JP7151312B2 - control system - Google Patents

Description

The present invention relates to a control device and control system for determining the presence or absence of an abnormality in a monitored object.

In various production sites, FA (Factory Automation) technology for controlling machines and devices to be controlled using a control device such as a PLC (Programmable Controller) is widely used. At production sites, there is a need to improve facility operating rates through predictive maintenance for machines and equipment. Predictive maintenance is a form of maintenance that determines the presence or absence of any abnormalities that occur in the machines and devices to be monitored, and performs maintenance work such as maintenance and replacement before the equipment must be stopped. means.

As a technique for determining the presence or absence of an abnormality in a monitored object, for example, a semiconductor manufacturing equipment management system disclosed in Japanese Patent Application Laid-Open No. 7-282146 (Patent Document 1) is known. In this semiconductor manufacturing equipment management system, a management computer manages the operation timing of the entire equipment from I/O (input/output) data obtained from various sensors attached to the semiconductor manufacturing equipment. In addition, the management computer compares the operational timing of the entire system with normal conditions and generates an alarm if the timing difference exceeds the management criteria.

JP-A-7-282146

A determination method for determining whether or not there is an abnormality in the monitoring target differs depending on the type of monitoring target. Therefore, a coordinator (hereinafter referred to as a "field engineer") needs to adjust the method of determining the presence or absence of an abnormality according to the type of object to be monitored. Adjustment of the determination method is, for example, adjustment of the determination criteria.

On the other hand, a designer who designs a control device for controlling a controlled object including a monitored object designs a user program that defines the operation of the control device. The designer defines a program (abnormality determination program) for determining the presence or absence of an abnormality in the monitored object in the user program.

Field engineers and designers are generally different. Furthermore, the timing for the field engineer to adjust the determination method differs from the timing for the designer to design the user program. Therefore, it is difficult for field engineers to adjust the judgment method while considering the influence on the user program, and it is difficult for the designer to design the user program while considering the influence on the adjustment of the judgment method. rice field.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide a control device and a control system that facilitate adjustment of a judgment method by a field engineer and facilitate design of a user program by a designer. is to provide

According to one example of the present disclosure, the control device includes a control calculation unit and an abnormality determination unit. The control calculation unit uses state values periodically collected from the controlled object to execute control calculations for controlling the controlled object. The anomaly determination unit calculates a score indicating a result of comparison between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set, and compares the calculated score with the preset learning data set. The presence or absence of an abnormality in the monitored object is determined based on the determined determination criteria. The abnormality determination section is implemented by defining instructions in a user program for controlling the controlled object. The control device further includes an updating unit that receives the update learning data set and the update criterion from the outside and updates the currently set learning data set and the criterion, respectively.

According to this disclosure, the designer of the control device can design the user program so that the presence or absence of an abnormality in the monitoring target is determined using the provisionally created learning data set and the determination criterion information. Furthermore, after the user program is designed, the field engineer adjusts the learning data set and the criterion information so that the presence or absence of anomalies in the monitoring target can be determined with high accuracy, and then sends the adjusted learning data set and the criterion information. You can provide it to the update department. Thus, according to the above-described control device, the field engineer can easily adjust the determination method, and the designer can easily design the user program.

In the above disclosure, the feature amount to be generated is predetermined for each monitoring target. A learning data set and a criterion are set according to the generated feature amount. According to this disclosure, it is possible to set the feature amount, the learning data set, and the criterion according to the monitoring target.

In the above disclosure, the abnormality determination section is implemented by defining commands in the form of function blocks in the user program. According to this disclosure, a designer can easily design a user program that determines the presence or absence of an abnormality in a monitored object using function blocks.

In the above disclosure, a function block, a provisional learning data set, and a provisional determination criterion are associated and created in advance for each monitoring target. By defining a function block, a provisional learning data set and provisional judgment criteria corresponding to the function block are set. According to this disclosure, by defining function blocks, a designer can easily set a temporary learning data set and a temporary determination criterion for determining the presence or absence of an abnormality in a monitoring target.

In the above disclosure, the abnormality determination unit generates feature quantities according to designation information that designates the types of state values and the types of feature quantities. The update unit receives specification information for update from the outside and updates the currently set specification information. According to this disclosure, a field engineer can appropriately adjust the feature amount suitable for determining whether there is an abnormality in the monitoring target.

According to one example of the present disclosure, a control system includes a control calculation section and an abnormality determination section. The control computation unit uses state values periodically collected from the controlled object to execute control computation for controlling the controlled object. The anomaly determination unit calculates a score indicating a result of comparison between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set, and compares the calculated score with the preset learning data set. The presence or absence of an abnormality in the monitored object is determined based on the determined determination criteria. The abnormality determination unit is implemented by being defined within a user program for controlling the controlled object. The control system further comprises a determiner and an updater. The determination unit determines a learning data set for updating and a determination criterion for updating using the state value or the generated feature value related to the monitored object. The updating unit receives the update learning data set and the update criterion, and updates the currently set learning data set and the criterion, respectively.

This disclosure also makes it easier for field engineers to adjust the determination method and easier for designers to design user programs.

In the above disclosure, the control system further includes a programming unit that creates a user program. The program creation unit creates a user program using function blocks that define commands for realizing the abnormality determination unit, and sets a temporary learning data set and a temporary determination criterion associated with the function blocks. According to this disclosure, by defining function blocks, a designer can easily set a temporary learning data set and a temporary determination criterion for determining the presence or absence of an abnormality in a monitoring target.

Advantageous Effects of Invention According to the present disclosure, it becomes easier for a field engineer to adjust a determination method and for a designer to easily design a user program.

1 is a schematic diagram showing a configuration example of a control device according to an embodiment; FIG. 1 is a schematic diagram showing an example of the overall configuration of a control system according to an embodiment; FIG. FIG. 3 is a schematic diagram showing a hardware configuration example of a design tool device; FIG. 3 is a schematic diagram showing a software configuration example of a design tool device; FIG. 10 is a diagram showing an example of a correspondence relationship between function blocks and attached data for determining whether or not there is an abnormality in a monitoring target; FIG. 3 is a diagram showing an example of function blocks for determining whether or not there is an abnormality in a monitored object; FIG. 7 is a diagram showing an example of a user program created using the function blocks shown in FIG. 6; FIG. 2 is a block diagram showing a hardware configuration example of a control device according to the present embodiment; FIG. 2 is a block diagram showing a hardware configuration example of a support device according to the present embodiment; FIG. FIG. 2 is a block diagram showing a main software configuration example of the control system according to the present embodiment; FIG. 11 is a block diagram showing an overview of functional modules included in the analysis tool shown in FIG. 10; FIG. FIG. 4 is a schematic diagram for explaining the basic concept of abnormality determination processing of the control system according to the present embodiment; FIG. 4 is a schematic diagram outlining the procedure of abnormality determination processing of the control system according to the present embodiment; 4 is a flow chart showing the procedure of abnormality determination processing executed in the control device according to the present embodiment; FIG. 4 is a schematic diagram showing the contents of analysis processing according to the present embodiment; FIG. 16 is a schematic diagram visually showing an overview of the processes (a) to (c) shown in FIG. 15; 4 is a flow chart showing an example of a processing procedure of the support device according to the present embodiment; FIG. 18 is a schematic diagram showing an example of a user interface screen provided to a field engineer in step S6 of FIG. 17; 18 is a schematic diagram showing an example of a user interface screen provided to the user in step S8 of FIG. 17; FIG. It is a block diagram showing an example of main software composition of a control system concerning a modification. 21 is a block diagram showing an overview of functional modules included in the analysis tool shown in FIG. 20; FIG. It is a schematic diagram which shows visually the outline|summary of the analysis process which concerns on a modification. It is a flow chart which shows a processing procedure of a support device concerning a modification. FIG. 24 is a schematic diagram showing an example of a user interface screen provided by the support device to the field engineer in step S108 of FIG. 23; FIG. 11 is a schematic diagram for explaining a process of evaluating the importance of feature amounts executed by the analysis tool of the support device according to the modification;

Embodiments according to the present invention will be described below with reference to the drawings. In the following description, identical parts and components are given identical reference numerals. Their names and functions are also the same. Therefore, detailed description of these will not be repeated.

§1 Application Example First, an example of a scene to which the present invention is applied will be described with reference to FIG. FIG. 1 is a schematic diagram showing a configuration example of a control device according to this embodiment. The control device 100 is, for example, a PLC (Programmable Logic Controller). As shown in FIG. 1 , the control device 100 includes a control calculation section 10 , an abnormality determination section 20 and an update section 30 .

The control computation unit 10 uses state values periodically collected from the controlled object to execute control computation for controlling the controlled object.

In this specification, the "state value" includes a value that can be observed in any controlled object (including: monitored object). "State values" are, for example, physical values that can be measured by any sensor, ON/OFF states of relays and switches, command values such as position, speed, and torque given to the servo driver by the PLC, and variable values used by the PLC for calculation. and so on.

The abnormality determination unit 20 calculates a score indicating a comparison result between the learning data set 41 and the feature amount generated from the state values related to the monitored object among the state values collected by the control calculation unit 10 . The abnormality determination unit 20 generates feature amounts according to the feature amount designation information 43 . The feature amount designation information 43 is information for designating the type of state value and the type of feature amount (or the method of generating the feature amount (for example, calculation algorithm)), and is set in advance for each monitoring target. The number of feature quantities generated by the abnormality determination unit 20 is one or more. The abnormality determination unit 20 determines whether or not there is an abnormality in the monitoring target based on the calculated score and a criterion (typically a threshold value) indicated by the criterion information. The learning data set 41 and the criterion information 42 are set in advance according to the feature amount to be generated.

As used herein, "score" means a value that indicates the degree of likelihood that a set of one or more feature values is an outlier or anomaly. The score is calculated such that the higher the value, the higher the possibility of being an abnormal value (however, the higher the possibility of being an abnormal value, the smaller the score may be). A method of calculating the score will be described in detail later.

Abnormality determination unit 20 is implemented by defining commands in a user program for controlling a controlled object. The user program is arbitrarily designed according to the object to be controlled.

The update unit 30 receives a learning data set for update and determination criterion information from an external support device, and updates the currently set learning data set 41 and determination criterion information 42, respectively.

As described above, the control device 100 according to the present embodiment includes the abnormality determination section 20 that is realized by defining commands in the user program. The control device 100 further includes an update unit 30 that updates the learning data set 41 and the determination reference information 42 used when the abnormality determination unit 20 determines whether there is an abnormality in the monitored object. Thereby, the designer of the control device 100 can design the user program so that the presence or absence of an abnormality in the monitored object is determined using the provisionally created learning data set and the determination criterion information. That is, the designer can design the user program without considering the adjustment of the abnormality determination method by the field engineer.

Furthermore, after the user program is designed, the field engineer adjusts the learning data set and the criterion information so that the presence or absence of anomalies in the monitoring target can be determined with high accuracy, and then sends the adjusted learning data set and the criterion information. It may be provided to the update unit 30 . As a result, the presence or absence of an abnormality in the monitoring target is determined using the adjusted learning data set and the determination criterion information.

As described above, according to the control device 100 of the present embodiment, the field engineer can easily adjust the determination method, and the designer can easily design the user program.

§2 Configuration example <2-1. Overall configuration example of control system>
Embodiments of the present invention will be described in detail with reference to the drawings. FIG. 2 is a schematic diagram showing an example of the overall configuration of the control system according to this embodiment.

As shown in FIG. 2, the control system 1 according to the present embodiment includes, as main components, a control device 100 that controls a controlled object, a design tool device 200 that is connected to the control device 100, and a control device 100. and a support device 300 connected to.

A design tool device 200 is a tool for designing the control device 100 . The design tool device 200 creates a user program for controlling the controlled object according to the operation by the designer.

A designer uses the design tool device 200 in various situations, such as a phase of creating a user program, a phase of checking the operation of the user program, and a phase of correcting bugs in the user program.

The user program defines an instruction (hereinafter referred to as an "abnormality determination instruction") for determining the presence or absence of an abnormality in the monitored object. The design tool device 200 sets, in the control device 100, the created user program, the feature amount designation information, the provisional learning data set, and the provisional determination criterion information that are associated with the abnormality determination instruction in the user program. .

The control device 100 may be embodied as a kind of computer, such as a PLC. The control device 100 is connected via a field network 2 to one or a plurality of field devices placed on a controlled object, and is connected via another field network 3 to one or a plurality of operation display devices 400 . Note that the operation display device 400 is an optional configuration and not an essential configuration of the control system 1 .

Control targets may include various types of machines and devices. The controlled object is, for example, a uniaxial actuator or an air cylinder. In the example shown in FIG. 2, the control device 100 is connected to servo drivers 619 and 629 included in such controlled objects. Servo drivers 619 and 629 control the rotation speed and torque of servo motors 618 and 628, respectively. Furthermore, the control device 100 is connected to a notification device 18 via an I/O (input/output) unit 16 .

Field networks 2 and 3 preferably employ industrial networks. EtherCAT (registered trademark), EtherNet/IP (registered trademark), DeviceNet (registered trademark), CompoNet (registered trademark), and the like are known as such industrial networks.

The control device 100 operates according to the system program and the user program set by the design tool device 200 . The control calculation unit 10 and the abnormality determination unit 20 shown in FIG. 1 are implemented by executing the system program and the user program. That is, the control device 100 executes the following processes (1) to (3) according to the user program.
(1) A process of periodically collecting state values from a controlled object.
(2) A process of generating a feature amount from the state values related to the monitored object among the collected state values according to the feature amount designation information set by the design tool device 200 .
(3) Calculate a score indicating a comparison result between the generated feature amount and the currently set learning data set, and based on the calculated score and the currently set determination criterion information (threshold value), Processing to determine whether there is an abnormality in the monitoring target.

In the process (1) above, the control device 100 collects state values such as the torque values of the servo motors 618 and 628, the current values of the servo drivers 619 and 629, and the voltage values of the servo drivers 619 and 629, for example.

Control device 100 further receives a learning data set and determination criterion information for update from support device 300, and updates the currently set learning data set and determination criterion information, respectively. The updating is performed by the updating unit 30 shown in FIG. Therefore, when the update has not been performed even once, in the above process (3), the provisional learning data set and the provisional determination criterion information set by the design tool device 200 are used to determine whether or not there is an abnormality in the monitoring target. is determined. When the update learning data set and the determination criterion information are received from the support device 300, in the above process (3), the presence or absence of an abnormality in the monitoring target is determined using the update learning data set and the determination criterion information. be done.

When the control device 100 determines that there is an abnormality in the monitored object, the control device 100 may notify the fact by any method. In the example shown in FIG. 2, the control device 100 notifies by blinking and/or ringing the notification device 18 connected via the I/O unit 16 . The notification method is not limited to the notification device 18, and any indicator, voice output device, voice synthesizer, e-mail, notification to any terminal, or the like can be used.

The support device 300 generates a learning data set for update and determination criterion information so that determination of the presence or absence of an abnormality in the monitoring target is appropriately performed. The support device 300 acquires from the control device 100 a feature amount generated from the state value related to the monitoring target, and performs an analysis process on the acquired feature amount to obtain a learning data set for update and a determination criterion. Generate information. The support device 300 provides the control device 100 with the generated update learning data set and the criterion information. Thereby, the learning data set and the criterion information set in the control device 100 are updated.

<2-2. Hardware configuration example of the design tool device>
FIG. 3 is a schematic diagram showing a hardware configuration example of the design tool device 200. As shown in FIG. The design tool device 200 is typically composed of a general-purpose computer. From the standpoint of maintainability at the manufacturing site where the control device 100 is arranged, the design tool device 200 is preferably a portable notebook personal computer.

The design tool device 200 of the example shown in FIG. 3 includes a storage unit 201 and a CPU 202 that executes various programs including an operating system (OS). The storage unit 201 includes a ROM (Read Only Memory) 204 that stores BIOS and various data, a RAM 206 that provides a work area for storing data necessary for the execution of programs by the CPU 202, and programs executed by the CPU 202. and a hard disk (HDD) 208 that nonvolatilely stores such as.

The design tool device 200 further includes an operation unit 203 including a keyboard 210 and a mouse 212 operated by the designer to input instructions to the design tool device 200, and a display 214 for presenting information to the designer. The design tool device 200 includes a communication interface 218 for communicating with the controller 100 or the like. Communication interface 218 may include a USB communication module to communicate with a USB interface (not shown) provided by controller 100 .

The design tool device 200 includes an optical recording medium reader 216 for reading from the optical recording medium 8 a support program for providing a development support environment stored therein.

These components of the design tool device 200 are communicably connected to each other via an internal bus 220 .

FIG. 3 shows a configuration example in which necessary functions are provided by a processor such as the CPU 202 executing a program. , ASIC or FPGA). In this case, virtualization technology may be used to execute a plurality of OSs with different purposes in parallel, and necessary applications may be executed on each OS.

<2-3. Software configuration example of the design tool device>
FIG. 4 is a schematic diagram showing a software configuration example of the design tool device 200. As shown in FIG. Instruction codes included in the software shown in FIG. 4 are read out at appropriate timing, provided to CPU 202 of design tool device 200, and executed. Further, the software shown in FIG. 4 is included in the support program stored in the optical recording medium 8 and provided.

As shown in FIG. 4, the design tool device 200 is equipped with an OS 240, a programming application 250, a function block library 280 and an attached data storage unit 290. FIG. The design tool device 200 runs an OS 240 and provides an environment in which a programming application 250 can be executed.

The function block library 280 is a library of function blocks (FBs), which are program parts for the PLC used by the programming application 250 . The function block library 280 includes function blocks 282 that are created in advance for each monitored object and used to determine whether there is an abnormality in the monitored object.

The attached data storage unit 290 stores attached data 292 created in advance for each function block for determining the presence or absence of an abnormality in the monitoring target.

FIG. 5 is a diagram showing an example of a correspondence relationship between function blocks and attached data for determining whether or not there is an abnormality in a monitored object. In the example shown in FIG. 5, a function block 282a and attached data 292a are associated, and a function block 282b and attached data 292b are associated. The function blocks 282a and 282b are program parts prepared in advance for determining whether there is an abnormality in the monitoring targets "A" and "B", respectively.

The attached data 292a includes the feature amount designation information 43a for generating the feature amount used for determining the presence or absence of an abnormality in the monitoring target "A", the temporary learning data set 41a corresponding to the feature amount designation information 43a, and the and criteria information 42a. The attached data 292b includes the feature amount designation information 43b for generating the feature amount used for determining the presence or absence of an abnormality in the monitoring target "B", the temporary learning data set 41b corresponding to the feature amount designation information 43b, and the and criteria information 42b.

In addition, hereinafter, each of the feature quantity designation information 43a, 43b, . Similarly, when the learning data sets 41a, 41b, . . . are not particularly distinguished, each of the learning data sets 41a, 41b, . When the criteria information 42a, 42b, . . . are not particularly distinguished, each of the criteria information 42a, 42b, .

In this way, the function block 282, the provisional learning data set 41, and the provisional criterion information 42 are created in advance for each monitoring target.

The feature quantity designation information 43 is appropriately set according to the monitored object, and designates one or a plurality of feature quantities. For example, if the object to be monitored is a ball screw servomotor, feature quantity designation information is set that designates the torque of the servomotor as the type of state value and designates the average value as the feature quantity.

The learning data set 41 is a set of learning data including feature values that can be taken when the monitored object is normal.

The determination criterion information 42 is information indicating a criterion (typically a threshold value) for determining whether or not there is an abnormality in the monitored object. For example, if the score indicating the comparison result between the feature amount and the learning data set 41 is larger than a threshold value, it is determined that there is an abnormality, and if the score is smaller than the threshold value, it is determined that there is no abnormality.

In the following description, a form using a threshold value as a typical example of the determination criterion will be illustrated, but the determination criterion is not limited to the threshold value. It may be a prescribed condition.

At the stage of creating the attached data 292, since the monitoring target is not actually operated, the provisional learning data set 41 and the criterion information 42 are provisionally created by simulation or the like.

Returning to FIG. 4, programming application 250 includes editor 252 , compiler 254 , debugger 256 , GUI (Graphical User Interface) module 258 , simulator 260 and data storage unit 270 . Each module included in the programming application 250 is typically distributed as a support program stored in the optical recording medium 8 (see FIG. 3) and installed in the design tool device 200 .

Editor 252 provides input and editing functions for creating the source program of user program 130 . More specifically, the editor 252 provides a function of creating the source program of the user program 130 by the designer operating the keyboard 210 and the mouse 212, as well as a saving function and an editing function of the created source program. The editor 252 creates the source program of the user program 130 using function blocks 282 selected from the function block library 280 according to the designer's operation.

Compiler 254 provides a function of compiling a source program and generating user program 130 in a program format executable by control device 100 .

Debugger 256 provides a function for debugging the source program of user program 130 . Contents of this debugging include operations such as partially executing a range specified by the designer in the source program and tracing temporal changes in variable values during execution of the source program.

The GUI module 258 has a function of providing a user interface screen for the designer to input various data and parameters. The user interface screen is displayed on display 214 (see FIG. 3). The GUI module 258 provides a user interface screen based on the operation received by the operation unit 203 and the user program 130 .

The simulator 260 builds an environment for simulating program execution in the control device 100 in the design tool device 200 .

The created user program 130 is stored in the data storage unit 270 . When the creation of the user program 130 is completed, the user program 130 stored in the data storage unit 270 is set in the control device 100 via the communication interface 218 (see FIG. 3). At this time, the attached data 292 corresponding to the function block 282 defined in the user program 130 is read from the attached data storage unit 290 and the attached data 292 is also set in the control device 100 .

In this way, the design tool device 200 operates as a program creating unit that creates the user program 130 using the function block 282 that defines commands for determining whether there is an abnormality in the monitored object. The design tool device 200 then sets the provisional learning data set 41 and the provisional criterion information 42 associated with the function block 282 in the control device 100 . That is, the designer can provisionally set the learning data set 41 and the criterion information 42 used for determining whether or not there is an abnormality in the monitoring target by selecting the function block 282 according to the monitoring target. can be done.

In the example shown in FIG. 4, it is assumed that the function block library 280 and the attached data storage unit 290 are installed in the design tool device 200. FIG. However, function block library 280 and attached data storage unit 290 may be implemented in a server device that can be connected to design tool device 200 via a network.

FIG. 6 is a diagram showing an example of a function block for determining whether there is an abnormality in the monitored object. In the function block 282 of the example shown in FIG. 6, the state value (axis variable) to be monitored is input to the input terminal "Axis". When "TRUE" is input to the input terminal "Enable", the function block 282 is executed to determine whether or not there is an abnormality in the monitored object. The determination result is output from the output terminal "Error".

FIG. 7 is a diagram showing an example of a user program created using the function blocks shown in FIG. In the example user program shown in FIG. 7, when a judgment result indicating that there is an abnormality is output from the output terminal "Error" of the function block 282, the notification device 18 (see FIG. 2) indicated by "Error Lamp" lights up. .

By using the design tool device 200, the designer can create a user program using the function blocks 282 associated with the temporarily created attached data 292. FIG. Therefore, the designer can create the user program 130 as shown in FIG. 7 even before the field engineer adjusts the method of determining whether there is an abnormality in the monitored object. Furthermore, the designer can use the debugger 256 and the simulator 260 to correct bugs in the created user program 130 and to check the operation of the created user program 130 .

<2-4. Hardware configuration example of control device>
FIG. 8 is a block diagram showing a hardware configuration example of the control device according to this embodiment. The control device 100 of the example shown in FIG. , a USB (Universal Serial Bus) controller 112, a memory card interface 114, an internal bus controller 122, field bus controllers 118, 120, an I/O unit 124, .

The processor 102 reads various programs stored in the storage 108, develops them in the main memory 106, and executes them, thereby realizing the control calculation unit 10, the abnormality determination unit 20, and the update unit 30 (see FIG. 1). The chipset 104 controls data transmission between the processor 102 and each component.

The storage 108 contains a system program 150 for realizing basic functions, a user program 130 for controlling a controlled object, feature quantity designation information 43, a learning data set 41, and criterion information 42. Stored. The user program 130, the feature quantity designation information 43, the learning data set 41, and the criterion information 42 are provided from the design tool device 200 and stored in the storage 108 as described above.

The host network controller 110 controls exchange of data with other devices via the host network 6 . USB controller 112 controls data exchange with design tool device 200 or support device 300 via a USB connection.

The memory card interface 114 is configured such that a memory card 116 can be attached/detached, and data can be written to the memory card 116, and various data (user program, trace data, etc.) can be read from the memory card 116. ing.

The internal bus controller 122 is an interface that exchanges data with the I/O units 124, .

Fieldbus controller 118 controls data exchange with other devices via field network 2 . Similarly, the fieldbus controller 120 controls data exchange with other devices via the field network 3 .

FIG. 8 shows a configuration example in which necessary functions are provided by the processor 102 executing a program. (Application Specific Integrated Circuit) or FPGA (Field-Programmable Gate Array), etc.). Alternatively, the main part of the control device 100 may be implemented using hardware conforming to a general-purpose architecture (for example, an industrial personal computer based on a general-purpose personal computer). In this case, virtualization technology may be used to execute a plurality of OSs with different purposes in parallel, and necessary applications may be executed on each OS.

<2-5. Example of hardware configuration of support device>
Support device 300 according to the present embodiment is realized, for example, by executing a program using hardware conforming to a general-purpose architecture (for example, a general-purpose personal computer). From the standpoint of maintainability at the manufacturing site where the control device 100 is arranged, the support device 300 is preferably a portable notebook personal computer.

FIG. 9 is a block diagram showing a hardware configuration example of the support device 300 according to this embodiment. The support device 300 in the example shown in FIG. 9 includes a processor 302 such as a CPU or MPU, an optical drive 304, a main storage device 306, a secondary storage device 308, a USB controller 312, a local network controller 314, an input A portion 316 and a display portion 318 are included. These components are connected via internal bus 320 .

The processor 302 reads various programs stored in the secondary storage device 308, develops them in the main storage device 306, and executes them, thereby realizing various processing including analysis processing as described later.

The secondary storage device 308 is composed of, for example, an HDD (Hard Disk Drive) or an SSD (Flash Solid State Drive). The secondary storage device 308 typically contains a PLC interface program 324 for exchanging data related to the abnormality detection function with the control device 100, an analysis program 326 for realizing analysis processing, and an OS 328. is stored. The secondary storage device 308 may store necessary programs other than the programs shown in FIG.

The support device 300 has an optical drive 304 . A program stored therein is read from a recording medium 305 (for example, an optical recording medium such as a DVD (Digital Versatile Disc)) that stores a computer-readable program non-transitory, and is stored in a secondary storage device 308 or the like. installed on.

Various programs executed by the support device 300 may be installed via the computer-readable recording medium 305, or may be installed by being downloaded from a server device or the like on the network. Also, the functions provided by the support device 300 according to the present embodiment may be realized by using some of the modules provided by the OS.

The USB controller 312 controls data exchange with the control device 100 via a USB connection. Local network controller 314 controls the exchange of data to and from other devices over any network.

An input unit 316 is composed of a keyboard, a mouse, and the like, and receives user operations. A display unit 318 includes a display, various indicators, a printer, and the like, and outputs processing results from the processor 302 and the like.

FIG. 9 shows a configuration example in which necessary functions are provided by the processor 302 executing a program. Alternatively, it may be implemented using an FPGA, etc.).

<2-6. Software Configuration Example/Function Configuration Example of Control Device and Support Device>
Next, an example of software configuration and an example of functional configuration of main devices constituting the control system 1 according to the present embodiment will be described. FIG. 10 is a block diagram showing a main software configuration example of the control system according to this embodiment.

Referring to FIG. 10, control device 100 includes a user program 130, an updating unit 30, and a time series database (TSDB: Time Series Data Base) 140 as its main software configuration.

The control device 100 manages state values collected from the field in the form of variables 44, and the variables 44 are updated at predetermined intervals.

The user program 130 includes control operation instructions 132 for controlling the controlled object and instructions defined by function blocks 282 . As described above, the user program 130 is created by the designer 4 using the design tool device 200 and provided from the design tool device 200 to the control device 100 .

The control operation instruction 132 includes an operation instruction for controlling the controlled object. The control computation unit 10 shown in FIG. 1 is implemented by the processor 102 of the control device 100 executing the control computation instructions 132 .

The instructions defined in function block 282 include a series of instructions for determining whether there is an anomaly to be monitored. Abnormality determination unit 20 shown in FIG. 1 is implemented by processor 102 of control device 100 executing instructions defined by function block 282 . Therefore, by using the function block 282, the user program 130 that implements the abnormality determination unit 20 can be easily created. Instructions defined by the function block 282 include a feature generation instruction 134 , a write instruction 135 , a score calculation instruction 136 and a determination instruction 137 .

The feature quantity generation command 134 generates a feature quantity of a type specified by the feature quantity specification information 43 from the state value of the type specified by the feature quantity specification information 43 (e.g., average value, maximum value, minimum value over a predetermined period of time, etc.).

The write instruction 135 includes an instruction to write the generated feature quantity to the time series database 140 .

The feature amounts sequentially written to the time-series database 140 are output to the support device 300 as the feature amount data 45 .

The score calculation instruction 136 converts the distance (e.g., Euclidean distance) between the generated feature quantity and the feature quantity group indicated by the learning data set 41 into a score indicating the comparison result between the generated feature quantity and the learning data set 41. Includes instructions to calculate as

The judgment instruction 137 compares the calculated score with the threshold indicated by the judgment reference information 42, judges that there is an abnormality when the score exceeds the threshold, and judges that there is an abnormality when the score does not exceed the threshold. Includes instructions for judging that there is no abnormality.

The update unit 30 is implemented by the processor 102 of the control device 100 executing the system program 150 (see FIG. 8). Upon receiving update learning data set 51 and determination criterion information 52 from support device 300 , updating unit 30 updates learning data set 41 and determination criterion information 42 currently set in control device 100 . That is, the update unit 30 updates the contents of the currently set learning data set 41 to the contents of the learning data set 51 for updating. Furthermore, the update unit 30 updates the content of the currently set determination criterion information 42 to the content of the determination criterion information 52 for updating.

On the other hand, the support device 300 includes an analysis tool 330 and a PLC interface 332 as main functional components.

The analysis tool 330 is a determination unit that determines the update learning data set 51 and the criterion information 52 using the feature amount data 45 composed of the feature amount generated in the control device 100 according to the operation of the field engineer 5. works as Analysis tool 330 is typically implemented by processor 302 of support device 300 executing analysis program 326 .

The PLC interface 332 takes charge of the processing of acquiring the feature amount data 45 from the control device 100, the processing of transmitting the determined learning data set 51 for update and the determination criterion information 52 to the control device 100, and the like. PLC interface 332 is typically implemented by processor 302 of support device 300 executing PLC interface program 324 .

FIG. 11 is a block diagram outlining functional modules included in the analysis tool 330 shown in FIG. Referring to FIG. 11, analysis tool 330 of support device 300 includes user interface 340, file management module 350, screen generation module 360, analysis module 370, and analysis library 380 as main functional configurations. .

The user interface 340 accepts settings from the user and performs overall processing for providing various types of information to the user. As a specific implementation, the user interface 340 has a script engine 342, reads a setting file 344 containing a script describing necessary processing, and executes the set processing.

The file management module 350 includes a data input function 352 that reads data from a specified file or the like, and a data generation function 354 that generates a file containing generated data or the like.

The screen generation module 360 includes a line graph generation function 362 that generates a line graph based on input data, etc., and a parameter adjustment function 364 that changes various parameters in response to user operations. The line graph generation function 362 may update the line as the parameters are changed. The line graph generation function 362 and the parameter adjustment function 364 refer to the graph library 366 to perform necessary processing.

The analysis module 370 is a module that implements main processing of the analysis tool 330 and has a parameter determination function 376 . The parameter determination function 376 executes processing for determining a learning data set and a threshold value, which are parameters required for the abnormality determination processing.

Analysis library 380 includes libraries for functions included in analysis module 370 to perform processing. More specifically, analysis library 380 includes anomaly determination engine 386 utilized by parameter determination function 376 . The processing executed by the abnormality determination engine 386 is substantially the same as the processing executed according to the score calculation instruction 136 and determination instruction 137 (see FIG. 10) of the function block 282 defined in the user program 130. FIG.

In this way, the abnormality determination engine 386 of the support device 300 uses the feature quantity (feature quantity data 45) provided from the time-series database 140 of the control device 100 to substantially The same abnormality determination processing is executed.

In the control system 1 according to the present embodiment, both the control device 100 and the support device 300 are provided with an environment in which the same abnormality determination process can be realized. In such an environment, the abnormality determination processing in the control device 100 can be reproduced in the support device 300 , and as a result, the support device 300 can determine the abnormality determination processing to be executed in the control device 100 .

More specifically, the analysis module 370 of the support device 300 determines the learning data set 51 for updating and the determination criterion information 52 based on the determination result of the abnormality determination engine 386 included in the analysis library 380 .

<2-7. Overview of Abnormality Judgment Processing>
Next, an overview of abnormality determination processing employed by the control system according to the present embodiment will be described.

In the present embodiment, when monitoring target data is evaluated as an outlier in a statistically obtained data set, it is determined as an abnormal value.

FIG. 12 is a schematic diagram for explaining the basic idea of the abnormality determination process of the control system according to this embodiment. In the example shown in FIG. 12, the feature quantity specifying information 43 specifies each of n feature quantities. Referring to FIG. 12, first, positions corresponding to feature amounts 1, 2, 3, . . . Define the set of coordinate values for the plotted locations as the normal set.

Then, feature quantities 1, 2, 3, . (corresponding to “input value” in FIG. 12).

Finally, based on the degree of deviation of the input value from the normal value group on the n-dimensional space, it is determined whether there is an abnormality in the monitoring target at the sampling timing corresponding to the input value. The group of normal values in FIG. 12 corresponds to the "model" indicating the monitoring target.

Methods for judging abnormality based on the degree of deviation include a method for judging the presence or absence of an abnormality based on the shortest distance from each point to the normal value group (k-nearest neighbor method), and an iForest (isolation forest) method using a score calculated from the path length.

FIG. 13 is a schematic diagram outlining the procedure of abnormality determination processing of the control system 1 according to the present embodiment. Referring to FIG. 13, assume that a set of state values is collected from the monitored object at arbitrary sampling timing. The presence or absence of an abnormality in the monitored object is determined at this sampling timing.

First, among the plurality of state values that can be collected from the monitoring target, the state values 1, 2, 3, . . . , n of the type specified by the information 43 are generated.

Note that a plurality of feature amounts may be generated from the same state value. For convenience of explanation, a configuration using at least four feature amounts is shown, but at least one feature amount is sufficient in the abnormality determination process according to the present embodiment.

A score is then calculated from one or more feature quantities. Then, the calculated score is compared with the threshold indicated by the criterion information 42 (see FIG. 10) to determine whether or not there is an abnormality in the monitored object.

In the abnormality determination process according to the present embodiment, a feature amount is generated from time-series data of state values over a predetermined period (hereinafter also referred to as a "frame"), and the generated feature amount is used to Calculate the score.

FIG. 14 is a flowchart showing the procedure of abnormality determination processing executed by the control device according to the present embodiment. Each step shown in FIG. 14 is typically implemented by processor 102 of control device 100 executing a program (system program and user program).

Referring to FIG. 14, when a predetermined frame start condition is satisfied (YES in step S50), control device 100 starts collecting one or more predetermined state values (step S52). Thereafter, when a predetermined frame end condition is satisfied (YES in step S54), control device 100 selects the type of state value specified by feature amount specifying information 43 among the state values collected during the period of the frame. A feature amount of the type specified by the feature amount specifying information 43 is generated from the time-series data of (step S56). Then, the control device 100 calculates a score based on the generated one or more feature amounts (step S58).

More specifically, the score is calculated using one or more generated feature values and the currently set learning data set 41 .

Subsequently, the control device 100 determines whether or not the calculated score exceeds the threshold indicated by the currently set determination criterion information 42 (step S60). If the calculated score exceeds the threshold (YES in step S60), control device 100 notifies the monitored object of abnormality (step S62). If the calculated score does not exceed the threshold (NO in step S60), control device 100 determines that the monitored object is normal (step S64). Then, the processing after step S50 is repeated.

<2-8. Outline procedure of analysis processing>
Next, a schematic procedure of analysis processing according to the present embodiment will be described. FIG. 15 is a schematic diagram showing the contents of analysis processing according to the present embodiment. FIG. 16 is a schematic diagram visually showing an overview of the processes (a) to (c) shown in FIG.

Referring to FIG. 15, analysis processing according to the present embodiment mainly includes (a) data input processing, (b) visualization/labeling processing, and (c) threshold determination processing.

Referring to FIG. 16, feature amount data 45 representing feature amounts generated by control device 100 is provided to support device 300 ((a) data input processing). In the example shown in FIG. 16, the feature amount data 45 includes a cycle count indicating the number of times of processing in the monitoring target and feature amounts 1, 2, 3, . . . , n.

Subsequently, the support device 300 visualizes the feature amount and labels the set of feature amounts at each sampling timing ((b) visualization/labeling process). Visualization of the features is basically performed by the support device 300, and all or part of the labeling may be performed by the user.

As a more specific example, the user sets whether the state of the monitoring target is "normal" or "abnormal" for each sampling timing while referring to the feature amount visualized in the form of a graph or the like. .

Next, a learning data set 51 for updating is generated based on the set of feature values labeled "normal", and based on the learning data set 51 and the set of feature values at each sampling timing, , the score at each sampling timing is calculated. Then, referring to the calculated score, a threshold value for determining abnormality is determined ((c) determination of threshold value).

Through the above procedure, the update learning data set 51 and the criterion information 52 (information indicating the threshold value) are generated. The generated update learning data set 51 and criterion information 52 are provided from the support device 300 to the control device 100 . The control device 100 updates the contents of the currently set learning data set 41 and the criterion information 42 to the contents of the updating learning data set 51 and the criterion information 52 .

As described above, when the design tool device 200 creates the user program 130 that defines commands for determining the presence or absence of an abnormality in the monitored object, the provisionally determined temporary learning data set 41 and the determination criteria Information 42 is set in the control device 100 . Therefore, although the user program 130 can be executed, the presence or absence of abnormality in the monitoring target cannot be accurately determined because the provisional learning data set 41 and the determination criterion information 42 are used. However, support device 300 generates update learning data set 51 and determination criterion information 52 by performing analysis processing. That is, the field engineer can use the support device 300 to appropriately adjust the learning data set and threshold. As a result, the abnormality determination process is executed using the adjusted learning data set and the threshold value, and the accuracy of determining whether there is an abnormality in the monitoring target can be improved.

Note that the processes (a) to (c) shown in FIG. 16 can be repeatedly executed as appropriate, and it is also possible to sequentially update the learning data set and the criterion information according to the state of the monitored object. .

<2-9. Example of processing procedure of support device>
FIG. 17 is a flow chart showing an example of the processing procedure of the support device according to this embodiment. Referring to FIG. 17, support device 300 first activates analysis tool 330 in response to an operation by the field engineer (step S2), and causes analysis tool 330 to read feature amount data 45 (step S4).

Subsequently, the analysis tool 330 visualizes changes in feature amounts according to selection operations by the field engineer. The field engineer labels the visualized changes in feature amounts as normal and/or abnormal (step S6) (corresponding to the visualization/labeling process (b) in FIG. 15).

FIG. 18 is a schematic diagram showing an example of a user interface screen 500 provided to the field engineer in step S6 of FIG. Referring to FIG. 18, user interface screen 500 visualizes changes in feature amounts. Typically, on the user interface screen 500, changes in feature values over time are graphed, and the field engineer refers to this graph to set the abnormal or normal range. The abnormal range and normal range set by the field engineer may be set based on information on whether the monitoring target is actually abnormal or operating normally. You may make it set the change of a feature-value arbitrarily. That is, the abnormal range and the normal range set on the user interface screen 500 define the "abnormal" or "normal" state output by the abnormality determination process according to the present embodiment. It is not necessarily the same whether the data is abnormal or normal.

More specifically, user interface screen 500 includes a selection acceptance area 502 for feature quantities, a graph display area 506 and a histogram display area 512 .

The selection reception area 502 displays a list indicating feature amounts specified by the feature amount designation information 43 , and the user selects an arbitrary feature amount on the list displayed in the selection reception area 502 .

A graph display area 506 displays a graph 508 showing changes in the selected feature amount. The graph 508 may be divided by time-series data for each sampling, or by units of processing to be monitored (for example, units of processing work).

The histogram display area 512 displays a histogram showing the distribution of changes in the selected feature amount. By checking the histogram displayed in the histogram display area 512, it is possible to know the main range of the selected feature amount.

The user can set a normal range and/or an abnormal range of data for changes in the feature amount (graph 508) displayed in the graph display area 506. FIG. More specifically, user interface screen 500 includes labeling tool 514 . The labeling tool 514 includes a normal label setting button 516 , an abnormal label setting button 517 , and a label setting range specification button 518 .

The field engineer selects the normal label setting button 516 or the abnormal label setting button 517 according to whether the label to be assigned is normal or abnormal, then selects the label setting range specification button 518, and continues to display the graph. An operation (for example, a drag operation) is performed to specify a target area of the display area 506 . As a result, the specified label is assigned to the specified area.

FIG. 18 shows an example in which an abnormal range 510 is set. The feature amount of the sampling timing included in the abnormal range 510 is labeled as "abnormal", and the other feature amount is labeled as "normal". In this way, the analysis tool 330 has a function of assigning at least one of “normal” and “abnormal” labels to a specific range of data series of a plurality of feature quantities generated in accordance with user operations. may

A normal range and/or an abnormal range of data may be set for the histogram displayed in the histogram display area 512 as well.

Referring to FIG. 17 again, next, analysis tool 330 executes threshold value determination processing according to the operation of the field engineer (step S8). The process of step S8 corresponds to (c) threshold determination process shown in FIG.

In this step S8, a default threshold value is set in advance. The field engineer adjusts the threshold while checking the index value indicating the judgment accuracy.

FIG. 19 is a schematic diagram showing an example of a user interface screen 520 provided to the user in step S8 of FIG. Referring to FIG. 19, user interface screen 520 accepts selection of a threshold value to be used for abnormality determination processing.

More specifically, user interface screen 520 includes graph display area 526 . A graph display area 526 displays a graph 528 showing changes in scores calculated based on the feature amount at each sampling timing indicated by the feature amount data 45 and the learning data set. As the learning data set, feature amount data at a plurality of sampling timings labeled as "normal" in step S8 is used.

A threshold setting slider 534 is arranged in association with the graph display area 526 . In conjunction with the operation of the threshold setting slider 534, the set threshold varies and the position of the threshold display bar 535 displayed in the graph display area 526 also varies. In this way, threshold setting slider 534 accepts threshold setting for the score displayed in graph display area 526 .

In graph display area 526, numerical display 530 indicating the index value and numerical display 532 indicating the set threshold value are arranged. The values of the numerical displays 530 and 532 are also updated in conjunction with the user's operation on the threshold setting slider 534 .

The index value indicated by the numerical display 530 is a value that indicates how accurate determination is possible with the currently set threshold value. More specifically, according to the current settings, the index value is the feature quantity included in the normal range set by the user (that is, the feature quantity labeled as “normal”) and the feature quantity set by the user. It shows how accurately it can be distinguished from the feature quantity included in the abnormal range (that is, the feature quantity labeled “abnormal”).

The index value is calculated, for example, by obtaining an AUC (Area Under the Curve) based on an ROC (Receiver Operating Characteristic) curve defined by a TruePositive axis and a FalsePositive axis.

For example, using data labeled "normal" and data labeled "abnormal", a value obtained by subtracting the probability of misjudgment from 100% is calculated as an index value. An erroneous determination means that data labeled as "normal" is determined as "abnormal", or data labeled as "abnormal" is determined as "normal".

The field engineer appropriately sets the threshold while checking the shape of the graph 528 displayed in the graph display area 526 and the index value indicated by the numerical display 530 .

When reset button 536 of user interface screen 520 is selected, user interface screen 520 is reset to a default state.

When the update data generation button 538 of the user interface screen 520 is selected, the learning data set 51 for updating and the criterion information 52 are generated according to the contents set at that time.

Referring to FIG. 17 again, when user interface screen 500 shown in FIG. 18 and user interface screen 520 shown in FIG. YES), the analysis tool 330 generates the update learning data set 51 and the criterion information 52 (step S12). That is, the analysis tool 330 generates the updated learning data set 51 and the criterion information 52 according to the labeling and threshold adjustment by the field engineer. Specifically, the analysis tool 330 generates a learning data set 51 for updating, which uses the feature amount of each sampling timing labeled "normal" as learning data. In addition, analysis tool 330 generates updated criteria information 52 indicating the adjusted thresholds.

Then, the update learning data set 51 and the criterion information 52 generated in step S12 are transmitted from the support device 300 to the control device 100 (step S14), and the learning data set set in the control device 100 41 and criteria information 42 are updated.

In addition, the field engineer periodically checks the index value and the like, and if the index value is bad, the field engineer operates the support device 300 to change the learning data set 41 and the criterion information 42 set in the control device 100. It may be updated as appropriate. When the index value is bad, the frequency of detecting anomalies is relatively high even though the monitoring target is in a normal state, and/or the monitoring target is not in a normal state even though it is detected as normal. is relatively high.

<2-10. Variation>
In the above description, the type of feature quantity generated for each monitoring target is fixed. However, depending on the state of the object to be monitored, the type of feature quantity suitable for determining whether there is an abnormality may vary. Therefore, the feature designation information 43 may be updated along with the learning data set 41 and the criterion information 42 . After the provisional feature quantity designation information 43 provisionally determined at the time of creating the user program 130 is set in the control device, the field engineer can also update the contents of the feature quantity designation information 43 using the support device. The control system will be explained.

(2-10-1. Software configuration example/functional configuration example of control device and support device)
FIG. 20 is a block diagram showing a main software configuration example of a control system according to a modification. 20 differs from the control device 100 shown in FIG. 10 in that an updating unit 30A is included instead of the updating unit 30. In FIG. Further, the instructions defined in function block 282 include write instruction 135A instead of write instruction 135. FIG.

Write instructions 135A include instructions to write variables 44 (state values) collected from fields to time series database 140. FIG.

State values sequentially written to the time-series database 140 are output as the raw data 46 to the support device 300A.

In addition to the processing of the updating unit 30 shown in FIG. 10, the updating unit 30A, upon receiving the updating feature amount designation information 53 from the support device 300, updates the feature amount designation information 43 currently set in the control device 100A. process. That is, similar to the updating unit 30, the updating unit 30A updates the contents of the currently set learning data set 41 and the determination criterion information 42 to the contents of the learning data set 51 for updating and the determination criterion information 52, respectively. . Furthermore, the update unit 30A updates the content of the currently set feature amount designation information 43 to the content of the feature amount designation information 53 for updating.

20 differs from the support device 300 shown in FIG. 10 in that an analysis tool 330A is included instead of the analysis tool 330. The support device 300A shown in FIG.

Analysis tool 330A analyzes raw data 46 consisting of state values collected in control device 100A according to operation of field engineer 5, learning data set 51 for updating, criterion information 52, and feature quantity designation information 53. to decide.

FIG. 21 is a block diagram showing an overview of functional modules included in analysis tool 330A shown in FIG. Referring to FIG. 21, analysis tool 330A differs from analysis tool 330 shown in FIG. 11 in that it includes analysis module 370A and analysis library 380A instead of analysis module 370 and analysis library 380, respectively. The analysis module 370A differs from the analysis module 370 in that it further includes a feature quantity generation function 372 and a feature quantity selection function 374 . The analysis library 380A is different in that it further includes a feature quantity generation library 382 used by the feature quantity generation function 372 and a feature quantity selection library 384 used by the feature quantity selection function 374 .

A feature quantity generation function 372 generates a feature quantity from time-series data of arbitrary state values included in the raw data 46 . The feature amount generation library 382 stores algorithms for generating various types of feature amounts (for example, average values, maximum values, minimum values, etc. over a predetermined period of time).

The feature quantity selection function 374 executes a process of selecting a feature quantity to be used in the abnormality determination process and a process of receiving the selection of the feature quantity.

(2-10-2. Outline procedure of analysis processing)
FIG. 22 is a schematic diagram visually showing an overview of analysis processing according to the modification. Referring to FIG. 22, the analysis processing according to the modification mainly includes (a) data input processing, (b) feature amount generation processing, (c) visualization/labeling processing, (d) feature amount selection processing, ( e) including a threshold determination process;

Referring to FIG. 22, raw data 46, which is time-series data of state values collected by control device 100A, is provided to support device 300A ((a) data input processing). Raw data 46 includes one or more state values at each sampling timing.

The support device 300A uses the input raw data 46 to generate one or more feature amounts ((b) feature amount generation). Generation of this feature quantity is achieved using the feature quantity generation function 372 of the analysis tool 330A. Usually, multiple types of feature quantities are generated. In the example shown in FIG. 22, feature amounts 1, 2, 3, . . . , N are generated.

Subsequently, the support device 300A visualizes the feature amount and labels the set of feature amounts at each sampling timing ((c) visualization/labeling process). The (c) visualization/labeling process according to the modification is described in <2-8. This is the same as the (b) visualization/labeling process described in General procedure of analysis process>.

Subsequently, one or a plurality of feature amounts to be used for abnormality determination are selected from among the plurality of feature amounts generated from the collected state values ((d) feature amount selection). In the example shown in FIG. 22, four feature amounts 1, 2, k, and n are selected.

Next, a learning data set 51 for updating is generated based on the set of features labeled as "normal" and selected. Furthermore, a score at each sampling timing is calculated based on the learning data set 51 and the set of feature amounts selected at each sampling timing. Then, referring to the calculated score, a threshold value for determining abnormality is determined ((e) determination of threshold value).

Through the procedure described above, the update learning data set 51, the criterion information 52 (information indicating the threshold value), and the feature quantity designation information 53 are generated. The generated update learning data set 51, criterion information 52, and feature quantity designation information 53 are provided from the support device 300A to the control device 100A. The control device 100A updates the contents of the currently set learning data set, criterion information and feature quantity designation information to the contents of the learning data set 51 for updating, the criterion information 52 and the feature quantity designation information 53 .

(2-10-3. Example of processing procedure of support device)
FIG. 23 is a flowchart illustrating a processing procedure of the support device according to the modification; Compared to the flowchart shown in FIG. 17, the flowchart shown in FIG. 23 further includes steps S30 and S32, and steps S104, S108, S112 and S114 instead of steps S4, S8, S12 and S14. differ.

Referring to FIG. 23, support device 300A activates analysis tool 330A in response to the field engineer's operation (step S2), and causes analysis tool 330A to read raw data 46 (step S104).

Subsequently, the analysis tool 330A performs data cleansing on the read raw data 46 (step S30). Data cleansing is a process of deleting unnecessary data included in the raw data 46 . For example, among the time-series data included in the raw data 46, state values with zero variance (that is, state values that do not fluctuate at all) are deleted. The data cleansing process may be automatically executed by the analysis tool 330A, or the analysis tool 330A may present candidate status values to be deleted, and the field engineer may explicitly select the deletion target. good.

Furthermore, the visualized state values and the like may be referred to so that the field engineer can manually delete the state values determined to be unnecessary or incorrect data. That is, the support device 300 accepts the selection of state values to be excluded from the generation of the feature amount from the raw data 46 .

After that, the analysis tool 330A generates one or more feature amounts based on the state values included in the raw data 46 after data cleansing (step S32). More specifically, feature quantity generation function 372 of analysis tool 330A generates a plurality of feature quantities from raw data 46 . In step S32, as many types of feature values as possible may be generated (corresponding to (b) feature value generation in FIG. 22).

Subsequently, the analysis tool 330A visualizes changes in the feature amount according to the field engineer's selection operation, and labels the visualized feature amount changes as normal and/or abnormal according to the field engineer's operation. (step S6) (corresponding to (c) visualization/labeling process in FIG. 22). Regarding step S6, <2-9. The detailed description is omitted here, since the description has been given in the example of the processing procedure of the support device).

Next, the analysis tool 330A determines the feature amount and the threshold value used for abnormality determination according to the operation of the field engineer (step S108). The process of step S108 corresponds to (d) feature amount selection process and (e) threshold value determination process shown in FIG.

FIG. 24 is a schematic diagram showing an example of a user interface screen provided by the support device 300A to the field engineer in step S108 of FIG. Referring to FIG. 24, user interface screen 520A mainly accepts selection of one or a plurality of feature quantities to be used for abnormality determination processing, and also accepts selection of threshold values to be used for abnormality determination processing.

More specifically, user interface screen 520A includes a selection acceptance area 522 for feature quantities and a graph display area 526 .

The selection reception area 522 corresponds to a user interface that receives selection of one or more feature amounts used in the abnormality determination process from among the plurality of generated feature amounts. More specifically, the selection reception area 522 displays a list showing the content of the generated feature amount, and the field engineer selects a check box 524 corresponding to an arbitrary feature amount on the displayed list. By checking, the feature amount used for abnormality determination processing is selected.

The feature quantities displayed in the selection reception area 522 are arranged so that those estimated to have high importance are ranked higher based on the results of analysis by the feature quantity selection function 374 (see FIG. 21) of the analysis tool 330A. may be listed in That is, in the selection reception area 522, the display order of the plurality of generated feature amounts may be determined according to the rank determined by the procedure described later.

Also, in the initial selection reception area 522, a feature amount pre-selected by the feature amount selection function 374 may be selected as a default. That is, in the selection receiving area 522, a predetermined number of feature amounts among the plurality of generated feature amounts may be selected and displayed in accordance with the determined rank. FIG. 24 shows a state in which two feature quantities are selected.

Since the details of the graph display area 526 are as described above with reference to FIG. 19, detailed description thereof will be omitted. Graph 528 displayed in graph display area 526 indicates changes in score calculated based on one or more feature amounts selected by checking check boxes 524 in selection reception area 522 .

When update data generation button 538 on user interface screen 520A is operated (YES in step S10), analysis tool 330A updates learning data set 51, criterion information 52, and The feature quantity designation information 53 is generated (step S112). That is, the analysis tool 330A generates feature quantity designation information 53 for designating one or a plurality of feature quantities selected by checking the check box 524 shown in FIG. Furthermore, the analysis tool 330A generates a learning data set 51 for updating, which uses, as learning data, the feature amount specified by the feature amount specifying information 53 among the feature amounts at each sampling timing labeled "normal". . In addition, analysis tool 330A generates updated criteria information 52 indicating the adjusted thresholds.

Then, the update learning data set 51, the criterion information 52, and the feature amount designation information 53 generated in step S112 are transmitted from the support device 300A to the control device 100A (step S114), and set in the control device 100A. The learning data set 41, the criterion information 42, and the feature quantity designation information 43 are updated.

(2-10-4. Calculation and ranking of importance of feature amount)
FIG. 25 is a schematic diagram for explaining the process of evaluating the importance of the feature quantity executed by the analysis tool of the support device according to the modification. Referring to FIG. 25, feature quantity selection function 374 of analysis tool 330A calculates the importance of each feature quantity using a plurality of methods. FIG. 25 shows, as an example, an example of evaluation by three methods of kurtosis, logistic regression, and decision tree.

The kurtosis stored in the evaluation value column 702 is a value obtained by evaluating the sharpness of the frequency distribution for the data series of the target feature quantity 700 . The higher the kurtosis, the sharper the peak of the frequency distribution and the broader the tail of the distribution. As a statistic used for anomaly detection, the higher the kurtosis, the more useful, that is, the more important.

Alternatively, the standard deviation of the frequency distribution of the data series of the target feature amount may be used as the evaluation value. In this case, it can be determined that the larger the standard deviation, the more the feature amount changes and the higher the abnormality detection capability (that is, the more important).

The logistic regression stored in the evaluation value column 704 applies an arbitrary logistic function to the target feature data series to search for parameters that define the logistic function that maximizes the likelihood. The likelihood corresponding to the finally searched parameters is regarded as the importance. In other words, it can be considered that a feature amount that can be estimated with higher accuracy by an arbitrary logistic function has a higher priority.

Typically, logistic regression can perform parameter search and likelihood calculation for each feature amount.

The decision tree stored in the evaluation value column 706 applies a classification tree to the data series of the feature amount of interest, and uses its classification ability as the importance. CART, C4.5, ID3, and the like are known as decision tree algorithms, and any algorithm may be used.

Thus, typically, the importance is calculated according to at least the kurtosis of the feature data series, the likelihood obtained by performing logistic regression on the feature data series, and the decision tree algorithm. including the importance of

As described above, the value indicating the degree of importance of each feature amount is calculated by a plurality of methods, and the result of integrating the respective results is stored in the evaluation value column 708 . Based on each evaluation value stored in the evaluation value column 708, each feature amount is ranked (rank column 710).

<2-11. Note>
As described above, the present embodiment includes the following disclosures.

(Configuration 1)
a control calculation unit (10) for executing control calculations for controlling the controlled object using state values periodically collected from the controlled object;
calculating a score indicating a comparison result between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set (101); an abnormality determination unit (20) for determining the presence or absence of an abnormality in the monitored object based on the determination criteria (102),
The abnormality determination unit (20) is realized by defining commands in a user program (130) for controlling the controlled object,
An update unit (30) that receives an update learning data set (301) and an update criterion (302) from the outside and updates the currently set learning data set (101) and the criterion (102), respectively. A controller (100, 100A) further comprising:

(Configuration 2)
The generated feature amount is predetermined for each monitoring target,
The control device (100, 100A) according to configuration 1, wherein a learning data set (101) and a criterion (102) are set according to the generated feature amount.

(Composition 3)
The control device (100, 100A) according to configuration 1, wherein the abnormality determination unit (20) is realized by defining commands in the form of function blocks (282) in the user program (130).

(Composition 4)
The function block (282), the provisional learning data set (101), and the provisional criterion (102) are associated with each monitoring target and created in advance,
The control device according to configuration 3, wherein the provisional learning data set (101) and the provisional criterion (102) corresponding to the function block (282) are set by defining the function block (282). (100, 100A).

(Composition 5)
The abnormality determination unit (20) generates the feature amount according to designation information (103) that designates the type of state value and the type of feature amount,
The control device (100A) according to configuration 1, wherein the update unit (30) receives update designation information (303) from the outside and updates currently set designation information (103).

(Composition 6)
A control system (1),
a control calculation unit (10) for executing control calculations for controlling the controlled object using state values periodically collected from the controlled object;
calculating a score indicating a comparison result between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set (101); an abnormality determination unit (20) for determining the presence or absence of an abnormality in the monitored object based on the determination criteria (102),
The abnormality determination unit (20) is implemented by being defined in a user program (130) for controlling the controlled object,
The control system (1) further comprises:
a determination unit (330, 330A) that determines a learning data set (301) for update and a determination criterion (302) for update using the state value related to the monitoring target or the generated feature quantity;
An update unit (30) that receives the update learning data set (301) and the update criterion (302) and updates the currently set learning data set (101) and the criterion (102), respectively. A control system (10) comprising:

(Composition 7)
Further comprising a program creation unit (200) for creating the user program (130),
The program creation unit (200)
Creating the user program (130) using a function block (282) defining instructions for realizing the abnormality determination unit (20),
7. Control system (1) according to configuration 6, wherein a provisional training data set (101) and a provisional criterion (102) associated with said function block (282) are established.

Although the embodiments of the present invention have been described, the embodiments disclosed this time should be considered as illustrative in all respects and not restrictive. The scope of the present invention is indicated by the scope of claims, and is intended to include all modifications within the meaning and scope of equivalence to the scope of claims.

1 control system 2, 3 field network 4 designer 5 field engineer 6 host network 8 optical recording medium 10 control arithmetic unit 16, 124 I/O unit 18 notification device 20 abnormality determination unit 30 , 30A update unit, 41, 41a, 41b, 51 learning data set, 42, 42a, 42b, 52 criterion information, 43, 43a, 43b, 53 feature amount designation information, 45 feature amount data, 46 raw data, 100, 100A control device, 102,302 processor, 104 chipset, 106 main memory, 108 storage, 110 upper network controller, 112,312 USB controller, 114 memory card interface, 116 memory card, 118,120 fieldbus controller, 122 internal bus controller 130 user program, 132 control operation instruction, 134 feature amount generation instruction, 135, 135A write instruction, 136 score calculation instruction, 137 determination instruction, 140 time series database, 150 system program, 200 design tool device, 201 storage unit, 203 operation unit, 206 RAM, 210 keyboard, 212 mouse, 214 display, 216 optical recording medium reader, 218 communication interface, 220, 320 internal bus, 250 programming application, 252 editor, 254 compiler, 256 debugger, 258 GUI module, 260 simulator, 270 data storage section, 280 function block library, 282, 282a, 282b function block, 290 accessory data storage section, 292, 292a, 292b accessory data, 300, 300A support device, 304 optical drive, 305 recording medium, 306 Main storage device, 308 secondary storage device, 314 local network controller, 316 input unit, 318 display unit, 324 PLC interface program, 326 analysis program, 330, 330A analysis tool, 332 PLC interface, 340 user interface, 342 script engine, 344 setting file, 350 file management module, 352 data input function, 354 data generation function, 360 screen generation module, 362 line graph generation function, 3 64 parameter adjustment function, 366 graph library, 370, 370A analysis module, 372 feature quantity generation function, 374 feature quantity selection function, 376 parameter determination function, 380, 380A analysis library, 382 feature quantity generation library, 384 feature quantity selection library, 386 abnormality determination engine, 400 display device, 500, 520, 520A user interface screen, 502, 522 selection reception area, 506, 526 graph display area, 508, 528 graph, 510 abnormality range, 512 histogram display area, 514 labeling tool , 516 Normal label setting button, 517 Abnormal label setting button, 518 Label setting range designation button, 524 Check box, 530, 532 Numeric display, 534 Setting slider, 535 Display bar, 536 Reset button, 538 Update data generation button, 618, 628 Servo Motor, 619, 629 Servo Driver.

Claims (6)

A control system,
a control calculation unit that executes control calculations for controlling the controlled object using state values periodically collected from the controlled object;
calculating a score indicating a comparison result between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set, and determining the calculated score and the preset criterion; and an abnormality determination unit that determines the presence or absence of an abnormality in the monitoring target based on,
The abnormality determination unit is realized by defining commands in a user program for controlling the controlled object,
The control system further comprises:
a determination unit that determines a learning data set for updating and a determination criterion for updating using the state value related to the monitoring target or the generated feature amount;
an updating unit that receives the update learning data set and the update determination criteria and updates the currently set learning data set and the determination criteria, respectively;
The decision unit
providing a user interface screen including a graph showing temporal changes in the generated feature quantity;
Receiving the setting of the normal range in the graph,
Determining the feature amount included in the normal range as the update learning data set ,
The decision unit
Further accepting the setting of the abnormal range in the graph,
receiving the update criteria on the user interface screen;
A control system that displays, on the user interface screen, an index value indicating determination accuracy between the feature amount included in the normal range and the feature amount included in the abnormal range when using the update determination criteria . The generated feature amount is predetermined for each monitoring target,
2. The control system according to claim 1, wherein a learning data set and a criterion are set according to said generated feature amount. 2. The control system according to claim 1 , wherein said abnormality determination section is implemented by defining commands in the form of function blocks in said user program. The function block, the provisional learning data set, and the provisional determination criteria are associated with each monitoring target and created in advance,
4. The control system of claim 3 , wherein defining the function block establishes a provisional learning data set and provisional criteria for the function block. A control system,
a control calculation unit that executes control calculations for controlling the controlled object using state values periodically collected from the controlled object;
calculating a score indicating a comparison result between a feature amount generated from the state values related to the monitored object among the collected state values and a preset learning data set, and determining the calculated score and the preset criterion; and an abnormality determination unit that determines the presence or absence of an abnormality in the monitoring target based on,
The abnormality determination unit is realized by defining commands in a user program for controlling the controlled object,
The control system further comprises:
a determination unit that determines a learning data set for updating and a determination criterion for updating using the state value related to the monitoring target or the generated feature amount;
an updating unit that receives the update learning data set and the update determination criteria and updates the currently set learning data set and the determination criteria, respectively;
The decision unit
providing a user interface screen including a graph showing temporal changes in the generated feature quantity;
Receiving the setting of the normal range in the graph,
Determining the feature amount included in the normal range as the update learning data set,
The abnormality determination unit generates the feature amount according to designation information that designates the type of the state value and the type of the feature amount,
The decision unit
displaying, on the user interface screen, a graph showing temporal changes in the score calculated based on a feature amount selected from among a plurality of types of feature amounts;
generating specification information for updating in response to receiving specification of a feature quantity used by the abnormality determination unit from among the plurality of types of feature quantity;
The control system, wherein the update unit receives the designation information for updating and updates currently set designation information. further comprising a program creation unit that creates the user program,
The program creation unit
Creating the user program using a function block that defines instructions for realizing the abnormality determination unit,
6. The control system according to claim 1 , wherein a provisional learning data set and a provisional criterion associated with said function block are set.

Images