Abnormality determination system and method, data transmitting and receiving device, and motor control device

Abstract

The present invention relates to an abnormality determination system and method, a data transmission / reception device, and a motor control device, which perform stable abnormality determination and data collection without increasing the processing load of an upper control device. An abnormality determination system that determines an operation abnormality of a motor-driven machine includes: a servo amplifier that controls a motor that drives the motor-driven machine on the basis of a motor control command and is capable of detecting an operation abnormality of the motor-driven machine by comparing operation data acquired in association with the control of the motor with stored reference data; the upper controller is used for sending a motor control instruction to the servo amplifier; the data collection module is used for receiving and transmitting reference data and action data between the data collection module and the servo amplifier; and an edge server that transmits and receives the reference data and the operation data to and from the data collection module, and that is capable of detecting an operation abnormality of the motor-driven machine by comparing the operation data with the reference data.

Description

[0001] This application is a divisional application of the Chinese invention patent application with application date of March 15, 2018, application number 201810213893.6 and invention name “Abnormality Judgment System”. Technical Field

[0002] Embodiments of the present invention relate to an abnormality determination system, a data transceiver, a motor control device, and an abnormality determination method. Background Art

[0003] Patent Document 1 discloses a predicted abnormality sign diagnostic device that acquires time-series data from each of a plurality of devices included in an equipment device and diagnoses the presence or absence of a predicted abnormality sign based on the time-series data.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent No. 5480440. Summary of the Invention

[0007] However, in the above-mentioned prior art, a single abnormality sign diagnostic device acquires time-series data from multiple devices within each of multiple equipment. In other words, it processes a large amount of time-series data acquired from multiple devices to diagnose the presence of abnormality signs. This places an excessive processing burden on higher-level control devices, such as the abnormality sign diagnostic device, and can lead to unstable abnormality determination and data collection.

[0008] The present invention has been made in view of these problems, and its object is to provide an abnormality determination system, a data transceiver, a motor control device, and an abnormality determination method that can perform stable abnormality determination and data collection without increasing the processing burden of the host control device.

[0009] In order to solve the above-mentioned problem, according to one aspect of the present invention, an abnormality judgment system is applied to judge the abnormal operation of a motor-driven machine, including: a motor control device, which controls the motor that drives the motor-driven machine based on a motor control instruction, and can detect the abnormal operation of the motor-driven machine by comparing the action data obtained in association with the control of the motor and the stored reference data; a host control device, which sends the motor control instruction to the motor control device; and a data transceiver, which sends and receives the reference data and the action data to and from the motor control device.

[0010] According to another aspect of the present invention, the data transceiver transmits reference data received from the data management device to the motor control device, and receives motion data acquired by the motor control device from the motor control device and transmits the motion data to the data management device.

[0011] In addition, according to another aspect of the present invention, a motor control device is used to control a motor that drives a motor-driven machine, and detect abnormal operation of the motor-driven machine by comparing observation-time action data and reference data, wherein the observation-time action data is obtained from the motor when the motor-driven machine is driven by observation, and the reference data is calculated based on normal-time action data obtained from the motor when the motor-driven machine is driven normally.

[0012] In addition, according to another aspect of the present invention, an abnormality judgment method is applied to judge the abnormality of a motor-driven machine driven by a motor, and the following is performed: obtaining the observation-time action data of the motor when the motor-driven machine is observed and driven; using the reference data generated by machine learning to judge the data abnormality of the observation-time action data; and judging the action abnormality of the motor-driven machine based on the acquisition method of the observation-time action data judged to be data abnormal.

[0013] According to the present invention, stable abnormality determination and data collection can be performed without increasing the processing burden on the host control device. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 FIG. 1 is a diagram showing a schematic system configuration of an abnormality determination system;

[0015] Figure 2 This is a diagram showing the contents of processing executed by each system device and the flow of information transmitted and received between each system device in the first preparation stage during normal driving;

[0016] Figure 3 is a diagram showing the contents of processing executed by each system device and the flow of information transmitted and received between each system device in the second preparation phase;

[0017] Figure 4 This is a diagram showing the content of processing executed by each system device and the flow of information sent and received between each system device during the application phase of observation driving;

[0018] Figure 5 FIG. 1 is a diagram showing an example of normal operation data and reference data;

[0019] Figure 6 This figure explains the comparison between the reference data of the torque command and the action data during observation.

[0020] Figure 7 This figure illustrates the relationship between the chi-square distribution, the data anomaly judgment threshold, and the Mahalanobis distance.

[0021] Figure 8 1. FIG. 1 is a diagram showing an example of observation-time motion data and a data abnormality detection state in the case of determining an operational abnormality due to aging;

[0022] Figure 9 1 is a flowchart showing a control procedure for abnormality determination processing for performing data abnormality determination and operation abnormality determination;

[0023] Figure 10 2 is a diagram showing a system configuration in which a data collection module and an edge server are integrated into one;

[0024] Figure 11 1 is a diagram showing a system configuration in which a data collection module and a servo amplifier are integrated into one;

[0025] Figure 12 This diagram shows a system configuration in which a data collection module and a servo amplifier are integrated and transmit and receive data with an edge server via wireless communication. DETAILED DESCRIPTION

[0026] Hereinafter, one embodiment will be described with reference to the drawings.

[0027] 1: Overall structure of the abnormality judgment system

[0028] Reference Figure 1 , an example of the overall structure of the abnormality judgment system involved in this embodiment is described.

[0029] Figure 1 The system structure diagram of the abnormality judgment system is shown. The abnormality judgment system of this embodiment is a system that drives and controls a machine system such as a production machine installed in a factory, etc., and acquires its motion data and detects abnormal motion. Figure 1 As shown, the abnormality determination system 100 includes a motor-driven device 1 , a servo amplifier 2 , a host controller 3 , an edge server 4 , a data collection module 5 , and a host instruction device 7 .

[0030] The motor-driven machine 1 is a machine system that is an object of various abnormalities related to its drive that are judged by the abnormality judgment system 100. The motor-driven machine 1 has a plurality of (four in the example shown) motors 12 and a drive mechanism (not specifically shown) driven by these motors 12. The motors 12 include encoders 11. Each motor 12 is coordinated and controlled in a different operation mode, thereby forming a multi-axis machine system that works as a whole. In the example of this embodiment, each motor 12 is a rotary electric motor, and the encoder 11 is a sensor that optically detects the rotational position of the motor 12 and outputs it. In addition, each motor 12 is not limited to a rotary type. In addition, a direct-acting so-called linear motor can also be used. In this case, the linear motor has a position detection sensor such as a linear scale instead of the encoder 11 (not specifically shown). In this embodiment, the motor-driven device 1 further includes an external sensor 13 capable of detecting various state quantities related to the drive control of each motor 12. The external sensor 13 (abbreviated as "sensor" in the figure) capable of detecting various state quantities (such as vibration, device temperature, ambient temperature, and ambient humidity) of each corresponding motor 12 and motor-driven device 1 is connected to each of the encoders 11 via a so-called domain network, such as Σ-LINK (registered trademark), so that information can be transmitted and received. Furthermore, the motor-driven device 1 is not limited to a multi-axis drive control system as in the illustrated example; a single-axis drive control system is also possible (not specifically illustrated).

[0031] The servo amplifiers 2 (motor control devices) are provided so as to correspond to the plurality of motors 12, respectively, and have the function of supplying drive power generated based on motor control commands input from the host controller 3 (host control device) described later to the corresponding motors 12 and performing drive control. In the example of this embodiment, these servo amplifiers 2 also have the function of sequentially acquiring two time-series data as motion data: a torque command generated during the supply of drive power and an output speed generated based on the output position of the motor 12 outputted from the encoder 11, and transmitting these data to an external device (see the following). Figure 2 、 Figure 4 ). In addition, in this embodiment, each servo amplifier 2 also has the following function: by comparing the acquired motion data with the pre-stored reference data, it can detect the abnormal operation of the motor-driven device 1. The function of detecting the abnormal operation will be described in detail later.

[0032] The host controller 3 (host control device) has the function of sequentially generating and outputting motor control commands, such as output position commands for each motor 12, to cause the motor-driven machine 1 to perform a desired time-sequential driving operation. Furthermore, the host controller 3's motor control command generation and output functions are triggered to start and stop, and various parameters are set based on host control commands input from a host command device 7, described later. In this embodiment, the host controller 3 and each servo amplifier 2 are connected to each other via a so-called domain network, such as MECHATROLINK (registered trademark), specifically designed for signal communication between machine control system devices, to enable information transmission and reception.

[0033] The edge server 4 (data management device) is composed of, for example, a desktop general-purpose personal computer and has the function of managing and controlling the entire abnormality judgment system based on input operations (access) from the user. Specifically, the edge server 4 stores and manages the motion data obtained from each servo amplifier 2 by the data collection module 5 described later, which is connected via ETHERNET (Ethernet, a registered trademark) as shown in the example shown in the figure, and in particular generates reference data for detecting data abnormalities based on the motion data during normal driving as described later. In addition, the generation process of this reference data will be described in detail later. In addition, the edge server 4 is connected to the external cloud server 6 via a wide area network NW such as the Internet to transmit and receive information, so that the stored motion data, abnormality detection information, etc. can be sent to the cloud server 6. In addition, in the case of a network method that emphasizes security, there are also cases where it is not connected to the cloud server 6.

[0034] The data collection module 5 (data transceiver) has the function of relaying the transmission and reception of various data, notification information, and the like between each servo amplifier 2 and the edge server 4. In this embodiment, the data collection module 5 is connected to the MECHATROLINK (registered trademark) of the domain network as a terminal on the same level as each servo amplifier 2, enabling parallel and real-time transmission and reception of various data and the like with the servo amplifier 2. Furthermore, in this embodiment, the data collection module 5 is directly connected to the edge server 4 via ETHERNET (registered trademark), enabling rapid transmission and reception of various data and the like. This allows the data collection module 5 to smoothly relay the transmission and reception of various data between the servo amplifier 2 and the edge server 4, accommodating differences in communication standards and processing cycles between the servo amplifier 2 and the edge server 4. Specifically, in this embodiment, there is a difference: data transmission characteristics on the general domain network side (servo amplifier 2 side) are characterized by a short access cycle and a narrow transmission bandwidth, while on the ETHERNET (registered trademark) side (edge ​​server 4 side) are characterized by a long access cycle and a wide bandwidth. In contrast, the data collection module 5 appropriately performs data buffering and synchronization control of transmission and reception, thereby enabling smooth transmission and reception of large amounts of data even between communication standards having different data transmission characteristics.

[0035] The upper-level command device 7, comprised of, for example, a general-purpose personal computer or a PLC (Programmable Logic Controller), has the function of managing the operation and shutdown of the entire motor-driven machine 1. Specifically, the upper-level command device 7 monitors the operating status of the motor-driven machine 1 by referring to the operating data and abnormality determination results stored by the edge server 4, and transmits upper-level control instructions reflecting this operating status to the upper-level controller 3. These upper-level control instructions are instructions to the upper-level controller 3 that instruct it to generate motor control instructions, start and stop output (i.e., to instruct the operation and shutdown of the entire motor-driven machine 1), and set various parameters.

[0036] 2: Features of this implementation

[0037] In a system that typically controls the drive of a motor-driven device 1, a host controller 3 generates and transmits motor control commands. The servo amplifiers 2 that receive these commands then control the motors 12 that drive the motor-driven device 1 based on the motor control commands. This allows for sequence control of the motor-driven device 1 in an operating mode that combines multiple motor control commands in a time-sequential manner. Furthermore, when a motor-driven device 1 is driven by multiple motors 12, a single host controller 3 generates and transmits motor control commands to the servo amplifiers 2 corresponding to each motor 12. This allows for sequence control of the entire motor-driven device 1, enabling coordinated drive of each motor 12.

[0038] On the other hand, when the motor-driven device 1 is used for a long time, it is prone to malfunctions due to aging and other factors. In recent years, there has been a demand for an malfunction detection function that can quickly detect subtle malfunctions and implement rapid fault countermeasures before such malfunctions accumulate and cause a major failure of the motor-driven device 1 as a whole. Furthermore, from the perspective of data-driven technology development, there is an increasing need to continuously acquire large amounts of sensor data detected sequentially from various sensors in real time, transmit it to a higher-level data management device (in this example, the edge server 4 or cloud server 6), and store it.

[0039] However, the processing resources (processing power) of the CPU possessed by the upper controller 3 are limited, and it is difficult to perform the following processing: in parallel with the normal processing of generating and sending motor control instructions for each of the multiple servo amplifiers 2 as described above, one upper controller 3 simultaneously processes the abnormality judgment processing of the driving action of each motor 12 and the sequential acquisition and storage of sensor data.

[0040] In contrast, in the abnormality diagnosis system 100 of this embodiment, the servo amplifier 2 controls the motor 12 based on motor control commands received from the host controller 3. It then detects abnormalities in the operation of the motor-driven device 1 by comparing the motion data acquired in connection with the control of the motor 12 with pre-stored reference data. Furthermore, a data collection module 5 is provided between the servo amplifier 2 and the servo amplifier 2, which transmits and receives reference data and motion data.

[0041] In the abnormality judgment system 100, the acquisition of the motion data of the motor 12 corresponding to each servo amplifier 2 and the detection of the motion abnormality can be processed separately and in parallel. In addition, the data collection module 5 can receive and manage the motion data acquired by each servo amplifier 2 at once. In this case, since each servo amplifier 2 is connected to the motor 12 by Hotelling's T 2 Since the servo amplifier 2 can simply detect abnormal operation by simply comparing the motion data with reference data, the servo amplifier 2 can execute within the processing resources of a typical CPU. Furthermore, since the dedicated data collection module 5 can receive and manage the motion data sequentially acquired from each servo amplifier 2, even if the motor-driven device 1 is driven by multiple motors 12, the host controller 3 only needs to perform the normal processing of generating and sending motor control commands to each servo amplifier 2, without increasing the processing load. These functions are described in detail below.

[0042] 3: About the information sent and received in each stage

[0043] First, the abnormality judgment system 100 of this embodiment needs to go through two preparation stages before it can judge the abnormal state of the operation of the motor drive machine 1. Specifically, it can go through the first preparation stage of acquiring the normal operation data described later in each servo amplifier 2 and the second preparation stage of generating reference data based on the acquired normal operation data and saving it in each servo amplifier 2, and then execute the application stage of performing normal drive control on the motor drive machine 1 for the first time and making abnormality judgment. The following describes the processing of each system device and the flow of information at each stage with reference to the accompanying drawings. In addition, in order to avoid complication of the diagram, the following diagram is shown. Figures 2 to 4 In the above, omit Figure 1 An illustration of the external sensor 13 etc. is shown.

[0044] 3-1: First preparation stage

[0045] First, the first preparation stage is performed when the motor-driven device 1 is being driven normally. This normal driving state can be, for example, a state where the motor-driven device 1 has been assembled and fully adjusted after manufacturing, with little operational abnormality, and is assured of operating as designed (initial operation or test operation).

[0046] Figure 2 The contents of the processing executed by each system device and the flow of information sent and received between each system device in the first preparation stage during such normal driving are shown. Figure 2 In the embodiment of the present invention, the upper command device 7 initially sends an upper control command to the upper controller 3 to start the operation of the entire motor-driven machine 1. Receipt of this upper control command serves as a trigger for the upper controller 3 to individually send motor control commands to each servo amplifier 2 (only one is shown in the figure for simplicity; the same applies hereinafter).

[0047] Each servo amplifier 2 refers to the detection information from its corresponding encoder 11 and generates drive power based on the received motor control command, supplying it to the corresponding motor 12. Meanwhile, during this first preparation phase, each servo amplifier 2 sequentially acquires two time-series data sets, for example, the torque command generated during the drive power supply and the motor 12 output speed calculated based on the detection information from the encoder 11, as a set of normal operation data. Upon receiving an instruction to acquire the operation data from the data collection module 5, the servo amplifier 2 transmits all of this normal operation data to the data collection module 5. The data collection module 5 separates the received normal operation data for each servo amplifier 2 and transmits it to the edge server 4. The edge server 4 stores the received normal operation data corresponding to each servo amplifier 2. Furthermore, in this embodiment, the normal operation data acquired during a predetermined period while the motor-driven device 1 operates in a predetermined determination operation mode (predetermined operation mode) is separated as normal operation data used as a judgment reference, and is acquired, transmitted, and stored. Furthermore, in this embodiment, multiple types of determination operation modes are set. The determination operation mode may be an operation mode operated during normal use of the motor-driven device 1 or an operation mode operated only in the first preparation stage (ie, an operation mode dedicated to generation of reference data described later).

[0048] As described above, in this first preparation phase, all normal operation data acquired by each servo amplifier 2 of the motor-driven device 1 during normal operation is stored separately for each servo amplifier 2 in the edge server 4. Furthermore, among the stored normal operation data, data acquired during drive control in the aforementioned judgment operation mode is specifically identified as normal operation data used as a judgment reference.

[0049] 3-2: Second preparation stage

[0050] The second preparation phase is performed after it is determined that the edge server 4 has sufficiently accumulated normal operation data of each servo amplifier 2 in the first preparation phase. In the second preparation phase, the host controller 3 does not issue motor control commands and the motor-driven device 1 is stopped.

[0051] Figure 3 The contents of the processing performed by each system device in such a second preparation phase and the flow of information sent and received between each system device are shown. Figure 3In the example, the edge server 4 generates reference data for each corresponding servo amplifier 2 based on the normal operation data used as the judgment standard stored for each servo amplifier 2, and transmits the data to the data collection module 5. The data collection module 5 transmits the received reference data to the corresponding servo amplifier 2. Each servo amplifier 2 stores the received reference data.

[0052] As described above, in the second preparation phase, respective reference data are generated based on the normal operation data for determination reference acquired for each servo amplifier 2 , and these reference data are stored in the corresponding servo amplifier 2 .

[0053] 3-3: Application stage

[0054] This application phase is performed during observation driving after the corresponding reference data has been stored in all servo amplifiers 2 in the second preparation phase. Consider, for example, driving in a state (practical application) where the motor-driven device 1 is used for a sufficiently long period of time and may experience operational abnormalities.

[0055] Figure 4 The contents of the processing performed by each system device in such an application phase and the flow of information sent and received between each system device are shown. Figure 4 In the process, the upper command device 7 initially sends an upper control command to the upper controller 3. This upper control command instructs the upper controller 3 to start the operation of the entire motor-driven device 1. Receipt of this upper control command triggers the upper controller 3 to send a motor control command to each servo amplifier 2. Each servo amplifier 2 refers to the detection information from the encoder 11 and generates drive power based on the motor control command, supplying it to the motor 12. This performs normal drive control of the motor-driven device 1.

[0056] Meanwhile, during the application phase, each servo amplifier 2 sequentially acquires two pieces of motion data, for example, the torque command and the output speed of the motor 12, as a set of observation-time motion data, and transmits all of this observation-time motion data to the data collection module 5. The data collection module 5 separates the received observation-time motion data for each servo amplifier 2 and transmits the data to the edge server 4. The edge server 4 stores the received observation-time motion data corresponding to each servo amplifier 2.

[0057] Furthermore, in the present embodiment, each servo amplifier 2 compares observed operational data acquired during a predetermined period of operation of the motor-driven device 1 in the aforementioned determination operation mode with stored reference data, thereby determining whether an operational abnormality has occurred in the motor-driven device 1. Specifically, the determination of operational abnormality involves determining operational abnormality of the movable portion driven by the motor 12 corresponding to the servo amplifier 2 and its surrounding portions within the device configuration of the motor-driven device 1.

[0058] In addition, in the example of this embodiment, the specific judgment of the action abnormality is carried out in two stages: judgment of data abnormality and judgment of action abnormality. That is, by directly comparing the observation action data obtained in the drive control of the judgment action mode with the reference data, it is judged whether a data abnormality occurs in the observation action data. And, based on the acquisition method of the observation action data judged as data abnormal, it is judged whether an action abnormality has finally occurred. The details of the judgment of these data abnormalities and the judgment of action abnormality will be described later. And, when it is judged that an action abnormality has occurred, the servo amplifier 2 performs a predetermined notification process, and specially distinguishes the observation action data judged as action abnormality as abnormal action data and sends it to the data collection module 5, and saves it in the edge server 4. In addition, as mentioned above, normal observation action data other than abnormal action data is also always sent to the data collection module 5 and saved in the edge server 4.

[0059] As described above, in this application phase, all observation-time motion data is separated for each servo amplifier 2 and stored in the edge server 4 in parallel with normal drive control of the motor-driven device 1. Furthermore, when an abnormality occurs in the movable portion driven by each motor 12 or in the surrounding area, the corresponding servo amplifier 2 detects the abnormality and issues an appropriate notification. Furthermore, the observation-time motion data in the corresponding judgment mode is stored in the edge server 4 as abnormal motion data.

[0060] 4: About the method of abnormal judgment

[0061] Hereinafter, a method of determining (detecting) an abnormality of the motor-driven device 1 by cooperation between the servo amplifiers 2 and the edge server 4 will be described in detail.

[0062] The state quantities detectable by the servo amplifier 2 include the torque input to the motor 12, the speed and position output by the motor 12, and detection information from the external sensor 13. In particular, torque also reflects the influence of the reaction force on the motor-driven device 1 side when position / speed control is performed. Therefore, by continuously observing the torque, it is believed that operational abnormalities such as aging can be detected. In this embodiment, machine learning based on statistical methods is used as a method for detecting changes based on the observed waveform.

[0063] However, the anomalies detected by the aforementioned machine learning are only those that can be directly determined based on instantaneously acquired data. In contrast, in a machine system like the motor-driven device 1, the position of the mechanism changes in a very short period of time, and depending on the conditions, the mechanism undergoes both abnormal and normal parts during continuous micro-displacements. Therefore, it is necessary to identify operational anomalies such as aging in all locations. Furthermore, considering the entire machine system, it is inappropriate to determine anomalies solely through statistical methods.

[0064] Therefore, in the abnormality determination system 100 of this embodiment, abnormal conditions directly determined from data through machine learning are defined as data abnormalities. Furthermore, abnormal conditions corresponding to aging or oscillation in the motor-driven device 1 are separately defined as operational abnormalities. These data abnormalities and operational abnormalities are treated separately. Furthermore, the abnormality determination system 100 acquires time-series data related to the input and output of the motor 12 during the operation of the motor-driven device 1 as operational data, and determines data abnormalities based on this operational data. Based on this, the operational abnormality of the motor-driven device 1 is determined based on the acquisition method (acquisition time, acquisition frequency, acquisition frequency, acquisition combination, etc.) of the operational data determined to be a data abnormality. The respective methods for determining data abnormalities and operational abnormalities will be described in the following order.

[0065] 5: About data anomaly judgment

[0066] 5-1: Data anomaly judgment based on machine learning

[0067] Normally, people observe waveforms to determine whether something is normal or abnormal based on experience. Machine learning is a method that expresses this experience in mathematical form and analyzes it on a computer. The basic idea behind change detection methods based on machine learning is to create a normal distribution for a baseline data set (the normal motion data described above). This serves as the baseline data, and the data acquired during the application phase (the observed motion data described above) is then checked to see if it deviates from the normal distribution of the baseline data.

[0068] When determining data anomalies, it is possible to consider the case where the baseline data is based on the premise that all data are normal, and the case where the baseline data with labels of normal and abnormal are mixed. However, when applied to the aging of mechanical parts, it is difficult to prepare abnormal baseline data in advance, so it is realistic to consider the premise that all baseline data are normal. Therefore, in the example of this embodiment, if Figure 5 As shown, the reference data is generated based on the normal operation data determined to be all normal in terms of data as described above. Figure 5 The curve graph of the benchmark data shown in FIG is a graph that combines the sample mean μ, the sample covariance matrix Σ, and the data abnormality judgment threshold a described later. th It is shown in the same time series graph as the normal operation data.

[0069] In order to judge the deviation from the normal distribution, Figure 6 As shown in FIG, a threshold value for judging data anomaly is set at the end of the normal distribution. It is sufficient to confirm that the motion data is far away from the data anomaly judgment threshold value relative to the center (average value, expected value) of the normal distribution during observation. In the example of this embodiment, as described above, motion data is obtained from two time series data, torque command and motor output speed, and data anomaly is judged corresponding to each of these multiple motion data (in Figure 6 The figure only shows the torque command, and the reference data is shown in the figure for easy understanding. Figure 5 The reference data shown is different from the simplified content).

[0070] 5-2: About Hotelling's T 2 Law

[0071] In this embodiment, Hotelling's T is applied as a change detection method based on machine learning. 2 Method. Hotelling's T 2 The method is a multivariate analysis method for observing the changing waveforms of a variety of data in parallel, and the processing is performed in the following (step 1) to (step 6).

[0072] (Step 1) Determine the false alarm rate

[0073] Data contains both normal and abnormal data. The false alarm rate α is an indicator of how much data deviates from the normal distribution and is considered abnormal. For example, if the false alarm rate is 1%, α = 0.01. Furthermore, in the theory of probability and statistics, all data are normal if the false alarm rate is set to 0. Therefore, the false alarm rate α is not theoretically set to 0.

[0074] (Step 2) Calculate the chi-square distribution

[0075] Assuming the degrees of freedom M and the scaling factor s = 1, the chi-square distribution is calculated according to the following formula. The degrees of freedom M is a parameter that specifies the number of independent benchmark data types (the number of variables in the multivariate analysis described above; in this embodiment, two types are used, namely torque command and motor output speed, so M = 2).

[0076]

[0077] Here, Γ represents the gamma function and is defined by the following equation.

[0078]

[0079] (Step 3) Calculate the data anomaly judgment threshold

[0080] Based on the false alarm rate α determined in the above (step 1) and the chi-square distribution calculated in the above (step 2), the data abnormality judgment threshold a that satisfies the following formula is calculated: th .

[0081]

[0082] (Step 4) Calculate the sample mean and sample covariance matrix

[0083] Based on the benchmark data serving as normal data, the sample mean μ (the upper mark (hat) is omitted in the text, and the same applies below) and the sample covariance matrix Σ (the upper mark is omitted in the text, and the same applies below) are calculated according to the following formula.

[0084]

[0085] Here, x (n) It is the reference data of the nth type.

[0086] (Step 5) Calculate the Mahalanobis distance

[0087] Based on the sample mean μ and sample covariance matrix Σ calculated in the above (step 4), and the detected observation data x', the Mahalanobis distance a(x') is calculated according to the following formula.

[0088]

[0089] (Step 6) Compare the data anomaly judgment threshold and Mahalanobis distance

[0090] The data abnormality judgment threshold a calculated in the above (step 3) is th Compare with the Mahalanobis distance a(x') calculated in the above (step 5). th In the case (a(x')>a th), it is determined that the observation data used in the above (step 5) is in a data abnormal state.

[0091] like Figure 7 As shown in the figure, the chi-square distribution is a probability distribution that changes according to the degree of freedom M. It is suitable for application in multivariate analysis due to its so-called reproducibility. For example, as in the example of this embodiment, when the number of action data of the types of variables (torque command, motor output speed) is two, the chi-square distribution is used with the degree of freedom M = 2 and the number of variables (torque command, motor output speed) = 1. Figure 7 The solid line in the figure shows the chi-square distribution. In this chi-square distribution, the Mahalanobis distance a(x') is greater than the data abnormality judgment threshold a corresponding to the false alarm rate α. th If the value is large, it is considered that a data anomaly has occurred in the action data used to calculate the Mahalanobis distance a(x'). That is, in a multivariate analysis with two types of variables, the threshold a can be determined based on the data anomaly. th The degree of abnormality (how far from normal) of the multivariate data based on the combination of the two data is determined by comparing it with the unary Mahalanobis distance a(x'). In addition, the sample mean μ and the sample covariance matrix Σ are used when calculating the Mahalanobis distance a(x') to offset the influence of the correlation between the normal distributions of the two motion data. In addition, the Hotelling T with the degree of freedom M = 1 can be applied separately according to the type of motion data. 2 The data anomaly judgment of the law.

[0092] 5-3: Specific data anomaly judgment

[0093] For example, if machine learning isn't used to detect data anomalies, a normal distribution and a data anomaly threshold must be created for each moment, and the normal distribution must be calculated even for the motion data during observation. Calculating the normal distribution requires calculating the mean and standard deviation, but the complexity of calculating the standard deviation makes it impractical to perform this calculation in real time during motion data acquisition. Furthermore, the data anomaly threshold is set sequentially relative to the normal distribution during motion data acquisition, resulting in a different value at each moment.

[0094] Therefore, in order to solve the above-mentioned problem, in the example of this embodiment, machine learning is used and the processing is as follows.

[0095] (Preparation: Edge Server)

[0096] 1: Get multiple normal action data.

[0097] 2: Calculate the sample mean μ and sample covariance matrix Σ based on the normal action data group.

[0098] 3: Calculate the data anomaly judgment threshold a based on the false alarm rate α and chi-square distribution th .

[0099] (Data abnormality judgment: servo amplifier)

[0100] 1: Get the action data during observation.

[0101] 2: Calculate the Mahalanobis distance a(x') for the motion data during observation.

[0102] 3: If the Mahalanobis distance a(x') exceeds the data anomaly judgment threshold a th , then it is judged as data anomaly.

[0103] In this method utilizing machine learning, instead of calculating the normal distribution, the sample mean μ, the sample covariance matrix Σ, and the Mahalanobis distance a(x') are calculated. Since these calculations are simple arithmetic operations, they do not cause a large processing load even when they are calculated sequentially in short cycles during the actual operation of the motor-driven machine 1 for a long period of time. In addition, the data abnormality judgment threshold a th The calculation formula is complicated, but since it is a constant that does not depend on time, it only needs to be calculated once in advance.

[0104] As described above, in the example of this embodiment, high-load processing, such as calculating the sample mean, sample covariance matrix, and data anomaly determination threshold as baseline data based on normal-time motion data, is performed by the edge server 4, which has high versatility and high CPU processing resources. Meanwhile, low-load processing is performed separately by the servo amplifier 2, which has low remaining CPU processing resources. This low-load processing calculates the Mahalanobis distance based on the sample mean, sample covariance matrix, and observed-time motion data, and determines whether a data anomaly has occurred in the observed-time motion data at that moment by comparing the data anomaly determination threshold and the Mahalanobis distance (i.e., comparing the baseline data with the observed-time motion data). In this embodiment, by distributing processing in this way, data anomaly determination can be performed functionally and in real time on each axis, even in the application phase of a motor-driven machine 1 driven by multiple axes.

[0105] 6: About abnormal action judgment

[0106] The above-described data anomaly determination allows for a binary determination of whether the data at the time of acquisition of the observed motion data is abnormal (i.e., abnormal / normal). However, as described above, even if a data anomaly is determined once, it should not be considered that an operational anomaly has occurred in the entire device system. Furthermore, if a data anomaly occurs multiple times, the nature of the operational anomaly can be estimated once based on the pattern of occurrence. In this embodiment, based on the observation that the frequency of data anomalies increases with aging, if the frequency of data anomalies exceeds a predetermined value, it is determined that an operational anomaly related to aging has occurred in the motor-driven device 1.

[0107] Specifically, if Figure 8 As shown, when the number of observation motion data acquired during the driving period of the judgment action mode is set to 1024 points, if the number of data abnormality detections is low (16 points in the example shown), it is determined that the action is normal, and if the number of data abnormality detections is high (235 points in the example shown), it is determined that the action is abnormal. Furthermore, if an action abnormality is determined to have occurred, all observation motion data acquired during the driving period of the judgment action mode is sent to the data collection module 5 as abnormal motion data.

[0108] In addition, if the causal relationship between the characteristics of abnormal operation and the method of judging data abnormality can be understood, it is also possible to judge that the types of machine abnormalities other than the above-mentioned aging are types of abnormal operation. For example, when the action data are judged to be abnormal by observing both the torque instruction and the motor output speed, the oscillation of the motor-driven machine 1 can be judged to be a type of abnormal operation. As a result, the user knows that when improving the machine abnormality, it is only necessary to deal with the oscillation, thereby improving convenience. In addition, when the data abnormality is judged by the torque instruction and the data abnormality is not judged by the motor output speed, the interference suppression (high friction) in the motor-driven machine 1 can be judged as a type of abnormal operation. In addition, when the data abnormality is not judged by the torque instruction and the data abnormality is judged by the motor output speed, the machine shaking in the motor-driven machine 1 can be judged as a type of abnormal operation.

[0109] 7: Specific control process

[0110] An example of the specific control flow for determining an operational abnormality due to aging will be described in detail below. Figure 9This flowchart illustrates the control process for abnormality determination processing, which determines data anomalies and operational anomalies. This flowchart is executed by the CPU (not specifically shown) of the servo amplifier 2 during observation drive (application phase) of the motor-driven device 1, which can generate data anomalies. The abnormality determination processing shown in this flowchart is executed in parallel with the servo amplifier 2's normal processing (not specifically shown) of supplying drive power based on motor control commands received from the host controller 3. Specifically, this processing is executed only while receiving motor control commands corresponding to the determined operational mode.

[0111] First, in step S105, the CPU of the servo amplifier 2 receives a motor control command corresponding to the determined operating mode (the operating mode for normal use or the operating mode for creating reference data) from the host controller 3 and determines whether to start driving through normal processing. The CPU then waits in a loop until driving in the determined operating mode begins.

[0112] Next, the process proceeds to step S110, where the CPU of the servo amplifier 2 obtains the observed action data of all variables (torque instructions and motor output speed) at each predetermined time such as the system cycle during the entire process of executing the judgment action mode, and saves it to a storage device such as RAM (not specifically shown).

[0113] Next, the process proceeds to step S115 , where the CPU of the servo amplifier 2 calculates the Mahalanobis distance a(x′) at each time point based on the sample mean μ and sample covariance matrix Σ stored in advance in the second preparation stage and the observation-time motion data set acquired in step S110 .

[0114] Next, the process proceeds to step S120, where the CPU of the servo amplifier 2 compares the Mahalanobis distances a(x') at each time point calculated in step S115 to see if they exceed the data abnormality judgment threshold a stored in advance in the second preparation stage. th In other words, it is determined whether the observation-time motion data at each time point acquired in the above step S110 is in a data abnormal state.

[0115] In step S125, the CPU of the servo amplifier 2 determines whether the frequency of data abnormality determinations (frequency of acquisition of observation-time motion data determined to be abnormal) in the observation-time motion data set for the single-stage motion mode determination acquired in step S110 is greater than a predetermined value (predetermined threshold value). In other words, the CPU determines whether an operational abnormality due to aging has occurred. If the frequency of data abnormality determinations is less than the predetermined value, the determination is not satisfied, and the process proceeds to step S130. In other words, it is determined that no operational abnormality due to aging has occurred.

[0116] In step 130, the CPU of the servo amplifier 2 transmits all of the observed-time operation data acquired in the judgment operation mode to the data collection module 5 as normal operation data, and shifts to step S140.

[0117] On the other hand, in the judgment in step S125 described above, in a case where the data abnormality judgment frequency is greater than the predetermined value, the judgment is satisfied, and shifts to step S135. In other words, the operation abnormality is considered to have occurred due to aging.

[0118] In step 135, the CPU of the servo amplifier 2 transmits and notifies the edge server 4 of the result of the judgment that the operation abnormality has occurred in the motor-driven machine 1 via the data collection module 5, and transmits all of the observed-time operation data acquired in the judgment operation mode to the data collection module 5 as abnormal operation data, and shifts to step S140.

[0119] In step S140, the CPU of the servo amplifier 2 judges whether the judgment operation mode has been repeatedly executed for a predetermined number of times determined in advance. In a case where the execution of the predetermined number of times has not ended, the judgment is not satisfied, and returns to step S110 described above to repeat the same process.

[0120] On the other hand, in a case where the predetermined number of times has been executed, the judgment is satisfied to end the flow.

[0121] With the flow of the abnormality judgment processing described above, even in the servo amplifier 2 in which the CPU power is relatively low, the data abnormality judgment processing (steps 5, 6) in which the load of the calculation processing is relatively small can be performed, and the resource burden of the entire abnormality judgment system 100 can be reduced. Further, in the example of the control flow described above, after all of the observed-time operation data of the judgment operation mode is acquired, the data abnormality of each time point is judged by summarizing, and so-called batch processing is performed, but is not limited thereto. In addition thereto, so-called real-time processing (not particularly shown) in which the data abnormality is sequentially judged when the observed-time operation data is acquired at each time point can also be performed.

[0122] 8: Effects based on the present embodiment

[0123] As described above, according to the abnormality judgment system 100 of the present embodiment, the servo amplifier 2 controls the motor 12 based on the motor control command received from the upper-level controller 3, and detects the operation abnormality of the motor-driven machine 1 by comparing the operation data acquired in association with the control of the motor 12 and the reference data saved in advance. In addition, the data collection module 5 that transmits and receives the reference data and the operation data between the servo amplifier 2 is provided.

[0124] In the abnormality determination system 100, the acquisition of the operation data associated with the control of each motor 12 corresponding to each servo amplifier 2 and the detection of the operation abnormality can be distributed and processed in parallel, and the data collection module 5 can receive and manage the operation data acquired by each servo amplifier 2. At this time, in each servo amplifier 2, since the operation abnormality can be simply detected by, for example, the Holtzlin T 2 method, etc., only with the comparison processing of the operation data and the reference data, the processing can be performed within the allowable range of the processing resources in the CPU of the servo amplifier 2. In addition, since the dedicated data collection module 5 can receive and manage the operation data sequentially acquired by each servo amplifier 2, even if the motor-driven machine 1 is configured by a plurality of motors 12, the upper controller 3 can perform only the usual processing for generating and transmitting the motor control command of each servo amplifier 2, and the processing load is not increased. As a result, the processing load of the upper controller 3 is not increased, and stable abnormality determination and data collection can be performed.

[0125] In addition, in the present embodiment, the edge server 4 is also provided, which transmits and receives the reference data and the operation data between the data collection module 5, and can detect the operation abnormality of the motor-driven machine 1 by comparing the operation data and the reference data. By providing such an edge server 4, the user can access the edge server 4, and can perform the review of the operation data associated with the control of each motor 12 and the monitoring of the operation abnormality. In addition, the edge server 4 with a large processing resource can share the processing load of the generation of the reference data with a large processing load as a preliminary preparation for the detection of the operation abnormality, so that the processing load of the entire system can be distributed. In addition, the edge server 4 transmits the detection state of the operation abnormality and the saved operation data to the cloud server 6, so that the detection of the operation abnormality and the collection of the operation data with respect to a plurality of motor-driven machines 1 can be managed on the cloud server 6 side.

[0126] In addition, in the present embodiment, in particular, the servo amplifier 2 determines the data abnormality of the operation data using the reference data generated by machine learning, and determines the operation abnormality based on the acquisition method of the operation data determined as the data abnormality, and transmits the operation data acquired when the motor-driven machine 1 is detected as the operation abnormality to the data collection module 5 as the abnormal operation data. Thus, even in the CPU of the servo amplifier 2 with a small remaining processing resource, the determination of the operation abnormality can be simply performed only with the comparison processing of the acquired operation data and the reference data. In addition, the abnormal operation data distinguished from the other normal operation data is received by the data collection module 5 and the edge server 4, so that the servo amplifier 2 of the acquisition source of the abnormal operation data can be identified as detecting the operation abnormality, and the abnormal operation data can be reviewed and monitored differently from the other operation data.

[0127] Furthermore, in this embodiment, reference data is calculated for each of a plurality of judgment operation modes based on normal operation data acquired during normal operation of the motor-driven device 1. Thus, since only reference data corresponding to the judgment operation mode (an operation mode for normal operation or an operation mode specifically used for reference data creation) needs to be stored and used, the processing burden on the CPU and the memory capacity burden in the servo amplifier 2 can be significantly reduced.

[0128] Furthermore, in the above embodiment, although not specifically illustrated, the servo amplifier 2 may transmit the acquired motion data to the data collection module 5 upon receiving an instruction to acquire motion data via the data collection module 5. In this case, the servo amplifier 2 may receive the acquisition instruction and transmit the motion data via the data collection module 5 according to an appropriate time schedule managed by, for example, an external edge server 4.

[0129] Furthermore, in this embodiment, the data collection module 5 transmits reference data received from the edge server 4 to the servo amplifier 2, and receives motion data acquired by the servo amplifier 2 from the servo amplifier 2 and transmits it to the edge server 4. This allows the dedicated data collection module 5 to sequentially receive and manage motion data acquired from each servo amplifier 2. Consequently, even if the motor-driven device 1 is a multi-axis device driven by multiple motors 12, the host controller 3 only needs to perform the normal processing of generating and sending motor control commands to each servo amplifier 2, without increasing the processing load.

[0130] Furthermore, in this embodiment, the servo amplifier 2 controls the motor 12 that drives the motor-driven device 1 and detects abnormal operation of the motor-driven device 1 by comparing observation-time motion data acquired from the motor 12 (encoder 11) during observation-time driving of the motor-driven device 1 with reference data calculated based on normal-time motion data acquired from the motor 12 during normal driving of the motor-driven device 1. This allows the servo amplifier 2, even in a CPU with limited processing resources due to the execution of normal motor control processing, to simply determine abnormal operation by comparing the acquired motion data with the reference data. Furthermore, since only the stored, capacity-saving reference data is used, the processing burden on the servo amplifier 2's CPU and the memory capacity burden can be significantly reduced.

[0131] Furthermore, in this embodiment, a method for determining an operational abnormality in the motor-driven device 1 is implemented as follows: observation-time operational data of the motor 12 during observation driving of the motor-driven device 1 is acquired, reference data generated through machine learning is used to determine data abnormalities in the observation-time operational data, and the operational abnormality of the motor-driven device 1 is determined based on the manner in which the observed-time operational data determined to be abnormal is acquired. Thus, when performing a diagnostic for operational abnormality in the motion system device control of the motor-driven device 1, a distinction is made between data abnormalities and operational abnormalities, and the operational abnormality is determined based on the manner in which the data abnormality occurs (the manner in which the abnormal operational data, which is determined to be abnormal, is acquired). As a result, operational abnormalities of the entire motor-driven device 1 can be determined more efficiently, in detail, and clearly, without being affected by subtle changes in the data abnormality. Furthermore, since operational abnormalities can be detected simply by comparing the reference data generated through machine learning with the observation-time operational data, the processing load on the CPU of the servo amplifier 2 executing this abnormality determination method can be reduced.

[0132] Furthermore, the method for acquiring abnormal motion data used as a reference when determining abnormal motion is not limited to the aforementioned acquisition frequency. Various acquisition methods (determination methods) can be applied, including acquisition timing, acquisition frequency, and combinations of abnormal motion data, depending on the target's abnormal motion.

[0133] In addition, in this embodiment, in particular, the reference data is based on a predetermined data abnormality judgment threshold and a sample mean value and a sample covariance matrix calculated from normal operation data acquired when the motor-driven device 1 is normally driven. When judging data abnormality, the following is performed: the Mahalanobis distance is calculated based on the sample mean value, the sample covariance matrix, and the observed operation data, and the data abnormality of the observed operation data is judged by comparing the data abnormality judgment threshold with the Mahalanobis distance. As a result, Hotelling T is able to be performed through so-called "supervised learning." 2 The machine learning method can improve the reliability of data anomaly judgment. 2 The data anomaly detection method based on this method, particularly for the sequential comparison of reference data processed by servo amplifier 2 with observed motion data (calculation of the Mahalanobis distance and comparison of this Mahalanobis distance with the data anomaly detection threshold), exhibits a significantly lower computational load than other machine learning methods, such as deep learning. Therefore, even in the servo amplifier 2's CPU, which has limited spare processing resources, highly reliable motion anomaly detection can be achieved in parallel with the normal processing of supplying drive power to the motor.

[0134] 9: Variation

[0135] In addition, the disclosed embodiments are not limited to the above, and various modifications can be made without departing from the scope of the subject matter and technical concept. Figure 1 Corresponding Figures 10 to 12 As shown in the system configuration diagram, even if the device mode of the data collection module 5 is changed, the same effect as the above embodiment can be achieved. Figures 10 to 12 The above is omitted Figure 1 An illustration of the external sensor 13 etc. is shown.

[0136] Figure 10 The following illustrates a system configuration in which data collection module 5A is integrated with edge server 4. In this case, data collection module 5A is assembled into edge server 4 in the form of an expansion board or peripheral device. General-purpose PCs, such as edge server 4, typically lack terminals or processing capabilities for direct connection to a domain network such as MECHATROLINK (registered trademark). Therefore, in this variation, data collection module 5A functions as an interface, enabling data transmission and reception with edge server 4 via an internal bus and shared memory. This eliminates the need for network paths such as ETHERNET (registered trademark) between edge server 4 and data collection module 5A, simplifying overall system wiring and improving the speed of data transmission and reception via the internal bus and shared memory.

[0137] in addition, Figure 11 The following figure shows a system configuration in which the data collection module 5B is integrated with the servo amplifier 2. In this case, the data collection module 5B is incorporated into the servo amplifier 2 in the form of an expansion board or peripheral device. This eliminates the need for a network path between the servo amplifier 2 and the data collection module 5B, simplifying overall system wiring and increasing the speed of data transmission and reception via internal buses and shared memory. Furthermore, if the machine system includes multiple servo amplifiers 2, the data collection module 5B can be integrated into any one of the servo amplifiers 2. In this case, other servo amplifiers 2 connected to the servo amplifier 2 via the local area network simply transmit their acquired motion data to the data collection module 5B via the local area network and the servo amplifier 2.

[0138] Furthermore, in this manner, the data collection module 5 is assembled into the servo amplifier 2, especially as Figure 12As shown, a method in which the data collection module 5C transmits and receives data with the edge server 4 via wireless communication is effective. Specifically, in the illustrated example, the data collection module 5C is connected to the USB port of the servo amplifier 2 using a so-called USB dongle. This data collection module 5C and the edge server 4 transmit and receive data via a wireless LAN (local area network) such as Wi-Fi (wireless LAN, a registered trademark) or Bluetooth (a registered trademark). This allows data to be transmitted and received with the edge server 4 via wireless communication via the data collection module 5C, even if the servo amplifier 2 is located in a location where wiring is difficult to achieve.

[0139] Furthermore, in the above-described embodiment and various variations, the edge server 4 calculates the reference data, but this is not limiting. For example, the data collection module 5 may calculate the reference data. Furthermore, the servo amplifier 2 and motor 12 are not limited to separate, independent configurations. Even when using integrated, so-called amplifier-integrated motors (a servo amplifier, motor, and encoder are integrated, not specifically shown), the data transmission and reception methods and abnormality detection methods described in the above-described embodiment and various variations can be applied. Furthermore, the amplifier-integrated motor in this case corresponds to the motor and motor control device described in each of the claims. Furthermore, in the above-described embodiment and various variations, two types of data, torque command and motor output speed, are acquired as operational data for abnormality detection. However, other data related to the control of the motor 12 (various commands, status values, or status quantity data detected by external sensors 13, etc.) may also be acquired and a combination of these data may be used as operational data.

[0140] In the above description, when terms such as "perpendicular," "parallel," and "planar" are used, these terms are not strictly defined. Specifically, these terms "perpendicular," "parallel," and "planar" are intended to mean "substantially perpendicular," "substantially parallel," and "substantially planar," allowing for design and manufacturing tolerances and errors.

[0141] In addition, in the above description, when the words "same," "identical," "equal," or "different" are used in relation to external dimensions, sizes, shapes, positions, etc., these words are not strictly speaking intended to mean "same," "identical," "equal," or "different," allowing for tolerances and errors in design and manufacturing, and are intended to mean "substantially the same," "substantially the same," "substantially equal," or "substantially different."

[0142] In addition, in addition to the above, the methods of the above embodiment and each modified example can also be used in combination as appropriate. In addition, the above embodiment and each modified example are not listed one by one, and various changes can be added and implemented within the scope of the gist thereof.

[0143] Explanation of symbols

[0144] 1 motor driven machine

[0145] 2 Servo amplifiers (motor control devices)

[0146] 3. Host controller (host control device)

[0147] 4Edge server (data management device)

[0148] 5, 5A, 5B, 5C data collection module (data transceiver) 7 upper command device

[0149] 11 encoders

[0150] 12 motors

[0151] 100 abnormality judgment system.

Claims

1. An abnormality detection system for detecting abnormality in the operation of a motor-driven device, characterized in that: include: a motor control device that controls a motor that drives the motor-driven device based on a motor control instruction and is capable of detecting abnormal operation of the motor-driven device by comparing operation data acquired in association with the control of the motor with stored reference data; A host control device sends the motor control instruction to the motor control device; as well as The data transceiver transmits and receives the reference data and the operation data to and from the motor control device.

2. The abnormality judgment system according to claim 1, wherein: The device further includes a data management device that transmits and receives the reference data and the operation data to and from the data transceiver, and can detect abnormal operation of the motor-driven device by comparing the operation data with the reference data.

3. The abnormality judgment system according to claim 1 or 2, characterized in that: The motor control device uses the reference data generated by machine learning to determine the data abnormality of the motion data, and determines the motion abnormality based on the acquisition method of the motion data determined to be data abnormal, and sends the motion data acquired when the motor drive machine is detected as having a motion abnormality as abnormal motion data to the data transceiver device.

4. The abnormality judgment system according to claim 1 or 2, characterized in that: The reference data is calculated corresponding to a plurality of predetermined operation patterns based on normal operation data acquired when the motor-driven device is normally driven.

5. The abnormality judgment system according to claim 1 or 2, characterized in that: Upon receiving the instruction to obtain the motion data via the data transceiver, the motor control device transmits the obtained motion data to the data transceiver.

6. The abnormality judgment system according to claim 2, wherein: The data transceiver device and the data management device are integrated into one.

7. The abnormality judgment system according to claim 1 or 2, characterized in that: The data transceiver device is integrated with the motor control device.

8. The abnormality judgment system according to claim 2, wherein: The data transceiver is integrated with the motor control device and performs transmission and reception with the data management device via wireless communication.

9. A data transceiver device, characterized in that: Sending the reference data received from the data management device to the motor control device, and The motion data acquired by the motor control device is received from the motor control device and sent to the data management device.

10. A motor control device, characterized in that: A motor driving a motor-driven machine is controlled, and abnormal operation of the motor-driven machine is detected by comparing observation-time operation data obtained from the motor when the motor-driven machine is driven for observation, and baseline data calculated based on normal-time operation data obtained from the motor when the motor-driven machine is driven normally.

11. A method for determining an abnormality of a motor-driven machine driven by a motor, characterized in that: implement: acquiring observation-time motion data of the motor when the motor-driven device is driving under observation; using benchmark data generated by machine learning to determine data anomalies in the motion data during the observation; as well as The abnormality in the operation of the motor-driven device is determined based on the acquisition method of the observation-time operation data determined to be abnormal.

12. The abnormality determination method according to claim 11, wherein: The reference data is a sample average value and a sample covariance matrix, and the sample average value and the sample covariance matrix are calculated based on a predetermined data abnormality judgment threshold and normal operation data obtained when the motor drive device is normally driven. When the data is judged to be abnormal, execute: Calculating the Mahalanobis distance based on the sample mean, the sample covariance matrix, and the observation time data; and The data anomaly of the data during the observation action is determined by comparing the data anomaly determination threshold and the Mahalanobis distance.

Images