Control system, control method, and program

Abstract

There is a need for a control system that inputs information undetectable by sensors into a large-scale language model and then has the large-scale language model perform a logical inference task. [Solution] The control system 100 for the industrial physical device 200 includes: an information acquisition unit that acquires setting information, measurement information, and non-measurement information of the industrial physical device 200; an inference unit that uses a large-scale language model 10 to generate evaluation information and basis information that underlies the execution results of the industrial physical device 200 based on the constraints of the setting information, measurement information, and non-measurement information; a data registration unit that registers the evaluation information and basis information as new knowledge information in the database 20; a plan generation unit that uses a large-scale language model 10 to generate plan information for driving the industrial physical device 200 based on the knowledge information registered in the database 20; and a drive generation unit that uses a large-scale language model 10 to generate a control program or control command for driving the industrial physical device 200 based on the plan information.

Description

Technical Field

[0001] The present disclosure relates to a control system, a control method, and a program for industrial physical devices.

Background Art

[0002] There is known a control system that controls industrial physical devices such as robots, machine tools, and measuring devices using a large language model (LLM). For example, Patent Document 1 discloses a technique for generating a dataset of robot paths using an LLM, performing a simulation, and generating a collision-free robot path that does not collide with obstacles.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In a conventional control system using a large language model, the input data to the large language model is limited to detection data of sensors such as measurement data and image data, and the large language model only performs a statistical classification task. Therefore, there is a need for a control system that inputs information that cannot be detected by sensors, such as device information (asset information) of measuring instruments, environmental information, and past defect information (knowledge information), to the large language model and makes the large language model perform a logical inference task.

Means for Solving the Problems

[0005] One example aspect that can be provided by the present disclosure is a control system for industrial physical devices, An information acquisition unit that acquires setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device, An inference unit that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that serves as the basis for said evaluation information, A data registration unit registers the evaluation information and the basis information generated by the inference unit as new knowledge information in a database, associating it with the setting information, the measurement information, and the non-measurement information. A planning generation unit that uses the large-scale language model to generate planning information for driving the industrial physical device based on the knowledge information registered in the database, A drive generation unit that generates a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, It is equipped with.

[0006] One example of an embodiment that may be provided by this disclosure is a method for controlling an industrial physical device, wherein a computer... An acquisition step of acquiring setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device; An evaluation step that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that forms the basis of said evaluation information, A registration step in which the evaluation information and the basis information generated by the evaluation step are registered in a database as new knowledge information, associated with the setting information, the measurement information, and the non-measurement information, A planning step of generating planning information for driving the industrial physical device based on the knowledge information registered in the database using the large-scale language model, A generation step of generating a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, It is equipped with.

[0007] One example of an embodiment that may be provided by this disclosure is a program executable by a processor installed in a control system for an industrial physical device, On the computer, An acquisition step of acquiring setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device; An evaluation step that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that forms the basis of said evaluation information, A registration step in which the evaluation information and the basis information generated by the evaluation step are registered in a database as new knowledge information, associated with the setting information, the measurement information, and the non-measurement information, A planning step of generating planning information for driving the industrial physical device based on the knowledge information registered in the database using the large-scale language model, A generation step of generating a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, Make it run.

[0008] According to the configurations described in each of the above examples, a large-scale language model is used to generate evaluation information and justification information regarding the execution results of industrial physical devices based on setting information, measurement information, and non-measurement information. Planning information for driving industrial physical devices is generated based on the justification information, and a control program or control command for driving industrial physical devices is generated based on the planning information. As a result, a control program or control command is generated by the logical inference of the large-scale language model, enabling highly accurate control of industrial physical devices. [Brief explanation of the drawing]

[0009] [Figure 1] This is a schematic diagram of the overall configuration, including a control system for an industrial physical device according to an embodiment of the present disclosure. [Figure 2] This block diagram shows the configuration of a large-scale language model according to an embodiment of this disclosure. [Figure 3] This is a block diagram showing the configuration of a control system according to an embodiment of the present disclosure. [Figure 4] This is a schematic diagram of a three-dimensional measuring device according to a first embodiment of the present disclosure. [Figure 5] This is an example of the processing flow of a control system according to the first embodiment of the present disclosure. [Modes for carrying out the invention]

[0010] Examples of embodiments will be described in detail below with reference to the attached drawings. In the drawings referenced below, the scale has been changed as necessary to make the elements shown recognizable.

[0011] Figure 1 is a schematic diagram of the overall configuration including the control system for an industrial physical device according to the embodiment of this disclosure. The overall configuration of the industrial physical device includes a control system 100, a Large Language Model (LLM) 10, a database 20, an industrial physical device 200, and a sensor 300.

[0012] The control system 100 is a control device for driving the industrial physical device 200, and is composed of, for example, a PC, a smartphone, or a tablet. The control system 100 acquires the setting information of the industrial physical device 200, the measurement information acquired by the sensor 300, and the non-measurement information that cannot be acquired by the sensor 300, generates a control program or a control command, and drives the industrial physical device 200.

[0013] The industrial physical device 200 is an edge device that measures physical data (for example, dimensions, images, waveforms, positions) of an object or performs direct operations (for example, machining, welding) at an industrial site such as a factory, infrastructure, or logistics. The industrial physical device 200 is, for example, a three-dimensional measuring device, an image measuring instrument, an electron microscope (SEM), an oscilloscope, an electrical property measuring device, an environmental measuring device, a machine tool (lathe, machining center, milling machine, etc.), a welding device, an injection molding machine, or a press device. The industrial physical device 200 is driven according to a control program or a control command transmitted from the control system 100.

[0014] The setting information of the industrial physical device 200 is information that a user sets in the control system 100 when driving the industrial physical device 200. Specifically, the setting information is information about the object of the industrial physical device 200 (for example, CAD drawings, dimensional information, material information) and specifications for the object (for example, tolerances, allowable values), and includes the constraint conditions when driving the industrial physical device 200.

[0015] The sensor 300 detects data (for example, dimensions, images, waveforms, positions, rigidity) related to the result of driving the industrial physical device 200 by a control program or a control command, and notifies the control system 100 of the detection result as measurement information. When the industrial physical device 200 is an edge device that measures physical data, the sensor 300 may be incorporated inside the industrial physical device 200. Also, the sensor 300 may be a measuring instrument capable of displaying the above detection result.

[0016] Non-measurement information is information that cannot be acquired by the sensor 300 and includes knowledge information 21, device information, or environmental information. The knowledge information 21 is, for example, past trend information, defect information, manual information, and recorded information of an operator who operates the industrial physical device 200, and is recorded in the database 20. The device information is, for example, ID information, characteristic information, calibration information, and reliability information of the industrial physical device 200, the sensor 300, measuring instruments, etc. The environmental information is information related to the environment where the industrial physical device 200, the sensor 300, etc. are installed, and is, for example, temperature information, humidity information, air pressure information, air flow information, vibration information, noise information, particle concentration information, cleanliness information, vacuum degree information, electromagnetic environment information.

[0017] The control system 100 is configured to be connectable to the large language model 10 and the database 20 via the communication network N. The communication network N is, for example, Bluetooth (registered trademark), the Internet, Wi-Fi (registered trademark), Li-Fi (registered trademark), a mobile communication system (3G, 4G, 5G, 6G, etc.), a wireless LAN, or a wired LAN.

[0018] The large language model 10 is, for example, GPT (registered trademark), Gemini, Claude, Llama. When the large language model 10 receives a prompt that is an inference request from the control system 100, it generates various text data according to the prompt and transmits it to the control system 100. A prompt is a command for giving instructions or questions to a large language model. The large language model 10 is recorded in the first external server.

[0019] Database 20 is a database in which knowledge information 21 is recorded and includes at least some of the functions of RAG (Retrieval Augmented Generation). When database 20 receives a search request from the control system 100, it searches the knowledge information 21 in accordance with the search request and sends the search results to the control system 100. Database 20 is configured by a second external server. Note that the first external server in which the large-scale language model 10 is recorded and the second external server that constitutes database 20 may be the same external server.

[0020] Figure 2 is a block diagram showing the configuration of a large-scale language model 10 according to an embodiment of the present disclosure. The large-scale language model 10 consists of, for example, a first large-scale language model 11, a second large-scale language model 12, and a third large-scale language model 13. The first large-scale language model 11 is a model mainly for evaluating the execution results of an industrial physical device 200. The second large-scale language model 12 is a model mainly for creating an execution plan for the industrial physical device 200. The third large-scale language model 13 is a model mainly for generating a control program or control command for the industrial physical device 200. The control program is, for example, a ladder program. The control command is, for example, a speed command or a position command. The first large-scale language model 11, the second large-scale language model 12, and the third large-scale language model 13 may each include a multimodal LLM.

[0021] By using these three large-scale language models (the first large-scale language model 11, the second large-scale language model 12, and the third large-scale language model 13), it becomes possible to autonomously perform tasks related to "evaluation," "planning," and "execution." In other words, the first large-scale language model 11, the second large-scale language model 12, and the third large-scale language model 13 are used to construct an AI agent.

[0022] When the industrial physical device 200 is executed, its execution results are detected as measurement information by the sensor 300. The first large-scale language model 11 generates evaluation information regarding the execution results of the industrial physical device 200 and supporting information that underlies the evaluation information, based on the measurement information, constraints included in the configuration information, and non-measurement information. The evaluation information indicates whether the measurement results of physical data related to the execution results of the industrial physical device 200 (e.g., dimensions, images, waveforms, position, stiffness) and the object resulting from the direct operation of the industrial physical device 200 (e.g., machining, welding) meet the constraints and specifications. The evaluation information is expressed, for example, as a binary value of OK / NG, or on a 10-point scale from 1 to 10 (1 being the lowest accuracy and 10 being the highest accuracy). For example, if the industrial physical device 200 performs a measurement of the dimensions of an object and the evaluation information is NG, the supporting information includes information explaining which parts of the object's dimensions do not meet which specifications and to what extent.

[0023] The second large-scale language model 12 generates planning information for driving the industrial physical device 200 based on the rationale information generated by the first large-scale language model 11. For example, suppose the industrial physical device 200 measures multiple dimensions of an object from a predetermined reference point, and the evaluation information is NG, and the rationale information includes information explaining that the multiple dimensions do not meet the specifications to the same extent. In that case, the second large-scale language model 12 infers that the predetermined reference point is not in an appropriate position, and generates planning information to change the reference point.

[0024] The third large-scale language model 13 generates a control program or control command for driving the industrial physical device 200 based on the planning information generated by the second large-scale language model 12.

[0025] Figure 3 is a block diagram showing the configuration of a control system according to an embodiment of the present disclosure. The control system 100 includes an information acquisition unit 101, an inference unit 102, a data registration unit 103, a plan generation unit 104, a drive generation unit 105, and a drive command unit 106. The information acquisition unit 101, the inference unit 102, the data registration unit 103, the plan generation unit 104, the drive generation unit 105, and the drive command unit 106 are interconnected by a bus B.

[0026] The information acquisition unit 101 acquires setting information of industrial physical devices 200 set by the user, measurement information transmitted from sensors 300, and non-measurement information (equipment information, environmental information, and knowledge information 21 transmitted from database 20, which are set by the user). The information acquisition unit 101 may include a UI operation screen in which the user can set setting information, equipment information, and environmental information.

[0027] The inference unit 102 sends a prompt to the first large-scale language model 11, causing the first large-scale language model 11 to generate evaluation information and justification information regarding the execution results of industrial physical devices based on the constraints of the configuration information, measurement information, and non-measurement information, and then acquires the evaluation information and justification information.

[0028] The data registration unit 103 registers the evaluation information and rationale information obtained by the inference unit 102 as new knowledge information in the database 20, associating it with setting information, measurement information, and non-measurement information.

[0029] The planning generation unit 104 sends a prompt to the second large-scale language model 12, causing the second large-scale language model 12 to generate planning information for driving the industrial physical device 200 based on the underlying information, and then acquires the planning information.

[0030] The drive generation unit 105 sends a prompt to the third large-scale language model 13, causing the third large-scale language model 13 to generate a control program or control command for driving the industrial physical device 200 based on the planning information, and then obtains the control program or control command.

[0031] The drive command unit 106 transmits the control program or control command acquired by the drive generation unit 105 to the industrial physical device 200 to drive it.

[0032] Although not shown in the diagram, the control system 100 includes processors such as a CPU (Central Processing Unit), MPU (Micro Processing Unit), and GPU (Graphics Processing Unit), storage, ROM (Read Only Memory), and RAM (Random Access Memory). The processor is configured to load a specified program from various programs embedded in the storage or ROM onto the RAM and to execute various processes in cooperation with the RAM. The storage is a storage device such as an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory, and is configured to store programs and various data (such as application document data).

[0033] Figure 4 is a schematic diagram of a three-dimensional measuring device 200A according to a first embodiment of the present disclosure. The three-dimensional measuring device 200A is an example of an industrial physical device 200.

[0034] The three-dimensional measuring device 200A acquires the coordinate values ​​of the measurement points of the workpiece W, which is the object to be measured, and performs measurement of the three-dimensional shape of the workpiece W. The three-dimensional measuring device 200A comprises a table 210, a right Y carriage 220R, a left Y carriage 220L, an X guide 230, an X carriage 240, a Z carriage 250, and a probe head 260 including a probe 261.

[0035] Table 210 has a right Y carriage 220R erected at one end of its upper surface in the X-axis direction, and a left Y carriage 220L erected at the other end. The right Y carriage 220R and the left Y carriage 220L are supported by the table 210 so as to be movable in the Y-axis direction. The X guide 230 has one end in the X-axis direction supported by the right Y carriage 220R, and the other end in the X-axis direction supported by the left Y carriage 220L. The X carriage 240 is supported by the X guide 230 so as to be movable in the X-axis direction. The Z carriage 250 is supported by the X carriage 240 so as to be movable along the Z-axis direction.

[0036] The probe head 260 is attached to the lower end of the Z carriage 250. The probe head 260 can be driven in five directions: the X axis, Y axis, Z axis, rotation axis around the X axis, and rotation axis around the Z axis. The probe 261 is, for example, a contact probe. When the probe head 260 is driven, the probe 261 comes into contact with the workpiece W, which is the object to be measured, and the coordinate values ​​of the contact point can be obtained.

[0037] In Figure 4, the workpiece W is rectangular in shape and has a groove in the Z direction at its upper end. The three-dimensional measuring device 200A measures the width D from one side P1 in the X direction of the groove to the other side P2, using P1 as the reference point. In this case, the probe 261 is brought into contact with side P1 to calculate the XYZ coordinates of side P1. Then, the probe 261 is brought into contact with side P2 to calculate the XYZ coordinates of side P2. The width D is calculated from the difference between the XYZ coordinates of side P1 and side P2. In this example, the probe 261 corresponds to the sensor 300 shown in Figure 1.

[0038] Figure 5 shows an example of the processing flow of a control system according to the first embodiment of the present disclosure. First, the information acquisition unit 101 acquires information on the workpiece W, which is the object to be measured by the three-dimensional measuring device 200A (e.g., CAD drawings, dimensional information, material information), and specifications for the workpiece W (e.g., tolerances, permissible values) as setting information. The information acquisition unit 101 also acquires non-measurement information, including equipment information (e.g., ID information, characteristic information, calibration information, reliability information of the three-dimensional measuring device 200A) and environmental information (e.g., temperature information, humidity information, atmospheric pressure information, airflow information, vibration information, noise information of the environment in which the three-dimensional measuring device 200A is installed) (step S1).

[0039] Next, the drive command unit 106 transmits an initial control program or control command recorded in storage or the like to the three-dimensional measuring device 200A to drive it. In the example shown in Figure 4, the probe 261 is brought into contact with side P1 to calculate the XYZ coordinates of side P1, and the probe 261 is brought into contact with side P2 to calculate the XYZ coordinates of side P2 (step S2).

[0040] Next, the three-dimensional measuring device 200A calculates the width D from side P1 to side P2 from the difference between the two XYZ coordinates, and the information acquisition unit 101 acquires the width D as measurement information (step S3).

[0041] Next, the information acquisition unit 101 acquires past trend information, malfunction information, manual information, and record information of operators operating the industrial physical device 200 from the database 20 as knowledge information 21 (step S4).

[0042] Next, the inference unit 102 sends a prompt to the first large-scale language model 11 containing constraints for the setting information, measurement information, and non-measurement information. Based on this information, the first large-scale language model 11 evaluates the execution result of the three-dimensional measuring device 200A (measured width D calculated from the difference in XYZ coordinates of side P1 and side P2) and generates evaluation information, as well as the basis for the evaluation. In the example in Figure 4, for example, the specification for width D is a dimension of 50.0 mm and a tolerance of ±3%. Therefore, the measured value of width D must fall within the range of 48.5 mm to 51.5 mm. If the measured value of width D obtained in step S3 is 51.8 mm, it does not fall within the range of 48.5 mm to 51.5 mm and does not satisfy the specification, so the evaluation information is NG. The basis information in this case is a violation of the specification for width D for workpiece W (width D exceeds 51.5 mm by 0.3 mm). The inference unit 102 obtains evaluation information and justification information generated by the first large-scale language model 11 (step S5).

[0043] Next, the data registration unit 103 associates the above evaluation information and supporting information regarding the three-dimensional measuring device 200A and registers it in the database 20 as new knowledge information, associated with setting information, measurement information, and non-measurement information (step S6). In this way, new knowledge information is accumulated in the database 20 each time the three-dimensional measuring device 200A is driven.

[0044] Next, if the evaluation information is NG (NO in step S7), the planning generation unit 104 sends a prompt to the second large-scale language model 12 that includes knowledge information with the new knowledge information added. The second large-scale language model 12 generates planning information for driving the three-dimensional measuring device 200A based on the knowledge information (step S8). Specifically, it is assumed that the knowledge information includes trend information or operator record information such as "For workpiece W, when measuring the width D from side P1 to the other side P2 with side P1 as the reference point, the measured value tends to be larger than the actual dimension." In that case, the second large-scale language model 12 generates planning information to measure the width D from side P2 to side P1, for example, with side P2 as the reference point. The planning generation unit 104 acquires the planning information generated by the second large-scale language model 12.

[0045] Next, the drive generation unit 105 sends a prompt containing planning information to the third large-scale language model 13. The third large-scale language model 13 generates a control program or control command for driving the three-dimensional measuring device 200A based on the planning information (step S9). Specifically, for example, it generates a control program or control command that calculates the XYZ coordinates of side P2 by bringing the probe 261 into contact with side P2, the XYZ coordinates of side P1 by bringing the probe 261 into contact with side P1, and measures the width D from the difference between the XYZ coordinates of side P1 and side P2. The drive generation unit 105 acquires the control program or control command generated by the third large-scale language model 13.

[0046] Next, the drive command unit 106 drives the three-dimensional measuring device 200A with a control program or control command generated by the third large-scale language model 13 (step S10).

[0047] Then, repeat step S5 above. In step S7, if the evaluation information is OK (YES in step S7), terminate the processing flow.

[0048] As described above, in the control system 100 according to this embodiment, a large-scale language model is made to perform a logical inference task based on measured and non-measured information. Therefore, it becomes possible to evaluate the execution results of industrial physical devices and to plan their operation, taking into account not only the state of sensors and measuring instruments, the environment such as temperature and humidity in the measurement environment, but also past control cases and know-how that are difficult to put into words. Consequently, high-precision control of industrial physical devices becomes possible.

[0049] Each of the configurations referenced in the preceding explanation is merely an example to facilitate understanding of this disclosure. Each configuration example may be modified or combined with other configuration examples as appropriate within the scope of the intent of this disclosure.

[0050] The large-scale language model 10 according to the embodiment of this disclosure consists of three models: a first large-scale language model 11, a second large-scale language model 12, and a third large-scale language model 13, but the number of models is not limited to three. The first large-scale language model 11, the second large-scale language model 12, and the third large-scale language model 13 may be the same large-scale language model.

[0051] Furthermore, although the large-scale language model 10 in the first embodiment of this disclosure is recorded on a first external server and the database 20 is configured on a second external server, the large-scale language model 10 and the database 20 may also be recorded and configured within the control system 100.

[0052] Furthermore, while the industrial physical device 200 in the first embodiment of this disclosure is a three-dimensional measuring device 200A, it may also be an image measuring instrument, an electron microscope (SEM), an oscilloscope, an electrical characteristic measuring instrument, an environmental measuring instrument, a robot, a machine tool, or a welding device. The industrial physical device 200 may also include multiple such devices.

[0053] Furthermore, the control system 100 according to the embodiment of this disclosure may have a system configuration of SaaS (Software as a Service), PaaS (Platform as a Service), IaaS (Infrastructure as a Service), BaaS (Backend as a Service), or DaaS (Desktop as a Service). [Explanation of symbols]

[0054] 10 Large-scale language models 11. The First Large-Scale Language Model 12. Second Large-Scale Language Model 13. Third Large-Scale Language Model 20 Databases 21 Knowledge Information 100 control systems 101 Information Acquisition Department 102 Reasoning part 103 Data Registration Department 104 Planning Generation Unit 105 Drive generation unit 106 Drive command unit 200 Industrial Physical Devices 200A three-dimensional measuring device 300 sensors

Claims

1. A control system for industrial physical devices, An information acquisition unit that acquires setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device, An inference unit that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that serves as the basis for said evaluation information, A data registration unit registers the evaluation information and the basis information generated by the inference unit as new knowledge information in a database, associating it with the setting information, the measurement information, and the non-measurement information. A planning generation unit that uses the large-scale language model to generate planning information for driving the industrial physical device based on the knowledge information registered in the database, A drive generation unit that generates a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, A control system equipped with the following features.

2. The control system according to claim 1, wherein the equipment information includes at least one of the characteristic information, calibration information, and reliability information of the measuring instrument.

3. The control system according to claim 1, wherein the environmental information includes at least one of the following: temperature, humidity, atmospheric pressure, airflow, vibration, noise, particle concentration, cleanliness, vacuum level, and electromagnetic environment information in the measurement environment.

4. The control system according to claim 1, wherein the sensor or measuring instrument measures parameters relating to the execution result of the industrial physical device by the control program or control command generated by the drive generation unit.

5. The control system according to claim 1, wherein the knowledge information includes at least one of past trend information relating to the control of the industrial physical device, malfunction information relating to the control of the industrial physical device, and manual information relating to the control of the industrial physical device.

6. The inference unit generates the evaluation information and the rationale information using the first large-scale language model. The aforementioned plan generation unit generates plan information using a second large-scale language model, The control system according to claim 1, wherein the drive generation unit generates the control program or the control command using a third large-scale language model.

7. A method for controlling industrial physical devices, wherein a computer, An acquisition step of acquiring setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device; An evaluation step that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that forms the basis of said evaluation information, A registration step in which the evaluation information and the basis information generated by the evaluation step are registered in a database as new knowledge information, associated with the setting information, the measurement information, and the non-measurement information, A planning step of generating planning information for driving the industrial physical device based on the knowledge information registered in the database using the large-scale language model, A generation step of generating a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, A control method comprising:

8. A program executable by a processor installed in a control system for industrial physical devices, On the computer, An acquisition step of acquiring setting information of the industrial physical device, measurement information acquired by a sensor or measuring instrument, and non-measurement information that cannot be acquired by the sensor or measuring instrument, including knowledge information, equipment information, or environmental information related to the control of the industrial physical device; An evaluation step that uses a large-scale language model to generate evaluation information regarding the execution results of the industrial physical device based on the constraints of the setting information, the measurement information, and the non-measurement information, and the basis information that forms the basis of said evaluation information, A registration step in which the evaluation information and the basis information generated by the evaluation step are registered in a database as new knowledge information, associated with the setting information, the measurement information, and the non-measurement information, A planning step of generating planning information for driving the industrial physical device based on the knowledge information registered in the database using the large-scale language model, A generation step of generating a control program or control command for driving the industrial physical device based on the planning information using the large-scale language model, A program that executes something.

Images