Diagnostic device for electric motor and diagnostic system for electric motor

Abstract

A diagnostic device (60) for an electric motor comprises: a detection unit (20) that detects a current signal and a voltage signal of an electric motor (1); and a control unit (50) that detects the presence or absence of an abnormality in the electric motor (1) on the basis of at least the current signal. The control unit (50): detects sideband components on the basis of a first frequency spectrum obtained through frequency analysis of the current signal; performs first control for removing, from the first frequency spectrum, a first sideband component, among the detected sideband components, in which the frequency changes by a first preset amount or more in accordance with the load fluctuation of the electric motor (1); and detects the presence or absence of an abnormality in the electric motor (1) on the basis of the first frequency spectrum from which the first sideband component has been removed.

Description

Electric motor diagnostic device and electric motor diagnostic system

[0001] The present disclosure relates to a diagnostic device for an electric motor and a diagnostic system for an electric motor.

[0002] Electric motors are used to power equipment and machinery used in production lines in the process industry. Their loads include pumps, compressors, blowers, and industrial robots, and their range of applications is expanding, with demand for them on the rise. To ensure stable and continuous operation of equipment and machinery, electric motors must operate continuously and healthily. However, not all electric motors are operated in appropriate environments. It is not uncommon for them to operate in high-stress environments, such as high temperatures, high humidity, heavy loads, corrosion, and wear. Traditionally, the condition of equipment and machinery has been diagnosed by maintenance department inspectors using their five senses through time-based maintenance (TBM). However, particularly important electric motors require periodic diagnosis for faults. Therefore, such inspections by inspectors present a significant challenge in terms of cost.

[0003] Therefore, interest in condition-based maintenance (CBM) technology for electric motors is growing. Diagnosis of inverter-driven electric motors is currently achieved by attaching various measuring instruments, such as sensors, to each electric motor. Measuring instruments include torque meters, speed sensors, and acceleration / vibration sensors. However, this is not practical for application to motor control centers that centrally manage hundreds or thousands of motors. Therefore, there is a need for a system that can diagnose inverter-driven electric motors from information such as current and voltage that can be measured at a motor control center, without using measuring instruments such as specialized sensors that require additional installation, and that ensures reliability, productivity, and soundness.

[0004] Diagnosis of a motor using current is called motor current signature analysis (MCSA), and it is possible to diagnose the motor from the results of analyzing the frequency components of the current. However, depending on the environment in which the motor is installed, there may be a lot of noise superimposed on the power supply, which could lead to false detection. Therefore, in order to improve the accuracy of motor diagnosis, the following diagnostic support devices have been disclosed.

[0005] That is, a conventional diagnostic support device serving as a diagnostic device for an electric motor includes an acquisition unit that acquires the q-axis current of a power conversion device that controls a rotating machine, an amplitude calculation unit that calculates the amplitude for each frequency related to the q-axis current from the acquired q-axis current, and an output unit that outputs information related to the amplitude for each frequency, and includes a filter unit that removes predetermined frequencies between the power conversion device and the amplitude calculation unit (see, for example, Patent Document 1).

[0006] International Publication No. WO2019 / 198563

[0007] The above-mentioned conventional diagnostic support device as a diagnostic device for an electric motor focuses on the q-axis current of the power conversion device and uses a method of improving diagnostic accuracy by removing specific noise components using a filter unit such as a low-pass filter or a high-pass filter. However, while this method can remove noise in a specific frequency band using a filter unit such as a low-pass filter or a high-pass filter, there is a problem in that it may not be possible to accurately remove inverter noise caused by the inverter that drives the electric motor.

[0008] The present disclosure discloses techniques for solving the above-described problems, and aims to provide an electric motor diagnostic device and an electric motor diagnostic system that can accurately diagnose the presence or absence of an abnormality in an electric motor.

[0009] An electric motor diagnostic device according to the present disclosure includes a detection unit that detects a current signal of a current flowing through an electric motor and a voltage signal of a voltage applied to the electric motor, and a control unit that detects the presence or absence of an abnormality in the electric motor based on at least the detected current signal, wherein the control unit detects sideband components of a power supply frequency of the electric motor based on a first frequency spectrum obtained by frequency analysis of the current signal, performs first control to remove from the first frequency spectrum a first sideband component, the first sideband component being a sideband component whose frequency varies by more than a first set amount in response to load fluctuations on the electric motor, and detects the presence or absence of an abnormality in the electric motor based on the first frequency spectrum from which the first sideband component has been removed.An electric motor diagnostic system according to the present disclosure includes the electric motor diagnostic device configured as described above, and a notification unit that is connected to the electric motor diagnostic device via a network and that displays a diagnosis result from the electric motor diagnostic device to an administrator.

[0010] According to the electric motor diagnostic device and electric motor diagnostic system of the present disclosure, an electric motor diagnostic device and electric motor diagnostic system can be obtained that can accurately diagnose the presence or absence of an abnormality in an electric motor.

[0011] 11A is a block diagram showing a schematic configuration of an electric motor diagnostic system according to embodiment 1; FIG. 11B is a block diagram showing an example of hardware of a calculation processing unit according to embodiment 1; FIG. 11C is a block diagram showing the configuration of a calculation processing unit of an electric motor diagnostic device according to embodiment 1; FIG. 11D is a flow chart showing control of a calculation processing unit according to embodiment 1; FIG. 11E is a flow chart showing control of a noise determination process performed by a noise determination unit of a calculation processing unit according to embodiment 1; FIG. 11F is a flow chart showing control of a load fluctuation noise determination process performed by a noise determination unit of a calculation processing unit according to embodiment 1; FIG. 11G is a flow chart showing control of a load fluctuation noise determination execution process performed by a noise determination unit of a calculation processing unit according to embodiment 1; FIG. 11H is a diagram showing a first frequency spectrum for explaining the floor noise estimation process performed by the noise determination unit of a calculation processing unit according to embodiment 1; FIG. 11H is a diagram showing a histogram for explaining the floor noise estimation process performed by the noise determination unit of a calculation processing unit according to embodiment 1; FIG. 11H is a diagram showing a first frequency spectrum for explaining the floor noise estimation process performed by the noise determination unit of a calculation processing unit according to embodiment 1; 12A is a diagram showing a first frequency spectrum at a load factor of 29.9% for explaining the load fluctuation noise estimation process performed by the noise determination unit of the arithmetic processing unit according to embodiment 1. FIG. 12B is a diagram showing a first frequency spectrum at a load factor of 42.2% for explaining the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit according to embodiment 1. FIG. 12C is a diagram showing a first frequency spectrum at a load factor of 55.2% for explaining the load fluctuation noise estimation process performed by the noise determination unit of the arithmetic processing unit according to embodiment 1. FIG. 12D is a diagram showing a first frequency spectrum at a load factor of 75.3% for explaining the load fluctuation noise estimation process performed by the noise determination unit of the arithmetic processing unit according to embodiment 1. FIG. 12E is a diagram showing a first frequency spectrum at a load factor of 88.3% for explaining the load fluctuation noise estimation process performed by the noise determination unit of the arithmetic processing unit according to embodiment 1.16A is a diagram showing the relationship between the load factor and the frequency of a characteristic frequency component, and the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16B is a diagram showing the relationship between the load factor and the frequency of a characteristic frequency component, and the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16C is a diagram showing the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16D is a diagram showing the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16E is a diagram showing the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16F is a diagram showing the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1. FIG. 16G is a diagram showing the relationship between the load factor and the frequency of an inverter noise component according to embodiment 1.

[0012] Embodiment 1. Figure 1 is a block diagram showing a schematic configuration of an electric motor diagnostic system 100 according to embodiment 1. The electric motor diagnostic system 100 includes an electric motor diagnostic device 60 and a notification unit 70, and diagnoses the presence or absence of abnormalities such as bearing damage in an electric motor 1 in a main circuit 10 using the diagnostic device 60, and notifies a maintenance technician of the diagnosis results via the notification unit 70. The following description uses an example in which the electric motor diagnostic system 100 monitors one electric motor 1, but it is also possible to monitor a plurality of electric motors 1.

[0013] As shown in Figure 1, a main circuit 10 connected to a power system K includes a motor 1 such as a three-phase induction motor as a load, a circuit breaker 2 that interrupts the electric circuit, an electromagnetic contactor 3 that opens and closes the electric circuit using an electromagnet, and a detector 4 that transforms the current flowing through the electric circuit to the motor 1 and the voltage applied to the motor 1. A mechanical equipment 6 as a load is connected to the motor 1, and the motor 1 drives the mechanical equipment 6.

[0014] The motor diagnostic device 60 includes a measurement circuit 20 as a detection unit and an arithmetic processing unit 50 as a control unit. The measurement circuit 20 detects the current signal and voltage signal transformed by the detector 4. The arithmetic processing unit 50 uses the current signal or voltage signal input from the measurement circuit 20 to detect the presence or absence of an abnormality in the motor 1 and a load such as mechanical equipment 6.

[0015] The configuration of the notification unit 70 will be described. The notification unit 70 includes an output circuit unit 71, a drive circuit 72, a display unit 73, and a communication circuit 74. The output circuit unit 71 outputs signals indicating an abnormal state of the electric motor 1, warnings, and other signals input from the arithmetic processing unit 50 to the outside to notify a manager. The drive circuit 72 outputs a control signal for opening and closing the electromagnetic contactor 3 based on the signal indicating the state of the electric motor 1 input from the arithmetic processing unit 50. The display unit 73 notifies a supervisor, for example, by displaying an alarm indicating an abnormal state when an abnormality of the electric motor 1 is detected based on the signal indicating the state of the electric motor 1 input from the arithmetic processing unit 50. The communication circuit 74 outputs information such as the diagnosis results of the electric motor 1 input from the arithmetic processing unit 50 to an external monitoring device 80.

[0016] The notification unit 70 is not limited to the configuration including all of the output circuit unit 71, drive circuit 72, display unit 73, and communication circuit 74 as described above. For example, the notification unit 70 may be configured to include only the display unit 73. Furthermore, for example, the notification unit 70 may be configured to include a speaker that outputs sound or the like, as long as it has the function of notifying the monitor of the calculation results in the calculation processing unit 50.

[0017] The external monitoring device 80 is configured by a PC (personal computer) or the like, and appropriately receives information from the arithmetic processing unit 50 via the communication circuit 74 to monitor the operating status of the electric motor diagnostic system 100. The connection between the external monitoring device 80 and the electric motor diagnostic system 100 may be by means of a cable or wirelessly. Alternatively, a network may be constructed between a plurality of electric motor diagnostic systems 100, and the external monitoring device 80 may be connected via the Internet.

[0018] Next, the schematic configuration of the arithmetic processing unit 50 will be described. The arithmetic processing unit 50 includes a processor 51, a storage device 52 as a storage device, and a rating information setting circuit 53. The rating information setting circuit 53 receives input of the power supply frequency, rated output, rated current, number of poles, rated rotation speed, etc. of the electric motor 1. The rating information input to the rating information setting circuit 53 is stored in the storage device 52. The rating information is information that can be easily obtained by looking at a catalog of the manufacturer of the electric motor 1 or a nameplate attached to the electric motor 1. Note that, if there are multiple electric motors 1 to be diagnosed, the rating information of all of the electric motors 1 to be diagnosed must be input in advance to the rating information setting circuit 53.

[0019] FIG. 2 is a diagram illustrating an example of hardware of the arithmetic processing unit 50. As described above, the arithmetic processing unit 50 serving as a control device includes a processor 51, a storage device 52, and a rating information setting circuit 53. The storage device 52 includes a volatile storage device such as a random access memory and a non-volatile auxiliary storage device such as a flash memory, both not shown. Alternatively, a hard disk auxiliary storage device may be provided instead of the flash memory. The processor 51 executes a program input from the storage device 52. In this case, the program is input to the processor 51 from the auxiliary storage device via the volatile storage device. The processor 51 may also execute a program using data recorded in the storage device 52 or the rating information setting circuit 53. The processor 51 may output data such as calculation results to the volatile storage device of the storage device 52, or may store the data in the auxiliary storage device via the volatile storage device.

[0020] Next, the detailed configuration and operation of the arithmetic processing unit 50 will be described. FIG. 3 is a block diagram showing the configuration of the arithmetic processing unit 50 of the motor diagnostic device 60 according to embodiment 1. FIG. 4 is a flow diagram showing control of the arithmetic processing unit 50 according to embodiment 1. FIG. 5 is a flow diagram showing control of the noise determination process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. FIG. 6 is a flow diagram showing control of the floor noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. FIG. 7 is a flow diagram showing control of the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. FIG. 8 is a flow diagram showing control of the load fluctuation noise determination execution process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. FIG. 9 is a diagram showing a first frequency spectrum for explaining the floor noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. FIG. 10 is a diagram showing a first frequency spectrum for explaining the floor noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. Fig. 11A is a histogram illustrating the floor noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. Fig. 11B is a first frequency spectrum illustrating the floor noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. Fig. 12A is a first frequency spectrum illustrating the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1 when the load factor of the electric motor 1 is 29.9%. Fig. 12B is a first frequency spectrum illustrating the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1 when the load factor of the electric motor 1 is 42.2%. Fig. 12C is a first frequency spectrum illustrating the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1 when the load factor of the electric motor 1 is 55.2%.Fig. 12D is a diagram showing a first frequency spectrum when the load factor of the electric motor 1 is 75.3%, for illustrating the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. Fig. 12E is a diagram showing a first frequency spectrum when the load factor of the electric motor 1 is 88.3%, for illustrating the load fluctuation noise estimation process performed by the noise determination unit 530 of the arithmetic processing unit 50 according to embodiment 1. Fig. 13 is a diagram showing the relationship between the load factor and the frequencies of the characteristic frequency components (fs±fr) and the relationship between the load factor and the frequencies of the inverter noise components.

[0021] As shown in FIG. 3 , the processor 51 of the calculation processing unit 50 includes a load factor calculation unit 501, a sampling frequency calculation unit 502, an FFT (Fast Fourier Transform) analysis unit 503, a normalization processing unit 504, a peak detection calculation unit 506, a rotational frequency band detection unit 507, a rotational frequency spectrum value detection unit 508, a frequency axis conversion calculation unit 509, an averaging calculation unit 511, a rotational frequency σ value calculation unit 512, a threshold calculation unit 513, a sideband wave extraction unit 520, a noise determination unit 530, a matrix selection unit 540, an FFT analysis result 541, a normal waveform estimation unit 550, and an abnormal state comparison unit 560.

[0022] The storage device 52 also includes a rotational frequency spectrum value moving average buffer 590 , a correction value data storage unit 591 , and a normal state storage unit 592 .

[0023] Diagnosis of the electric motor 1 can be broadly divided into two phases: a learning phase in which the initial state (normal state) of the electric motor 1 is learned, and a diagnosis phase in which the presence or absence of an abnormality is diagnosed. The flow diagram shown in Fig. 4 shows control in the learning phase, in which the initial state of the electric motor 1 is first learned as the normal state of the electric motor 1. Then, in the diagnosis phase described below, the calculation processing unit 50 diagnoses the presence or absence of an abnormality relatively based on the difference from the normal state.

[0024] <Learning Phase> When the control of the arithmetic processing unit 50 is started, the arithmetic processing unit 50 first detects that the operation of the electric motor 1 has started (step S101). Next, the arithmetic processing unit 50 acquires a current signal of the current flowing through the electric motor 1 and a voltage signal of the voltage applied to the electric motor 1 via the measurement circuit 20 (step S102).

[0025] Next, the load factor calculation unit 501 calculates the load factor of the electric motor 1 based on the acquired current signal and voltage signal (step S103). Next, the sampling frequency calculation unit 502 calculates the power supply frequency fs based on the input current signal or voltage signal, and calculates the sampling frequency that determines the measurement interval by the measurement circuit 20.

[0026] Next, the FFT analysis unit 503 performs FFT analysis on the input current signal to derive a first frequency spectrum by decomposing the current signal into frequency components (step S104). Next, the normalization processing unit 504 performs normalization processing on the derived first frequency spectrum (step S105).

[0027] Next, the peak detection calculation unit 506 detects all peaks in the normalized first frequency spectrum (step S106). Below, we will explain the peak value processing step (step S106A) that processes the peaks detected by the peak detection calculation unit 506.

[0028] The rotational frequency band detection unit 507 extracts a peak point due to the rotational frequency fr of the electric motor 1 from the detected multiple peak points. Specifically, the rotational frequency band detection unit 507 determines the rotational frequency fr of the electric motor 1 from the rated rotation speed recorded in the storage device 52. The rotational frequency band detection unit 507 then extracts peak points (fs±fr) that occur at positions spaced apart by the frequency of the determined rotational frequency fr on both sides of the power supply frequency fs of the electric motor 1. The rotational frequency band detection unit 114 determines the frequency band in which these peak points (fs±fr) exist, including any error, as the rotational frequency band W.

[0029] Next, the rotational frequency spectral value detection unit 508 detects spectral values ​​that are peak values ​​of signal intensity at peak locations (fs±fr) caused by the selected rotational frequency fr. Next, the frequency axis conversion calculation unit 509 performs a correction process to convert the frequency axis of the peak locations (fs±fr) caused by the rotational frequency fr in the first frequency spectrum derived multiple times to match, for example, the load factor of the electric motor 1 when no load is applied. The frequency axis conversion calculation unit 509 stores the spectral values ​​of the multiple peak locations (fs±fr) whose frequency axes have been corrected and aligned in the moving average buffer 590.

[0030] Next, the averaging calculation unit 511 averages the spectral values ​​of the multiple peak locations (fs±fr) where the frequency axis is aligned, which are stored in the moving average buffer 590. For example, the averaging calculation unit 511 performs the averaging process ten times, thereby reducing the spectral value of a peak location that occurs only once to one-tenth the original value, thereby reducing the base noise.

[0031] Next, the rotational frequency σ value calculation unit 512 uses the stored value in the moving average buffer 121 to calculate the variation σ of the spectrum value at the peak point (fs±fr) caused by the rotational frequency fr.

[0032] Next, the threshold calculation unit 513 determines a threshold to be used for abnormality diagnosis based on the variation σ of the spectrum value at the peak point (fs±fr). Specifically, since the difference value between the spectrum value of the power supply frequency fs and the spectrum value at the peak point (fs±fr) correlates with the degree of deterioration of the electric motor 1, the threshold calculation unit 513 can determine the threshold based on the difference value in this way.

[0033] Next, based on the peak locations (fs±fr) in the rotational frequency band W averaged by the averaging calculation unit 511, the sideband wave extraction unit 520 extracts whether there are peak locations on both sides of the power supply frequency fs in frequency regions other than this rotational frequency band W.

[0034] Hereinafter, the peak points (fs±fr) in the rotational frequency band W that occur on both sides of the power supply frequency fs, and the peak points (fs±A·fr) that occur in frequency bands other than the rotational frequency band W, will both be referred to as sideband wave components. As mentioned above, the spectral values ​​of the sideband wave components (fs±fr) are related to the degree of deterioration of the electric motor 1 and can be used in diagnosing whether or not there is an abnormality in the electric motor 1. Therefore, in the following explanation, these sideband wave components (fs±fr) may be referred to as characteristic frequency components (fs±fr).

[0035] The above is the process performed in step S106A for the peak location in the first frequency spectrum.

[0036] The following describes control in the noise determination step performed by the noise determination unit 530. The noise determination unit 530 performs a noise determination step (step S107) of detecting noise in the first frequency spectrum. As shown in Fig. 5 , the noise determination step includes a floor noise estimation step (step S130), a load fluctuation noise estimation step (step S131), and a voltage noise estimation step (step S132).

[0037] First, the floor noise estimation step (step S130) will be described with reference to Fig. 6 and Fig. 9 to Fig. 11. Generally, a first frequency spectrum derived based on a current signal of the electric motor 1 driven by an inverter contains a large amount of floor noise due to, for example, a timing error in inverter switching, and the signal strength of this floor noise is very strong. Therefore, this floor noise is superimposed on the characteristic frequency components (fs±fr) in the first frequency spectrum used for determining an abnormality in the electric motor 1, as described below, which may result in an erroneous diagnosis of an abnormality.

[0038] To prevent misdiagnosis, a method is required to distinguish between the floor noise and the characteristic frequency components (fs±fr) used to determine an abnormality in the motor 1. This floor noise estimation process sets a signal strength determination threshold value that distinguishes between the characteristic frequency components (fs±fr) and the floor noise.

[0039] First, the noise determination unit 530 acquires the first frequency spectrum derived by the FFT analysis unit 503 (step S130A). As shown in Fig. 9, in the first frequency spectrum, pulsation can be confirmed in which the signal strength increases as the frequency decreases, mainly between signal strengths of -70 dB and -100 dB.

[0040] Next, the noise determination unit 530 calculates the pulsating component of this floor noise by deriving a second regression equation showing the relationship between the frequency components in this first frequency spectrum and the signal strength through regression analysis that performs low-order polynomial approximation (step S130B).

[0041] Next, the noise determination unit 530 performs a first control to subtract the calculated pulsating component of the floor noise, i.e., the signal strength of each frequency component determined by the second regression equation, from the first frequency spectrum (step S130C). This subtraction process removes the pulsating component of the floor noise from the first frequency spectrum.

[0042] As shown in Figure 10, it can be confirmed that the pulsation, whose signal strength increases with decreasing frequency, has been removed. In this example, a third-order polynomial approximation was performed using the least squares method. Using this polynomial approximation makes it possible to estimate dynamic pulsation.

[0043] Next, based on the first frequency spectrum from which the pulsating component of the floor noise has been removed, the noise determination unit 530 generates a histogram showing the data density as the frequency of occurrence of frequency components for each signal strength, as shown in Fig. 11A. In the example of Fig. 11A, the noise determination unit 530 sets a data density (3%) in which the signal strength decreases by a set value toward the higher signal strength side, using the signal strength (near -90 dB) at which the data density as the frequency of occurrence is maximized (8.5%) in this histogram as a reference (step S130D). Note that the reference signal strength, the set value, the data density, and the like are not limited to the values ​​shown above, and can be set to any value depending on the control state of the motor, the installation environment, and the like.

[0044] Next, the noise determination unit 530 sets the signal strength at this data density (3%) as the first signal strength (-80 dB) (step S130E), and in the first control, removes the signal strength of frequency components having a signal strength smaller than this first signal strength (-80 dB) from the first frequency spectrum (step S130F).

[0045] The signal strength where floor noise occurs has a high data density on the histogram. Therefore, conversely, a signal strength band with a low data density can be considered a signal strength band with a high S / N ratio (Signal to noise ratio) relative to the floor noise. Therefore, the noise determination unit 530 subtracts the signal strength of frequency components having a signal strength smaller than the first signal strength (-80 dB) from the first frequency spectrum in this way. This allows for processing that prevents erroneous detection due to the influence of floor noise when detecting the presence or absence of an abnormality in the electric motor 1, as described below, by using only peaks that occur at or above the first signal strength (-80 dB).

[0046] In the above description, the noise determination unit 530 performed both the process of removing the pulsating component of the floor noise using the second regression equation and the process of removing the pulsating component of the floor noise using the data density in the first control. However, this is not limited to this, and only one of the processes of removing the pulsating component of the floor noise using the second regression equation and the process of removing the pulsating component of the floor noise using the data density may be performed on the first frequency spectrum. Furthermore, if the floor noise is estimated to be small, these two processes may not be performed.

[0047] Next, the load fluctuation noise estimation step (step S131) ​​shown in Fig. 5 will be described with reference to Fig. 7, Fig. 8, and Fig. 12. First, the noise determination unit 530 acquires the first frequency spectrum derived by the FFT analysis unit 503, and also acquires the load factor of the electric motor 1 calculated by the load factor calculation unit 501 when the first frequency spectrum was derived (steps S131A and S131B).

[0048] Next, the noise determination unit 530 checks whether a setting flag, which will be described later, is set (step S131C). If the flag is not set (step S131C, YES), the noise determination unit 530 acquires the sideband components extracted by the sideband extraction unit 520 (step S131D).

[0049] Next, the noise determination unit 530 stores the frequencies of the multiple side band components that occur in response to the load fluctuation of the electric motor 1 in a table in the storage device 52 together with the load factor at the time of acquisition (step S131E).

[0050] By repeating steps S131A to S131E, the frequencies of the multiple sideband components extracted in the first frequency spectrum are stored in correspondence with the load factor of the motor 1 at the time of extraction. Note that the flow chart in Fig. 7 does not show the control flow for repeating steps S131A to S131E.

[0051] Next, after the frequencies of the sideband wave components at different load rates have been stored in the table, the noise determination unit 530 extracts sideband wave components that disappear, are newly generated, or change frequency in accordance with changes in the load rate based on the stored table, and records these in the storage device 52 (step S131F).

[0052] Once the number of sideband wave components corresponding to the set load factor has been extracted (step S131G<YES), the noise determination unit 530 estimates a first regression equation using the load factor as an explanatory variable based on these sideband wave components, using the least squares method or the like (step S131H). Inverter noise generated by load fluctuations of the motor 1, such as overmodulation noise, has a characteristic in which the frequency changes linearly in proportion to an increase in the load factor, with a negative slope, by more than a first set amount, as shown in Fig. 13. In the following description, this inverter noise may be referred to as load fluctuation noise as a first sideband wave component.

[0053] The noise determination unit 530 compares the actual frequency with the frequency of the load fluctuation noise determined by the estimated simple regression equation, and evaluates the estimation accuracy of the simple regression equation based on the relative error. If the estimated simple regression equation satisfies a certain level of accuracy, the noise determination unit 530 determines the regression coefficients in the simple regression equation.

[0054] If the relative error is not satisfied, i.e., if the frequencies of the sideband components actually detected at each load factor deviate significantly from the frequencies determined by the simple regression equation (step S131I, NO), a decrease in accuracy may be due to an insufficient number of sideband components used to estimate the simple regression equation. In this case, the noise determination unit 530 determines whether the number of sideband components for each load factor used to estimate the simple regression equation is equal to or less than a set number. If the number of sideband components for each load factor is equal to or less than the set number (step S131L, NO), the process returns to step S131A, and the frequencies of the sideband components for each different load factor are acquired.

[0055] Furthermore, if the frequency actually detected at each load factor has a large residual deviation from the frequency determined by the simple regression equation, and if the number of sideband wave components for each load factor used in estimating the simple regression equation is greater than the set number (step S131L: YES), it is considered that the frequency of the detected sideband wave component contains a large error. In this case, one method is to set an upper limit on the frequency of the sideband wave component used in the estimation and re-estimate the regression coefficient, but if this estimation is completely difficult, the data for the corresponding load factor from the load factor data recorded in the table in S131E is deleted (step S131M), and the regression coefficient is re-estimated.

[0056] In this way, the noise determination unit 530 performs at least one of detecting the side band wave component for each load factor and erasing the data of the side band wave component stored in the storage device 52 until a regression coefficient is derived that minimizes the residual error of the frequency actually detected at each load factor from the frequency determined by the simple regression equation. In this way, the first value is accurately estimated as a first set amount, which is the slope of the simple regression equation and indicates the amount of change in frequency of the load fluctuation noise.

[0057] The noise determination unit 530 calculates and records the regression coefficients for each power supply frequency fs (step S131J), and when the number of calculations reaches a predetermined number (step S131K: YES), sets a flag to complete the calculation of the regression coefficients in the simple regression equation for estimating the load fluctuation noise at that power supply frequency fs (step S131N). The flag indicates that the derivation of the simple regression equation for estimating the load fluctuation noise has been completed for each power supply frequency fs of the electric motor 1.

[0058] In step S131C, if the flag is set for the current power supply frequency fs of the electric motor 1 (step S131C: NO), the load fluctuation noise can be determined to be superimposed on the characteristic frequency (fs±fr), so the process proceeds to the load fluctuation noise determination execution process (step S140).

[0059] Next, the load fluctuation noise determination execution step (step S140) will be described with reference to FIG. 8 . The noise determination unit 530 acquires the first frequency spectrum derived in S104 and also acquires the load factor of the electric motor 1 at the time the first frequency spectrum was derived (steps S140A and S140B). Next, the noise determination unit 530 determines whether a flag for the current power supply frequency fs of the electric motor 1 based on the first frequency spectrum has been set, i.e., whether a regression coefficient for estimating the load fluctuation noise at the power supply frequency fs has been derived (step S140C).

[0060] Note that, although steps S140A to S140C correspond to steps S131A to S131C in Fig. 7 described above, deviations may occur in the power supply frequency fs due to inverter control, etc. Therefore, in order to improve the estimation accuracy of the load fluctuation noise, the presence or absence of a flag is determined based on the power supply frequency fs in the first frequency spectrum derived most recently in this manner.

[0061] If the flag is set for the current power supply frequency fs (step S140C: YES), the noise determination unit 530 references the regression coefficients stored in S131J (step S140D).Then, based on a simple regression equation using the estimated regression coefficients, the noise determination unit 530 estimates at which frequency the load fluctuation noise occurs for the current power supply frequency fs and load factor (step S140E).

[0062] For example, according to Fig. 13, at a load factor of 90%, the frequency of the load fluctuation noise (inverter noise) is near 30 Hz, and it can be estimated that it is superimposed on a characteristic frequency component (fs±fr) of approximately 30 Hz. Note that if the flag is not set at the current power supply frequency fs in step S140C (step S140C: NO), the load fluctuation noise determination execution process ends and the process returns to step S131A shown in Fig. 7. This completes the load fluctuation noise determination execution process.

[0063] Next, the voltage noise estimation step (step S132) shown in Fig. 5 will be described with reference to Fig. 14. Fig. 14 is a diagram showing the voltage noise estimation step according to the first embodiment. If the first sideband component, which is load fluctuation noise, cannot be estimated in the first spectrum peak derived by the current FFT, the FFT analysis unit 503 also performs a voltage FFT.

[0064] The noise determination unit 530 compares the analysis results of the current FFT with the voltage FFT, and if a sideband component generated in the voltage FFT also occurs in the current FFT, it detects this sideband component as a first sideband component that is load fluctuation noise.

[0065] Next, the noise determination unit 530 calculates six types of feature quantities: the frequency, signal strength, load factor, and power supply frequency fs of the load fluctuation noise occurring in the current FFT, and the frequency and signal strength of the load fluctuation noise occurring in the voltage FFT.

[0066] The noise determination unit 530 completes the noise determination process after going through the floor noise estimation process (step S130), the load fluctuation noise estimation process (step S131), and the voltage noise estimation process (step S132) shown in FIG.

[0067] As described above, the noise determination unit 530 performs processing to determine a signal strength threshold that allows for distinguishing between floor noise and frequency components used for abnormality determination, thereby removing the floor noise. After removing the floor noise in this manner, the noise determination unit 530 detects sideband wave components whose frequencies change with fluctuations in the load factor of the motor. If there is the aforementioned correlation between the amount of change in the load factor and the amount of change in the frequency of the sideband wave components, the noise determination unit 530 performs processing to determine that the sideband wave components are load fluctuation noise as first sideband wave components.

[0068] However, the noise determination process is not limited to control that requires the processes in steps S130 and S132. For example, the load fluctuation noise estimation process in step S131 may be performed without performing the process of removing floor noise in step S130. Alternatively, the voltage noise estimation process in step S132 may not be performed.

[0069] 4 , the normal waveform estimation step (step S108) performed by the normal waveform estimation unit 550 after the noise determination step (step S107) will be described. In the normal waveform estimation step (step S108), the normal waveform estimation unit 550 performs a first control to subtract the signal strength of the load fluctuation noise estimated in the noise estimation step (step S107) from the first frequency spectrum.

[0070] Since the load fluctuation noise is estimated using a regression equation, even if the signal strength of the characteristic frequency component (fs±fr) and the signal strength of the load fluctuation noise are superimposed in the same frequency band depending on the load rate, it is possible to accurately remove only the signal strength of the load fluctuation noise from the first frequency spectrum.

[0071] The following describes the case where normal waveform estimation unit 550 performs the normal waveform estimation step (step S108) using AI (Artificial Intelligence). FIG. 15 is a schematic diagram of a normal waveform estimation unit 550 according to embodiment 1 having an AI function. FIG. 16A is a diagram showing the feature values ​​of input data at a load factor of 29.9% as determined by normal waveform estimation unit 550 according to embodiment 1. FIG. 16B is a diagram showing the feature values ​​of input data at a load factor of 42.2% as determined by normal waveform estimation unit 550 according to embodiment 1. FIG. 16C is a diagram showing the feature values ​​of input data at a load factor of 55.2% as determined by normal waveform estimation unit 550 according to embodiment 1. In this diagram, load fluctuation noise is indicated as N. FIG. 17 is a diagram showing AI control performed by normal waveform estimation unit 550 according to embodiment 1. FIG. 18 is a diagram showing a neural network in normal waveform estimation unit 550 according to embodiment 1. FIG. 19 is a diagram showing a database storing correction values ​​corresponding to load fluctuations according to embodiment 1. FIG. 20 is a flow chart showing the control of the abnormal state comparison unit of the arithmetic processing unit according to the first embodiment.

[0072] As mentioned above, inverter noise, which is load fluctuation noise, has a relatively strong correlation with the frequency of occurrence relative to load fluctuations. Signal strength also has a certain degree of correlation with the degree of noise occurrence, but this value varies depending on the surrounding environment being diagnosed and the influence of inverter PWM control, etc. Therefore, an advanced regression estimation method using AI (artificial intelligence) is used to estimate the signal strength of only the inverter noise.

[0073] In particular, even when inverter noise, which is load fluctuation noise, is superimposed on or close to the characteristic frequency components (fs±fr) used to determine abnormalities in the motor 1 depending on fluctuations in the load factor of the motor 1, it is possible to remove only the inverter noise components from the first frequency spectrum with even greater accuracy.

[0074] As shown in FIG. 15, the normal waveform estimation unit 550 includes a data acquisition unit 550A, a model generation unit 550B, and a trained model storage unit 550C.

[0075] The data acquisition unit 550A acquires, as learning data, at least one (input 1) of the frequency, signal strength, load factor, power supply frequency fs of the load fluctuation noise in the first frequency spectrum (current FFT), and the frequency and signal strength of the load fluctuation noise generated in the second frequency spectrum (voltage FFT) shown in FIG. 16 , and the signal strength (input 2: correct answer) of the characteristic frequency component (fs±fr), which is a sideband component in the first frequency spectrum from which the load fluctuation noise has been removed.

[0076] 17 , the model generation unit 550B learns the signal strength (output 1) of a feature frequency, which is a sideband component in the second frequency spectrum from which load fluctuation noise has been removed, based on training data created based on a combination of input 1 and input 2 (correct answer) output from the data acquisition unit 550A. That is, it generates a trained model that infers an optimal output from input 1 and input 2 (correct answer). Here, the training data is data in which input 1 and input 2 (correct answer) are associated with each other.

[0077] The model generation unit 550B learns the C output by so-called supervised learning, for example, according to a neural network model. Here, supervised learning refers to a technique in which a learning device is provided with pairs of input and result (label) data, and the device learns the features of the learning data and infers the result from the input.

[0078] A neural network is composed of an input layer consisting of multiple neurons, an intermediate layer (hidden layer) consisting of multiple neurons, and an output layer consisting of multiple neurons. The intermediate layer may be one layer or two or more layers. For example, in a three-layer neural network as shown in FIG. 18, when multiple inputs are input to the input layer (X1-X3), the values ​​are multiplied by weights W1 (w11-w16) and input to the intermediate layer (Y1-Y2), and the results are further multiplied by weights W2 (w21-w26) and output from the output layer (Z1-Z3). This output result varies depending on the values ​​of weights W1 and W2.

[0079] In the present disclosure, the neural network learns the output by so-called supervised learning in accordance with learning data created based on a combination of input 1 and input 2 (correct answer) acquired by data acquisition unit 550 A. That is, the neural network learns by adjusting weights W1 and W2 so that the result output from the output layer after input 1 is input to the input layer approaches input 2 (correct answer).

[0080] The model generation unit 550B generates and outputs a trained model by performing the above-described learning. The trained model storage unit 550C stores the trained model output from the model generation unit 550B.

[0081] Next, the normal waveform estimation unit 550 uses the learned model based on the input feature quantities of the frequency, signal strength, load factor, power supply frequency fs of the load fluctuation noise in the first frequency spectrum (current FFT), the frequency of the load fluctuation noise occurring in the second frequency spectrum (voltage FFT), and signal strength (input 1) to estimate the signal strength of the feature frequency component (fs±fr), which is a sideband component in the first frequency spectrum from which the load fluctuation noise component has been removed.

[0082] In this embodiment, the case where supervised learning is applied to the learning algorithm used by the model generation unit has been described, but the present invention is not limited to this. As for the learning algorithm, reinforcement learning, unsupervised learning, semi-supervised learning, or the like can also be applied in addition to supervised learning.

[0083] Here, the signal strength of the sideband wave component in the first frequency spectrum from which the load fluctuation noise component, which is the input 2 (correct answer), has been removed may be derived based on a first regression equation that includes the load factor of the motor 1 as an explanatory variable.

[0084] In this way, the input feature quantities are six types: the signal strength and frequency of the sideband components generated in the voltage FFT that also occur at the same frequency in the current FFT; the signal strength and frequency of the voltage FFT; the inverter power supply frequency fs at that time; and the load factor. By subtracting the spectrum peak of the load fluctuation noise caused by the voltage FFT calculated by inference using the AI ​​from the analysis results of the current FFT, a current FFT waveform from which the load fluctuation noise has been removed is generated by the AI. In this way, the normal waveform estimation unit 550 may perform first control that removes the load fluctuation noise component from the first frequency spectrum by using AI.

[0085] Examples of AI used for time series regression problems include LSTM (Long short-term memory), GRU (Gated Recurrent Unit), XGBoost, and LightGBM (Light Gradient Boosting Machine), but other AI methods may also be used, and estimation may be performed using any method other than AI, such as statistical methods.

[0086] After the above steps, the normal waveform estimation unit 550 determines the first frequency spectrum from which the load fluctuation noise has been removed as the frequency spectrum to be used in subsequent abnormality diagnosis (step S109). Note that if the noise determination unit 530 determines in step S107 that no load fluctuation noise is present in the first frequency spectrum, the normal waveform estimation step of step S108 is not performed.

[0087] Next, the averaging calculation unit 511 extracts the signal intensity of the characteristic frequency components (fs±fr) based on the frequency components of the first frequency spectrum from which the load fluctuation noise components have been removed (step S110), and, as shown in FIG. 19 , performs averaging processing of the characteristic frequency components (fs±fr) multiple times at each load factor for each power supply frequency fs (step S111).

[0088] The averaging calculation unit 511 accumulates the averaged values ​​multiple times for a certain period of time and performs initial learning as a confirmed value in the normal state (step S112). The normal state storage unit 592 stores the first frequency spectrum from which the signal intensity of the load fluctuation noise component has been subtracted as a measured value in the normal state for use in determining an abnormality in the electric motor 1, which will be described later.

[0089] After the initial learning is completed, the matrix selection unit 540 calculates, based on the derived average values, matrix correction values, which are correction values ​​for the feature frequency components (fs±fr) for each load factor, and a determination threshold for abnormal vibration (step S113). The matrix correction values ​​are correction values ​​that correct the signal strength of the detected feature frequency components (fs±fr) to the signal strength at the first power supply frequency fs and the first load factor that are set according to the power supply frequency fs and the load factor at which the signal strength was detected. The matrix selection unit 540 creates a database of these correction values ​​and stores a determination threshold used for abnormality determination in the correction value data accumulation unit 591 based on the standard deviation of the averaged feature frequencies. The stored database of correction values ​​may be, for example, a database in which correction values ​​are entered in the AVE column shown in FIG. 19 , as long as a correction value corresponding to each load factor is set for each power supply frequency fs.

[0090] Next, the diagnosis start flow will be described with reference to Fig. 20. The processes from confirming the start of operation of the electric motor 1 (step S101) to step S111, in which the signal strength of the characteristic frequency component (fs±fr) for each power supply frequency fs and load factor is calculated, are the same as those in the initial learning flow described above.

[0091] The abnormal state comparison unit 560 performs a correction matrix position selection step (step S212) in which the correction matrix values ​​learned in the initial learning flow are read from the set database as correction values ​​for the current load factor, and reads the correction values ​​and judgment thresholds for the current load factor. That is, the abnormal state comparison unit 560 performs correction control to correct the signal strength of the characteristic frequency component (fs±fr) at the current load factor to the signal strength at the set first power supply frequency fs and first load factor. Then, the abnormal state comparison unit 560 performs an initial-current value comparison (step S213) in which the initial learned value is compared with the corrected measurement value.

[0092] The abnormal state comparison unit 560 compares the value stored in the normal state storage unit 592 with the current value and diagnoses whether or not there is an abnormality in the motor based on the judgment threshold (step S129). Note that the judgment threshold used for abnormality judgment may be the judgment threshold derived in step S106A, but when it is estimated that load fluctuation noise is superimposed on the characteristic frequency component, it is preferable to use the judgment threshold derived in step S113.

[0093] The electric motor diagnostic device of this embodiment configured as described above comprises a detection unit that detects a current signal of a current flowing through an electric motor and a voltage signal of a voltage applied to the electric motor, and a control unit that detects the presence or absence of an abnormality in the electric motor based on at least the detected current signal, wherein the control unit detects sideband wave components of a power supply frequency of the electric motor based on a first frequency spectrum obtained by frequency analysis of the current signal, performs first control to remove from the first frequency spectrum a first sideband wave component, which is a sideband wave component whose frequency changes by more than a first set amount in response to load fluctuations on the electric motor, and detects the presence or absence of an abnormality in the electric motor based on the first frequency spectrum from which the first sideband wave component has been removed.

[0094] Through extensive research, the inventors discovered that overmodulation noise, which is likely to occur when a motor is operating under high load and undergoes overmodulation, occurs as a sideband wave, and its frequency changes depending on the load fluctuation (degree of overmodulation). Furthermore, the amount of change in this frequency varies above a certain value in response to the change in the load factor. Therefore, it was discovered that it is possible to determine whether or not the noise is present based on the correlation between the load factor and frequency.

[0095] The control unit of this embodiment performs first control to remove from the first frequency spectrum sideband wave components, of the sideband wave components obtained by frequency analysis, sideband wave components whose frequencies change by more than a first set amount in response to load fluctuations of the motor. That is, the control unit detects sideband wave components whose frequencies change by more than the set first change amount in response to load fluctuations as noise components and performs first control to remove them from the first frequency spectrum. In this way, by accurately detecting and removing noise components generated by load fluctuations of the motor, which could not be identified in the past, it is possible to dramatically improve the accuracy of diagnosing the presence or absence of an abnormality in the motor.

[0096] Furthermore, although the characteristic frequency used to determine whether the motor is abnormal varies depending on the load factor, the amount of change is very small. Therefore, by setting the first set amount as the amount of change in frequency, it becomes possible to accurately distinguish between noise components and the characteristic frequency.

[0097] Furthermore, based on the motor diagnosis results estimated with high accuracy as described above, the supervisor can grasp the state of the motor via a notification unit such as a display unit, and can perform efficient maintenance work on the motor.

[0098] Furthermore, in the electric motor diagnostic device of this embodiment configured as described above, the control unit detects, from among the plurality of sideband wave components in the first frequency spectrum, the first sideband wave component whose frequency changes by more than the first set amount in response to load fluctuations of the electric motor, based on a first regression equation having the load factor of the electric motor as an explanatory variable.

[0099] The frequency of the noise components generated when the motor is operating in overmodulation changes depending on the load fluctuation of the motor. Therefore, in the first frequency spectrum, the frequency band of the characteristic frequency components used to diagnose abnormalities in the motor may overlap with the frequency band of the noise components. In this case, erroneous detection may occur when diagnosing whether or not there is an abnormality in the motor.

[0100] The electric motor diagnostic device of this embodiment detects the first sideband wave component, which is a noise component, based on a first regression equation that uses the load factor of the electric motor as an explanatory variable. In this manner, the regression equation is used to identify the frequency characteristics of the first sideband wave component, which is a noise component. By identifying the frequency characteristics of the first sideband wave component using the first regression equation in a frequency band where the first sideband wave component is not superimposed on the characteristic frequency component, the signal strength of the first sideband wave component in a frequency band where the first sideband wave component is superimposed on the characteristic frequency component can be estimated. In this way, even when the first sideband wave component is superimposed on the characteristic frequency component depending on the load state of the electric motor, it is possible to accurately remove only the noise component from the first frequency spectrum, enabling accurate diagnosis of the electric motor regardless of the load state of the electric motor.

[0101] Furthermore, in the electric motor diagnostic device of the present embodiment configured as described above, when the relationship between the load factor of the electric motor and the frequency of the sideband wave component is represented by a simple regression equation represented by a linear function as the first regression equation, and a regression coefficient indicating a slope in the simple regression equation is equal to or greater than a first value as the first set amount having a set polarity, the control unit detects, among the plurality of sideband wave components in the first frequency spectrum, the sideband wave component having the relationship between the load factor and the frequency that satisfies the simple regression equation as the first sideband wave component.

[0102] As a result of extensive research by the inventors, it was found that overmodulation noise, such as inverter noise, which is likely to occur when an electric motor is operating under high load, sometimes has a characteristic change in frequency, where the frequency increases linearly in proportion to an increase in the load factor. It was also found that the characteristic frequency used to determine an abnormality in the electric motor decreases slightly linearly in proportion to an increase in the load factor.

[0103] In this manner, the control unit of this embodiment sets the frequency characteristics of the first frequency component, which is a noise component, to satisfy a simple regression equation represented by a linear function, with the regression coefficient being equal to or greater than a first value having a set polarity. Then, of the detected sideband components, those that satisfy this simple regression equation are detected as first sideband components, which are noise components, and are removed from the first frequency spectrum. In this way, the frequency characteristics of the first frequency component, which is a noise component, can be estimated in advance with greater accuracy, and it becomes possible to distinguish between noise components and characteristic frequency components with greater accuracy based on the frequency change identification and the difference in the direction of the frequency change, thereby determining an abnormality.

[0104] Furthermore, in the electric motor diagnostic device of this embodiment configured as described above, the control unit executes at least one of detecting the sideband wave components for each load factor of the electric motor, or erasing the frequencies of the sideband wave components for each load factor recorded in the memory unit, until the regression coefficient in the simple regression equation is derived such that the residual error of the frequency of the sideband wave components detected at each load factor from the frequency determined by the simple regression equation is minimized.

[0105] Depending on the setting environment of the electric motor, the control state of the inverter that drives the electric motor, and the like, noise is often superimposed on the first frequency spectrum. The control unit of this embodiment repeatedly detects sideband components used in estimating the simple regression equation or deletes data on those sideband components until a regression coefficient in the simple regression equation is derived that minimizes the residual deviation of the detected sideband components from the frequency determined by the simple regression equation. As a result, if there are insufficient sideband components to use in estimating the simple regression equation, data on those sideband components can be acquired until a sufficient number is obtained for the estimation. Furthermore, for example, frequency components of external noise that occur only once are removed in the estimation of the simple regression equation, enabling the simple regression equation to be set with high accuracy.

[0106] Furthermore, in the electric motor diagnostic device of the present embodiment configured as described above, the control unit includes: a data acquisition unit that acquires learning data including at least one of the frequency, signal strength, the power supply frequency, and the load factor of the electric motor of the sideband wave component in the first frequency spectrum, and the signal strength of the sideband wave component in the first frequency spectrum from which the first sideband wave component has been removed; and a model generation unit that, in the first control, uses the learning data to generate a trained model for inferring the signal strength of the sideband wave component in the first frequency spectrum from which the first sideband wave component has been removed, based on at least one of the frequency, signal strength, the power supply frequency, and the load factor of the sideband wave component in the first frequency spectrum.

[0107] In this way, by performing control using AI in the first control that removes the first frequency component from the first frequency spectrum, it is possible to remove the influence of factors such as the noise environment and the influence of inverter control, and to remove the noise component from the first frequency spectrum with high precision.

[0108] Furthermore, in the electric motor diagnostic device of this embodiment configured as described above, the control unit derives a second regression equation indicating the relationship between the frequency components and the signal strength in the first frequency spectrum by regression analysis that performs polynomial approximation, and in the first control, removes the signal strength of the first sideband wave component and the signal strength at each of the frequency components determined by the second regression equation from the first frequency spectrum.

[0109] In this way, noise components such as floor noise whose frequency characteristics do not change linearly are estimated by regression analysis using polynomial approximation. This allows noise components that do not have a unidirectional increase and decrease characteristic to be estimated dynamically with high accuracy, enabling even more accurate diagnosis of the motor.

[0110] Although exemplary embodiments are described in this disclosure, the various features, aspects, and functions described in the embodiments are not limited to specific embodiments, but may be applied to the embodiments alone or in various combinations. Therefore, countless variations not illustrated are anticipated within the scope of the technology disclosed in this specification. For example, variations in, addition to, or omission of at least one component are included.

[0111] 1 Electric motor, 20 Measurement circuit (detection unit), 50 Arithmetic processing unit (control unit), 52 Storage device (storage unit), 550A Data acquisition unit, 550B Model generation unit, 60 Electric motor diagnostic device, 70 Notification unit (display unit), 100 Electric motor diagnostic system.

Claims

1. The system comprises a detection unit that detects a current signal of the current flowing through the motor and a voltage signal of the voltage applied to the motor, and a control unit that detects whether or not there is an abnormality in the motor based on at least the detected current signal. The control unit, Based on a first frequency spectrum obtained by frequency analysis of the current signal, the sideband component of the power supply frequency of the motor is detected. From the detected sideband component, a first control is performed to remove from the first frequency spectrum a first sideband component whose frequency changes by a first set amount or more in response to load fluctuations of the motor. Based on the first frequency spectrum from which the first sideband component has been removed, the presence or absence of an abnormality in the motor is detected. A diagnostic device for electric motors.

2. The control unit, Based on a first regression equation in which the load factor of the motor is used as an explanatory variable, a first sideband component is detected among a plurality of sideband components in the first frequency spectrum whose frequency changes by a first set amount or more in response to the load fluctuation of the motor. The diagnostic device for an electric motor according to claim 1.

3. The control unit, When the relationship between the load factor of the motor and the frequency of the sideband component is expressed by a simple regression equation represented by a linear function as the first regression equation, and the regression coefficient representing the slope in the simple regression equation is greater than or equal to a first value as the first set quantity having a set polarity, then among the plurality of sideband components in the first frequency spectrum, the sideband component having the relationship between the load factor and the frequency that satisfies the simple regression equation is detected as the first sideband component. The diagnostic device for an electric motor according to claim 2.

4. The control unit, Until the regression coefficient in the simple regression equation is derived that minimizes the residual deviation of the frequency of the sideband component detected at each load factor from the frequency determined by the simple regression equation, at least one of the following is performed: detection of the frequency of the sideband component for each load factor of the motor, or erasure of the frequency of the sideband component for each load factor recorded in the storage unit. The motor diagnostic device according to claim 3.

5. The control unit, A data acquisition unit that acquires learning data including at least one of the frequency, signal strength, power supply frequency, and motor load factor of the sideband component in the first frequency spectrum, and the signal strength of the sideband component in the first frequency spectrum from which the first sideband component has been removed. In the first control, a model generation unit generates a trained model for inferring the signal intensity of the sideband component in the first frequency spectrum from which the first sideband component has been removed, based on the training data and at least one of the frequency, signal intensity, power supply frequency, and load factor of the sideband component in the first frequency spectrum. Equipped with, The diagnostic device for an electric motor according to claim 1.

6. The control unit, A data acquisition unit that acquires learning data including at least one of the frequency, signal strength, power supply frequency, and motor load factor of the sideband component in the first frequency spectrum, and the signal strength of the sideband component in the first frequency spectrum from which the first sideband component has been removed. In the first control, a model generation unit generates a trained model for inferring the signal intensity of the sideband component in the first frequency spectrum from which the first sideband component has been removed, based on the training data and at least one of the frequency, signal intensity, power supply frequency, and load factor of the sideband component in the first frequency spectrum. Equipped with, The diagnostic device for an electric motor according to claim 2.

7. The control unit, A data acquisition unit that acquires learning data including at least one of the frequency, signal strength, power supply frequency, and motor load factor of the sideband component in the first frequency spectrum, and the signal strength of the sideband component in the first frequency spectrum from which the first sideband component has been removed. In the first control, a model generation unit generates a trained model for inferring the signal intensity of the sideband component in the first frequency spectrum from which the first sideband component has been removed, based on the training data and at least one of the frequency, signal intensity, power supply frequency, and load factor of the sideband component in the first frequency spectrum. Equipped with, The motor diagnostic device according to claim 3.

8. The control unit, A data acquisition unit that acquires learning data including at least one of the frequency, signal strength, power supply frequency, and motor load factor of the sideband component in the first frequency spectrum, and the signal strength of the sideband component in the first frequency spectrum from which the first sideband component has been removed. In the first control, a model generation unit generates a trained model for inferring the signal intensity of the sideband component in the first frequency spectrum from which the first sideband component has been removed, based on the training data and at least one of the frequency, signal intensity, power supply frequency, and load factor of the sideband component in the first frequency spectrum. Equipped with, The diagnostic device for an electric motor according to claim 4.

9. The signal intensity of the sideband component in the first frequency spectrum from which the first sideband component has been removed, which serves as the learning data, is derived based on a first regression equation that includes the load factor of the motor as an explanatory variable. The motor diagnostic device according to claim 5.

10. The control unit, A second regression equation showing the relationship between frequency components and signal intensity in the first frequency spectrum was derived by regression analysis using polynomial approximation. In the first control, the signal intensity of the first sideband component and the signal intensity of each frequency component determined by the second regression equation are removed from the first frequency spectrum. A diagnostic device for an electric motor according to any one of claims 1 to 9.

11. The control unit, Based on the first frequency spectrum, a histogram is generated showing the frequency of occurrence of each frequency component for each signal intensity, and in the first control, the signal intensity of frequency components having a signal intensity smaller than a first signal intensity set based on the signal intensity that has the maximum frequency of occurrence in the histogram is removed from the first frequency spectrum. A diagnostic device for an electric motor according to any one of claims 1 to 9.

12. The control unit, Of the sideband components in the second frequency spectrum obtained by frequency analysis of the aforementioned voltage signal, In the first frequency spectrum, the sideband component occurring at the same frequency is defined as the first sideband component, and this first sideband component is removed from the first frequency spectrum in the first control. A diagnostic device for an electric motor according to any one of claims 1 to 9.

13. The control unit, Based on the first frequency spectrum from which the first sideband component has been removed, the average value of the signal intensity of the sideband component over multiple cycles at each load factor of the motor is derived for each power supply frequency. The system includes a database containing correction values ​​that, based on the derived average value, correct the signal intensity of the detected sideband component to the signal intensity at a set first power supply frequency and first load factor, according to the power supply frequency and motor load factor at which the signal intensity was detected. Based on the database, correction control is performed to correct the signal intensity of the detected sideband component to the signal intensity at the first power supply frequency and the first load factor. A diagnostic device for an electric motor according to any one of claims 1 to 9.

14. The control unit, Based on the signal strength of the sideband component corrected by the correction control, the presence or absence of an abnormality in the electric motor is detected. The motor diagnostic device according to claim 13.

15. A diagnostic device for an electric motor according to any one of claims 1 to 9, The system includes a notification unit that connects to the motor diagnostic device via a network and notifies the administrator of the diagnostic results obtained by the motor diagnostic device, Electric motor diagnostic system.