Drive device, diagnostic device, and diagnostic method

Abstract

The drive device drives its load through a motor. The drive device includes an inverter, a controller, and a state recognition unit. The inverter drives a motor. The controller controls the inverter on the basis of a command value, and outputs information indicating the operating condition of the motor based on the control. The state recognition unit detects an abnormality of the motor and a mechanical part loaded on the motor on the basis of a machine learning model learned in advance by a learning process including information of vibration of the motor in learning data, the information of vibration of the motor, and the operating condition.

Description

Technical Field

[0001] Embodiments of the present invention relate to a driving device, a diagnostic device, and a diagnostic method. Background Art

[0002] The drive device drives the load through the shaft output of the electric motor. The drive device is used under various operating conditions depending on its intended use. The rotational stability of the electric motor driven by the drive device is sometimes affected by fluctuations in the load. Furthermore, if the mechanical components of the motor and its load change over time, this may cause a decrease in the rotational stability of the motor. It is sometimes difficult to determine whether such a decrease in the rotational stability of the motor is within the range of fluctuations allowed under appropriate operating conditions or is due to abnormalities caused by, for example, changes in the mechanical components over time.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Laid-Open No. 2022-20512 Summary of the Invention

[0006] Problems to be solved by the invention

[0007] An object of the present invention is to provide a drive device, a diagnostic device, and a diagnostic method that further improve the accuracy of estimating abnormalities in a mechanical portion.

[0008] Means for solving problems

[0009] A drive device according to an embodiment drives its load via an electric motor. The drive device includes an inverter, a controller, and a state recognition unit. The inverter drives the electric motor. The controller controls the inverter based on a command value and outputs information indicating the operating status of the electric motor as a result of the control. The state recognition unit detects abnormalities in the mechanical components of the electric motor and its load based on a machine learning model pre-learned by including motor vibration information in learning data, the motor vibration information, and the operating status. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 It is a structural diagram of a drive system according to an embodiment.

[0011] Figure 2 It is a schematic structural diagram of a driving device according to an embodiment.

[0012] Figure 3 This is a functional block diagram showing an example of a driving device according to the embodiment.

[0013] Figure 4 This is a functional block diagram showing an example of a control device according to the embodiment.

[0014] Figure 5 It is a diagram for explaining a rolling process in a hot rolling line according to an embodiment.

[0015] Figure 6 It is a diagram for explaining a rolling process in a hot rolling line according to an embodiment.

[0016] Figure 7 This is a diagram for explaining the definition (mode division) of the operation modes according to the embodiment.

[0017] Figure 8 This is a diagram for explaining vibration observed when the vehicle is driven in a specific operation mode according to the embodiment.

[0018] Figure 9 It is a diagram for explaining the compression processing of the characteristics of the frequency components according to the embodiment.

[0019] Figure 10A This is a schematic diagram of the structure of the machine learning model of the implementation method.

[0020] Figure 10B It is a structural diagram of the machine learning model of the implementation method. DETAILED DESCRIPTION

[0021] Hereinafter, a driving device, a diagnostic device, and a diagnostic method according to embodiments will be described with reference to the drawings.

[0022] In the following description, the drive system for isochronous control will be referred to simply as the drive system. Structures with identical or similar functions will be designated with the same reference numerals. Repeated descriptions of these structures may be omitted. Furthermore, electrical connections may be referred to simply as "connections."

[0023] (Implementation Method)

[0024] Figure 1 It is a configuration diagram of the drive system 1 according to the embodiment.

[0025] The drive system 1 includes, for example, a host device 10 and drive devices 30A to 30D. The motors 21A to 21D are examples of motors to be controlled by the drive system 1 .

[0026] Figure 1 An example of manufacturing equipment related to the drive system 1 is also shown.

[0027] Figure 1The drive system 1 in the manufacturing facility shown includes, for example, rolling stands STA to STD for conveying objects. The rolling stand STA is provided with a motor 21A, the second rolling stand ST2 is provided with a motor 21B, the third rolling stand ST3 is provided with a motor 21C, and the rolling stand STD is provided with a motor 21D. The rolling stands STA to STD are driven by the power of the motors 21A to 21D, respectively.

[0028] The motor 21A is driven by the drive unit 30A. The motor 21B is driven by the drive unit 30B. The motor 21C is driven by the drive unit 30C. The motor 21D is driven by the drive unit 30D. When the motors 21A to 21D are not distinguished, they are simply referred to as the motor 21. When the drive units 30A to 30D are not distinguished, they are simply referred to as the drive unit 30. The motors 21 are operated simultaneously by the drive units 30, and the power output by the motors 21 is used to drive the rolling stand STA to the rolling stand STD respectively. In this way, an object that is relatively long in the first direction is transported at least in the extension direction (first direction) of the object. This manufacturing equipment can be applied to the rolling process of steel plates (objects), for example. In the following description, an application example of steel plates to the rolling process is shown, and a detailed example thereof is described.

[0029] The host device 10 (a higher-level controller) sends command values to each drive device 30 via the network NW and uses the command values to control each drive device 30, thereby adjusting the operating conditions of each motor 21 when conveying an object. The host device 10 controls each drive device 30 to adjust (synchronize) the rotational speed of the motor 21, thereby maintaining a good conveying state for the object.

[0030] Reference Figures 2 to 4 , a more specific example of each driving device 30 is described.

[0031] Figure 2 It is a schematic configuration diagram of the driving device 30 according to the embodiment. Figure 3 This is a functional block diagram showing an example of the driving device 30 according to the embodiment. Figure 4 This is a functional block diagram showing an example of the control device 31 according to the embodiment.

[0032] The driving device 30 includes, for example, a control device 31 and a state recognition unit 32 .

[0033] The control device 31 includes, for example, a power converter 311, a control unit 312, and a communication processing unit 313 ( Figure 4 )、Storage unit 314( Figure 4 STRAGE. ) and input / output unit 315 ( Figure 4 ).

[0034] The power converter 311 supplies electric power for transporting an object to a winding (not shown) of the electric motor 21 .

[0035] In order to use DC power POW to control the motor 21 of the AC motor, the power converter 311 serves as a so-called inverter to convert the DC power POW into AC power. The above is an example and is not limited to this. For example, the motor 21 may also be a DC motor. In order to use AC power POW to control the motor 21 of the DC motor, the power converter 311 serves as a so-called converter to convert the AC power POW into DC power. The specifications of the power converter 311 are determined by the specifications of the motor 21 and the specifications of the power supply, and therefore, it may be a type of power converter other than the above. In the following description, the case where the motor 21 is an AC motor is described.

[0036] Power converter 311( Figure 2 The records in it are PC. ) For example, Figure 2 The plurality of semiconductor switches and their drive circuits (not shown) are, for example, formed into a full-bridge or half-bridge configuration. The circuit configuration of power converter 311 is not limited thereto and may be modified as appropriate. The plurality of semiconductor switches may be, for example, IGBTs (Insulated Gate Bipolar Transistors) or FETs (Field-Effect Transistors), but are not limited to these types. Other types of semiconductor switches may also be used, and diodes may be appropriately combined.

[0037] The communication processing unit 313 can communicate with the host device 10 via the network NW and supply various information acquired from the host device 10 to the control unit 312. The communication processing unit 313 can also communicate with the state recognition unit 32 to acquire recognition results from the state recognition unit 32.

[0038] The storage unit 314 is implemented by ROM, RAM, HDD, flash memory, etc. The storage unit 314 is allocated a storage area for storing various setting information for making the control unit 312 function, basic programs such as OS, application programs, and the like.

[0039] The input / output unit 315 receives, for example, information on the position of the shaft of the motor 21 and the output state of the power converter 311 as input information, and outputs a gate signal GP. The input / output unit 315 may include, for example, a display unit such as a liquid crystal display that displays various information, and an operation detection unit. The display unit and operation detection unit may also be configured as a touch panel in which they are combined.

[0040] The control unit 312 transmits the gate signal GP to the power converter 311 via the input / output unit 315 to control the power converter 311 .

[0041] The control unit 312 includes functions for executing software programs, which will be described later. The control unit 312 is connected to a bus BUS together with a communication processing unit 313 , a storage unit 314 , and an input / output unit 315 .

[0042] The control unit 312 performs some or all of its functions by executing a software program. The software program for implementing the functions of the control unit 312 may be pre-stored in the storage unit 314 or downloaded to the storage unit 314 from an external device (not shown), a removable storage medium, or via a communication line.

[0043] For example, Figure 3 As shown, the state recognition unit 32 obtains information indicating the operating status, such as speed and torque, from the control device 31 .

[0044] The speed mentioned above may be any of a speed reference specified by the host device 10 , a speed detection value detected by a speed sensor (not shown), and a speed estimation value generated in the control device 31 .

[0045] The above-mentioned torque may be a torque reference generated based on the above-mentioned speed and the like.

[0046] Furthermore, the state recognition unit 32 obtains the vibration detection results of the sensor 21S. When an accelerometer is used for the sensor 21S, the vibration can be detected as acceleration. The sensor 21S can output the detection results of the acceleration components in three orthogonal axial directions. The type of sensor 21S is not limited to this.

[0047] The sensor 21S is disposed on a housing of the motor 21 or a stand on which the motor 21 is disposed, and detects vibration at the location where the sensor 21S is disposed.

[0048] The state recognition unit 32 includes an operation mode recognition unit 321 (operation mode), an FFT operation processing unit 322 (FFT), a machine learning model 323, a state determination unit 324 (state determination), and a storage unit (not shown).

[0049] The operation mode identification unit 321 obtains information such as speed and load (torque) from the control device 31 through communication, and identifies the operation mode based on the information such as speed and load (torque).

[0050] The FFT operation processing unit 322 converts the information (vibration information) supplied from the sensor 21S into time-series information and performs an FFT operation on the time-series information (vibration information) and performs preceding and subsequent processing. The FFT operation and preceding and subsequent processing will be described later.

[0051] The machine learning model 323 includes a neural network formed by forming a plurality of layers. The machine learning model 323 is learned in advance using learning data to construct the machine learning model 323 that obtains desired characteristics.

[0052] The state determination unit 324 determines the state based on input data to the machine learning model 323, output data generated by the machine learning model 323, results recognized by the machine learning model 323, and the like.

[0053] The details about them are described later.

[0054] The state recognition unit 32 includes, for example, a function of executing a calculation process described later.

[0055] The storage unit of the state recognition unit 32 is implemented by ROM, RAM, HDD, SSD, flash memory, etc. The storage unit of the state recognition unit 32 can be configured to be provided inside the state recognition unit 32 or as a portion of the storage unit 314. When each component of the state recognition unit 32 acquires data or temporarily stores calculation progress or calculation results, it writes to a specified area of the storage unit and stores the data.

[0056] Each component of the state recognition unit 32 reads this data as needed and performs the desired processing. For example, a program for implementing the functions of the state recognition unit 32 may be stored in the storage unit of the state recognition unit 32 or in the storage unit 314. Some or all of the functions of the state recognition unit 32 are implemented by executing the software program.

[0057] Reference Figure 5 and Figure 6 , the rolling process in the hot rolling line of the embodiment is described.

[0058] Figure 5 and Figure 6 It is a diagram for explaining a rolling process in a hot rolling line according to an embodiment.

[0059] In the hot rolling line, the rolling mill is Figure 1 ) and other components, pulling the reddish iron through the gap between two rollers (cylinders) to achieve the desired thickness. At this point, the speed and torque of the motor 21 used to transport the iron vary depending on the material and finishing method, and the operating mode. The following examples illustrate the differences between materials A through C.

[0060] exist Figure 5 The timing chart in the upper section (a) in FIG. 1 shows in parallel the case where three materials having different rolling process requirements are subjected to the rolling process.

[0061] Material A becomes an operation mode in which the speed of the motor 21 is high and the torque (load) is small. (See Figure 6 Scenario 1 in .)

[0062] Material B becomes an operation mode in which the speed of the motor 21 is slow and the torque (load) is large. (See Figure 6 Scenario 2 in .)

[0063] Material C is in an operation mode in which the speed of the motor 21 is high and the torque (load) is of a medium magnitude.

[0064] exist Figure 5 In the timing chart of the lower section (b), the time axis is enlarged to show the primary amount of the rolling process of the above-mentioned material B.

[0065] At time t0, the material B starts to be applied to the rolls of the rolling stand STA. Thereafter, the material B is moved in a predetermined moving direction without applying a load (torque) to the motor 21 until time t1.

[0066] At time t1, the motor 21 of the rolling stand STA moves the material B in a predetermined moving direction while applying a predetermined load (torque).

[0067] From time t1 to time t2, the speed and load (torque) of the motor 21 in the rolling stand STA are both increased. At time t2, the speed and load (torque) of the motor 21 in the rolling stand STA are adjusted to be constant, and the rolling process of the material B is carried out in this state.

[0068] The material B is rolled, and when it reaches the vicinity of the rear edge of the material (time t4), the load (torque) of the motor 21 is set to 0, and the speed is gradually reduced.

[0069] In such a rolling process, vibrations occur in the mechanism portion of the rolling stand STA.

[0070] This vibration occurs even when no abnormality occurs in the rolling stand STA, and the magnitude of this vibration tends to increase while a load (torque) is applied.

[0071] However, if the rolling process is performed with an abnormality in the mechanism of the rolling stand STA, the above-mentioned vibration tends to become larger, but it is sometimes difficult to recognize the occurrence of the abnormality based on the magnitude (amplitude) of the vibration.

[0072] Therefore, in this embodiment, vibration generated while both the speed and the load (torque) are adjusted to be constant from time t2 to time t4 is detected, and the data is recorded as time-series information.

[0073] like Figure 6 As shown, the vibration components vary depending on the speed of the motor 21 and the magnitude of the load.

[0074] For example, the frequency component of the vibration in the case of the high-speed, low-torque operation mode (Case 1) and the frequency component of the vibration in the case of the low-speed, high-torque operation mode (Case 2) are compared.

[0075] It can be seen that in the frequency spectrum, the distribution of frequencies where energy is concentrated is in different regions.

[0076] The drive device 30 utilizes the fact that the frequency spectra differ when the operation mode is changed in the operation status identification process. This makes it easier to evaluate the operation status compared to the comparative example that does not utilize this feature.

[0077] Reference Figure 7 and Figure 8 , the evaluation of the operating status using the characteristics of each operating mode of the embodiment will be described.

[0078] Figure 7 This is a diagram for explaining the definition (mode division) of the operation modes according to the embodiment.

[0079] like Figure 7 As shown, the speed and load (torque) are assigned to two orthogonal axes, and the coordinate space including the two axes is divided into a plurality of cells.

[0080] For example, regarding the speed axis, the range from 0% to 100% of the rated speed is divided into five equally spaced units. Regarding the load (torque) axis, the range from 0% to 150% (which exceeds 100% of the rated load) is divided into five equally spaced units. By dividing the coordinate space containing the two axes in this manner, the coordinate space can be divided into 25 cells.

[0081] Each of the cells is associated with an "operation mode," and identification information is assigned to each of the associated cells. For example, identification information from 0 to 24 is assigned to each pair of a cell and an "operation mode."

[0082] Figure 8 This is a diagram for explaining vibration observed when the vehicle is driven in a specific operation mode according to the embodiment.

[0083] like Figure 8 As shown, the characteristics of the vibration observed when the vehicle is driven in a specific operation mode are different from those of the vibration observed when the vehicle is driven in other operation modes.

[0084] Figure 8 (a) in the figure indicates the selected specific operation mode.

[0085] The operating mode shown here is identified by the number 19 ( Figure 7) is identified. It corresponds to the fourth cell from the bottom on the torque axis and the fifth cell from the bottom on the time axis.

[0086] Figure 8 (b) in FIG. 1 shows vibration data observed during operation in this operation mode.

[0087] The vibration data is obtained by sampling at a data number suitable for FFT processing and a sampling period corresponding to the required frequency resolution. The horizontal axis is assigned time, and the vertical axis is assigned signal magnitude.

[0088] The data within the range extracted within the time window based on the above conditions is used as the object of the FFT operation. In addition, a window function such as a Hanning window can also be applied to the sampled data before the FFT operation. The processing of the window function is an example of the processing before the FFT operation.

[0089] Perform FFT operation on the data of the sampling result or the data with the window function applied to obtain Figure 8 The spectrum shown in (c) in FIG. The horizontal axis is assigned frequency, and the vertical axis is assigned signal magnitude.

[0090] However, increasing the number of data processed by the FFT operation improves the frequency resolution, but the computational load also increases accordingly. Limiting the number of data processed by the FFT operation reduces the frequency resolution, and it is necessary to appropriately detect the energy of frequency components that do not match the frequency specified by the frequency resolution.

[0091] Reference Figure 9 , the compression processing of the characteristics of the frequency components of the implementation method is described.

[0092] Figure 9 It is a diagram for explaining the compression processing of the characteristics of the frequency components according to the embodiment.

[0093] Generally speaking, increasing the frequency resolution of the FFT calculation results improves the accuracy of converting time-series data into frequency-domain information. However, the determination of similarity of the spectra becomes stricter, resulting in a lower similarity evaluation.

[0094] Therefore, the FFT operation processing unit 322 can slightly reduce the strictness of the similarity evaluation by reducing the number of poles included in the frequency spectrum of the FFT operation processing result calculated by itself, so as to avoid missing similar determinations.

[0095] For example, the FFT operation processing unit 322 may compress the waveform of the frequency spectrum by calculating the sum of squares of several consecutive frequency components in the FFT operation processing result.

[0096] exist Figure 9In the example shown, the sum of the squares of six consecutive frequency components is calculated and plotted together in the same graph. This compression operation allows the FFT-processed spectrum to be used to generate a waveform representing the characteristics of that spectrum. As can be seen, the number of poles is significantly reduced compared to the original spectrum waveform.

[0097] Reference Figure 10A and Figure 10B , the machine learning model 323 of the implementation method is explained.

[0098] Figure 10A This is a schematic structural diagram of the machine learning model 323 and its surroundings showing the implementation method.

[0099] Figure 10B It is a structural diagram of the machine learning model 323 of the implementation method.

[0100] The machine learning model 323 , for example, includes an encoder 3231 , a decoder 3232 , and an encoder 3233 .

[0101] The encoder 3231 compresses spectrum data (referred to as data X in the following description) which is output information from the FFT operation processing unit 322, and outputs the compressed data to the encoder 3233 and the decoder 3232. The compressed data is an example of a feature amount.

[0102] The decoder 3232 decodes the compressed data (feature amount) input from the encoder 3231 to generate reconstructed data, and outputs the reconstructed data (data Xest) to the state determination unit 324. The combination of the encoder 3231 and the decoder 3232 is an example of an autoencoder.

[0103] Encoder 3233 processes the compressed data (features) input from encoder 3231 to generate final layer data and intermediate layer data. The final layer data corresponds to the output value of the neural network, and the intermediate layer data corresponds to the data of all layers other than the final layer data and the input layer data.

[0104] For example, you can also Figure 10B The machine learning model 323 is constructed as shown.

[0105] The encoder 3231 is configured as a neural network including an intermediate layer 32311 located after the input layer and an output layer 32310. The number of intermediate layers and the number of neurons in each layer can be appropriately determined as needed.

[0106] The middle layer 32311 includes a fully connected layer (Dense) 32311D and an activation function processing unit 32311A at the subsequent stage. The activation function illustrated here is ReLU (Rectified Linear Unit). The type of activation function can be selected appropriately.

[0107] The output layer 3221O includes a fully connected layer 3231OD and an activation function processing unit 3231OA at the subsequent stage. The activation function illustrated here is a ReLU. The type of activation function can be selected appropriately. The output from the activation function processing unit 3231OA becomes the output of the encoder 3231.

[0108] Decoder 3232 is configured as a neural network including intermediate layer 32321, intermediate layer 32322, and output layer 3232O located after output layer 3221O of encoder 3231. The number of intermediate layers in decoder 3232 and the number of neurons in each layer can be appropriately determined as needed.

[0109] The intermediate layer 32321 includes a fully connected layer 32321D and an activation function processing unit 32321A at the subsequent stage. The activation function illustrated here is ReLU. The type of activation function can be selected appropriately.

[0110] The intermediate layer 32322 includes a fully connected layer 32322D and an activation function processing unit 32322A at the subsequent stage. The activation function illustrated here is ReLU. The type of activation function can be selected appropriately.

[0111] The output layer 3222O includes a fully connected layer. The output from the fully connected layer 3232OD becomes the output of the decoder 3232.

[0112] Encoder 3233 is configured as a neural network including intermediate layer 32331, intermediate layer 32332, and output layer 3233O located after output layer 3221O of encoder 3231. The number of intermediate layers in encoder 3233 and the number of neurons in each layer can be appropriately determined as needed.

[0113] The intermediate layer 32331 includes a fully connected layer 32331D and an activation function processing unit 32331A at the subsequent stage. The activation function illustrated here is ReLU. The type of activation function can be selected appropriately.

[0114] The intermediate layer 32332 includes a fully connected layer 32332D and an activation function processing unit 32332A at the subsequent stage. The activation function illustrated here is ReLU. The type of activation function can be selected appropriately.

[0115] The output layer 3222O includes a softmax function operation process. The output from the fully connected layer 3232OD becomes the output of the decoder 3232. The result of the softmax function operation process becomes a probability.

[0116] The machine learning model 323 is previously trained using appropriate learning data and is thus capable of reproducing the input data (data X) of the machine learning model 323. In the machine learning model 323, the decoder 3232 outputs data Xest estimated from the data X. If the machine learning model 323 has been properly trained for the predetermined data X, the data Xest approximates the data X.

[0117] Therefore, a predetermined evaluation function can be used to determine the similarity between data X and data Xest. For example, the mean square error (MSE) of data Xest relative to data X can be used as an evaluation function. The state determination unit 324 in the subsequent stage of the machine learning model 323 can calculate the mean square error (MSE) of data Xest relative to data X and use the result to evaluate the normality of the state represented by data X.

[0118] The machine learning model 323 is previously trained using appropriate training data to recognize data X input to the machine learning model 323, and the encoder 3233 outputs the recognition result. If the machine learning model 323 is appropriately trained for the predetermined data X, the data X can be recognized based on the recognition result.

[0119] As described above, the machine learning model 323 is configured to include an autoencoder (encoder 3231 and decoder 3232).

[0120] The state determination unit 324 at the subsequent stage of the machine learning model 323 can utilize a desired determination process from among several types of determinations.

[0121] For example, the first determination process may include a determination (first determination) based on the input information (data X) to the autoencoder and the output information (data Xest) generated by the autoencoder based on the input information (data X). In this case, the state determination unit 324 can perform the first determination based on the input information (data X) to the autoencoder and the output information (data Xest) generated by the autoencoder based on the input information (data X).

[0122] When the machine learning model 323 is configured to include an encoder 3233 (estimation unit) in addition to the above-mentioned autoencoder, the state determination unit 324 can also perform the following second determination process and third determination process.

[0123] For example, the second determination process may include a determination (second determination) based on the identification information (operation mode) of the operating status of the drive device 30 and the estimated information generated by the encoder 3233. In this case, the state determination unit 324 can perform the second determination based on the identification information of the operating status of the drive device 30 and the estimated information generated by the encoder 3233.

[0124] The third determination process is a combination of the first and second determination processes. In this case, the state determination unit 324 can perform the first and second determinations and detect an abnormality in the motor 21 and the mechanical portion of the load based on the results of the first and second determinations.

[0125] Next, the learning of the machine learning model 323 is explained.

[0126] Vibration data observed using the drive system 1 in a good state and vibration data observed using the drive system 1 in a bad state can be used as learning data for pre-learning the machine learning model 323. The "bad state" can be a state that simulates a typical failure or degradation to evaluate the phenomena that occur in that situation.

[0127] More specifically, the results of FFT calculations on vibration data detected when the rolling stand STA and the like of the drive system 1 are operating normally and the results of FFT calculations on vibration data detected when operating in a faulty state can be used as learning data.

[0128] Furthermore, the machine learning model 323 of the embodiment corresponds to each operation mode. The machine learning model 323 can be trained for each operation mode using learning data corresponding to the operation mode to form a machine learning model for each operation mode.

[0129] In this learning process, an evaluation indicator that has a better evaluation value based on an evaluation function such as the mean square error (MSE) of the data Xest relative to the data X and is recognized as an evaluation based on a prescribed operating mode through evaluation by encoder 3233 can be used as a more appropriate evaluation indicator.

[0130] According to the above embodiment, the drive device 30 drives the load via the motor 21. The drive device 30 includes a power converter 311 (inverter) of the control device 31, a control unit 312 (controller), and a state recognition unit 32.

[0131] The power converter 311 is an inverter that drives the electric motor.

[0132] The control unit 312 is a controller that controls the power converter 311 based on a command value to drive the electric motor 21 and outputs information indicating the operating status of the electric motor 21 based on the control.

[0133] The state recognition unit 32 detects abnormalities in the mechanical parts of the motor 21 and its load based on the machine learning model 323 pre-learned by including information indicating the rotational stability of the motor 21 in the learning data, the information indicating the rotational stability of the motor 21, and the operating status.

[0134] Thus, the drive device 30 can further improve the accuracy of estimating abnormalities in the mechanical portion.

[0135] Furthermore, the information indicating the operating status of the electric motor 21 may include any one of a speed reference based on a command value, a torque reference, and a detected speed.

[0136] By defining multiple operating modes that can be identified using any one of the speed reference, torque reference, and detection speed based on the command value, the state recognition unit 32 can use the machine learning model 323 corresponding to the multiple operating modes to detect abnormalities in the mechanical parts of the motor 21 and its load.

[0137] The information indicating the rotational stability of the motor 21 includes information on the magnitude of each frequency component of the spectrum. This spectrum may be based on the spectrum included in the time series information obtained by detecting the vibration of the motor 21.

[0138] As described above, the speed and required torque of the motor 21 in the drive device 30 vary depending on the operating conditions. Even in the case of rolled iron, the operating conditions vary depending on the material, such as high speed and low torque or slow speed and high torque.

[0139] The vibration of the motor 21 and the like varies depending on the operating conditions.

[0140] In the comparative example, when detecting the occurrence of a mechanical abnormality in the motor 21 or the like based on the magnitude (peak value) of the vibration of the motor 21 or the like, it is necessary to set the threshold level used for this determination to a high level to avoid misjudging various operating conditions as abnormalities. In such a method of determining based on the magnitude of vibration, an abnormality cannot be detected until a state of large vibration occurs.

[0141] In contrast, the drive device 30 according to the embodiment can recognize the operating status in a patterned manner, and thus effectively utilizes information on the operating status to perform deterioration diagnosis of the mechanical portion.

[0142] For example, in speed control of the drive device 30 , a speed command is input from the host device 10 or externally, a torque command is generated to follow the speed command, and a required torque is output to follow the torque command, thereby continuing operation.

[0143] Furthermore, the characteristics of data such as the current flowing through the motor 21 and the vibration of the motor 21 vary depending on the operating conditions of the drive device 30. For example, if we focus on the relationship between the speed and torque of the motor 21 regarding the operating conditions, as the torque increases, the amplitude generally increases, while as the speed changes, the frequency component and amplitude change. The operating mode can be defined in a manner that can recognize differences in load (torque) and speed.

[0144] To detect abnormalities in motor 21, vibration data of motor 21 is captured by drive device 30. Based on the operating mode of drive device 30 and the vibration data, a machine learning model 323 is created that learns the waveform obtained by frequency analysis of the vibration. If the detected data differs from the output of machine learning model 323 (autoencoder), an abnormality is detected, and information indicating that an abnormality has been detected or a deterioration trend is expected is output.

[0145] The vibration generated varies depending on the installation conditions of each device (machine), so data after each device is installed and operated can be obtained and used as learning data to implement the learning process of the machine learning model 323.

[0146] The result of the FFT operation processing based on the vibration data of the motor 21 (FFT data) is provided to Figure 10B The IN terminal of the machine learning model 323, from Figure 10B The OUT1 terminal of the FFT data can be obtained. Figure 10B The estimated result of the operation mode is obtained from the OUT2 terminal of the controller.

[0147] State determination unit 324 performs analysis based on the aforementioned FFT calculation results (FFT data), reproduced data from the FFT data, and the estimated operating mode. State determination unit 324 can diagnose the deterioration trend of drive device 30 based on the aforementioned analysis results, including differences from the determination of the operating mode based on the command value, and the reproduced data.

[0148] According to at least one embodiment described above, a drive device drives its load via an electric motor. The drive device includes an inverter, a controller, and a state recognition unit. The inverter drives the electric motor. The controller controls the inverter based on a command value and outputs information indicating the operating status of the electric motor based on the control. The state recognition unit detects an abnormality in the mechanical portion of the electric motor and its load based on a learning model pre-learned by including information about the vibration of the electric motor in learning data, the vibration information of the electric motor, and the operating status. This further improves the accuracy of estimating an abnormality in the mechanical portion.

[0149] The functions realized by the components such as the control unit 312 and the state recognition unit 32 described in this specification may also be installed in a circuit (circuitry) or a processing circuit (processing circuitry). A circuit (circuitry) or a processing circuit (processing circuitry) may include a general-purpose processor, a special-purpose processor, an integrated circuit, ASICs (Application Specific Integrated Circuits), an FPGA (Field Programmable Gate Array), a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an existing circuit and / or a combination thereof that is programmed to realize the functions described. The processor that realizes the above functions includes transistors and other circuits and is considered to be a circuit (circuitry) or a processing circuit (processing circuitry). The processor for realizing the above functions may include a programmable processor that executes a program stored in a memory and / or a programmable device that can be reconstructed according to data stored in the memory, or it may be itself.

[0150] In this specification, a circuit, unit, or means is hardware programmed to implement the described function or hardware that performs the function. This hardware may also be all hardware disclosed in this specification, or all hardware that is programmed to implement the described function or is known as hardware that performs the function.

[0151] In the case where the hardware is a processor of a type regarded as a circuit, the circuit, means or unit is a combination of hardware and software for constituting the hardware and / or processor.

[0152] While several embodiments of the present invention have been described, these embodiments are provided as examples and are not intended to limit the scope of the invention. These embodiments may be implemented in various other ways, and various omissions, substitutions, and modifications may be made without departing from the gist of the invention. These embodiments and their variations are included within the scope and gist of the invention, and are also included within the invention described in the claims and their equivalents.

[0153] For example, the control device 31 and the state recognition unit 32 may be configured as separate processors, or the control device 31 and the state recognition unit 32 may share a common processor to perform respective processes.

[0154] Description of Reference Numerals

[0155] 1…Drive system

[0156] 21, 21A, 21B, 21C, 21D…motor

[0157] 30, 30A, 30B, 30C, 30D…Drive device

[0158] 31…Control device

[0159] 32…Status recognition unit

[0160] 311…Power converter (inverter)

[0161] 312…Control unit (controller)

[0162] 321…Operation mode recognition unit (operation mode)

[0163] 322…FFT calculation processing unit (FFT)

[0164] 323…Machine Learning Model

[0165] 324…Status determination unit (status determination)

[0166] 3231, 3233... encoder

[0167] 3232...decoder

Claims

1. A drive device that drives a load through an electric motor, comprising: Inverter, driving the electric motor; a controller that controls the inverter based on a command value to drive the motor and outputs information indicating an operating condition of the motor based on the control; and The state recognition unit detects abnormalities in the motor and the mechanical parts of its load based on a machine learning model pre-learned by including information indicating the rotational stability of the motor in a learning process of learning data, the information indicating the rotational stability of the motor, and the operating status.

2. The driving device according to claim 1, wherein: The information indicating the operating status of the electric motor includes any one of a speed reference, a torque reference, and a detected speed based on the command value.

3. The driving device according to claim 2, wherein: A plurality of operation modes are defined that can be identified using any one of a speed reference, a torque reference, and a detection speed based on the command value. The state recognition unit detects abnormalities in mechanical parts of the motor and its load using the machine learning models corresponding to the plurality of operation modes.

4. The driving device according to claim 1, wherein: The machine learning model includes an autoencoder.

5. The driving device according to claim 4, wherein: The state recognition unit performs a first determination based on input information to the autoencoder and output information generated by the autoencoder based on the input information.

6. The driving device according to claim 4, wherein: The machine learning model includes an estimating unit that identifies a feature quantity generated by the autoencoder.

7. The driving device according to claim 6, wherein: The state recognition unit performs a second determination based on the identification information of the operating status and the estimation information generated by the estimation unit.

8. The driving device according to claim 1, wherein: The machine learning model includes an autoencoder and an estimating unit that recognizes a feature quantity generated by the autoencoder. The state recognition unit implements a first judgment based on input information to the autoencoder and output information generated by the autoencoder based on the input information, and a second judgment based on the identification information of the operating condition and the estimation information generated by the estimation unit, and detects abnormalities in the mechanical parts of the motor and its load based on the results of the first judgment and the results of the second judgment.

9. The driving device according to claim 1, wherein: The information indicating the rotational stability of the motor includes information on the magnitude of each frequency component of a frequency spectrum of time-series information obtained by detecting vibration of the motor.

10. A diagnostic device for a drive system, the drive system comprising: Inverter, driving the electric motor; and a controller that controls the inverter based on a command value to drive the motor and outputs information indicating an operating condition of the motor based on the control, wherein the drive system drives its load via the motor, The diagnostic device includes a state recognition unit that detects abnormalities in the mechanical parts of the motor and its load based on a machine learning model that has been pre-learned by including information indicating the rotational stability of the motor in a learning process of learning data, information indicating the rotational stability of the motor, and the operating status. The diagnostic apparatus according to claim 10 , wherein: The information indicating the operating status of the electric motor includes any one of a speed reference, a torque reference, and a detected speed based on the command value.

12. The diagnostic apparatus according to claim 11, wherein: A plurality of operation modes are defined that can be identified using any one of a speed reference, a torque reference, and a detection speed based on the command value. The state recognition unit detects abnormalities in mechanical parts of the motor and its load using the machine learning models corresponding to the plurality of operation modes.

13. The diagnostic apparatus according to claim 10, wherein: The machine learning model includes an autoencoder.

14. The diagnostic apparatus according to claim 13, wherein: The state recognition unit performs a first determination based on input information to the autoencoder and output information generated by the autoencoder based on the input information.

15. The diagnostic apparatus according to claim 13, wherein: The machine learning model includes an estimating unit that identifies a feature quantity generated by the autoencoder.

16. The diagnostic apparatus according to claim 15, wherein: The state recognition unit performs a second determination based on the identification information of the operating status and the estimation information generated by the estimation unit.

17. The diagnostic apparatus according to claim 10, wherein: The machine learning model includes an autoencoder and an estimating unit that recognizes a feature quantity generated by the autoencoder. The state recognition unit implements a first judgment based on input information to the autoencoder and output information generated by the autoencoder based on the input information, and a second judgment based on the identification information of the operating condition and the estimation information generated by the estimation unit, and detects abnormalities in the mechanical parts of the motor and its load based on the results of the first judgment and the results of the second judgment.

18. A diagnostic method for a drive system, the drive system comprising: Inverter, driving the electric motor; and a controller that controls the inverter based on a command value to drive the motor and outputs information indicating an operating condition of the motor based on the control, wherein the drive system drives its load via the motor, The diagnostic method includes a process of detecting abnormalities in the mechanical parts of the motor and its load based on a machine learning model that has been pre-learned by including information indicating the rotational stability of the motor in a learning process of learning data, information indicating the rotational stability of the motor, and the operating condition.

19. The diagnostic method according to claim 18, wherein A plurality of operation modes are defined that can be identified using any one of a speed reference, a torque reference, and a detection speed based on the command value. Abnormalities in mechanical parts of the motor and its load are detected using the machine learning models corresponding to the plurality of operating modes.

20. The diagnostic method according to claim 19, wherein The machine learning model identifies the autoencoder and the features generated by the autoencoder, The diagnostic method implements a first judgment based on input information to the autoencoder and output information generated by the autoencoder based on the input information, and a second judgment based on identification information of the operating condition and the generated estimated information, and detects abnormalities in the mechanical parts of the motor and its load based on the results of the first judgment and the results of the second judgment.

Images