Control device, control system, and control method

Abstract

To provide a control device and the like that can achieve stable operation of a control target device while incorporating AI processing.SOLUTION: A control device performs feedback control of a control target device. The control device has a first processing unit that uses AI processing to acquire a first output value related to the feedback control of the control target device, and a second processing unit that uses non-AI processing to acquire a second output value related to the feedback control. The control device has a correction unit that corrects a selection ratio of the first output value and the second output value based on the reliability of the first output value, and a control unit that outputs a control signal for feedback controlling the control target device based on the selection ratio.SELECTED DRAWING: Figure 1A

Description

[Technical Field]

[0001] The present invention relates to a control device, a control system, and a control method. [Background technology]

[0002] For example, a known plant optimization support device has a machine learning calculation unit built into a control device that has a feedback control unit that adjusts the manipulated variable that operates the plant process of the controlled equipment in accordance with the difference between the controlled variable target value and the controlled variable measurement value.

[0003] The machine learning calculation unit generates a machine learning model by learning using accumulated measurement values ​​of the controlled equipment, and estimates a controlled variable by referring to the generated machine learning model and inputting measurement values ​​currently acquired from the plant process. The machine learning calculation unit determines a controlled variable target value using the difference between a set target value, which is a target value of a predetermined controlled variable, and the estimated controlled variable. As a result, the feedback control unit can feedback control the controlled equipment. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2018-106561 [Patent Document 2] Patent Publication No. 2021-152702 Summary of the Invention [Problem to be solved by the invention]

[0005] However, machine learning models are learned inductively through data, and the reality is that they do not necessarily produce appropriate outputs for untrained data, such as extrapolated data. In particular, plants require stable operation of controlled equipment. Even if efficient results can be obtained in many cases by incorporating machine learning models, stable operation may not be guaranteed in a small number of cases. Therefore, there is a demand for control devices that can achieve stable operation of controlled equipment while incorporating AI (artificial intelligence) processing such as machine learning models.

[0006] On one hand, the objective is to provide a control device that can realize stable operation of controlled equipment while incorporating AI processing. [Means for solving the problem]

[0007] In one aspect, the control device disclosed herein performs feedback control of a control target device. The control device includes a first processing unit that uses AI (Artificial Intelligence) processing to acquire a first output value related to the feedback control of the control target device, and a second processing unit that uses non-AI processing to acquire a second output value related to the feedback control. The control device further includes a correction unit that corrects a selection ratio of the first output value and the second output value based on the reliability of the first output value, and a control unit that outputs a control signal for feedback controlling the control target device based on the selection ratio. [Effects of the Invention]

[0008] According to one embodiment, stable operation of controlled equipment can be achieved while introducing AI processing. [Brief explanation of the drawings]

[0009] [Figure 1A] FIG. 1A is an explanatory diagram illustrating an example of a control system according to a first embodiment. [Figure 1B]FIG. 1B is a flowchart showing an example of processing operations related to control processing of the control device. [Figure 2] FIG. 2 is an explanatory diagram illustrating an example of a control system according to the second embodiment. [Figure 3] FIG. 3 is an explanatory diagram illustrating an example of a control system according to a third embodiment. [Figure 4] FIG. 4 is an explanatory diagram illustrating an example of a control system according to a fourth embodiment. [Figure 5] FIG. 5 is an explanatory diagram illustrating an example of a control system according to a fifth embodiment. [Figure 6] FIG. 6 is an explanatory diagram illustrating an example of a hardware configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of the control device and the like disclosed in the present application will be described in detail with reference to the drawings. Note that the disclosed technology is not limited to these embodiments. Furthermore, the embodiments described below may be modified as appropriate within the scope of the present invention.

[0011] Example 1 FIG. 1A is an explanatory diagram illustrating an example of a control system 1 according to a first embodiment. The control system 1 illustrated in FIG. 1A is a feedback control system having a control device 2 and a control target device 3. The control target device 3 is a device that is subject to feedback control, such as a boiler. The control device 2 is a feedback control device that feedback controls the control target device 3. The control device 2 has an input unit 11, a control unit 12, an AI (Artificial Intelligence: hereinafter simply referred to as AI) sensor 13, a non-AI sensor 14, a reliability determination unit 15, and a correction unit 16.

[0012] The input unit 11 inputs a target value and a feedback value, and inputs a correction value, which is the difference between the target value and the feedback value, to the control unit 12. The control unit 12 calculates an operation amount for controlling the controlled device 3 based on the correction value, which is the difference between the target value and the feedback value, and a set algorithm, and outputs a control signal corresponding to the calculated operation amount to the controlled device 3. The controlled device 3 is driven and controlled in accordance with the control signal from the control unit 12.

[0013] The AI ​​sensor 13 is a first processing unit such as a first sensor that uses AI processing to acquire a first output value related to feedback control of the control target device 3. The AI ​​sensor 13 is a sensor that uses input of photographed images of the control target device 3, for example, the inside of a boiler, and a neural network to estimate the temperature, pressure, etc. of the boiler from the conditions of multiple locations inside the boiler, and detects the estimated temperature, pressure, etc. as a first output value.

[0014] The non-AI sensor 14 is a second processing unit such as a second sensor that does not use AI processing and acquires a second output value related to feedback control of the control-target device 3. The non-AI sensor 14 is a sensor that actually measures the temperature and pressure of the control-target device 3, for example, a boiler, and detects the actually measured temperature and pressure as a second output value.

[0015] The reliability determination unit 15 calculates the reliability of the first output value, which is the output of the AI ​​sensor 13, and determines the reliability of the AI ​​sensor 13 based on the calculated reliability. The reliability determination unit 15 includes a calculation unit 21 and a determination unit 22. The calculation unit 21 quantifies the reliability of the first output value of the AI ​​sensor 13 to calculate the reliability of the first output value. Methods for calculating the reliability include, for example, a method that uses only the first output value obtained from the AI ​​sensor 13, and a method that uses the first output value obtained from the AI ​​sensor 13 and other data.

[0016] Methods that use only the first output value obtained from the AI ​​sensor 13 include various methods such as a simple output of the first output value of the AI ​​sensor 13, distribution information of the first output value of the AI ​​sensor 13, a distribution of intermediate features in training data, a reconstruction error such as an autoencoder, and a dropout mechanism. Note that the distribution information of the first output value is not limited to an estimated value as the output of the AI ​​sensor 13, but also includes a variance value. An example of a method that uses the first output value obtained from the AI ​​sensor 13 and other data is a method that uses Pearson's product-moment correlation coefficient.

[0017] The determination unit 22 determines whether the AI ​​sensor 13 is reliable based on the reliability of the first output value calculated by the calculation unit 21. The determination unit 22 sets a reference value, such as a predetermined threshold value or a predetermined amount of change over time, and determines whether the AI ​​sensor 13 is reliable based on the result of comparing the reliability with the reference value.

[0018] The determination unit 22 determines whether the reliability of the first output value is equal to or greater than a predetermined threshold, and if the reliability of the first output value is equal to or greater than the predetermined threshold, determines that the AI ​​sensor 13 is reliable. If the reliability of the first output value is less than the predetermined threshold, the determination unit 22 determines that the AI ​​sensor 13 is unreliable.

[0019] The determination unit 22 also determines whether the time-series change in the reliability of the first output value is equal to or greater than a predetermined change, and determines that the AI ​​sensor 13 is reliable if the time-series change in the reliability of the first output value is less than the predetermined change. The determination unit 22 also determines that the AI ​​sensor 13 is unreliable if the time-series change in the reliability of the first output value is equal to or greater than the predetermined change.

[0020] The correction unit 16 corrects the selection ratio of the first output value and the second output value based on the determination result of the reliability of the AI ​​sensor 13. The correction unit 18 includes a selection unit 23 and an output control unit 24. If the AI ​​sensor 13 is reliable, the selection unit 23 selects only the first output value from the AI ​​sensor 13 from the first output value and the second output value. If the AI ​​sensor 13 is unreliable, the selection unit 23 selects only the second output value from the non-AI sensor 14 from the first output value and the second output value.

[0021] The output control unit 24 inputs a feedback value to the input unit 11 using the first output value of the AI ​​sensor 13 or the second output value of the non-AI sensor 14 selected by the selection unit 23 as an output value.

[0022] Then, the input unit 11 inputs a correction value, which is the difference between the target value and the feedback value from the correction unit 16, to the control unit 12. Then, the control unit 12 calculates an adjustment amount for the controlled device 3 based on the correction value and the setting algorithm, and outputs a control signal according to the calculated adjustment amount to the controlled device 3. As a result, the controlled device 3 can achieve stable operation.

[0023] 1B is a flow diagram showing an example of processing operations related to the control processing of the control device 2. The control processing shown in FIG. 1B is processing for controlling the controlled device 3 based on a first output value from the AI ​​sensor 13 or a second output value from the non-AI sensor 14. In FIG. 1B, the control device 2 determines whether or not the first output value has been acquired from the AI ​​sensor 13 (step S11). If the first output value has been acquired from the AI ​​sensor 13 (step S11: Yes), the calculation unit 21 in the control device 2 calculates the reliability of the first output value (step S12).

[0024] The determination unit 22 in the control device 2 determines whether the calculated reliability of the first output value is equal to or greater than a predetermined threshold (step S13). If the reliability of the first output value is equal to or greater than the predetermined threshold (step S13: Yes), the determination unit 22 determines that the AI ​​sensor 13 is reliable (step S14).

[0025] If the selection unit 23 in the control device 2 determines that the AI ​​sensor 13 is reliable, it selects the first output value from the AI ​​sensor 13 as a feedback value (step S15) and inputs the selected feedback value to the input unit 11 (step S16).

[0026] The input unit 11 in the control device 2 inputs a correction value, which is the difference between the input target value and the input feedback value, to the control unit 12 (step S17). The control unit 12 in the control device 2 calculates an adjustment amount for the controlled-target device 3 based on the correction value and the setting algorithm (step S18). The control unit 12 outputs a control signal corresponding to the calculated adjustment amount to the controlled-target device 3 (step S19), and ends the processing operation shown in FIG. 1B. As a result, the controlled-target device 3 can achieve stable operation.

[0027] Furthermore, if the reliability of the first output value is not equal to or greater than the predetermined threshold (step S13: No), that is, if the reliability of the first output value is less than the predetermined value, the determiner 22 determines that the AI ​​sensor 13 is unreliable (step S20). If the selector 23 determines that the AI ​​sensor 13 is unreliable, the selector 23 selects the second output value from the non-AI sensor 14 as the feedback value (step S21), and proceeds to the process of step S16 in which the selector 23 inputs the selected feedback value to the input unit 11.

[0028] If the selector 23 has not acquired the first output value from the AI ​​sensor 13 (step S11: No), the selector 23 proceeds to the process of step S21, where it selects the second output value from the non-AI sensor 14 as the feedback value.

[0029] The control device 2 of the first embodiment selects the first output value of the AI ​​sensor 13 or the second output value of the non-AI sensor 14 based on the reliability of the first output value from the AI ​​sensor 13. The control device 2 outputs a control signal according to the selected first output value or second output value to the controlled device 3. As a result, the control device 2 can achieve stable operation of the controlled device 3 while introducing the AI ​​sensor 13 with machine learning capabilities.

[0030] If the reliability of the first output value is less than a predetermined threshold and the control device 2 determines that the AI ​​sensor 13 is unreliable, the control device 2 does not select the first output value of the AI ​​sensor 13 and instead selects the second output value of the non-AI sensor 14. If the reliability of the first output value is equal to or greater than a predetermined threshold and the control device 2 determines that the AI ​​sensor 13 is reliable, the control device 2 selects the first output value of the AI ​​sensor 13. The control device 2 outputs a control signal corresponding to the selected first or second output value to the controlled device 3. As a result, the control device 2 can achieve stable operation of the controlled device 3 while introducing an AI sensor 13 with machine learning capabilities.

[0031] If the time-series change in the reliability of the first output value is equal to or greater than a predetermined change amount and it is determined that the AI ​​sensor 13 is unreliable, the control device 2 does not select the first output value of the AI ​​sensor 13 and instead selects the second output value of the non-AI sensor 14. If the time-series change in the reliability of the first output value is less than a predetermined threshold and it is determined that the AI ​​sensor 13 is reliable, the control device 2 selects the first output value of the AI ​​sensor 13. The control device 2 outputs a control signal corresponding to the selected first output value or second output value to the controlled device 3. As a result, the control device 2 can achieve stable operation of the controlled device 3 while introducing an AI sensor 13 with machine learning capabilities.

[0032] In addition, the control device 2 in Example 1 is exemplified as selecting either the first output value of the AI ​​sensor 13 or the second output value of the non-AI sensor 14 based on the reliability of the first output value of the AI ​​sensor 13, but this is not limited to this, and an embodiment thereof will be described below as Example 2.

[0033] <Example 2> Fig. 2 is an explanatory diagram showing an example of a control system 1A according to a second embodiment. The control system 1A shown in Fig. 2 is a feedback control system having a control device 2A and a control target device 3. The control device 2A according to the second embodiment differs from the control device 2 according to the first embodiment in that the control device 2A has a storage unit 14A in which conservative sensor values ​​are stored in advance instead of the non-AI sensor 14.

[0034] The storage unit 14A is a second processing unit that stores a conservative sensor value in advance. The conservative sensor value is a feedback value that serves as a second output value that is set in advance so that the control-target device 3 operates normally.

[0035] The correction unit 16A corrects the selection ratio of the first output value and the second output value based on the determination result of the reliability of the AI ​​sensor 13. The correction unit 16A includes a selection unit 23A and an output control unit 24A. If the AI ​​sensor 13 is reliable, the selection unit 23A selects only the first output value from the AI ​​sensor 13 out of the first output value and the second output value. If the AI ​​sensor 13 is unreliable, the selection unit 23A selects only the second output value read from the storage unit 6A out of the first output value and the second output value.

[0036] The output control unit 24A inputs a feedback value to the input unit 11 using the first output value of the AI ​​sensor 13 selected by the selection unit 23A or the second output value of the storage unit 14A as an output value.

[0037] Then, the input unit 11 inputs a correction value, which is the difference between the target value and the feedback value from the correction unit 16A, to the control unit 12. Then, the control unit 12 calculates an adjustment amount for the controlled device 3 based on the correction value and the setting algorithm, and outputs a control signal according to the calculated adjustment amount to the controlled device 3. As a result, the controlled device 3 can achieve stable operation.

[0038] The control device 2A of the second embodiment selects the first output value of the AI ​​sensor 13 or the second output value of the memory unit 14A based on the reliability of the first output value from the AI ​​sensor 13. The control device 2A outputs a control signal according to the selected first output value or second output value to the controlled device 3. As a result, the control device 2A can achieve stable operation of the controlled device 3 while introducing the AI ​​sensor 13 with machine learning capabilities.

[0039] If the reliability of the first output value is less than a predetermined threshold and the control device 2A determines that the AI ​​sensor 13 is unreliable, the control device 2A does not select the first output value of the AI ​​sensor 13 and instead selects the second output value read from the storage unit 14A. If the reliability of the first output value is equal to or greater than a predetermined threshold and the control device 2A determines that the AI ​​sensor 13 is reliable, the control device 2A selects the first output value of the AI ​​sensor 13. The control device 2A outputs a control signal corresponding to the selected first or second output value to the controlled device 3. As a result, the control device 2A can achieve stable operation of the controlled device 3 while introducing an AI sensor 13 with machine learning capabilities.

[0040] If the time-series change in the reliability of the first output value is equal to or greater than a predetermined change amount and the control device 2A determines that the AI ​​sensor 13 is unreliable, the control device 2A does not select the first output value of the AI ​​sensor 13 and instead selects the second output value of the non-AI sensor 14. If the time-series change in the reliability of the first output value is less than a predetermined threshold and the control device 2A determines that the AI ​​sensor 13 is reliable, the control device 2A selects the first output value of the AI ​​sensor 13. The control device 2A outputs a control signal corresponding to the selected first or second output value to the controlled device 3. As a result, the control device 2A can achieve stable operation of the controlled device 3 while introducing an AI sensor 13 with machine learning capabilities.

[0041] In the first embodiment, the correction unit 16 selects either the first output value from the AI ​​sensor 13 or the second output value from the non-AI sensor 14 based on the reliability of the first output value of the AI ​​sensor 13. However, the present invention is not limited to this, and an embodiment thereof will be described below as a third embodiment.

[0042] Example 3 3 is an explanatory diagram showing an example of a control system 1B according to Example 3. Note that the same components as those in the control system 1 according to Example 1 are denoted by the same reference numerals, and redundant descriptions of the components and operations will be omitted.

[0043] 3 is a feedback control system having a control device 2B and a controlled device 3. The control device 2B of the third embodiment differs from the control device 2 of the first embodiment in that a correction unit 16B corrects the selection ratio between the first output value of the AI ​​sensor 13 and the second output value of the non-AI sensor 14 based on the reliability of the first output value of the AI ​​sensor 13, and selects a hybrid mixture of the first output value and the second output value.

[0044] The correction unit 16B corrects the selection ratio between the first output value of the AI ​​sensor 13 and the second output value of the non-AI sensor 14 based on the reliability determination result of the AI ​​sensor 13, and selects by hybridly mixing the first output value and the second output value. The correction unit 16B includes a calculation unit 25B, a selection unit 23B, and an output control unit 24B. The calculation unit 25B calculates the selection ratio between the first output value and the second output value based on the difference between the reliability of the first output value and a predetermined threshold. The selection ratio is calculated, for example, as a weighted average of the second output value of the non-AI sensor 14 and the first output value of the AI ​​sensor 13. The selection unit 23B selects by hybridly mixing the first output value and the second output value based on the calculated selection ratio.

[0045] The output control unit 24BB calculates an output value based on the selection ratio of the first output value of the AI ​​sensor 13 and the second output value of the non-AI sensor 14 selected by the selection unit 23B, and inputs a feedback value to the input unit 11 as the calculated output value.

[0046] Then, the input unit 11 inputs a correction value, which is the difference between the target value and the feedback value from the correction unit 16B, to the control unit 12. Then, the control unit 12 calculates an adjustment amount for the controlled device 3 based on the correction value and the setting algorithm, and outputs a control signal according to the calculated adjustment amount to the controlled device 3. As a result, the controlled device 3 can achieve stable operation.

[0047] The control device 2B of the third embodiment calculates a selection ratio between the first output value of the AI ​​sensor 13 and the second output value of the non-AI sensor 14 based on the reliability of the first output value of the AI ​​sensor 13. Furthermore, the control device 2B calculates an output value based on the selection ratio between the first output value and the second output value, and inputs a feedback value of the calculated output value to the input unit 11. The control device 2B outputs a control signal according to the selected first output value and second output value to the controlled device 3. As a result, the control device 2B can achieve stable operation of the controlled device 3 while introducing an AI sensor 13 with machine learning functionality.

[0048] In the control device 2 of Example 1, the reliability of the AI ​​sensor 13 is determined based on the reliability of the first output value of the AI ​​sensor 13. However, the present invention is not limited to the AI ​​sensor 13. For example, the reliability of the AI ​​processing unit may be determined based on the reliability of the first output value, which is the output of the AI ​​processing unit used in the control unit 12. Such an embodiment will be described below as Example 4.

[0049] Example 4 FIG. 4 is an explanatory diagram showing an example of a feedback control system 1C according to a fourth embodiment. The control system 1C shown in FIG. 4 is a feedback control system having a control device 2C and a control target device 3. The control target device 3 is, for example, a device such as a motor. The control target device 3 is driven and controlled in response to a control signal from the control device 2C. The control device 2C has a sensor 17 and a control unit 30. The sensor 17 is a sensor that detects an output signal from the control target device 3, for example, the rotation speed of the motor. The control unit 30 drives and controls the control target device 3, for example, by outputting a control signal to the control target device 3 to change the rotation speed of the motor.

[0050] The control unit 30 has a reliability determination unit 31, an AI processing unit 32, a non-AI processing unit 33, and a correction unit 34. The AI ​​processing unit 32 is a first control unit within the first processing unit that outputs a first control signal, which is a first output value for controlling the control-target device 3, using the sensor value acquired from the sensor 17 and a predetermined AI model.

[0051] The non-AI processing unit 33 is a second control unit within the second processing unit that outputs a second control signal, which is a second output value for controlling the control-target device 3, based on the sensor value acquired from the sensor 17. The non-AI processing unit 33 may issue an alarm if it is unable to output the second control signal.

[0052] The reliability determination unit 31 determines whether the AI ​​processing unit 32 is reliable based on a first output value that is an output of the AI ​​processing unit 32. The reliability determination unit 31 includes a calculation unit 21C and a determination unit 22C. The calculation unit 21C calculates the reliability of the first output value of the AI ​​processing unit 32.

[0053] The determination unit 22C determines whether the AI ​​processing unit 32 is reliable based on the reliability of the first output value of the AI ​​processing unit 32 calculated by the calculation unit 21C. The determination unit 22C sets a reference value, such as a threshold value or a predetermined amount of change over time, and determines whether the AI ​​processing unit 32 is reliable based on the result of comparing the reliability with the reference value.

[0054] The determination unit 22C determines whether the reliability of the first output value is equal to or greater than a predetermined threshold, and if the reliability of the first output value is equal to or greater than the predetermined threshold, determines that the AI ​​processing unit 32 is reliable. If the reliability of the first output value is less than the predetermined threshold, the determination unit 22C determines that the AI ​​processing unit 32 is unreliable.

[0055] Furthermore, the determination unit 22C determines whether the time-series change in the reliability of the first output value is equal to or greater than a predetermined change, and if the time-series change in the reliability of the first output value is less than the predetermined change, determines that the AI ​​processing unit 32 is reliable. If the time-series change in the reliability of the first output value is equal to or greater than the predetermined change, the determination unit 22C determines that the AI ​​processing unit 32 is unreliable.

[0056] The correction unit 34 corrects the selection ratio of the first output value and the second output value based on the reliability determination result of the AI ​​processing unit 32. The correction unit 34 has a selection unit 23C and an output control unit 24C.

[0057] When the AI ​​processing unit 32 is reliable, the selection unit 23C selects only the first output value from the AI ​​processing unit 32 from the first output value and the second output value. When the AI ​​processing unit 32 is not reliable, the selection unit 23C selects only the second output value from the non-AI processing unit 33 from the first output value and the second output value.

[0058] The output control unit 24C outputs the first control signal, which is the first output value of the AI ​​processing unit 32, or the second control signal, which is the second output value of the non-AI processing unit 33, selected by the selection unit 23C, as a control signal to the controlled device 3. As a result, the controlled device 3 can achieve stable operation.

[0059] The control device 2C of the fourth embodiment selects the first output value of the AI ​​processing unit 32 or the second output value of the non-AI processing unit 33 based on the reliability of the first output value of the AI ​​processing unit 32. The control device 2C outputs a control signal corresponding to the first control signal, which is the selected first output value, or the second control signal, which is the selected second output value, to the controlled device 3. As a result, the control device 2C can achieve stable operation of the controlled device 3 while introducing an AI processing unit 32 with a machine learning function.

[0060] When the reliability of the first output value is less than a predetermined threshold and it is determined that the AI ​​processing unit 32 is unreliable, the control device 2C does not select the first output value of the AI ​​processing unit 32 and instead selects the second output value of the non-AI processing unit 33. When the reliability of the first output value is equal to or greater than a predetermined threshold and it is determined that the AI ​​processing unit 32 is reliable, the control device 2C selects the first output value of the AI ​​processing unit 32. The control device 2C outputs a control signal corresponding to the selected first control signal, which is the first output value, or the selected second control signal, which is the second output value, to the controlled device 3. As a result, the control device 2C can achieve stable operation of the controlled device 3 while introducing an AI processing unit 32 with machine learning capabilities.

[0061] When the time-series change in the reliability of the first output value is equal to or greater than a predetermined change amount and it is determined that the AI ​​processing unit 32 is unreliable, the control device 2C does not select the first output value of the AI ​​processing unit 32 and instead selects the second output value of the non-AI processing unit 33. When the time-series change in the reliability of the first output value is less than a predetermined threshold and it is determined that the AI ​​processing unit 32 is reliable, the control device 2C selects the first output value of the AI ​​processing unit 32. The control device 2C outputs a control signal corresponding to the selected first control signal, which is the first output value, or the selected second control signal, which is the second output value, to the controlled device 3. As a result, the control device 2C can achieve stable operation of the controlled device 3 while introducing an AI processing unit 32 with machine learning capabilities.

[0062] The control device 2C in the above fourth embodiment illustrates a case in which either the first output value of the AI ​​processing unit 32 or the second output value of the non-AI processing unit 33 is selected and output based on the reliability of the first output value of the AI ​​processing unit 32. However, the control device 2C calculates a selection ratio by combining the first output value of the AI ​​processing unit 32 and the second output value of the non-AI processing unit 33 based on the reliability of the first output value of the AI ​​processing unit 32, and generates a control signal corresponding to the first output value of the AI ​​processing unit 32 and the second output value of the non-AI processing unit 33 according to the calculated selection ratio. The control device 2C may output the generated control signal to the controlled device 3, which can be changed as appropriate.

[0063] <Example 5> 5 is an explanatory diagram showing an example of a control system 1D according to a fifth embodiment. The control system 1D shown in FIG.

[0064] The control target device 3 is, for example, a device such as a motor. The control device 2D is a feedback control device that performs feedback control on the control target device 3. The control device 2D has a sensor 17 and a control unit 40. The sensor 17 is a sensor that detects an output signal from the control target device 3, for example, the rotation speed of the motor. The control unit 40 drives and controls the control target device 3, for example, by outputting a control signal to the control target device 3 to change the rotation speed of the motor.

[0065] The control unit 40 includes an input unit 41 , an AI processing unit 42 , a non-AI processing unit 43 , a reliability determination unit 44 , and a correction unit 45 .

[0066] The input unit 41 receives a target value and a feedback value, which is a sensor value detected by the sensor 17, and outputs a correction value, which is the difference between the target value and the feedback value, to the AI ​​processing unit 42 and the non-AI processing unit 43.

[0067] The AI ​​processing unit 42 is a first control unit within the first processing unit that uses AI to output a first control signal, which is a first output value obtained by AI processing, based on the correction value from the input unit 41 and the set AI algorithm.

[0068] The non-AI processing unit 43 is a second control unit within the second processing unit that outputs a second control signal, which is a second output value obtained by non-AI processing, based on the correction value from the input unit 41 and a set algorithm, without using AI.

[0069] The reliability determination unit 44 determines whether the AI ​​processing unit 42 is reliable based on the reliability of the first output value of the AI ​​processing unit 42. The reliability determination unit 44 includes a calculation unit 21D and a determination unit 22D. The calculation unit 21D calculates a first control signal, which is the first output value of the AI ​​processing unit 42.

[0070] The determination unit 22D determines whether the AI ​​processing unit 42 is reliable based on the reliability of the first output value of the AI ​​processing unit 42 calculated by the calculation unit 21D. The determination unit 22D sets a reference value, such as a threshold value or a predetermined amount of change over time, and determines whether the AI ​​processing unit 42 is reliable based on the result of comparing the reliability with the reference value.

[0071] The determination unit 22D determines whether the reliability of the first output value is equal to or greater than a predetermined threshold, and if the reliability of the first output value is equal to or greater than the predetermined threshold, determines that the AI ​​processing unit 42 is reliable. Furthermore, if the reliability of the first output value is less than the predetermined threshold, the determination unit 22D determines that the AI ​​processing unit 42 is unreliable.

[0072] Furthermore, the determination unit 22D determines whether the time-series change in the reliability of the first output value is equal to or greater than a predetermined change, and if the time-series change in the reliability of the first output value is less than the predetermined change, determines that the AI ​​processing unit 42 is reliable. Furthermore, if the time-series change in the reliability of the first output value is equal to or greater than the predetermined change, the determination unit 22D determines that the AI ​​processing unit 42 is unreliable.

[0073] The correction unit 45 corrects the selection ratio of the first output value and the second output value based on the reliability determination result of the AI ​​processing unit 42. The correction unit 45 has a selection unit 23D and an output control unit 24D.

[0074] When the AI ​​processing unit 42 is reliable, the selection unit 23D selects only the first control signal, which is the first output value from the AI ​​processing unit 42, from the first output value and the second output value. When the AI ​​processing unit 42 is not reliable, the selection unit 23D selects only the second control signal, which is the second output value from the non-AI-based control unit 43, from the first output value and the second output value.

[0075] The output control unit 24D outputs the first control signal, which is the first output value of the AI-based control unit 42, or the second control signal, which is the second output value of the non-AI-based control unit 43, selected by the selection unit 23D, as a control signal to the controlled-target device 3. As a result, the controlled-target device 3 can achieve stable operation.

[0076] The control device 2D of the fifth embodiment selects the first output value of the AI ​​processing unit 42 or the second output value of the non-AI processing unit 43 based on the reliability of the first output value of the AI ​​processing unit 42. The control device 2D outputs a control signal corresponding to the first control signal, which is the selected first output value, or the second control signal, which is the selected second output value, to the controlled device 3. As a result, the control device 2D can achieve stable operation of the controlled device 3 while introducing an AI processing unit 42 with a machine learning function.

[0077] If the reliability of the first output value is less than a predetermined threshold and it is determined that the AI ​​processing unit 42 is unreliable, the control device 2D does not select the first output value of the AI ​​processing unit 42 and instead selects the second output value of the non-AI processing unit 43. If the reliability of the first output value is equal to or greater than a predetermined threshold and it is determined that the AI ​​processing unit 42 is reliable, the control device 2D selects the first output value of the AI ​​processing unit 42. The control device 2D outputs a control signal corresponding to the selected first control signal, which is the first output value, or the selected second control signal, which is the second output value, to the controlled device 3. As a result, the control device 2D can achieve stable operation of the controlled device 3 while introducing an AI processing unit 42 with machine learning capabilities.

[0078] When the time-series change in the reliability of the first output value is equal to or greater than a predetermined change amount and it is determined that the AI ​​processing unit 42 is unreliable, the control device 2D does not select the first output value of the AI ​​processing unit 42 and instead selects the second output value of the non-AI processing unit 43. When the time-series change in the reliability of the first output value is less than a predetermined threshold and it is determined that the AI ​​processing unit 42 is reliable, the control device 2D selects the first output value of the AI ​​processing unit 42. The control device 2D outputs a control signal corresponding to the selected first control signal, which is the first output value, or the selected second control signal, which is the second output value, to the controlled device 3. As a result, the control device 2D can achieve stable operation of the controlled device 3 while introducing an AI processing unit 42 with machine learning capabilities.

[0079] The control device 2D of the fifth embodiment described above illustrates a case in which the first output value of the AI ​​processing unit 42 or the second output value of the non-AI processing unit 43 is selected and output based on the reliability of the first output value of the AI ​​processing unit 42. However, the control device 2D calculates a selection ratio between the first output value of the AI ​​processing unit 42 and the second output value of the non-AI processing unit 43 based on the reliability of the first output value of the AI ​​processing unit 42, and generates a control signal that reflects the first output value of the AI ​​processing unit 42 and the second output value of the non-AI processing unit 43 in accordance with the calculated selection ratio. The control device 2D may output the generated control signal to the controlled device 3, and this can be changed as appropriate.

[0080] For convenience of explanation, the control devices 2, 2A, and 2B of Examples 1 to 3 illustrate the case where the first output value and the second output value are the same measurement target, such as temperature, but the present invention is also applicable to cases where the first output value is the oxygen concentration of a boiler and the second output value is the temperature of the boiler. The first output value and the second output value may be output values ​​related to feedback control of the control target device 3, and can be changed as appropriate.

[0081] Furthermore, all or any part of the processing functions performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware using wired logic.

[0082] (Hardware) Next, an example of a hardware configuration for realizing the control devices 2, 2A, 2B, 2C, and 2D will be described. Fig. 6 is an explanatory diagram showing an example of the hardware configuration of a computer. The control devices 2, 2A, 2B, 2C, and 2D can be realized, for example, by a computer 90 shown in Fig. 6. As shown in Fig. 6, the computer 90 has a processor 91, a memory 92, a hard disk 93, and a communication device 94. The processor 91 is connected to the memory 92, the hard disk 93, and the communication device 94 via a bus.

[0083] The communication device 94 is a network interface card or the like, and is used for communication with the controlled device 3. The communication device 94 is used for communication with the AI ​​sensor 13 and the non-AI sensor 14. The hard disk 93 is an auxiliary storage device. The hard disk 93 stores various programs, including programs for realizing the functions of the control devices 2, 2A, 2B, 2C, and 2D.

[0084] The processor 91 reads out various programs stored in the hard disk 93, expands them into the memory 92, and executes them. In this way, the processor 91 realizes the functions of the control devices 2, 2A, 2B, 2C, and 2D. The functions of the control device 2 are the functions of the input unit 11, the control unit 12, the reliability determination unit 15, and the correction unit 16 shown in FIG. 1A. The functions of the control device 2A are the functions of the input unit 11, the control unit 12, the reliability determination unit 15, and the correction unit 16A shown in FIG. 2.

[0085] The functions of the control device 2B are the functions of the input unit 11, the control unit 12, the reliability determination unit 15, and the correction unit 16B shown in Fig. 3. The functions of the control device 2C are the functions of the reliability determination unit 31, the AI ​​processing unit 32, the non-AI processing unit 33, and the correction unit 34 shown in Fig. 4. The functions of the control device 2D are the functions of the input unit 41, the AI ​​processing unit 42, the non-AI processing unit 43, the reliability determination unit 44, and the correction unit 45 shown in Fig. 5. [Explanation of symbols]

[0086] 1, 1A, 1B, 1C, 1D control system 2, 2A, 2B, 2C, 2D control device 3 Controlled devices 13 AI Sensor 14 Non-AI sensors 14A Storage section 15, 31, 44 Reliability determination section 16, 16A, 16B, 34, 45 Correction section 23, 23A, 23B, 23C, 23D selection section 24, 24A, 24B, 24C, 24D Output control section

Claims

1. A control device that performs feedback control of a control target device, a first processing unit that acquires a first output value related to feedback control of the control-target device using AI (Artificial Intelligence) processing; a second processing unit that acquires a second output value related to the feedback control using non-AI processing; a correction unit that corrects a selection ratio of the first output value and the second output value based on the reliability of the first output value; a control unit that outputs a control signal for feedback-controlling the control target device based on the selection ratio; A control device comprising:

2. a calculation unit that quantifies the reliability of the first output value and calculates the reliability of the first output value; a determination unit that determines whether the first processing unit is reliable based on the reliability of the first output value, The correction unit 2. The control device according to claim 1, wherein the selection ratio is corrected based on a result of a determination as to whether the first processing unit is reliable.

3. The determination unit determining whether the reliability of the first output value is equal to or greater than a predetermined threshold, determining that the first processing unit is reliable if the reliability of the first output value is equal to or greater than the predetermined threshold, and determining that the first processing unit is unreliable if the reliability of the first output value is less than the predetermined threshold; The correction unit a selection unit that selects only the first output value as the selection ratio when it is determined that the first processing unit is reliable, and that selects only the second output value as the selection ratio when it is determined that the first processing unit is unreliable.

3. The control device according to claim 2, further comprising:

4. The determination unit determining whether a time-series change in the reliability of the first output value is equal to or greater than a predetermined change, determining that the first processing unit is reliable if the time-series change in the reliability of the first output value is less than the predetermined change, and determining that the first processing unit is unreliable if the time-series change in the reliability of the first output value is equal to or greater than the predetermined change; The correction unit a selection unit that selects only the first output value as the selection ratio when it is determined that the first processing unit is reliable, and that selects only the second output value as the selection ratio when it is determined that the first processing unit is unreliable.

3. The control device according to claim 2, further comprising:

5. The first processing unit includes: a first sensor that acquires an output of the control target device as the first output value using the AI ​​processing; The second processing unit includes: The control device according to any one of claims 1 to 4, characterized in that it has a second sensor that acquires the output of the controlled device as the second output value using the non-AI processing.

6. The first processing unit includes: a first sensor that acquires an output of the control target device as the first output value using the AI ​​processing; The second processing unit includes:

5. The control device according to claim 1, further comprising a storage unit that stores a preset output of the controlled device as the second output value.

7. The first processing unit includes: a first control unit that acquires, as the first output value, a first control signal that performs feedback control on the control-target device using the AI ​​processing; The second processing unit includes: The control device according to any one of claims 1 to 4, characterized in that it has a second control unit that acquires a second control signal that feedback controls the control target device using the non-AI processing as the second output value.

8. The control unit The first processing unit and the second processing unit are included, The first processing unit includes: a first control unit that acquires, as the first output value, a first control signal that performs feedback control on the control-target device using the AI ​​processing; The second processing unit includes: a second control unit that acquires, as the second output value, a second control signal that performs feedback control of the control-target device using the non-AI processing; The control unit The control device according to any one of claims 1 to 4, characterized in that the first control signal and the second control signal are output as the control signal based on the selection ratio of the first output value and the second output value.

9. A control system having a control target device and a control device that feedback controls the control target device, The control device a first processing unit that acquires a first output value related to feedback control of the control-target device using AI (Artificial Intelligence) processing; a second processing unit that acquires a second output value related to the feedback control using non-AI processing; a correction unit that corrects a selection ratio of the first output value and the second output value based on the reliability of the first output value; a control unit that outputs a control signal for feedback-controlling the control target device based on the selection ratio; A control system comprising:

10. A control device that performs feedback control of a controlled device, acquiring a first output value related to feedback control of the control-target device using AI (Artificial Intelligence) processing; obtaining a second output value related to the feedback control using a non-AI process; correcting a selection ratio of the first output value and the second output value based on the reliability of the first output value; Based on the selection ratio, a control signal for feedback control of the controlled device is output. A control method comprising: executing a process.

Images