Device state monitoring apparatus and device state monitoring method

Abstract

An apparatus state monitoring device (1) includes a feature quantity extraction section (11) that extracts a feature quantity of operation data obtained by measuring a state of an apparatus, an operation mode determination section (12) that determines whether an operation mode of the apparatus at the time of measuring the operation data of the apparatus is a learning completion mode in which a determination range of the state of the apparatus has been learned or an unlearned mode in which learning has not been performed, a feature quantity correction section (13) that corrects the feature quantity of the operation data corresponding to the operation mode determined as the unlearned mode in a manner corresponding to the learning completion mode, based on a relationship between the operation mode of the apparatus and the feature quantity of the operation data, and an apparatus state determination section (14) that determines the state of the apparatus based on the feature quantity of the operation data of the apparatus and the determination range of the state of the apparatus.

Description

Technical Field

[0001] This invention relates to an equipment status monitoring device and a method for monitoring equipment status. Background Technology

[0002] As a prior art for monitoring the state of equipment, there exists a technique that calculates the normal range of the equipment's state based on operating data obtained from measuring the normal state of the equipment, and monitors the equipment's state based on the degree of deviation from the normal range. For example, Patent Document 1 describes a diagnostic device for a workshop, which diagnoses the workshop as being in a normal state when the measurement signal obtained from measuring the workshop's state quantities is classified as a normal model, and diagnoses the workshop as being in an unknown state that has never been experienced before when the measurement signal is not classified as a normal model.

[0003] Patent Document 1: International Publication No. 2012 / 073289 Summary of the Invention

[0004] The diagnostic device for a workshop described in Patent Document 1 diagnoses the workshop as being in an unknown state when the measured signal is not classified as falling within the normal range of the workshop's state. Therefore, for example, if the equipment's operating data is not classified as a previously learned state, determining it to be in an unknown state presents the problem that it is impossible to determine whether the equipment is in a normal state, an abnormal state, or a state showing signs of abnormality.

[0005] The present invention was proposed to solve the above-mentioned problems. Its purpose is to obtain an equipment status monitoring device and equipment status monitoring method that can determine the status of the equipment even when the determination range of the equipment status uses operating data corresponding to the unlearned operating mode.

[0006] The device status monitoring apparatus of the present invention comprises: a feature quantity extraction unit that extracts feature quantities of operating data obtained by measuring the status of the device; an operating mode determination unit that determines whether the operating mode of the device when the operating data of the device is measured is a learned mode (learned and completed mode) or an unlearned mode (not learned mode); a feature quantity correction unit that, based on the relationship between the operating mode of the device and the feature quantities of the operating data, performs correction until the distribution of feature quantities of the operating data corresponding to the operating mode determined to be an unlearned mode is close to or overlaps with the distribution of feature quantities of the operating data corresponding to the learned mode; and a device status determination unit that determines the status of the device based on the feature quantities of the operating data of the device and the determination range of the device status.

[0007] The effects of the invention

[0008] According to the present invention, based on the relationship between the operating mode of the device and the feature quantities of the operating data, the feature quantities of the operating data of the device corresponding to the unlearned mode are corrected in a manner corresponding to the learning completion mode. The device state is then determined based on the feature quantities of the corrected operating data and the determination range of the device state. Therefore, the device state monitoring device according to the present invention can determine the device state even if the determination range of the device state uses operating data corresponding to the unlearned operating mode. Attached Figure Description

[0009] Figure 1 This is a block diagram showing the structure of the device status monitoring device involved in Embodiment 1.

[0010] Figure 2 This is a flowchart illustrating the device status monitoring method involved in Implementation 1.

[0011] Figure 3 It is a summary diagram that represents the distribution of characteristic quantities of equipment operation data and the range of equipment status determination.

[0012] Figure 4 This is a flowchart of example (1) of the process of correcting the feature quantities of the operating data corresponding to the unlearned mode.

[0013] Figure 5 It is a graph that represents the relationship between the operating mode command value and the characteristic quantities of the operating data of the equipment.

[0014] Figure 6 It is a graphical representation of the correction process for test data in the relationship between the characteristic quantities of the equipment's operating mode command values ​​and operating data.

[0015] Figure 7 This is a schematic diagram illustrating the process of correcting the difference between the feature distribution of the operating data corresponding to the unlearned mode and the feature distribution of the operating data corresponding to the learned mode.

[0016] Figure 8 This is a flowchart of example (2) showing the process of correcting the feature quantities of the operating data corresponding to the unlearned mode.

[0017] Figure 9A It is a graph representing the distribution of operational data calculated using the correction process (1) of the physical model. Figure 9B It is a graph representing the distribution of operational data calculated using the correction process (2) of the physical model. Figure 9CIt is a graph representing the distribution of operational data calculated using the correction process (3) of the physical model. Figure 9D It is a graph representing the distribution of operational data calculated using the correction process (4) of the physical model.

[0018] Figure 10A This is a block diagram illustrating the hardware structure for implementing the device status monitoring device described in Embodiment 1. Figure 10B This is a block diagram representing the hardware structure of the software that executes the functions of the device status monitoring device involved in Implementation Method 1.

[0019] Figure 11 This is a block diagram showing the structure of a modified example of the device status monitoring device according to Embodiment 1. Detailed Implementation

[0020] Implementation method 1.

[0021] Figure 1 This is a block diagram illustrating the structure of the equipment status monitoring device according to Embodiment 1. Figure 1 In this system, the equipment status monitoring device 1 monitors the equipment status using operational data obtained by measuring the equipment's status through sensors installed on the equipment. The monitored equipment is a device that repeatedly performs a series of actions indicated by a specified operating mode, such as an industrial robot. The operating mode is a predetermined series of actions, executed by setting command values ​​representing each action (e.g., acceleration, deceleration, or constant speed) to the equipment. Furthermore, the command values ​​for the operating mode may include, for example, command speed, command position, or command load.

[0022] Operational data measured from equipment operating in a certain operating mode is time-series data of the equipment's state measurements, and a physical relationship exists between it and the command values ​​of the operating mode. For example, if the monitored equipment is an industrial robot with a rotating mechanism, and the robot operates in an operating mode that rotates the rotating mechanism at a constant speed, the relationship between the commanded speed value that rotates the rotating mechanism at the constant speed and the average torque of the rotating mechanism rotating at that commanded speed value can be represented by a monotonically increasing function. Among the characteristic quantities of the operational data, there are general statistics such as the average, minimum, maximum, dispersion, or standard deviation of the measured values ​​shown in the operational data, or a power spectrum obtained by performing a high-speed Fourier transform (FFT).

[0023] As described above, the equipment status monitoring device 1 is effective for monitoring the status of equipment that exhibits a physical relationship between operating modes and characteristic quantities of operating data. Existing methods for monitoring equipment status typically use operating data obtained by measuring the equipment status as learning data to learn the normal range, abnormal range, and abnormality warning range of the equipment status, and determine the equipment status based on which range the characteristic quantity of the operating data (e.g., average value) belongs to.

[0024] In control equipment such as industrial robots, the operating mode is sometimes changed by altering the products manufactured by the equipment or by changing their specifications. In this case, the operating data measured to monitor the state of the equipment operating under the changed operating mode may not fall within any pre-learned range. In existing methods, if the operating data is not classified into the learned range as described above, it may be determined that the equipment is in an unknown state, or even if the equipment is in a normal state, it may be mistakenly determined to be in an abnormal state.

[0025] Therefore, the equipment status monitoring device 1 focuses on the situation where a physical relationship exists between the command value of the equipment's operating mode and the characteristic quantity of the operating data. Thus, even in an operating mode different from the learned operation mode, such as an operating mode where the equipment's state determination range is an unlearned operating mode (hereinafter referred to as the unlearned mode), the characteristic quantity of the corresponding operating data can be corrected in a manner corresponding to an operating mode where the equipment's state determination range is a learned operation mode (hereinafter referred to as the learned operation mode). Therefore, the equipment status monitoring device 1 can determine the equipment's state based on the characteristic quantity of the operating data corresponding to the unlearned mode and the equipment's state determination range.

[0026] The equipment status monitoring device 1 generates a learning model for each operating mode shown in the operating mode information. This model is obtained by learning the range of equipment status determination using the operating mode information contained in the learning data and the corresponding operating data. For example, a One-Class SVM is used to calculate the determination range. The equipment status monitoring device 1 selects the learning model corresponding to the operating mode contained in the test data from the generated learning models, inputs the feature values ​​of the operating data into the selected learning model, and thereby determines the equipment status shown by the operating data. The test data consists of the operating data measured by sensors from the monitored equipment and the corresponding operating mode information.

[0027] Furthermore, when the equipment status monitoring device 1 determines that the operating mode included in the test data is an unlearned mode, it corrects the feature quantities of the operating data corresponding to the unlearned mode based on the relationship between the equipment's operating mode and the feature quantities of the operating data, in a manner corresponding to the learned mode. Moreover, the equipment status monitoring device 1 determines the status of the equipment based on the corrected feature quantities of the operating data and the determination range of the equipment's status.

[0028] Equipment status monitoring device 1, such as Figure 1 As shown, the device includes a feature extraction unit 11, an operation mode determination unit 12, a feature correction unit 13, and a device status determination unit 14. The feature extraction unit 11 extracts feature quantities from the operation data obtained by measuring the device's status. For example, the feature extraction unit 11 inputs operation data measured by sensors from the device in units of a certain measurement cycle, and calculates the feature quantities of the input operation data for each measurement cycle. The feature quantities of the operation data may be, for example, statistical quantities such as the average, minimum, maximum, or dispersion of the operation data measured within the measurement cycle, or a power spectrum obtained by performing an FFT.

[0029] The operation mode determination unit 12 determines the operation mode of the equipment when the operation data of the equipment is measured, which is either a learned mode that has learned the range of equipment state determination or an unlearned mode that has not been learned. For example, the operation mode determination unit 12 compares the operation mode information contained in the test data with the operation mode information contained in the learning data, and thereby determines the operation mode information in the test data that does not match the operation mode information contained in the learning data as an unlearned mode.

[0030] The feature quantity correction unit 13 corrects the feature quantities of the operating data corresponding to the operating mode determined to be an unlearned mode, based on the relationship between the device's operating mode and the feature quantities of the operating data, in a manner corresponding to the learning completion mode. For example, the feature quantity correction unit 13 uses test data and learning data to learn the relationship between the device's operating mode and the feature quantities of the operating data. Based on the learned relationship, the feature quantity correction unit 13 corrects the feature quantities of the operating data corresponding to the operating mode determined to be an unlearned mode, in a manner corresponding to the learning completion mode. Alternatively, the feature quantity correction unit 13 can also use the device's physical model to estimate the operating data of the device in the unlearned mode, and correct the estimated operating data feature quantities in a manner corresponding to the learning completion mode, based on the relationship between the learning completion mode and the feature quantities of the operating data.

[0031] The equipment status determination unit 14 determines the status of the equipment based on the characteristic quantities of the equipment's operating data and the determination range of the equipment's status. For example, the equipment status determination unit 14 acquires a learning model that has been pre-learned regarding the determination range of the equipment's status, and inputs the equipment's operating data included in the test data into the acquired learning model. The learning model determines whether the equipment status shown by the input operating data belongs to the normal range, the abnormal range, or the range of signs of abnormality. The equipment status determination unit 14 outputs the determination result of the equipment status obtained through the learning model.

[0032] The device status monitoring method involved in Implementation 1 is as follows.

[0033] Figure 2 This is a flowchart illustrating the equipment status monitoring method according to Embodiment 1, showing a series of processes performed by the equipment status monitoring device 1. First, the feature extraction unit 11 extracts feature quantities from the operating data obtained by measuring the equipment status (step ST1). For example, the feature extraction unit 11 inputs the equipment operating data included in the test data and calculates the feature quantities of the input operating data for each measurement cycle.

[0034] The operation mode determination unit 12 determines whether the operation mode contained in the test data is an unlearned mode (step ST2). If it is determined that the operation mode contained in the test data is a learned mode (step ST2; NO), the equipment status monitoring device 1 proceeds to step ST4. Alternatively, if it is determined that the operation mode contained in the test data is an unlearned mode (step ST2; YES), the feature quantity correction unit 13 corrects the feature quantities of the operation data corresponding to the operation mode determined to be an unlearned mode based on the relationship between the equipment's operation mode and the feature quantities of the operation data, in a manner corresponding to the learned mode (step ST3).

[0035] The device status determination unit 14 determines the device status based on the characteristic values ​​of the device's operating data and the determination range of the device's status (step ST4). For example, if it is determined that the operating mode included in the test data is a learning completion mode, the device status determination unit 14 inputs the characteristic values ​​of the operating data corresponding to that operating mode into the learning model. The learning model determines whether the device status shown by the input operating data belongs to the normal range, the abnormal range, or the range of signs of abnormality. Furthermore, if it is determined that the operating mode included in the test data is an unlearned mode, the characteristic values ​​of the corrected operating data are input into the learning model to determine the device status.

[0036] Figure 3 It is a summary diagram representing the distribution of characteristic quantities of equipment operating data and the range for determining the equipment's state. Figure 3 In this context, characteristic quantity (1) and characteristic quantity (2) are characteristic quantities of operating data measured from equipment that operates through a common operating mode. For example, if the operating data is the torque of a rotating mechanism, then characteristic quantity (1) can be the average value of the torque, and characteristic quantity (2) can be the standard deviation of the torque. Ranges A, B, and C are the ranges for determining the state of the equipment. Range A shows the normal range of the equipment, range B shows the range of signs that the equipment will become abnormal, and range C shows the abnormal range of the equipment.

[0037] Ranges A, B, and C are pre-learned using training data. For example, the characteristic value da of operating data measured from equipment in normal condition belongs to range A. The characteristic value db of operating data measured from equipment indicating a precursor to an abnormal state belongs to range B. The characteristic value dc of operating data measured from equipment in an abnormal state belongs to range C.

[0038] If the characteristic quantity d1 of the operating data obtained as test data does not belong to any of the ranges A, B, and C, the operating mode determination unit 12 determines that the operating mode corresponding to the characteristic quantity d1 of the operating data is an unlearned mode. In this case, the characteristic quantity correction unit 13 corrects the characteristic quantity d1 of the operating data in such a way that it belongs to any of the ranges A, B, and C corresponding to the learned mode. For example, based on the relationship between the operating mode of the device and the characteristic quantity of the operating data, the characteristic quantity correction unit 13 determines that the distance between the characteristic quantity d1 of the operating data and range B is the shortest, and corrects the characteristic quantity d1 of the operating data to the characteristic quantity d2 of the operating data within range B. As a result, the device that obtained the characteristic quantity d1 of the operating data is determined to be in a state of impending abnormality.

[0039] The details of the process for correcting the characteristic quantities of the equipment's operating data are as follows.

[0040] Figure 4 This is a flowchart illustrating an example (1) of the process for correcting the characteristic quantities of operating data corresponding to the unlearned mode, showing a series of processes performed by the characteristic quantity correction unit 13. The characteristic quantity correction unit 13 learns the relationship between the operating mode of the equipment contained in the learning data and the characteristic quantities of the operating data (step ST1a). The physical relationship between the command value of the operating mode and the characteristic quantities of the operating data of the equipment being monitored in the equipment status monitoring device 1 is established. Figure 5 It is a graph representing the relationship between the command values ​​of the operating mode of a device and the characteristic quantities of the operating data. For example, in the operating mode of an industrial robot where the rotating mechanism rotates at a constant speed, the average torque of the rotating mechanism is monotonically increasing relative to the command speed values ​​representing each rotation speed.

[0041] exist Figure 5 In this context, the equipment's operating data *d* is time-series data of measured values ​​of the equipment's state corresponding to the operating mode command value after the learning completion mode, forming a distribution *e* for each operating mode command value. For example, when the operating mode command value is 500 (rpm), the operating data *d* is time-series data of the torque measured from a rotating mechanism rotating at 500 (rpm). The regression curve *D* is estimated by applying the least squares method to the average value of the operating data *d* calculated based on the distribution *e* for each operating mode command value. Figure 5 As shown, the regression curve D is a function that monotonically increases the characteristic quantity of the operating data relative to the operating mode command value. The characteristic quantity correction unit 13 uses the learning data to learn the regression curve D as described above.

[0042] Next, the feature quantity correction unit 13 calculates the difference between the feature quantity of the operating data corresponding to the learning completed mode and the feature quantity of the operating data corresponding to the unlearned mode (step ST2a). Figure 6 This is a graphical summary of the calibration process for test data, representing the relationship between the equipment's operating mode command values ​​and characteristic quantities of operating data d. For example, in Figure 6 In the test data, the operation mode instruction value P1 is not included in any operation mode instruction value that represents the learning completed mode, and therefore is the instruction value that represents the unlearned mode.

[0043] The feature quantity correction unit 13 determines the point on the regression curve D corresponding to the unlearned mode, i.e., the operation mode command value P1, as the feature quantity d1 of the operation data corresponding to the operation mode command value P1. Next, the feature quantity correction unit 13 determines the operation mode command value P2 in the learned mode and determines the point on the regression curve D corresponding to the operation mode command value P2, i.e., the feature quantity d2 of the operation data. The relationship shown by the regression curve D is valid between the operation mode command value P1 and its corresponding operation data feature quantity d1, and the relationship is valid between the operation mode command value P2 and its corresponding operation data feature quantity d2. Therefore, the feature quantity correction unit 13 calculates the difference E between the operation data feature quantity d1 and the operation data feature quantity d2.

[0044] Next, the feature quantity correction unit 13 uses the calculated difference E to correct the feature quantity distribution of the operating data corresponding to the unlearned mode (step ST3a). Figure 7 This is a schematic diagram illustrating the process of correcting the difference between the feature distribution of operational data corresponding to the unlearned mode and the feature distribution of operational data corresponding to the learned mode. For example... Figure 7As shown, let G1 be the distribution of feature quantity d1 of operation data corresponding to the instruction value P1 of the unlearned mode (i.e., operation mode) and F be the distribution of feature quantity d2 of operation data corresponding to the instruction value P2 of the learned mode (i.e., operation mode).

[0045] The feature quantity correction unit 13 corrects the distribution G1 of the operating data to be close to the distribution F of the characteristic quantity d2 of the operating data by using the difference E between the distribution G1 of the characteristic quantity d1 and the distribution F of the characteristic quantity d2 of the operating data, thereby correcting the distribution G1 to the distribution G2. The equipment state determination unit 14 compares the distribution G2 and the distribution F, and determines the state of the equipment based on the comparison result.

[0046] Figure 8 This is a flowchart of example (2) of the process of correcting the feature quantity of the operating data corresponding to the unlearned mode, showing a series of processes performed by the feature quantity correction unit 13.

[0047] The feature quantity correction unit 13 uses the physical model of the device to estimate the operating data contained in the learning data (step ST1b). The physical model is input with an operating mode command value, and the operating data corresponding to the input operating mode command value is estimated. The feature quantity correction unit 13 inputs the operating mode command value representing the learning completion mode into the physical model, and the operating data corresponding to the input learning completion mode is output from the physical model.

[0048] Furthermore, the characteristic quantity correction unit 13 calculates the characteristic quantities of the estimated distribution of the operating data. Figure 9A This is a graph representing the distribution of operational data calculated using the correction process (1) of the physical model, showing the distribution H1 of operational data corresponding to the learning completion mode estimated using the physical model. For example, the feature quantity correction unit 13 calculates the average value μ in the distribution H1 of operational data corresponding to the estimated learning completion mode. train and standard deviation σ train The calculation is then performed. Next, the feature quantity correction unit 13 calculates the difference Δd between the operating data corresponding to the estimated learning completion mode and the operating data measured from the device operating through the common learning completion mode (step ST2b).

[0049] Next, the feature quantity correction unit 13 uses the physical model of the device to estimate the operating data corresponding to the unlearned mode (step ST3b). For example, the feature quantity correction unit 13 inputs the operating parameter command value representing the unlearned mode into the physical model, and outputs the operating data corresponding to the input unlearned mode from the physical model. Figure 9BThis is a graph representing the distribution of operational data calculated using the correction process (2) of the physical model. The characteristic quantity correction unit 13 calculates the average value μ of the operational data corresponding to the estimated unlearned mode. test Calculations are performed. Furthermore, the characteristic quantity correction unit 13, as... Figure 9B As shown, the average value μ in the distribution H1 that generates the operational data corresponding to the learning completion mode is generated. train From the average value μ test The permuted distribution H2. Distribution I1 is the distribution of operational data measured from the equipment operating in the unlearned mode.

[0050] Next, the feature quantity correction unit 13 uses the feature quantity of the distribution H1 of the operating data corresponding to the estimated learning completion mode, the difference Δd between the operating data corresponding to the estimated learning completion mode and the measured operating data, and the operating data corresponding to the estimated non-learning mode to estimate the distribution I2 of the operating data corresponding to the non-learning mode (step ST4b). Figure 9C This is a graph representing the distribution of operating data calculated using the correction process (3) of the physical model. The feature quantity correction unit 13 uses the feature quantity of the distribution of operating data corresponding to the learning completion mode, which is estimated by using the physical model to calculate the distribution I1 composed of the measured values ​​of operating data corresponding to the unlearned mode, the operating data estimated by using the physical model, and the difference Δd between the measured data and the actual data, to interpolate the data of distribution I1, thereby calculating distribution I2.

[0051] The feature quantity correction unit 13 corrects the distribution I2 of the operating data corresponding to the unlearned mode in a manner corresponding to the learning completion mode (step ST5b). Figure 9D It is a graph representing the distribution of operational data calculated using the correction process (4) of the physical model. For example... Figure 9D As shown, the feature quantity correction unit 13 generates the average value μ in the distribution I2 of the operating data corresponding to the estimated unlearned mode. test From the average value μ train The replaced distribution I3. The equipment status determination unit 14 compares distribution H1 and distribution I3, and determines the status of the equipment based on the comparison result.

[0052] By using physical models to estimate the operating data of equipment, the amount of operating data that should be measured can be reduced.

[0053] The hardware structure for implementing the function of the equipment status monitoring device 1 is described below.

[0054] Figure 10AThis is a block diagram representing the hardware structure that implements the function of the equipment status monitoring device 1. Figure 10B This is a block diagram representing the hardware structure of the software that executes the functions of the device status monitoring device 1. Figure 10A and Figure 10B In this interface, input interface 100 is an interface for relaying the input of test data and learning data from the device. Output interface 101 is an interface for relaying the judgment results output from the device status determination unit 14 to the outside.

[0055] The functions of the feature quantity extraction unit 11, operation mode determination unit 12, feature quantity correction unit 13, and equipment status determination unit 14 in the equipment status monitoring device 1 are realized through a processing circuit. That is, the equipment status monitoring device 1 has the function of executing... Figure 2 The processing circuits for each step from ST1 to ST4 are shown. The processing circuits can be dedicated hardware, or they can be CPUs (Central Processing Units) that execute programs stored in memory.

[0056] In the processing circuit Figure 10A In the case of the dedicated hardware processing circuit 102 shown, the processing circuit 102 may be, for example, a single circuit, a composite circuit, a programmable processor, a parallel programmable processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof. The functions of the feature extraction unit 11, the operation mode determination unit 12, the feature correction unit 13, and the equipment status determination unit 14 of the equipment status monitoring device 1 can be implemented by each processing circuit, or these functions can be combined and implemented by a single processing circuit.

[0057] In the processing circuit Figure 10B In the case of the processor 103 shown, the functions of the feature quantity extraction unit 11, the operation mode determination unit 12, the feature quantity correction unit 13, and the device status determination unit 14 of the device status monitoring device 1 are implemented by software, firmware, or a combination of software and firmware. Furthermore, the software or firmware is described as a program and stored in the memory 104.

[0058] The processor 103 reads and executes the program stored in the memory 104, thereby realizing the functions of the feature quantity extraction unit 11, the operation mode determination unit 12, the feature quantity correction unit 13, and the equipment status determination unit 14 of the equipment status monitoring device 1. For example, the equipment status monitoring device 1 has a memory 104, which, when executed by the processor 103, stores... Figure 2The programs that are ultimately executed for each of the processes shown in steps ST1 to ST4 are stored. These programs cause the computer to execute the feature extraction unit 11, the operation mode determination unit 12, the feature correction unit 13, and the device status determination unit 14 in a sequence or method. The memory 104 is a computer-readable storage medium that stores programs for enabling the computer to function as the feature extraction unit 11, the operation mode determination unit 12, the feature correction unit 13, and the device status determination unit 14.

[0059] The memory 104 is, for example, non-volatile or volatile semiconductor memory such as RAM (Random Access Memory), ROM (Read Only Memory), flash memory, EPROM (Erasable Programmable Read Only Memory), EEPROM (Electrically-EPROM), disk, floppy disk, optical disk, compact disk, mini disk, DVD, etc.

[0060] The functions of the feature extraction unit 11, operation mode determination unit 12, feature correction unit 13, and equipment status determination unit 14 in the equipment status monitoring device 1 may be partially implemented by dedicated hardware, and the remaining part may be implemented by software or firmware. For example, the feature extraction unit 11 may be implemented by dedicated hardware, namely the processing circuit 102, and the operation mode determination unit 12, feature correction unit 13, and equipment status determination unit 14 may be implemented by the processor 103 reading and executing the program stored in the memory 104. As described above, the processing circuit can implement the above functions through hardware, software, firmware, or a combination thereof.

[0061] The description up to this point has shown a case where the equipment status monitoring device 1 determines the status of the equipment by acquiring a pre-generated learning model, but it may also have structural elements that generate a learning model. Figure 11 This is a block diagram showing the structure of a modified example of the equipment status monitoring device 1, namely, equipment status monitoring device 1A. Figure 11 In the middle, to and Figure 1 Identical structural elements are labeled with the same number, omitting redundant descriptions. For example, equipment status monitoring device 1A... Figure 11 As shown, it has a feature extraction unit 11, an operation mode determination unit 12, a feature correction unit 13, an equipment status determination unit 14, a classification unit 15, and a model generation unit 16.

[0062] The classification unit 15 classifies the operating data of the monitored equipment for each operating mode. For example, when measuring the operating data contained in the learning data from the equipment, the classification unit 15 classifies the operating data for each operating mode based on the instruction values ​​set for the equipment. The model generation unit 16 generates a learning model for each operating mode. This learning model uses the operating data classified for each operating mode to learn the range of equipment state determination. The equipment state determination unit 14 uses the feature values ​​of the corrected operating data and the learning model to determine the state of the equipment.

[0063] Furthermore, the functions of the feature extraction unit 11, operation mode determination unit 12, feature correction unit 13, equipment status determination unit 14, classification unit 15, and model generation unit 16 in the equipment status monitoring device 1A are implemented through a processing circuit. That is, the equipment status monitoring device 1A has a processing circuit that performs various processes, including the classification of operation data and the generation of a learning model. The processing circuit can be... Figure 10A The dedicated hardware processing circuit 102 shown can also be executed in... Figure 10B The processor 103 stores the program in the memory 104 shown.

[0064] As described above, in the device status monitoring device 1 according to Embodiment 1, based on the relationship between the device's operating mode and the characteristic quantities of the operating data, the characteristic quantities of the operating data of the device corresponding to the unlearned mode are corrected in a manner corresponding to the learning completion mode. The device status is then determined based on the characteristic quantities of the corrected operating data and the determination range of the device status. Therefore, the device status monitoring device 1 can determine the device status even when using operating data corresponding to the unlearned mode.

[0065] Furthermore, it is possible to modify any structural element of the implementation method or omit any structural element of the implementation method.

[0066] Industrial applicability

[0067] The equipment status monitoring device involved in this invention can be used, for example, for monitoring the status of industrial robots.

[0068] Explanation of the label

[0069] 1.1A Equipment Status Monitoring Device, 11 Feature Extraction Unit, 12 Operation Mode Determination Unit, 13 Feature Correction Unit, 14 Equipment Status Determination Unit, 15 Classification Unit, 16 Model Generation Unit, 100 Input Interface, 101 Output Interface, 102 Processing Circuit, 103 Processor, 104 Memory.

Claims

1. An apparatus state monitoring device, characterized by comprising: having: a feature quantity extraction section that extracts a feature quantity of operation data obtained by measuring a state of an apparatus; an operation mode determination section that determines whether an operation mode of the apparatus at the time of measurement of the operation data of the apparatus is a learning completion mode in which a determination range of the state of the apparatus has been learned or an unlearned mode in which learning has not been performed; a feature quantity correction section that performs correction to bring a distribution of the feature quantity of the operation data corresponding to the operation mode determined to be the unlearned mode close to or overlap with a distribution of the feature quantity of the operation data corresponding to the learning completion mode, based on a relationship between the operation mode of the apparatus and the feature quantity of the operation data; and an apparatus state determination section that determines the state of the apparatus based on the feature quantity of the operation data of the apparatus and the determination range of the state of the apparatus. having:

2. The apparatus state monitoring device according to claim 1, characterized by a classification section that classifies the operation data of the apparatus for each operation mode; and a model generation section that generates a learning model for each operation mode, the learning model learning the determination range of the state of the apparatus using the operation data classified for each operation mode, the apparatus state determination section determining the state of the apparatus using the feature quantity of the operation data and the learning model.

3. The apparatus state monitoring device according to claim 1 or 2, characterized in that the feature quantity correction section learns a relationship between the operation mode and the feature quantity of the operation data, and performs correction to bring a distribution of the feature quantity of the operation data corresponding to the operation mode determined to be the unlearned mode close to or overlap with a distribution of the feature quantity of the operation data corresponding to the learning completion mode, based on the learned relationship.

4. The apparatus state monitoring device according to claim 1 or 2, characterized in that the feature quantity correction section estimates the operation data of the apparatus for the unlearned mode using a physical model of the apparatus, and performs correction to bring a distribution of the feature quantity of the estimated operation data close to or overlap with a distribution of the feature quantity of the operation data corresponding to the learning completion mode, based on a relationship between the learning completion mode and the feature quantity of the operation data. having:

5. A method of device state monitoring, characterized by, a feature quantity extraction section that extracts a feature quantity of operation data obtained by measuring a state of an apparatus; an operation mode determination section that determines whether an operation mode of the apparatus at the time of measurement of the operation data of the apparatus is a learning completion mode in which a determination range of the state of the apparatus has been learned or an unlearned mode in which learning has not been performed; a feature quantity correction section that performs correction to bring a distribution of the feature quantity of the operation data corresponding to the operation mode determined to be the unlearned mode close to or overlap with a distribution of the feature quantity of the operation data corresponding to the learning completion mode, based on a relationship between the operation mode of the apparatus and the feature quantity of the operation data; and an apparatus state determination section that determines the state of the apparatus based on the feature quantity of the operation data of the apparatus and the determination range of the state of the apparatus. having: a classification section that classifies the operation data of the apparatus for each operation mode; and a model generation section that generates a learning model for each operation mode, the learning model learning the determination range of the state of the apparatus using the operation data classified for each operation mode, the apparatus state determination section determining the state of the apparatus using the feature quantity of the operation data and the learning model.

Images