastewater and outputs the captured images. The multiple devices 42 provided in the factory A system 4a also include a PLC that controls a separating agent feeding device that automatically feeds a separating agent into the separation tank.

[0115] Consider a case where a person determines the amount of separating agent to be dispensed. In this case, the worker visually checks the state of the wastewater in the separation tank before determining the amount of separating agent to be dispensed. Similar to the conditioner, the effectiveness of the separating agent varies depending on the temperature of the wastewater. Therefore, in this example, in the learning stage, the machine learning engine trains a learning model using the captured images and detected temperatures obtained by the camera as training data to generate a trained model that outputs the amount of separating agent to be dispensed. Then, in the inference stage, the machine learning engine uses the generated trained model to infer the amount of separating agent to be dispensed. That is, the machine learning engine uses the output information of the trained model when the captured images and detected temperature are input to the trained model as the amount of separating agent to be dispensed. Once the amount of separating agent to be dispensed is determined, the fog P / F 400 of the factory A system 4a notifies the device GW 41, which is connected to the PLC that controls the separating agent dispenser, of the determined amount. The device GW 41 sets the notified amount to the PLC that controls the separating agent dispenser. The PLC controls the separating agent feeding device so that a set amount of separating agent is fed into the separation tank.

[0116] The amount of separating agent dispensed may be inferred by the machine learning engine 510 in the on-premise system 5. In this case, the fog P / F 400 in the factory A system 4a transmits the captured image and detected temperature received from the device GW 41 to the on-premise system 5. In the on-premise system 5, the captured image and detected temperature are stored in the technical information DB 503d of the business division server 50 corresponding to the business division that includes the factory A. In the learning stage, the machine learning engine 510 trains a learning model using the captured image and detected temperature in the technical information DB 503d as learning data to generate a trained model that outputs the amount of separating agent dispensed. This generates a trained model for the factory A. After the trained model is generated, the machine learning engine 510 inputs the latest captured image and detected temperature in the technical information DB 503d into the trained model to calculate the amount of separating agent dispensed. The amount of separating agent dispensed calculated by the machine learning engine 510 is notified to the fog P / F 400 in the factory A system 4a. The fog P / F 400 sets the amount of separating agent to be added to the PLC 42b, which controls the separating agent adding device, via the device GW 41. As a result, the separating agent is added to the separation tank in the amount estimated by the machine learning engine 510. Note that the machine learning engine 510 may automatically retrain the trained model based on the captured images and detected temperatures stored in the technical information DB 503d.

[0117] When the machine learning engine 510 determines the amount of separating agent to be added for each of factories A to D, the machine learning engine 510 generates a trained model for factory A, a trained model for factory B, a trained model for factory C, and a trained model for factory D.

[0118] The amount of separating agent dispensed may also be inferred by the machine learning engine 101 of the cloud server 10. In this case, the fog P / F 400 of the factory A system 4a transmits the captured image and detected temperature received from the device GW 41 to the cloud P / F 100. The cloud P / F 100 stores the received captured image and detected temperature in the memory unit 153 of the cloud server 10. In the learning stage, the machine learning engine 101 trains a learning model using the captured image and detected temperature in the memory unit 153 as learning data, and generates a trained model that outputs the amount of separating agent dispensed. After the trained model is generated, the machine learning engine 101 inputs the latest captured image and detected temperature in the memory unit 153 into the trained model to calculate the amount of separating agent dispensed. The cloud P / F 100 notifies the fog P / F 400 of the factory A system 4a of the amount of separating agent dispensed. The fog P / F 400 sets the amount of separating agent to be added to the PLC 42b, which controls the separating agent adding device, via the device GW 41. As a result, the separating agent is added to the separation tank in the amount estimated by the machine learning engine 101. Note that the machine learning engine 101 may automatically re-learn the trained model based on the captured images and detected temperatures stored in the memory unit 153.

[0119] The amount of separating agent dispensed may also be inferred by the machine learning engine 111 of the cloud server 11. In this case, the fog P / F 400 of the factory A system 4a transmits the captured image and detected temperature received from the device GW 41 to the cloud P / F 110. The cloud P / F 110 stores the received captured image and detected temperature in the memory unit 153 of the cloud server 11. In the learning stage, the machine learning engine 111 trains a learning model using the captured image and detected temperature in the memory unit 153 as learning data, and generates a trained model that outputs the amount of separating agent dispensed. After the trained model is generated, the machine learning engine 111 inputs the latest captured image and detected temperature in the memory unit 153 into the trained model to calculate the amount of separating agent dispensed. The amount of separating agent dispensed calculated by the machine learning engine 111 is notified from the cloud P / F 110 to the fog P / F 400 of the factory A system 4a. The fog P / F 400 sets the amount of separating agent to be added to the PLC 42b, which controls the separating agent adding device, via the device GW 41. As a result, the separating agent is added to the separation tank in the amount estimated by the machine learning engine 101. Note that the machine learning engine 111 may automatically re-learn the trained model based on the captured images and detected temperatures stored in the memory unit 153.

[0120] For example, when an adjuster and a separating agent are added to the same treatment tank, the machine learning engine may generate a trained model that infers the amounts of adjuster and separating agent added. In this case, the machine learning engine trains the learning model using the detected pH, detected temperature, and photographed images as training data in the learning stage to generate a trained model that outputs the amounts of adjuster and separating agent added. Then, in the inference stage, the machine learning engine inputs the detected pH, detected temperature, and photographed images into the generated trained model to determine the amounts of adjuster and separating agent added.

[0121] Furthermore, in the learning stage, the machine learning engine may train the learning model without using the information 420 obtained by the factory system 4. Furthermore, in the statistical analysis model generation stage, the statistical analysis engine may generate the statistical analysis model without using the information 420 obtained by the factory system 4.

[0122] The information inferred by the machine learning engine is not limited to the above examples. Furthermore, the information sought by the statistical analysis engine is not limited to the above examples. For example, the machine learning engine may generate a trained model that infers whether or not a substrate manufactured on a production line has a manufacturing abnormality based on a captured image (information 420) of the substrate. Furthermore, the statistical analysis engine may generate a statistical analysis model that outputs the amount and timing of ink order based on, for example, the amount of ink consumed per unit time (information 420) and the remaining amount of ink (information 420).

[0123] Furthermore, the types of information to be inferred may differ among the multiple machine learning engines. In this case, the fog P / F 400 of the intranet factory system 4 may determine the destination of the received information 420 based on which machine learning engine will use the information 420. For example, if the information 420 is to be used by the machine learning engine 101 of the cloud P / F 100, the fog P / F 400 sends the information 420 to the cloud P / F 100. Also, if the information 420 is to be used by the machine learning engine 510 of the on-premise system 5, the fog P / F 400 sends the information 420 to the on-premise system 5. Similarly, the types of information required by the multiple statistical analysis engines may differ among the multiple statistical analysis engines.

[0124] Furthermore, a trained model that is the same as a trained model for a certain factory that is generated outside the intranet 2 may be constructed in the factory system 4 provided in the certain factory. FIG. 12 is a diagram showing an example of how a trained model 45 that is the same as a trained model for a certain factory that is generated outside the intranet 2 is constructed in the fog P / F 400 of the factory system 4 provided in the certain factory. The trained model 45 may be a first cloud-side trained model generated by the machine learning engine 101, or may be a second cloud-side trained model generated by the machine learning engine 111.

[0125] In the example of FIG. 12, the fog P / F 400 can input information 420 from the device GW 41 to the trained model 45 included in the fog P / F 400. This allows the fog P / F 400 to immediately obtain the inference results based on machine learning. This improves the real-time nature of processing. For example, if the trained model 45 is a model capable of identifying an abnormality in a manufacturing device, when an abnormality occurs in the manufacturing device, the fog P / F 400 can immediately stop the manufacturing device via the device GW 41 and the PLC 42b.

[0126] FIG. 13 is a diagram showing an example of how a trained model 45 is constructed in a device GW41. In the example of FIG. 13, information 420 obtained from a device 42 connected to the device GW41 is input to the trained model 45 in the device GW41. In the example of FIG. 13, the device GW41 can input the information 420 obtained from the device 42 to the trained model 45 included in the device GW41. This allows the device GW41 to immediately obtain an inference result based on machine learning. Therefore, compared to the example of FIG. 12, the real-time nature of processing is further improved.

[0127] When a trained model 45 identical to the first cloud-side trained model generated by the machine learning engine 101 of the cloud P / F 100 is constructed in the factory system 4, the cloud P / F 100 outputs data representing the first cloud-side trained model to the fog P / F 400 of the factory system 4. The fog P / F 400 constructs the trained model 45 within itself or in the device GW 41 based on the data representing the first cloud-side trained model. Note that when the trained model 45 is constructed in the factory system 4, the above-mentioned system change information may include identification information for identifying the trained model 45.

[0128] Furthermore, when a trained model 45 identical to the second cloud-side trained model generated by the machine learning engine 111 of the cloud P / F 110 is constructed in the factory system 4, the cloud P / F 110 outputs data representing the second cloud-side trained model. The data output from the cloud P / F 110 is input to the fog P / F 400 of the factory system 4. The fog P / F 400 constructs the trained model 45 within itself or within the device GW 41 based on the data representing the second cloud-side trained model. The fog P / F 400 of the factory D system 4d receives the data representing the second cloud-side trained model through the cloud server 10. Therefore, it can be said that the trained model 45 is constructed in the factory D system 4d through the cloud server 10.

[0129] Similarly, a trained model that is the same as a trained model for a certain factory (on-premise trained model) generated in the on-premise system 5 may be constructed within the factory system 4 installed in the certain factory.

[0130] Furthermore, a statistical analysis model 46 that is the same as a statistical analysis model for a certain factory that is generated outside the intranet 2 may be constructed in the factory system 4 provided in the certain factory. FIG. 14 is a diagram showing an example of how the statistical analysis model 46 is constructed in the fog P / F 400. FIG. 15 is a diagram showing an example of how the statistical analysis model 46 is constructed in the device GW 41. When the statistical analysis model 46 is constructed in the factory system 4, the system change information described above may include identification information for identifying the statistical analysis model 46.

[0131] Similarly, a statistical analysis model that is the same as a statistical analysis model for a certain factory (on-premise side statistical analysis model) generated in the on-premise system 5 may be constructed in the factory system 4 installed in the certain factory.

[0132] <Combining multiple models> The information processing system 1 of this example can also use a combination of multiple trained models. For example, a machine learning engine 510 within the intranet 2 inputs first information into a trained model to infer second information. Then, a machine learning engine 101 outside the intranet 2 inputs the second information and third information into a trained model to infer fourth information. Figure 16 shows this situation.

[0133] As shown in FIG. 16, first information is input to the trained model within the intranet 2. Then, second information and third information output from the trained model within the intranet 2 are input to the trained model outside the intranet 2. The trained model outside the intranet 2 outputs fourth information. The fourth information is input, for example, into the intranet 2. A specific example of the processing shown in FIG. 16 will be described below. As an example, the operation of the information processing system 1 when the order quantity and order timing of a certain chemical (hereinafter sometimes referred to as the first chemical) used in the wastewater treatment system of factory A is determined using the processing of FIG. 16 will be described below.

[0134] For example, the machine learning engine 510 in the intranet 2 generates an on-premise trained model that infers the provisional order quantity and order timing (second information) of a first drug (such as a regulator or separator). Also, the machine learning engine 101 outside the intranet 2 generates a first cloud-side trained model that infers the final order quantity and order timing of the first drug (fourth information).

[0135] For example, highly confidential information is input to the on-premises trained model as first information. For example, the current inventory amount of the first drug, ordering information for the first drug, and production plan information for factory A are input to the on-premises trained model. The ordering information includes, for example, the ordered amount of the first drug and the arrival time of the ordered first drug. The production plan information includes, for example, information indicating what products are planned to be manufactured in factory A, and in what quantities and at what time. The state of wastewater generated in factory A changes depending on the type of products manufactured in factory A and the production volume at factory A. Therefore, the production plan information is important information when determining the order amount and order timing of drugs.

[0136] In the inference stage, the machine learning engine 510 acquires the current inventory amount of the first drug, ordering information for the first drug, and production plan information for factory A from the production management information DB 503b of the business division server 50 corresponding to the business division that includes factory A. Then, the machine learning engine 510 inputs the current inventory amount of the first drug, ordering information for the first drug, and production plan information for factory A into the on-premises trained model to infer the provisional order amount and order timing for the first drug. The output information of the on-premises trained model, i.e., the provisional order amount and order timing for the first drug, is notified to the fog P / F 400 of the factory A system 4a. The fog P / F 400 notifies the cloud P / F 100 of the notified output information of the on-premises trained model.

[0137] The output information (second information) of the on-premise trained model is input to the first cloud-side trained model. Furthermore, for example, low confidentiality information is input to the first cloud-side trained model as third information. For example, temperature forecast information is used as the low confidentiality information. The cloud server 10 can acquire the temperature forecast information from, for example, a server outside the information processing system 1. Because the treatment tank into which the chemical agent is added is located outdoors, the temperature of the wastewater in the treatment tank also changes depending on the temperature. The effectiveness of the chemical agent changes depending on the temperature of the wastewater. For example, when the temperature of the wastewater is low, the effectiveness of the chemical agent is reduced. Therefore, when the temperature is low for several days, a large amount of chemical agent is required. Therefore, the temperature forecast information affects the amount and timing of ordering the chemical agent.

[0138] The machine learning engine 101 of the cloud P / F 100 inputs the temperature forecast information acquired by the cloud server 10 and the output information of the on-premise trained model into the first cloud-side trained model to infer the final order quantity and order timing of the first pharmaceutical. The cloud P / F 100 notifies the fog P / F 400 of the factory A system 4a of the inferred order quantity and order timing (fourth information). The cloud server 10 does not store the highly confidential inferred order quantity and order timing. The fog P / F 400 of the factory A system 4a notifies, for example, an order server in the on-premise system 5 of the order quantity and order timing of the first pharmaceutical. The order server performs ordering processing for the first pharmaceutical based on the order quantity and order timing of the first pharmaceutical.

[0139] As another example, consider a case where the amounts of the above-mentioned adjuster and separator added to the same treatment tank are calculated using multiple trained models in a wastewater treatment system in a factory equipped with an intra-factory system 4. Hereinafter, the intra-factory system 4 that is the subject of the explanation may be referred to as the target intra-factory system 4. Also, the factory equipped with the target intra-factory system 4 may be referred to as the target factory.

[0140] For example, the fog P / F 400 of the target intranet factory system 4 transmits the detected temperature from the device GW 41, which is not confidential, to the cloud server 10. The cloud server 10 stores the detected temperature from the fog P / F 400 in the storage unit 153. The fog P / F 400 also transmits the detected pH and captured images from the device GW 41, which are highly confidential, to the on-premise system 5 without transmitting them to the cloud server 10. In the on-premise system 5, the detected pH and captured images are stored in the technical information DB 503d of the business division server 50 corresponding to the business division that owns the target factory. The machine learning engine 510 in the on-premise system 5 generates an on-premise-side trained model that infers the provisional dosage amounts of the adjuster and the separating agent. The machine learning engine 101 outside the intranet 2 generates a first cloud-side trained model that infers the final dosage amounts of the adjuster and the separating agent.

[0141] In the inference stage, the machine learning engine 510 acquires the detected pH and captured images from the technical information DB 503d of the business division server 50. Then, the machine learning engine 510 inputs the detected pH and captured images into the on-premise trained model to infer the provisional amounts of adjuster and separator to be added. The output information of the on-premise trained model, i.e., the provisional amounts of adjuster and separator to be added, is notified to the fog P / F 400 of the target intranet factory system 4. The fog P / F 400 notifies the cloud P / F 100 of the notified output information of the on-premise trained model.

[0142] In the inference stage, the machine learning engine 101 of the cloud P / F 100 acquires the detected temperature from the storage unit 153. Then, the machine learning engine 101 inputs the detected temperature and the output information of the on-premise trained model received from the fog P / F 400 into the first cloud trained model to infer the final dosage amounts of the adjuster and the separator. The cloud P / F 100 notifies the fog P / F 400 of the target intra-intra factory system 4 of the inferred dosage amounts (fourth information). The cloud server 10 does not store the inferred dosage amounts, which are highly confidential. The fog P / F 400 of the target intra-intra factory system 4 notifies the device GW 41, to which the PLC 42b, which controls the adjuster dosage device and the separator dosage device, is connected, of the inferred dosage amounts. The device GW 41 sets the dosage amounts for the PLC 42b. The PLC 42b controls the adjuster dosage device so that the set dosage amount of adjuster is dispensed into the treatment tank. The PLC 42b controls the separating agent supply device so that the set amount of separating agent is supplied to the treatment tank. As a result, in the target intranet factory system 4, the device 42 is controlled based on the fourth information output from the first cloud-side trained model.

[0143] As described above, in the example of Figure 16, the first information within the intranet 2 is input to a trained model within the intranet 2, and the output information of the trained model within the intranet 2 is input to a trained model outside the intranet 2. This makes it possible to reduce the possibility that the first information within the intranet 2 will leak outside the intranet 2, while still using a trained model outside the intranet 2. This makes it possible to increase the confidentiality of the first information.

[0144] In addition, by not transmitting highly confidential information 420, such as photographed images, obtained from device 42 by fog P / F 400 of factory system 4 to cloud P / F 100, the possibility of highly confidential information 420 generated in factory system 4 leaking outside intranet 2 can be reduced.

[0145] In the above example, the trained model outside the intranet 2 in Figure 16 was the first cloud-side trained model, but it may also be a second cloud-side trained model provided in the machine learning engine 111.

[0146] The on-premise trained model in the intranet 2 may receive input of information within the intranet 2 and output information of the first cloud-side trained model outside the intranet 2. Figure 17 shows this situation.

[0147] As shown in FIG. 17, fifth information is input to the trained model outside the intranet 2. Then, sixth information output from the trained model outside the intranet 2 and seventh information within the intranet 2 are input to the trained model within the intranet 2. The trained model within the intranet 2 outputs eighth information. A specific example of the processing shown in FIG. 17 will be described below. As an example, the operation of the information processing system 1 when the order quantity and order timing of a certain chemical (hereinafter sometimes referred to as the second chemical) used in the wastewater treatment system of factory B is determined using the processing of FIG. 17 will be described below.

[0148] For example, the first cloud-side trained model outputs temperature prediction information (sixth information) based on information (fifth information) from a device outside the information processing system 1. Furthermore, the on-premise-side trained model infers the order quantity and order timing of the second drug (eighth information).

[0149] The machine learning engine 101 outputs the output information of the first cloud-side trained model, i.e., the temperature prediction information, to the fog P / F 400 in the factory B system 4b. The fog P / F 400 outputs the temperature prediction information to the on-premise system 5. In the on-premise system 5, the temperature prediction information is input to the machine learning engine 510.

[0150] In the inference stage, the machine learning engine 510 acquires the current inventory amount of the second drug, ordering information for the second drug, and production plan information for factory B (seventh information) from the production management information DB 503b of the business division server 50 corresponding to the business division that includes factory B. Then, the machine learning engine 510 inputs the current inventory amount of the second drug, ordering information for the second drug, production plan information for factory B, and temperature forecast information into the on-premises trained model, and infers the order amount and order timing for the second drug. The computer device 51 notifies the fog P / F 400 of the factory B system 4b of the order amount and order timing for the second drug calculated by the machine learning engine 510. The fog P / F 400 of the factory B system 4b notifies, for example, an order server in the on-premises system 5 of the order amount and order timing for the second drug. The order server performs an order process for the second drug based on the order amount and order timing for the second drug.

[0151] As another example, an example will be described in which the amounts of the above-mentioned adjuster and separating agent to be added to the same treatment tank in a wastewater treatment system of a factory provided with an intra-factory system 4 are calculated by the process of FIG.

[0152] For example, the fog P / F 400 of the target intranet factory system 4 transmits the detected temperature from the device GW 41 to the cloud server 10. The cloud server 10 stores the detected temperature from the fog P / F 400 in the memory unit 153. This detected temperature indicates the temperature of the wastewater in the treatment tank where the adjuster and the separator are added. The detected temperature (fifth information) in the memory unit 153 is input to the first cloud-side trained model of the cloud server 10. In addition, predicted information of the air temperature (fifth information) obtained from a server outside the information processing system 1 is input to the first cloud-side trained model. Then, the first cloud-side trained model outputs predicted information of the temperature of the wastewater in the treatment tank (sixth information). The output information of the first cloud-side trained model is notified to the fog P / F 400 of the target intranet factory system 4. The fog P / F 400 notifies the on-premise system 5 of the output information of the first cloud-side trained model. In the on-premise system 5, the output information of the first cloud-side trained model is input to the machine learning engine 510.

[0153] Here, the effectiveness of the adjuster and the separating agent varies depending on the temperature of the wastewater. Therefore, the predicted information on the wastewater temperature affects the amount of adjuster and separating agent added. For example, if it is predicted that the wastewater temperature will drop, it is possible to add more adjuster and separating agent. Therefore, in this example, the on-premises trained model infers the amount of adjuster and separating agent to be added based on the output information of the first cloud-side trained model, i.e., the predicted information on the wastewater temperature.

[0154] The fog P / F 400 of the target intranet factory system 4 transmits the detected pH and captured images from the device GW 41 to the on-premise system 5. In the on-premise system 5, the detected pH and captured images are stored in the technical information DB 503d of the business division server 50 corresponding to the business division that has the target factory. In the inference stage, the machine learning engine 510 acquires the detected pH and captured images from the technical information DB 503d of the business division server 50. Then, the machine learning engine 510 inputs the detected pH and captured images (seventh information) to the on-premise trained model. The machine learning engine 510 also inputs the output information of the first cloud-side trained model, i.e., predicted information on the wastewater temperature, to the on-premise trained model. The on-premise trained model outputs the dosage amounts of adjuster and separator (eighth information).

[0155] The amounts of adjusting agent and separating agent added are notified to the fog P / F 400 of the target intra-plant factory system 4. The fog P / F 400 notifies the notified amounts to the device GW 41 connected to the PLC 42b that controls the adjusting agent adding device and separating agent adding device. The device GW 41 sets the amounts to be added to the PLC 42b. The PLC 42b controls the adjusting agent adding device so that only the set amount of adjusting agent is added to the treatment tank. The PLC 42b also controls the separating agent adding device so that only the set amount of separating agent is added to the treatment tank.

[0156] As described above, in the example of FIG. 17, the sixth information output from a trained model outside the intranet 2 and the seventh information within the intranet 2 are input to a trained model within the intranet 2. This makes it possible to reduce the possibility that the seventh information within the intranet 2 will leak outside the intranet 2, while still using a trained model outside the intranet 2. This therefore makes it possible to increase the confidentiality of the seventh information.

[0157] In the above example, the trained model outside the intranet 2 in Figure 17 was the first cloud-side trained model, but it may also be a second cloud-side trained model provided in the machine learning engine 111.

[0158] 16 may be performed using information 420 obtained from one factory system 4 as first information, and the process shown in Fig. 17 may be performed using information 420 obtained from another factory system 4 as seventh information. Also, the process shown in Fig. 16 and the process shown in Fig. 17 may be performed using information 420 obtained from the same factory system.

[0159] Furthermore, the information processing system 1 may use a combination of multiple statistical analysis models, just as it uses a combination of multiple trained models. For example, in the configuration shown in Fig. 16, the trained model within the intranet 2 may be replaced with an on-premise statistical analysis model generated by the statistical analysis engine 520, and the trained model outside the intranet 2 may be replaced with a cloud-side statistical analysis model generated by the statistical analysis engine 121. Furthermore, in the configuration shown in Fig. 17, the trained model within the intranet 2 may be replaced with an on-premise statistical analysis model, and the trained model outside the intranet 2 may be replaced with a cloud-side statistical analysis model.

[0160] <Other examples of information processing systems> The fog P / F 400 of each factory system 4 may transmit information 420 obtained from the device 42 to the cloud P / F 100 regardless of whether the information is highly confidential. In this case, for example, if the information 420 from the factory system 4 is information with low confidentiality, the cloud-side fog P / F 400 of the cloud P / F 100 may store the information in the storage unit 153. On the other hand, if the information 420 from the factory system 4 is information with high confidentiality, the cloud-side fog P / F 400 may input the information to the on-premise system 5 without storing it in the storage unit 153. In the on-premise system 5, for example, the information 420 from the cloud-side fog P / F 400 is stored in the storage unit 53 of the business department server 50.

[0161] Furthermore, the information processing system 1 may not include cloud servers 11 and 12. Furthermore, the cloud P / F 100 of the cloud server 10 may include a statistical analysis engine. In this case, the statistical analysis engine included in the cloud P / F 100 may be used instead of the statistical analysis engine 520 or instead of the statistical analysis engine 121. Furthermore, the information processing system 1 may include only one of a machine learning engine and a statistical analysis engine within the intranet 2. Furthermore, the information processing system 1 may include only one of a machine learning engine and a statistical analysis engine outside the intranet 2. At least one of the cloud servers 10, 11, and 12 may be provided separately from the information processing system 1 and may not be included in the information processing system 1.

[0162] Furthermore, the information processing system 1 may be provided with an off-intranet factory system 4 other than the factory D system 4d. Furthermore, the off-intranet factory system 4 may not be a system provided by a specific company, but may be a system provided by an organization related to the specific company. For example, the off-intranet factory system 4 may be a system provided by a subsidiary of the specific company, or a system provided by an affiliated company of the specific company. Furthermore, the off-intranet factory system 4 may be a system provided by a partner company of the specific company. Furthermore, the factory D system 4d may be provided within the intranet 2. Furthermore, the information processing system 1 does not have to be provided with the off-intranet factory system 4.

[0163] Furthermore, instead of the factory system 4, the information processing system 1 may include a system that has a configuration similar to that of the factory system 4 and is installed at a location other than the factory. For example, the information processing system 1 may include a system installed at a hospital. Alternatively, the information processing system 1 may include a system installed at a university. Even when the information processing system 1 includes a system installed at a location other than the factory, the information processing system 1 can operate in the same manner as described above.

[0164] As described above, the information processing system 1 has been described in detail, but the above description is merely illustrative in all respects and does not limit the scope of this disclosure. Furthermore, the various examples described above can be combined and applied as long as they are not mutually contradictory. It is understood that countless examples not illustrated can be envisioned without departing from the scope of this disclosure. [Explanation of symbols]

[0165] 1,3 Information Processing Systems 4. Factory System 4a Factory A System 4b Factory B System 4c Factory C System 4d Factory D System 10 Servers 42 devices 42a Sensor 42b PLC 420 Information

Claims

1. A control unit is provided, The control unit acquiring a first captured image of the wastewater in the first tank and a first detected temperature of the wastewater in the first tank; The apparatus has a first trained model that calculates wastewater temperature prediction information, which is prediction information of the temperature of wastewater, based on air temperature prediction information, which is prediction information of the air temperature, and the first detected temperature, and a second trained model that calculates a first dosage amount of separating agent to be added into the first tank based on the wastewater temperature prediction information and the first captured image.

2. 10. The apparatus of claim 1, The control unit acquiring a second photographed image of wastewater in a second tank located at a different location from the first tank; The apparatus has a third trained model that determines a second amount of separating agent to be dispensed into the second tank based on the second captured image.

3. 3. The device according to claim 1 or claim 2, the device is located within an intranet; The second trained model calculates the first input amount based on the first captured image and the wastewater temperature prediction information obtained outside the intranet.

4. 3. The device according to claim 1 or claim 2, the device is located outside an intranet; The second trained model calculates the first input amount based on the first captured image and the wastewater temperature prediction information obtained within the intranet.

5. 5. An apparatus according to any one of claims 1 to 4, The control unit further acquires a detected pH of the wastewater in the first tank, The second trained model calculates the third dosage amount and the first dosage amount of pH adjuster to be added to the first tank based on the sewage temperature prediction information, the first captured image, and the detected pH.

6. An apparatus according to any one of claims 1 to 5; a separating agent supplying device that supplies a separating agent to the treatment tank according to the first supply amount determined by the device.

7. An apparatus according to claim 5; a conditioner dosing device that doses a conditioner into the treatment tank according to the third dose determined by the device.

Images