Abnormality sign detector, abnormality sign detection system, abnormality sign detection method, and abnormality sign detection program

Abstract

To improve the accuracy of abnormality sign detection at a plant.SOLUTION: An abnormality sign detector (110) is connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign at a plant. Control parts (501, 502) included in the abnormality sign detector (110) input internal data of the local devices (161, 171) and process data of the controller (150) to trained models (820, 1620), thereby determining the presence or absence of any abnormality sign of the local devices (161, 171) based on output data outputted from the trained models (820, 1620) and data to be monitored at the plant and then output results of the determination of the presence or absence of the abnormality sign.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to an abnormality sign detection device, an abnormality sign detection system, an abnormality sign detection method, and an abnormality sign detection program. [Background technology]

[0002] 2. Description of the Related Art There is known a technique for detecting an abnormality in a plant based on process data that is measured by various instrumentation devices installed in the plant and is used for controlling and monitoring the plant. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2022 / 070656 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the above technology does not provide sufficient detection accuracy for abnormal signs of individual local devices in a plant (for example, actuators such as control valves, and instrumentation devices such as flow meters) or abnormal signs of the plant itself. Note that the "abnormal sign" here refers to a state before an abnormality occurs, which is a slight change from a normal state.

[0005] The present disclosure aims to improve the accuracy of detecting signs of abnormality in a plant. [Means for solving the problem]

[0006] A first aspect of the present disclosure is a method for manufacturing a semiconductor device comprising: An abnormality sign detection device (110) that is communicatively connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), it is determined whether or not there is a sign of abnormality in the local devices (161, 171) based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; The determination result of the presence or absence of the abnormality sign is output.

[0007] According to the first aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in a local device in a plant.

[0008] In addition, the second aspect of the present disclosure is An abnormality sign detection device (110) that is communicatively connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of an abnormality in the plant (1800, 1900) is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; The determination result of the presence or absence of the abnormality sign is output.

[0009] According to the second aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in the plant itself.

[0010] A third aspect of the present disclosure is the anomaly sign detection device according to the first or second aspect, The data to be monitored is output data output from the local devices (161, 171).

[0011] A fourth aspect of the present disclosure is the anomaly sign detection device according to the third aspect, The trained models (820, 1620) are The local equipment (161, 171) is generated by performing a learning process using learning data (600, 1400) including internal data of the local equipment (161, 171), process data of the controller (150), and output data output by the local equipment (161, 171) when the local equipment and the plant are in a normal state.

[0012] A fifth aspect of the present disclosure is the anomaly sign detection device according to the third aspect, The local devices (161, 171) include actuators (161) and instrumentation devices (171) in the plant.

[0013] A sixth aspect of the present disclosure is the anomaly sign detection device according to the fifth aspect, When the local device (161, 171) is an actuator (161), The process data includes: Data indicating a physical quantity of an object controlled by the actuator (161) and measured by the instrumentation device (171); control data for the controller (150) to control the operation of the actuator (161); Including, The monitored data is data indicating the operation result of the actuator (161).

[0014] A seventh aspect of the present disclosure is the anomaly sign detection device according to the fifth aspect, When the local device (161, 171) is an instrumentation device (171), the process data includes control data for the controller (150) to control the operation of the actuator (161); The data of the monitored object is data indicating a physical quantity of the object controlled by the actuator (161) and measured by the instrumentation device (171).

[0015] An eighth aspect of the present disclosure is the anomaly sign detection device according to the third aspect, The controller (150) Superimposed data in which the internal data of the local device (161, 171) is superimposed on the output data output from the local device (161, 171) is obtained.

[0016] A ninth aspect of the present disclosure is the anomaly sign detection device according to the eighth aspect, The plant control system (140) includes a HART converter (162, 172); The superimposed data is HART data, and the controller (150) acquires the HART data from the HART converters (162, 172) connected to the local devices (161, 171).

[0017] A tenth aspect of the present disclosure is the anomaly sign detection device according to the third aspect, The control unit (501, 502) of the abnormality sign detection device (110) The process data is obtained from the controller (150) via a data server (120).

[0018] An eleventh aspect of the present disclosure is the anomaly sign detection device according to the sixth aspect, When the actuator (161) is an air-operated control valve (200), The process data includes: Data indicating the flow rate of an object controlled by the pneumatic control valve (200) and measured by a flow meter (1210); control data for the controller (150) to control the operation of the pneumatic control valve (200); Including, the monitored data is data indicating a valve opening degree of the pneumatic control valve (200); The internal data of the pneumatic control valve (200) includes a signal for controlling the operation of the pneumatic control valve (200) received from the controller (150), a signal indicating the output air pressure, and a signal indicating the supply air pressure.

[0019] A twelfth aspect of the present disclosure is the anomaly sign detection device according to the seventh aspect, When the instrumentation device (171) is a flow meter (1210), The process data includes data indicating a valve opening degree of a regulator valve (200), the data of the monitored object is data indicating a flow rate of the object controlled by the control valve (200) and measured by the flow meter (1210); The internal data of the flow meter (1210) includes the density of the object, the drive gain of the flow meter (1210), and the tube frequency.

[0020] A thirteenth aspect of the present disclosure is an abnormality sign detection system, The abnormality sign detection device (110) according to the ninth aspect; The plant control system (140) is communicatively connected to the abnormality sign detection device (110).

[0021] According to the thirteenth aspect of the present disclosure, it is possible to improve the accuracy of detecting signs of abnormality in a plant.

[0022] A fourteenth aspect of the present disclosure is an abnormality sign detection system, The abnormality sign detection device (110) according to the tenth aspect, The abnormality sign detection device (110) includes the data server (120) for acquiring the process data from the plant control system (140).

[0023] According to the fourteenth aspect of the present disclosure, it is possible to improve the accuracy of detecting signs of abnormality in a plant.

[0024] A fifteenth aspect of the present disclosure is an abnormality sign detection method, comprising: a control unit (501, 502) included in an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), it is determined whether or not there is a sign of abnormality in the local devices (161, 171) based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; A process is executed to output the result of the determination of the presence or absence of the abnormality sign.

[0025] According to the fifteenth aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in a local device in a plant.

[0026] A sixteenth aspect of the present disclosure is an abnormality sign detection method, comprising: a control unit (501, 502) included in an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of an abnormality in the plant (1800, 1900) is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; A process is executed to output the result of the determination of the presence or absence of the abnormality sign.

[0027] According to the sixteenth aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in the plant itself.

[0028] A seventeenth aspect of the present disclosure is an abnormality sign detection program, comprising: A control unit (501, 502) of an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), it is determined whether or not there is a sign of abnormality in the local devices (161, 171) based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; A process is executed to output the result of the determination of the presence or absence of the abnormality sign.

[0029] According to the seventeenth aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in a local device in a plant.

[0030] An eighteenth aspect of the present disclosure is an abnormality sign detection program, comprising: A control unit (501, 502) of an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of an abnormality in the plant (1800, 1900) is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; A process is executed to output the result of the determination of the presence or absence of the abnormality sign.

[0031] According to the eighteenth aspect of the present disclosure, it is possible to improve the detection accuracy when detecting a sign of abnormality in the plant itself. [Brief description of the drawings]

[0032] [Figure 1] FIG. 1 is a diagram illustrating an example of a system configuration of an abnormality sign detection system. [Diagram 2] FIG. 13 is a diagram showing an example of actuator internal data of the pneumatic control valve. [Diagram 3] FIG. 2 is a diagram showing an example of an abnormality location of an air-operated control valve. [Figure 4] FIG. 11 is a diagram illustrating an example of HART data. [Diagram 5] 2 is a diagram illustrating an example of a hardware configuration of an abnormality sign detection device. [Figure 6] FIG. 1 is a first diagram showing an example of learning data. [Figure 7] 1 is a first diagram showing an example of a functional configuration of an anomaly sign detection device in a learning phase. FIG. [Figure 8] 1 is a first diagram showing an example of a functional configuration of an anomaly sign detection device in an inference phase. FIG. [Figure 9] 13 is an example of a first flowchart showing the flow of a learning process and an abnormality sign detection process. [Figure 10] 11 is a diagram showing an example of a processing result of an abnormality sign detection process executed by an abnormality sign detection device of a comparative example. FIG. [Figure 11] 4A to 4C are diagrams illustrating an example of a processing result of an abnormality sign detection process executed by the abnormality sign detection device according to the first embodiment. [Figure 12] FIG. 11 is a diagram showing an example of internal data of an instrumentation device of a Coriolis flowmeter. [Figure 13] FIG. 2 is a diagram showing an example of an abnormality location of a Coriolis flowmeter. [Figure 14] FIG. 2 is a second diagram showing an example of learning data. [Figure 15] FIG. 2 is a second diagram showing an example of the functional configuration of the anomaly sign detection device in the learning phase. [Figure 16]FIG. 2 is a second diagram showing an example of a functional configuration of the anomaly sign detection device in the inference phase. [Figure 17] 13 is an example of a second flowchart showing the flow of a learning process and an abnormality sign detection process. [Figure 18] FIG. 1 is a first diagram showing an example of an abnormality location in a chemical plant. [Figure 19] FIG. 2 is a second diagram showing an example of an abnormality location in a chemical plant. [Figure 20] FIG. 3 is a third diagram showing an example of learning data. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0033] Hereinafter, each embodiment will be described with reference to the accompanying drawings. In this specification and the drawings, components having substantially the same functional configurations are denoted by the same reference numerals, and redundant description will be omitted.

[0034] [First embodiment] <System configuration of the anomaly detection system> First, a system configuration of an abnormality sign detection system according to a first embodiment will be described. Fig. 1 is a diagram showing an example of the system configuration of an abnormality sign detection system. As shown in Fig. 1, an abnormality sign detection system 100 includes an abnormality sign detection device 110, a data server 120, an OPC (OLE for Process Control) server 130, and a plant control system 140.

[0035] In the abnormality sign detection system 100, the data server 120 and the OPC server 130 are connected to each other via a network 181 so that they can communicate with each other. In addition, the OPC server 130 and the plant control system 140 are connected to each other via a network 182 so that they can communicate with each other.

[0036] The abnormality sign detection device 110 acquires control data, operation result data, actuator internal data, and measurement data (described in detail later) from the plant control system 140 via the data server 120 and the OPC server 130. Furthermore, in the learning phase, the abnormality sign detection device 110 generates learning data based on the acquired data, and performs a learning process on a model for detecting abnormality signs of local devices in the plant control system 140 using the generated learning data.

[0037] In addition, in the inference phase, the anomaly sign detection device 110 detects anomaly signs in the local device by inputting the acquired data into a trained model generated by performing a learning process.

[0038] The data server 120 collects various data (in the example of FIG. 1, control data, operation result data, actuator internal data, and measurement data) from various devices (in the example of FIG. 1, the OPC server 130) connected to the network 181. The various data collected by the data server 120 is analyzed by various devices for data analysis.

[0039] The OPC server 130 transmits and receives various data based on a data communication standard for process control. In the first embodiment, the OPC server 130 transmits data (control data, operation result data, actuator internal data, and measurement data) received from the plant control system 140 to the data server 120 based on the above data communication standard.

[0040] The plant control system 140 includes a DCS (Distributed Control System) controller 150 , an actuator 161 , a HART converter 162 , a power source 163 , instrumentation equipment 171 , and a HART converter 172 .

[0041] The DCS controller 150 has a control function for controlling the plant. Specifically, the DCS controller 150 acquires measurement data measured by instrumentation devices 171 installed in the plant, and calculates control data for controlling the actuator 161 so that the acquired measurement data approaches a target value. The DCS controller 150 also transmits the control data to the actuator 161 so that the actuator 161 operates based on the calculated control data.

[0042] Furthermore, the DCS controller 150 has a monitoring function for monitoring the plant. However, in the first embodiment, among the monitoring functions of the DCS controller 150, a function for detecting an abnormality sign of a local device (the actuator 161 in the first embodiment) is realized in the abnormality sign detection device 110. For this reason, the DCS controller 150 acquires data necessary for the abnormality sign detection device 110 to realize the function of detecting an abnormality sign, and transmits the data to the abnormality sign detection device 110 via the OPC server 130 and the data server 120.

[0043] Specifically, the DCS controller 150 acquires the calculated control data. The DCS controller 150 also receives HART data from the HART converter 162, and acquires the operation result data and actuator internal data contained in the HART data. The DCS controller 150 also acquires measurement data from the instrumentation device 171, and receives HART data from the HART converter 172, and acquires the instrumentation device internal data contained in the HART data. Furthermore, the DCS controller 150 transmits the acquired data (control data, operation result data, internal data (actuator internal data of the actuator 161 in the first embodiment), and measurement data) to the abnormality sign detection device 110.

[0044] Actuator 161 is any actuator (e.g., an air-type control valve) installed in the plant, and is an actuator that operates with power (e.g., air pressure) supplied by power source 163 (e.g., a compressor). Actuator 161 operates according to control data received from DCS controller 150. For example, if actuator 161 is an air-type control valve, it receives a valve opening as control data from DCS controller 150. Therefore, the air-type control valve operates by an amount corresponding to the difference between the received valve opening and the current valve opening, so as to achieve the received valve opening.

[0045] The actuator 161 outputs operation result data indicating the operation result (e.g., valve opening) and actuator internal data during operation (e.g., control signal, signal indicating supply air pressure, signal indicating output air pressure) to the HART converter 162.

[0046] The HART converter 162 generates HART data (an example of superimposed data) by superimposing the actuator internal data on the operation result data output from the actuator 161. In addition, the HART converter 162 transmits the generated HART data to the DCS controller 150.

[0047] The instrumentation device 171 is any instrumentation device (e.g., a Coriolis flowmeter) installed in a plant. The instrumentation device 171 measures a physical quantity of an object and transmits the measurement data (e.g., mass flow rate) to the DCS controller 150. The instrumentation device 171 also outputs internal data of the instrumentation device at the time of measurement (e.g., density, drive gain, tube frequency) to the HART converter 172.

[0048] The HART converter 172 generates HART data based on the internal instrument data output from the instrumentation device 171. In addition, the HART converter 172 transmits the generated HART data to the DCS controller 150.

[0049] <Actuator internal data> Next, a description will be given of the actuator internal data of the actuator 161. Here, a description will be given of the actuator internal data of an air-type control valve, which is an example of the actuator 161. Fig. 2 is a diagram showing an example of the actuator internal data of an air-type control valve.

[0050] 2, the pneumatic control valve 200 has a valve body 210 and a positioner 220. The positioner 220 has a control calculator 221, an electro-pneumatic converter 222, a flow amplifier 223, and an opening detector 224.

[0051] The control calculator 221 receives control data (valve opening (control value)) from the DCS controller 150 and also receives the valve opening (actual value) from the opening detector 224, and outputs a control signal current so that the control data (valve opening (control value)) and the valve opening (actual value) match.

[0052] The electro-pneumatic converter 222 uses a portion of the compressed air supplied from a compressor, which is an example of the power source 163 , to output an air pressure signal of an air pressure corresponding to the control signal current output from the control calculator 221 .

[0053] The flow amplifier 223 uses compressed air supplied from a compressor, which is an example of the power source 163, to output compressed air of an air pressure (output air pressure) corresponding to the air pressure signal output from the electro-pneumatic converter 222 to the valve body 210. This causes the valve body 210 to operate, and the valve opening degree is controlled.

[0054] The opening detector 224 measures the valve opening after the valve body 210 operates, and outputs the measured valve opening (actual measurement value) to the control calculator 221.

[0055] Positioner 220 is Data indicating the operation result of the pneumatic control valve 200: the valve opening measured by the opening detector 224; Actuator internal data: current value (control signal) of the control signal current output by the control calculator 221, · Actuator internal data: A signal indicating the supply air pressure, which is the air pressure of the compressed air supplied from a compressor, where the flow amplifier 223 is an example of the power source 163. · Actuator internal data: A signal indicating the output air pressure, which is the air pressure of the compressed air output by the flow amplifier 223 to the valve body 210. Outputs to the HART converter 162.

[0056] Note that the signal indicating the output air pressure includes one type of signal when the pneumatic control valve 200 is of the single-acting type, and two types of signals when the pneumatic control valve 200 is of the double-acting type.

[0057] <Explanation of abnormal parts of the actuator> Next, the locations where abnormalities occur in the actuator 161 will be explained. Here too, the abnormal parts of the pneumatic control valve 200, which is an example of the actuator 161, will be explained. Fig. 3 is a diagram showing an example of the abnormal parts of the pneumatic control valve.

[0058] In Fig. 3, reference numerals 301 to 303 indicate the main abnormal parts occurring in the pneumatic control valve 200. Specifically, reference numeral 301 indicates the location where snagging occurs in the gland packing part, reference numeral 302 indicates the location where corrosion occurs at the tip of the plug, and reference numeral 303 indicates the location where corrosion occurs in the seat ring. Note that although not shown in Fig. 3, the abnormal parts occurring in the pneumatic control valve 200 include, for example, malfunctions of the positioner, corrosion of the stem, etc.

[0059] Previously, the abnormalities occurring at these locations were not known until the components were disassembled during regular maintenance. However, according to the abnormality prediction detection device 110, the abnormality prediction can be accurately detected at an early stage. Therefore, according to the abnormality prediction detection device 110, it becomes possible to perform maintenance in a planned manner.

[0060] <Explanation of HART data> Next, the HART data generated by the HART converter 162 will be explained. Fig. 4 is a diagram showing an example of the HART data.

[0061] 4, reference numeral 410 indicates the change over time of the operation result data output from positioner 220, with the horizontal axis indicating time and the vertical axis indicating the operation result data (current value). For ease of explanation, the example of reference numeral 410 indicates a case in which the operation result data (current value) changes in proportion to the passage of time, from 4 [mA] to 20 [mA], as the change over time of the operation result data.

[0062] Furthermore, in Fig. 4, reference numeral 411 indicates that the actuator internal data (e.g., supply air pressure) value at each time output from the positioner 220 is converted into 6-bit binary data ("001100" in the example of Fig. 4) and then converted into a sine wave. Reference numeral 411 indicates that binary data=0 is converted into a sine wave with a frequency of 1200 Hz, and binary data=1 is converted into a sine wave with a frequency of 2400 Hz.

[0063] The HART converter 162 generates HART data by superimposing the actuator internal data (411) converted into a sine wave on the operation result data (410). In Fig. 4, reference numeral 420 indicates an example of the generated HART data.

[0064] The HART data generated by the HART converter 162 is transmitted to the DCS controller 150, and is separated and processed in the DCS controller to restore the operation result data and the actuator internal data.

[0065] In Fig. 4, reference numeral 430 indicates a change over time in the operation result data restored by the separation process performed by the DCS controller 150. Also, in Fig. 4, reference numeral 431 indicates a sine wave restored by the separation process performed by the DCS controller 150. The DCS controller 150 recognizes binary data based on the restored sine wave, thereby making it possible to acquire the value of the actuator's internal data (e.g., the supply air pressure).

[0066] <Hardware configuration of the anomaly detection device> Next, a hardware configuration of the abnormality sign detection device 110 will be described. Fig. 5 is a diagram showing an example of the hardware configuration of the abnormality sign detection device. As shown in Fig. 5, the abnormality sign detection device 110 has a processor 501, a memory 502, an auxiliary storage device 503, an I / F (Interface) device 504, a communication device 505, and a drive device 506. Note that each piece of hardware of the abnormality sign detection device 110 is connected to each other via a bus 507.

[0067] The processor 501 includes various arithmetic devices such as a CPU (Central Processing Unit), etc. The processor 501 reads various programs (for example, an abnormality sign detection program, etc.) onto the memory 502 and executes them.

[0068] The memory 502 has a main storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), etc. The processor 501 and the memory 502 form a so-called computer (also called a "control unit"), and the processor 501 executes various programs read onto the memory 502, causing the computer to realize various functions.

[0069] The auxiliary storage device 503 stores various programs and various information used when the various programs are executed by the processor 501 .

[0070] The I / F device 504 connects the operation device 511 and the display device 512 to the abnormality sign detection device 110 .

[0071] The communication device 505 is a device for connecting to a network (not shown) and communicating with the data server 120 and the like.

[0072] The drive device 506 is a device for setting the recording medium 513. The recording medium 513 here includes media that record information optically, electrically, or magnetically, such as a CD-ROM, a flexible disk, a magneto-optical disk, etc. The recording medium 513 may also include semiconductor memories that record information electrically, such as ROM, flash memory, etc.

[0073] The various programs to be installed in the auxiliary storage device 503 are installed, for example, by setting the distributed recording medium 513 in the drive device 506 and reading out the various programs recorded in the recording medium 513 by the drive device 506. Alternatively, the various programs to be installed in the auxiliary storage device 503 may be installed by being downloaded from a network (not shown) via the communication device 505.

[0074] <Explanation of learning data> Next, a description will be given of learning data used when the abnormality sign detection device 110 performs a learning process on a model for detecting an abnormality sign of a local device in the plant control system 140. Fig. 6 is a diagram showing an example of the learning data.

[0075] As shown in FIG. 6, learning data 600 has information items such as "input data" and "correct answer data."

[0076] "Input data" refers to data (explanatory variables) input to the model, and types of data include "process data" and "internal data." Process data is data used by the DCS controller 150 to control and monitor the plant. Internal data is data that is not used by the DCS controller 150 to control and monitor the plant, and is data held by local devices. Of the internal data, data held internally by the actuator 161 is actuator internal data, and internal data held internally by the instrumentation device 171 is instrumentation device internal data.

[0077] The “process data” includes the measurement data (mass flow rate (t)) measured by the Coriolis flowmeter and the control data (valve opening (t)) sent by the DCS controller 150 to the pneumatic control valve 200.

[0078] The “internal data” includes the actuator internal data output from the positioner 220 (the control signal (t), the supply air pressure (t), and the output air pressure (t)).

[0079] The "correct data" is the correct data (objective variables) that should be output from the model when data (explanatory variables) are input to the model. The "correct data" includes the operation result data (valve opening (t)) output from the positioner 220 when the pneumatic control valve 200 and the plant are in a normal state.

[0080] Note that the addition of "(t)" after the data name of each piece of data included in the learning data 600 indicates that the data is time-series data.

[0081] In this way, when performing learning processing on a model for detecting a sign of abnormality of the actuator 161 among the local devices in the plant control system 140, the learning data is Control data for the actuator 161, which is process data for instructing the actuator 161 to operate; Input data including at least actuator internal data, which is internal data held internally by the actuator 161; Correct answer data including operation result data representing the operation result of the actuator 161; It is composed of a combination of the following.

[0082] <Functional configuration of the anomaly detection device (learning phase)> Next, a functional configuration in the learning phase of the anomaly sign detection device 110 according to the first embodiment will be described. Fig. 7 is a first diagram showing an example of the functional configuration in the learning phase of the anomaly sign detection device.

[0083] As described above, an abnormality sign detection program is installed in the abnormality sign detection device 110. The abnormality sign detection device 110 according to the first embodiment executes the program to function as a learning unit 700 in the learning phase. The learning unit 700 includes an input data acquisition unit 710, a model 720, and a comparison and change unit 730.

[0084] The input data acquisition unit 710 reads out process data and internal data from the data server 120 as input data for the learning data 600. The process data of the learning data 600 includes measurement data (mass flow rate (t)) and control data (valve opening (t)). The internal data of the learning data 600 includes actuator internal data (control signal (t), supply air pressure (t), output air pressure (t)). Therefore, the input data acquisition unit 710 reads out these data from the data server 120.

[0085] Moreover, the input data acquisition unit 710 inputs the read process data and internal data to the model 720 .

[0086] The model 720 is an AI technology (e.g., a neural network) suitable for processing time-series data, and outputs output data when process data and internal data are input by the input data acquisition unit 710. The output data output by the model 720 is input to the comparison and change unit 730.

[0087] The comparison and change unit 730 reads out the correct answer data of the learning data 600 from the data server 120. Since the correct answer data of the learning data 600 includes the operation result data (valve opening degree (t)), the comparison and change unit 730 reads out the data.

[0088] Moreover, the comparison and change unit 730 calculates a loss by comparing the output data output from the model 720 with the read-out correct answer data, and updates the model parameters of the model 720 based on the calculated loss. In this manner, the anomaly sign detection device 110 performs a learning process on the model 720, and a learned model is generated.

[0089] <Functional configuration of the anomaly detection device (inference phase)> Next, a functional configuration in the inference phase of the anomaly sign detection device 110 according to the first embodiment will be described. Fig. 8 is a first diagram showing an example of the functional configuration in the inference phase of the anomaly sign detection device.

[0090] The anomaly sign detection device 110 executes the anomaly sign detection program, and thereby functions as an inference unit 800, a judgment unit 830, and an output unit 840 in the inference phase. The inference unit 800 includes an input data acquisition unit 810 and a trained model 820.

[0091] The input data acquisition unit 810 reads out process data and internal data from the data server 120. The process data includes measurement data (mass flow rate (t)) and control data (valve opening (t)), and the internal data includes actuator internal data (control signal (t), supply air pressure (t), output air pressure (t)). Therefore, the input data acquisition unit 810 reads out these data.

[0092] In addition, the input data acquisition unit 810 inputs the read process data and internal data to the trained model 820.

[0093] The trained model 820 is a trained model generated by performing a learning process on the model 720. The trained model 820 receives process data and internal data, and outputs prediction data (predicted valve opening (t)) and a 95% confidence interval of the prediction data. The prediction data (predicted valve opening (t)) and the 95% confidence interval of the prediction data output by the trained model 820 are input to the determination unit 830.

[0094] The determination unit 830 outputs the 95% confidence interval of the predicted data output from the trained model 820 to the output unit 840.

[0095] In addition, the judgment unit 830 compares the operation result data (valve opening (t)) read from the data server 120 with the 95% confidence interval of the predicted data output from the trained model 820, and judges whether the operation result data (valve opening (t)) exceeds the 95% confidence interval.

[0096] If the operation result data (valve opening (t)) does not exceed the 95% confidence interval (if there is an inclusive relationship), the judgment unit 830 judges that there is no sign of abnormality and outputs the judgment result to the output unit 840. On the other hand, if the operation result data (valve opening (t)) exceeds the 95% confidence interval (if there is no inclusive relationship), the judgment unit 830 judges that there is a sign of abnormality and outputs the judgment result to the output unit 840.

[0097] The output unit 840 outputs the 95% confidence interval of the predicted data output from the determination unit 830. The output unit 840 also outputs the determination result of the presence or absence of an abnormality sign output from the determination unit 830.

[0098] <Learning process and anomaly detection process flow> Next, the flow of the learning process and the abnormality sign detection process executed by the abnormality sign detection device 110 according to the first embodiment will be described. FIG. 9 is an example of a first flowchart showing the flow of the learning process and the abnormality sign detection process. Of these, FIG. 9(a) is a flowchart showing the flow of the learning process executed by the abnormality sign detection device 110 according to the first embodiment. Also, FIG. 9(b) is a flowchart showing the flow of the abnormality sign detection process executed by the abnormality sign detection device 110 according to the first embodiment.

[0099] (1) Learning process flow 9(a), when the anomaly sign detection device 110 according to the first embodiment starts a learning process in the learning phase, the anomaly sign detection device 110 acquires process data as input data in step S901. Specifically, the anomaly sign detection device 110 acquires measurement data (mass flow rate (t)) and control data (valve opening degree (t)), which are process data, from the data server 120.

[0100] In step S902, the abnormality sign detection device 110 acquires internal data as input data. Specifically, the abnormality sign detection device 110 acquires actuator internal data (control signal (t), supply air pressure (t), output air pressure (t)) which is internal data, from the data server 120.

[0101] In step S903, the abnormality sign detection device 110 acquires the operation result data (valve opening degree (t)) from the data server 120 as correct answer data.

[0102] In step S904, the anomaly sign detection device 110 generates learning data 600 in which the input data and the correct answer data are associated with each other, and stores the learning data 600 in the data server 120.

[0103] In step S905, the anomaly sign detection device 110 reads the learning data 600 from the data server 120 and performs learning processing to generate a trained model.

[0104] In step S906, the anomaly sign detection device 110 stores the trained model generated by performing the learning process.

[0105] (2) Anomaly detection process flow 9(b), when the abnormality sign detection device 110 starts the abnormality sign detection process in the inference phase, the abnormality sign detection device 110 acquires process data in step S911. Specifically, the abnormality sign detection device 110 acquires measurement data (mass flow rate (t)) and control data (valve opening degree (t)) which are process data from the data server 120.

[0106] In step S912, the abnormality sign detection device 110 acquires, from the data server 120, the actuator internal data (control signal (t), supply air pressure (t), output air pressure (t)), which is internal data.

[0107] In step S913, the anomaly sign detection device 110 inputs the acquired process data and internal data into the trained model 820, executes the trained model 820, and calculates predicted data (predicted valve opening (t)) and a 95% confidence interval for the predicted data.

[0108] In step S914, the abnormality sign detection device 110 acquires the operation result data (valve opening degree (t)) from the data server 120 as the monitored data.

[0109] In step S915, the abnormality sign detection device 110 outputs the calculated 95% confidence interval.

[0110] In step S916, the abnormality sign detection device 110 determines whether or not the acquired operation result data (valve opening degree (t)) exceeds the calculated 95% confidence interval (whether or not there is an inclusive relationship), thereby determining whether or not there is an abnormality sign.

[0111] In step S916, if the calculated 95% confidence interval is not exceeded (if there is an inclusive relationship), it is determined that there is no abnormality sign (NO in step S916), and the process proceeds to step S918.

[0112] On the other hand, if it is determined in step S916 that the calculated 95% confidence interval is exceeded (if there is no inclusion relationship), it is determined that there is a sign of an abnormality (YES in step S916), and the process proceeds to step S917.

[0113] In step S917, the abnormality sign detection device 110 outputs a determination that there is an abnormality sign (that an abnormality sign has been detected).

[0114] In step S918, the abnormality sign detection device 110 determines whether or not to continue the abnormality sign detection process. If it is determined in step S918 that the abnormality sign detection process is to be continued (YES in step S918), the process returns to step S911.

[0115] On the other hand, if it is determined in step S918 that the abnormality sign detection process should not be continued (NO in step S918), the abnormality sign detection process ends.

[0116] <Results of abnormality prediction detection process> Next, a description will be given of the processing result of the abnormality sign detection process executed by the abnormality sign detection device 110 according to the first embodiment. Note that here, first, the processing result of the abnormality sign detection process executed by the abnormality sign detection device of the comparative example will be described, and then the processing result of the abnormality sign detection process executed by the abnormality sign detection device 110 according to the first embodiment will be described.

[0117] (1) Comparative Example of Anomaly Sign Detection Device First, a process result of the abnormality sign detection process executed by the abnormality sign detection device of the comparative example will be described. A trained model is provided that has been trained using only process data used for controlling and monitoring the plant (i.e., without using internal data); Using the trained model, input only the process data (i.e., without inputting internal data) to calculate predicted data (predicted valve opening (t)) and a 95% confidence interval for the predicted data; Refers to the device.

[0118] Fig. 10 is a diagram showing an example of the results of an abnormality sign detection process executed by an abnormality sign detection device of a comparative example. In Fig. 10, reference numeral 1000 denotes a graph with time on the horizontal axis and valve opening on the vertical axis, in which the solid line represents the operation result data (valve opening (t)) and the dotted line represents the prediction data (predicted valve opening (t)).

[0119] 10, reference numeral 1010 denotes a graph showing an enlarged rectangular area in the graph shown by reference numeral 1000. As shown by reference numeral 1010, in the case of the anomaly sign detection device of the comparative example, the 95% confidence interval of the predicted data was wide.

[0120] (2) In the case of the abnormality sign detection device 110 according to the first embodiment Next, a processing result of the abnormality sign detection processing executed by the abnormality sign detection device 110 according to the first embodiment will be described.

[0121] Fig. 11 is a diagram showing an example of the results of an abnormality sign detection process performed by an abnormality sign detection device. In Fig. 11, reference numeral 1100 denotes a graph with time on the horizontal axis and valve opening on the vertical axis, where the solid line represents operation result data (valve opening (t)) and the dotted line represents prediction data (predicted valve opening (t)).

[0122] 11, reference numeral 1110 denotes a graph showing an enlarged rectangular area in the graph shown by reference numeral 1100. As shown by reference numeral 1110, in the case of the anomaly sign detection device 110 according to the first embodiment, the 95% confidence interval of the prediction data was narrow.

[0123] In other words, it has been verified that the abnormality detection device 110 according to the first embodiment makes it possible to improve the detection accuracy when detecting abnormality signs of an actuator 161 (e.g., an air-operated control valve 200) in a plant compared to the abnormality detection device of the comparative example.

[0124] <Summary> As is clear from the above description, the abnormality sign detection device 110 according to the first embodiment is communicably connected to a plant control system 140 including a local device (actuator 161) and a DCS controller 150, By inputting the internal data of the actuator 161 and the process data of the DCS controller 150 into the trained model 820, the 95% confidence interval output from the trained model 820 is compared with the operation result data, which is the data to be monitored in the plant. If the operation result data exceeds the 95% confidence interval (if there is no inclusion relationship), it is determined that there is an abnormality sign and an output is generated indicating that an abnormality sign has been detected.

[0125] In this way, by using the internal data of the actuator, the abnormality sign detection device 110 according to the first embodiment can improve the detection accuracy when detecting an abnormality sign of an actuator in a plant.

[0126] [Second embodiment] In the above first embodiment, the case has been described in which the local device that detects an abnormality sign among the local devices in the plant is an actuator, but the local device that detects an abnormality sign is not limited to an actuator and may be, for example, an instrumentation device. Hereinafter, the second embodiment will be described, focusing on the differences from the above first embodiment.

[0127] <Instrumentation device internal data> First, the internal data of the instrumentation device 171 will be described. Here, the internal data of the instrumentation device of a Coriolis flowmeter, which is an example of the instrumentation device 171, will be described. Fig. 12 is a diagram showing an example of the internal data of the instrumentation device of a Coriolis flowmeter.

[0128] As shown in FIG. 12, a Coriolis flowmeter 1210 includes a vibration generator 1211, a tube frequency detector 1212, and a flow rate converter 1213.

[0129] The vibration generator 1211 generates vibrations to resonate the tube 1214 through which the object flows. The tube frequency detector 1212 detects the frequency of the vibrations generated in the tube 1214. The flow rate converter 1213 calculates the mass flow rate of the object flowing through the tube 1214 based on the frequency of the tube 1214 detected by the tube frequency detector 1212.

[0130] Coriolis flowmeter 1210 is Measurement data: mass flow rate calculated by the flow rate converter 1213; to the DCS controller 150, Internal data of the instrument: Density of the object used when the flow converter 1213 calculates the mass flow rate, Internal data of the instrument: the drive gain when the vibration generator 1211 resonates the tube 1214, Instrument internal data: frequency of tube 1214 detected by tube frequency detector 1212; is output to the HART converter 172.

[0131] <Explanation of abnormality in instrumentation equipment> Next, a description will be given of a location where an abnormality occurs in the instrumentation device 171. Here, an abnormal location will be described in a Coriolis flowmeter, which is an example of the instrumentation device 171. Fig. 13 is a diagram showing an example of an abnormal location in the Coriolis flowmeter.

[0132] 13, tube 1214, through which an object flows in the direction of the thin arrow, resonates in the direction of the thick arrow, and tube frequency detector 1212 detects the frequency of the vibration generated in tube 1214. As a result, Coriolis flowmeter 1210 calculates the mass flow rate of the object flowing in tube 1214.

[0133] 13, reference numeral 1301 indicates a main abnormality location that occurs in the Coriolis flowmeter. Reference numeral 1301 indicates a location in the tube 1214 where foreign matter accumulates and causes clogging.

[0134] Conventionally, abnormalities occurring in such locations could not be detected until the equipment was disassembled for regular maintenance, but abnormality sign detection device 110 can accurately detect abnormality signs at an early stage. Therefore, abnormality sign detection device 110 makes it possible to perform maintenance in a planned manner.

[0135] <Explanation of learning data> Next, a description will be given of learning data used when the abnormality sign detection device 110 performs a learning process on a model for detecting an abnormality sign of a local device in the plant control system 140. Fig. 14 is a diagram showing an example of the learning data.

[0136] As shown in FIG. 14, learning data 1400 has information items such as "input data" and "correct answer data."

[0137] "Input data" refers to data (explanatory variables) that are input to the model, and types of data include "process data" and "internal data."

[0138] The “process data” includes the control data (valve opening (t)) sent by the DCS controller 150 to the pneumatic control valve 200.

[0139] “Internal data” includes instrumentation internal data (density, drive gain, tube frequency (t)) output from Coriolis flowmeter 1210.

[0140] The "correct data" is the correct data (objective variables) that should be output from the model when data (explanatory variables) are input to the model. The "correct data" includes the measurement data (mass flow rate (t)) output from the Coriolis flowmeter 1210 when the Coriolis flowmeter 1210 and the plant are in a normal state.

[0141] Note that the addition of "(t)" after the data name of each piece of data included in the learning data 1400 indicates that the data is time-series data.

[0142] In this way, when performing learning processing on a model for detecting a sign of abnormality in the instrumentation device 171 among the local devices in the plant control system 140, the learning data is Control data for the actuator 161, which is process data related to the physical quantity of the object to be measured by the instrumentation device 171; Input data including at least instrumentation device internal data, which is internal data held internally by the instrumentation device 171; Correct answer data including measurement data that is the operation result of the instrumentation device 171; It is composed of a combination of the following.

[0143] <Functional configuration of the anomaly detection device (learning phase)> Next, a functional configuration in the learning phase of the anomaly sign detection device 110 according to the second embodiment will be described. Fig. 15 is a second diagram showing an example of the functional configuration in the learning phase of the anomaly sign detection device.

[0144] The anomaly sign detection device 110 according to the second embodiment executes an anomaly sign detection program to function as a learning unit 1500 in the learning phase. The learning unit 1500 includes an input data acquisition unit 1510, a model 1520, and a comparison and change unit 1530.

[0145] The input data acquisition unit 1510 reads out process data and internal data from the data server 120 as input data for the learning data 1400. The process data of the learning data 1400 includes control data (valve opening (t)), and the internal data includes instrumentation device internal data (density, drive gain, tube frequency (t)). Therefore, the input data acquisition unit 1510 reads out these data.

[0146] Moreover, the input data acquisition unit 1510 inputs the read process data and internal data to the model 1520 .

[0147] The model 1520 is an AI technology (e.g., a neural network) suitable for processing time-series data, and outputs output data when process data and internal data are input by the input data acquisition unit 1510. The output data output from the model 1520 is input to the comparison and change unit 1530.

[0148] The comparing and changing unit 1530 reads out the correct answer data of the learning data 1400 from the data server 120. Since the correct answer data of the learning data 1400 includes the measurement data (mass flow rate (t)), the comparing and changing unit 1530 reads out the data.

[0149] Moreover, the comparison and change unit 1530 calculates a loss by comparing the output data output from the model 1520 with the read-out correct answer data, and updates the model parameters of the model 1520 based on the calculated loss. In this manner, the anomaly sign detection device 110 performs a learning process on the model 1520, and a learned model is generated.

[0150] <Functional configuration of the anomaly detection device (inference phase)> Next, a functional configuration in the inference phase of the anomaly sign detection device 110 according to the second embodiment will be described. Fig. 16 is a second diagram showing an example of the functional configuration in the inference phase of the anomaly sign detection device.

[0151] The anomaly sign detection device 110 executes the anomaly sign detection program, and thereby functions as an inference unit 1600, a judgment unit 1630, and an output unit 1640 in the inference phase. The inference unit 1600 includes an input data acquisition unit 1610 and a trained model 1620.

[0152] The input data acquisition unit 1610 reads out the process data and the internal data from the data server 120. The process data includes the control data (valve opening (t)), and the internal data includes the instrumentation internal data (density, drive gain, tube frequency (t)), so these data are read out.

[0153] In addition, the input data acquisition unit 1610 inputs the read process data and internal data to the trained model 1620.

[0154] The trained model 1620 is a trained model generated by performing a learning process on the model 1520. The trained model 1620 receives process data and internal data, and outputs prediction data (predicted mass flow rate (t)) and a 95% confidence interval of the prediction data. The prediction data (predicted mass flow rate (t)) and the 95% confidence interval of the prediction data output by the trained model 1620 are input to the determination unit 1630.

[0155] The determination unit 1630 outputs the 95% confidence interval of the predicted data output from the trained model 1620 to the output unit 1640.

[0156] In addition, the judgment unit 1630 compares the measurement data (mass flow rate (t)) read from the data server 120 with the 95% confidence interval of the predicted data output from the trained model 1620, and judges whether the measurement data (mass flow rate (t)) exceeds the 95% confidence interval.

[0157] If the measurement data (mass flow rate (t)) does not exceed the 95% confidence interval (if there is an inclusive relationship), the judgment unit 1630 judges that there is no sign of abnormality and outputs the judgment result to the output unit 1640. On the other hand, if the measurement data (mass flow rate (t)) exceeds the 95% confidence interval (if there is no inclusive relationship), the judgment unit 1630 judges that there is a sign of abnormality and outputs the judgment result to the output unit 1640.

[0158] The output unit 1640 outputs the 95% confidence interval of the predicted data output from the determination unit 1630. The output unit 1640 also outputs the determination result of the presence or absence of an abnormality sign output from the determination unit 1630.

[0159] <Learning process and anomaly detection process flow> Next, the flow of the learning process and the abnormality sign detection process executed by the abnormality sign detection device 110 according to the second embodiment will be described. FIG. 17 is an example of a second flowchart showing the flow of the learning process and the abnormality sign detection process. Of these, FIG. 17(a) is a flowchart showing the flow of the learning process executed by the abnormality sign detection device 110 according to the second embodiment. Also, FIG. 17(b) is a flowchart showing the flow of the abnormality sign detection process executed by the abnormality sign detection device 110 according to the second embodiment.

[0160] (1) Learning process flow 17(a), when the abnormality sign detection device 110 according to the second embodiment starts a learning process in the learning phase, the abnormality sign detection device 110 acquires process data as input data in step S1701. Specifically, the abnormality sign detection device 110 acquires control data (valve opening degree (t)), which is process data, from the data server 120.

[0161] In step S1702, the abnormality sign detection device 110 acquires internal data as input data. Specifically, the abnormality sign detection device 110 acquires internal data of the instrumentation device (density, drive gain, tube frequency (t)) from the data server 120.

[0162] In step S1703, the abnormality sign detection device 110 acquires the measurement data (mass flow rate (t)) from the data server 120 as the correct answer data.

[0163] In step S1704, the anomaly sign detection device 110 generates learning data 1400 in which the input data and the correct answer data are associated with each other, and stores the learning data in the data server 120.

[0164] In step S1705, the anomaly sign detection device 110 reads the learning data 1400 from the data server 120 and performs learning processing to generate a trained model.

[0165] In step S1706, the anomaly sign detection device 110 stores the trained model generated by performing the learning process.

[0166] (2) Anomaly detection process flow 17(b), when the abnormality sign detection device 110 according to the second embodiment starts the abnormality sign detection process in the inference phase, the abnormality sign detection device 110 acquires process data in step S1711. Specifically, the abnormality sign detection device 110 acquires control data (valve opening degree (t)), which is process data, from the data server 120.

[0167] In step S1712, the abnormality sign detection device 110 acquires from the data server 120 the internal data of the instrumentation device (density, drive gain, tube frequency (t)).

[0168] In step S1713, the anomaly sign detection device 110 inputs the acquired process data and internal data into the trained model 1620, thereby executing the trained model 1620 and calculating predicted data (predicted mass flow rate (t)) and a 95% confidence interval for the predicted data.

[0169] In step S1714, the abnormality sign detection device 110 acquires the measurement data (mass flow rate (t)) from the data server 120 as the data to be monitored.

[0170] In step S1715, the abnormality sign detection device 110 outputs the calculated 95% confidence interval.

[0171] In step S1716, the abnormality sign detection device 110 determines whether or not the acquired measurement data (mass flow rate (t)) exceeds the calculated 95% confidence interval (whether or not there is an inclusive relationship), thereby determining whether or not there is an abnormality sign.

[0172] In step S1716, if the calculated 95% confidence interval is not exceeded (if there is an inclusive relationship), it is determined that there is no abnormality sign (NO in step S1716), and the process proceeds to step S1718.

[0173] On the other hand, if it is determined in step S1716 that the calculated 95% confidence interval is exceeded (if there is no inclusion relationship), it is determined that there is an abnormality sign (YES in step S1716), and the process proceeds to step S1717.

[0174] In step S1717, the abnormality sign detection device 110 outputs a determination that there is an abnormality sign (that an abnormality sign has been detected).

[0175] In step S1718, the abnormality sign detection device 110 determines whether or not to continue the abnormality sign detection process. If it is determined in step S1718 that the abnormality sign detection process is to be continued (YES in step S1718), the process returns to step S1711.

[0176] On the other hand, if it is determined in step S1718 that the abnormality sign detection process should not be continued (NO in step S1718), the abnormality sign detection process ends.

[0177] <Summary> As is clear from the above description, the abnormality sign detection device 110 according to the second embodiment is communicably connected to a plant control system 140 including a local device (instrumentation device 171) and a DCS controller 150, By inputting the internal data of the instrumentation device 171 and the process data of the DCS controller 150 into the trained model 1620, the 95% confidence interval output from the trained model 1620 is compared with the measurement data, which is the data to be monitored in the plant. If the measurement data exceeds the 95% confidence interval (if there is no inclusion relationship), it is determined that there is an abnormality and an output is generated indicating that a sign of an abnormality has been detected.

[0178] In this way, by using internal data of the instrumentation equipment, the anomaly sign detection device 110 of the second embodiment can improve the detection accuracy when detecting anomaly signs of the instrumentation equipment in a plant.

[0179] [Third embodiment] In each of the above embodiments, the abnormality sign detection device 110 has been described as detecting an abnormality sign of each local device. However, the target for which the abnormality sign detection device 110 detects an abnormality sign is not limited to each local device. For example, the plant itself, to which multiple local devices are attached and which is controlled by the DCS controller 150, may be the target for which an abnormality sign is detected.

[0180] In this case, too, any of the process data used by the DCS controller 150 to control and monitor the plant is stored in the "process data" as input data for the learning data. Furthermore, any of the data stored in any of the local devices installed in the plant, which is not used by the DCS controller 150 to control and monitor the plant, is stored in the "internal data" as input data for the learning data. Furthermore, monitoring target data for monitoring for signs of abnormality in the plant itself is stored as correct answer data for the learning data. The monitoring target data for monitoring for signs of abnormality in the plant itself includes operation result data indicating the operation results of the plant (for example, measurement data that is process data indicating the amount (production amount) of a product produced by the plant).

[0181] Furthermore, an abnormality sign in the plant itself refers to an abnormality sign in equipment other than the local devices installed in the plant.

[0182] Fig. 18 is a first diagram showing an example of a schematic configuration of a chemical plant, which is an example of a chemical plant that continuously produces a predetermined product. As shown in Fig. 18, the chemical plant 1800 includes, as local devices, a plurality of actuators (reference numbers 1811 to 1814) and a plurality of instrumentation devices (reference numbers 1801 to 1807). The chemical plant 1800 also includes a plurality of pieces of equipment (reference numbers 1831 to 1834).

[0183] 18, abnormalities in the plant itself may include, for example, a clog in the upper part of equipment 1831 (see reference symbol 1841), a clog in the pipe connecting equipment 1831 and equipment 1833 (see reference symbol 1842), etc. For this reason, the abnormality sign detection device 110 detects these abnormality signs.

[0184] Fig. 19 is a second diagram showing an example of a schematic configuration of a chemical plant, which is an example of a chemical plant that produces a predetermined product in batches. As shown in Fig. 19, the chemical plant 1900 includes, as local devices, a plurality of actuators (reference numbers 1911-1913) and a plurality of instrumentation devices (reference numbers 1901-1904). The chemical plant 1900 also includes a plurality of pieces of equipment (reference numbers 1931-1932).

[0185] 19, abnormalities in the plant itself may include, for example, corrosion (see reference numeral 1941) inside equipment 1931, clogging (see reference numeral 1942) inside a pipe connecting equipment 1931 and equipment 1932, etc. For this reason, the abnormality sign detection device 110 detects these abnormality signs.

[0186] FIG. 20 is the third diagram showing an example of learning data, which is an example of learning data used when performing learning processing on a model for detecting signs of abnormality in the chemical plant 1800 or 1900 itself.

[0187] As shown in FIG. 20, learning data 2000 has "input data" and "correct answer data" as information items.

[0188] "Input data" refers to data (explanatory variables) that are input to the model, and types of data include "process data" and "internal data."

[0189] The “process data” includes control data for each actuator installed in the chemical plant 1800 or 1900 and measurement data for each instrumentation device installed in the chemical plant 1800 or 1900 .

[0190] "Internal data" includes actuator internal data held internally by each actuator installed in chemical plant 1800 or 1900, and instrumentation internal data held internally by each instrumentation device installed in chemical plant 1800 or 1900.

[0191] The "correct answer data" is the correct answer data (objective variables) that should be output from the model when data (explanatory variables) are input to the model. The "correct answer data" includes operation result data that indicates the operation results of the chemical plant 1800 or 1900 when the chemical plant 1800 or 1900 is in a normal state. The operation result data that indicates the operation results of the chemical plant 1800 or 1900 refers to, for example, measurement data that indicates the production amount of the chemical plant 1800 or 1900 measured by any of the instrumentation devices.

[0192] In this way, according to the third embodiment, the plant itself, to which a plurality of local devices are attached and which is controlled by the DCS controller 150, can be a target for detecting abnormality signs.

[0193] In the third embodiment, Fig. 18 shows a case where a flow meter is installed in a pipe to detect a sign of abnormality due to clogging of the pipe. However, depending on the plant, a case where a flow meter is not installed in the pipe may be assumed. In such a case where a pump and a pressure gauge are installed, for example, a sign of abnormality may be detected using control data for the pump and internal data of the instrumentation device held inside the pressure gauge (clogging index derived from pressure fluctuation).

[0194] Specifically, the “process data” of the learning data 2000 includes, for example, Control data (current data (t)) that controls the current applied to the pump when it is driven; The "internal data" of the learning data 2000 includes, for example, - Internal data of the instrumentation equipment output from the pressure gauge (clogging index (t) derived from pressure fluctuations), In addition, the "correct answer data" of the learning data 2000 may include, for example, Measurement data measured by a pressure gauge (pressure data (t)), may be included.

[0195] [Other embodiments] In the above first embodiment, the abnormality sign detection device 110 has been described as acquiring the control signal (t), supply air pressure (t), and output air pressure (t) as the actuator internal data of the pneumatic control valve 200. However, the actuator internal data of the pneumatic control valve 200 is not limited to these combinations, and may be other combinations. For example, if the pneumatic control valve 200 does not have sensors that measure the supply air pressure (t) and the output air pressure (t), these may be replaced by other actuator internal data that the positioner 220 has. Furthermore, if a sensor that measures nozzle exhaust pressure is included, the actuator internal data of the pneumatic control valve 200 may include nozzle exhaust pressure.

[0196] Similarly, in the above second embodiment, the abnormality sign detection device 110 has been described as acquiring density, drive gain, and tube frequency (t) as the internal data of the instrumentation device of the Coriolis flowmeter. However, the internal data of the instrumentation device of the Coriolis flowmeter is not limited to these combinations and may be other combinations.

[0197] In addition, in the above first embodiment, the anomaly sign detection device 110 has given an example of the learning data 600 in FIG. 6 as a combination of input data and supervised data, but the combination of input data and supervised data is not limited to this and may be other combinations.

[0198] Similarly, in the above second embodiment, the anomaly sign detection device 110 has given an example of the learning data 1400 in FIG. 14 as a combination of input data and supervised data, but the combination of input data and supervised data is not limited to this and may be other combinations.

[0199] Similarly, in the above third embodiment, the anomaly sign detection device 110 has given an example of the learning data 200 in FIG. 20 as a combination of input data and supervised answer data, but the combination of input data and supervised answer data is not limited to this and may be other combinations.

[0200] In the above first embodiment, the pneumatic control valve 200 has been described as an example of the actuator 161 of the local device, but the actuator 161 of the local device is not limited to the pneumatic control valve 200 and may be another type of control valve. The actuator 161 of the local device includes any driving device other than a control valve used to control the plant. Examples of the arbitrary driving device include a pump, a compressor, a heater, a cooler, a motor, and the like.

[0201] In addition, when the local device is a pump, the “process data” as the input data of the learning data 600 is, for example, Measurement data (flow data (t)) measured by a flowmeter that measures the flow rate of an object being transported through a pipe by a pump; Control data (current data (t)) that controls the current applied to the pump when it is driven; In addition, as input data for the learning data 600, the “internal data” may include, for example, Actuator internal data output from the control valve (valve opening (t)), In addition, the "correct answer data" of the learning data 600 may include Measurement data (pressure data (t)) measured by a pressure gauge that measures the pressure of the object being conveyed by the pump; may be included.

[0202] Similarly, in the above second embodiment, the Coriolis flowmeter 1210 has been described as an example of the instrumentation device 171 of the local device, but the instrumentation device 171 of the local device is not limited to the Coriolis flowmeter 1210. For example, the instrumentation device 171 of the local device may include a flow meter other than the Coriolis flowmeter 1210 (for example, a thermal flow meter, an area flow meter, a differential pressure flow meter, an ultrasonic flow meter, a positive displacement flow meter, a vortex flow meter, an electromagnetic flow meter). Alternatively, the instrumentation device 171 of the local device may include an instrumentation device other than a flow meter (for example, a liquid level gauge, a pressure gauge, a thermometer, an analyzer, etc.). The liquid level gauge may include various liquid level gauges such as a differential pressure type, a float type, a disperser type, a microwave type, a capacitance type, a tuning fork type, etc. Furthermore, the pressure gauge may include, for example, an electric pressure gauge. Furthermore, the analyzer may include, for example, a pH meter, an ORP meter (oxidation-reduction potential meter), a conductivity meter, a density meter, and the like.

[0203] In addition, in each of the above embodiments, the internal data of the local devices is configured to be transmitted to the DCS controller 150 via the HART converter 162 or the HART converter 172. However, the configuration of the plant control system 140 is not limited to this. For example, the internal data of the local devices may be transmitted to the DCS controller 150 via a converter other than the HART converter 162 or the HART converter 172, or the internal data of the local devices may be transmitted directly to the DCS controller 150.

[0204] Furthermore, in the learning phase of the anomaly sign detection system in each of the above embodiments, the anomaly sign detection device 110 acquires the data used to generate the learning data 600 or the learning data 1400 via the OPC server 130 and the data server 120. However, the configuration of the anomaly sign detection system in the learning phase is not limited to this, and the anomaly sign detection device 110 may be configured to acquire data without going through either or both of the OPC server 130 and the data server 120.

[0205] Similarly, the anomaly sign detection system in each of the above embodiments is configured such that, in the inference phase, the anomaly sign detection device 110 acquires data used for detecting an anomaly sign via the OPC server 130 and the data server 120. However, the configuration of the anomaly sign detection system in the inference phase is not limited to this, and the anomaly sign detection device 110 may be configured to acquire data without going through either or both of the OPC server 130 and the data server 120.

[0206] Further, the anomaly sign detection system in each of the above embodiments is configured to detect anomaly signs in a local device using one anomaly sign detection device 110, but may also be configured to detect anomaly signs in a local device using, for example, multiple anomaly sign detection devices 110.

[0207] Although the embodiments have been described above, it will be understood that various changes in form and details are possible without departing from the spirit and scope of the claims. [Explanation of symbols]

[0208] 100: Anomaly detection system 110: Abnormality detection device 120: Data Server 130: OPC Server 140: Plant control systems 150: DCS controller 161: Actuator 162: HART converter 163:Power source 171: Instrumentation equipment 172: HART converter 200: Pneumatic control valve 600: Training data 700: Learning Department 800: Reasoning part 820: Trained model 830: Judgment section 840: Output section 1210: Coriolis flowmeter 1400: Training data 1500: Learning Department 1600: Reasoning part 1620: Trained model 1630: Judgment section 1640: Output section 2000: Training data

Claims

1. An abnormality sign detection device (110) that is communicatively connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of abnormality in the local devices (161, 171) is determined based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; outputting the determination result of the presence or absence of the abnormality sign; An abnormality sign detection device (110).

2. An abnormality sign detection device (110) that is communicatively connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of an abnormality in the plant is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; outputting the determination result of the presence or absence of the abnormality sign; An abnormality sign detection device (110).

3. The data to be monitored is output data output from the local device (161, 171). The abnormality sign detection device (110) according to claim 1 or 2.

4. The trained model (820, 1620) is The learning data (600, 1400) includes internal data of the local devices (161, 171), process data of the controller (150), and output data output by the local devices (161, 171) when the local devices and the plant are in a normal state. The abnormality sign detection device (110) according to claim 3.

5. The local devices (161, 171) include actuators (161) and instrumentation devices (171) in the plant. The abnormality sign detection device (110) according to claim 3.

6. When the local device (161, 171) is an actuator (161), The process data includes: Data indicating a physical quantity of an object controlled by the actuator (161) and measured by the instrumentation device (171); control data for the controller (150) to control the operation of the actuator (161); Including, The monitored data is data indicating the operation result of the actuator (161). The abnormality sign detection device (110) according to claim 5.

7. When the local device (161, 171) is an instrumentation device (171), The process data includes control data for the controller (150) to control the operation of the actuator (161); The data of the monitored object is data indicating a physical quantity of the object controlled by the actuator (161) and measured by the instrumentation device (171). The abnormality sign detection device (110) according to claim 5.

8. The controller (150) acquiring superimposition data in which internal data of the local device (161, 171) is superimposed on output data output from the local device (161, 171); The abnormality sign detection device (110) according to claim 3.

9. The plant control system (140) includes a HART transducer (162, 172); The superimposed data is HART data, and the controller (150) acquires the HART data from the HART converter (162, 172) connected to the local device (161, 171). The abnormality sign detection device (110) according to claim 8.

10. The control unit (501, 502) of the abnormality sign detection device (110) acquiring the process data from the controller (150) via a data server (120); The abnormality sign detection device (110) according to claim 3.

11. When the actuator (161) is a pneumatic control valve (200), The process data includes: Data indicating the flow rate of an object controlled by the pneumatic control valve (200) and measured by a flow meter (1210); control data for the controller (150) to control the operation of the pneumatic control valve (200); Including, The monitored data is data indicating a valve opening degree of the air-operated control valve (200), The internal data of the air-operated regulating valve (200) includes a signal for controlling the operation of the air-operated regulating valve (200) received from the controller (150), a signal indicating an output air pressure, and a signal indicating a supply air pressure. The abnormality sign detection device (110) according to claim 6.

12. When the instrumentation device (171) is a flow meter (1210), The process data includes data indicating a valve opening degree of a regulator valve (200), The data of the monitored object is data indicating a flow rate of the object controlled by the control valve (200) and measured by the flow meter (1210); The internal data of the flow meter (1210) includes the density of the object, the drive gain of the flow meter (1210), and the tube frequency. The abnormality sign detection device (110) according to claim 7.

13. The abnormality sign detection device (110) according to claim 9, The plant control system (140) is communicatively connected to the abnormality sign detection device (110); An abnormality sign detection system (100) comprising:

14. The abnormality sign detection device (110) according to claim 10; The abnormality sign detection device (110) includes a data server (120) for acquiring the process data from the plant control system (140); An abnormality sign detection system (100) comprising:

15. a control unit (501, 502) included in an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in the plant, By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of abnormality in the local devices (161, 171) is determined based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; outputting the determination result of the presence or absence of the abnormality sign; An anomaly detection method that executes processing.

16. A control unit (501, 502) of an abnormality sign detection device (110) that is communicatively connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in the plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of abnormality in the plant (1800, 1900) is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; outputting the determination result of the presence or absence of the abnormality sign; An anomaly detection method that executes processing.

17. A control unit (501, 502) of an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in a plant, By inputting internal data of the local devices (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of abnormality in the local devices (161, 171) is determined based on output data output from the trained model (820, 1620) and data of a monitoring target in the plant; outputting the determination result of the presence or absence of the abnormality sign; Anomaly detection program to execute processing.

18. A control unit (501, 502) of an abnormality sign detection device (110) that is communicably connected to a plant control system (140) including local devices (161, 171) and a controller (150) and detects an abnormality sign in the plant, The control unit (501, 502) of the abnormality sign detection device (110) By inputting internal data of the local device (161, 171) and process data of the controller (150) into a trained model (820, 1620), the presence or absence of a sign of abnormality in the plant (1800, 1900) is determined based on output data output from the trained model (820, 1620) and data of a monitored object in the plant; outputting the determination result of the presence or absence of the abnormality sign; Anomaly detection program to execute processing.

Images