Condition diagnostic device, condition diagnostic method, and abnormality sign inference device for an electric motor

Abstract

Provided is a condition diagnosis device for an electric motor, wherein a characteristic peak extraction unit extracts peak intensities of sideband waves due to a belt transmission frequency from peak intensities of sideband waves detected by a calculation unit from a spectrum waveform, and a judgment unit compares the extracted peak intensities with a peak intensity threshold to perform condition diagnosis for the electric motor. This reduces the number of parameters required for diagnosis and shortens the time required for diagnosis.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to a condition diagnosis device, a condition diagnosis method, and an abnormality sign inference device for an electric motor. TECHNICAL BACKGROUND

[0002] An electric motor is used as a drive source for production line devices, mechanical equipment, and the like in an industrial plant and is indispensable in the industry. Therefore, the electric motor is required to ensure normal and stable operation at all times.

[0003] In traditional electric motor condition diagnosis, an inspector visually checks the operating condition and listens to the operating noise during operation to determine whether there is any abnormality. However, this method relies on the inspector's experience or intuition and therefore has large fluctuations in diagnosis accuracy. Furthermore, many electric motors are at high risk of sudden failure, which is why interest in continuous monitoring technologies is increasing.

[0004] Examples of technologies for continuously monitoring an electric motor include a method for diagnosing abnormalities by detecting the current applied to an electric motor. For example, Patent Document 1 discloses a method for detecting abnormalities in a rotating machine system, in which the current applied to an electric motor is measured and subjected to frequency analysis to detect an abnormality of the electric motor. CITATION LISTPATENT DOCUMENT

[0005] Patent Document 1: Japanese Patent Laid-Open Publication No. 2017-181437 SUMMARY OF THE INVENTION PROBLEM TO BE SOLVED BY THE INVENTION

[0006] In the above-mentioned method for detecting abnormalities in a rotating machinery system, the acquired current information about the electric motor is subjected to a fast Fourier transform, and a characteristic frequency is extracted from the resulting analysis result. Subsequently, a deterioration parameter is calculated from a peak value thereof to determine the degree of deterioration of an abnormality in a rotating machinery system.

[0007] However, for the deterioration parameter used to determine an abnormality in the rotating machine system, it is necessary to calculate a variety of parameters such as an effective current value, a three-phase current balance, and an individual current harmonic distortion factor.

[0008] The present disclosure has been created to solve the above problem.

[0009] An object of the present disclosure is to provide a condition diagnosis device, a condition diagnosis method, and an abnormality sign inference device for an electric motor, which can perform a condition diagnosis for the electric motor using a peak intensity of a sideband wave due to a natural frequency of a belt transmission portion as a parameter without having to calculate a plurality of parameters. MEANS TO SOLVE THE PROBLEM

[0010] A condition diagnosis device for an electric motor according to the present disclosure comprises: a current input unit that receives current data of the electric motor detected by a current detector; an analysis unit that performs frequency analysis of the current data received by the current input unit to calculate an analysis result; a calculation unit that detects a plurality of peak intensities of sideband waves with respect to a power supply frequency of the electric motor from the analysis result; a characteristic peak extraction unit that calculates a natural frequency of a belt transmission section based on evaluation information of the electric motor and extracts, from the plurality of peak intensities detected by the calculation unit, the peak intensity of the sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor as a reference;and a judgment unit that compares the peak intensity extracted by the characteristic peak extraction unit with a peak intensity threshold calculated from the normally operated electric motor to perform a diagnosis for the electric motor.;

[0011] A condition diagnosis method for an electric motor according to the present disclosure comprises the steps of: Receiving current data detected by a current detector from an electric motor; Performing a frequency analysis of the received current data to calculate an analysis result; Detecting a plurality of peak intensities of sideband waves with respect to a power supply frequency of the electric motor from the analysis result; Calculating a natural frequency of a belt transmission section based on evaluation information of the electric motor and Extracting, from the plurality of detected peak intensities, the peak intensity of the sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor as a reference; and Comparing the peak intensity to a peak intensity threshold calculated from the normally operating electric motor to perform a diagnosis for the electric motor.

[0012] An abnormality sign inference device for an electric motor according to the present disclosure comprises: a data acquisition unit that acquires current data and a diagnosis result for the electric motor associated with current data that has not been subjected to frequency analysis from the above-described state diagnosis device for the electric motor; a trained model generation unit that learns the current data acquired by the data acquisition unit based on the diagnosis result to generate a trained model; and a current data inference unit that derives an abnormality sign of the electric motor from the current data newly acquired by the data acquisition unit using the trained model and outputs an inference result. EFFECT OF THE INVENTION

[0013] In the condition diagnosis device and the condition diagnosis method for the electric motor according to the present disclosure, the characteristic peak extraction unit extracts peak intensities of sideband waves due to a belt transmission frequency from peak intensities of sideband waves detected by the calculation unit from a spectrum waveform, and the judgment unit compares the extracted peak intensities with a peak intensity threshold to perform condition diagnosis for the electric motor. This reduces the number of parameters required for diagnosis and shortens the time required for diagnosis.

[0014] In the abnormality sign inference device for the electric motor according to the present disclosure, before the electric motor condition diagnosis device diagnoses an abnormality sign from the current data, the abnormality sign inference device derives an abnormality sign from the current data and outputs the presence or absence of an abnormality sign to the monitoring device. This enables scheduled maintenance of the electric motor. BRIEF DESCRIPTION OF THE DRAWINGS [ Fig. 1] Fig. 1 shows an overall system configuration of a condition diagnosis device for an electric motor according to Embodiment 1. [ Fig. 2] Fig. 2 shows a hardware configuration of the condition diagnosis device for the electric motor according to Embodiment 1. [ Fig. 3] Fig. 3 shows a configuration of a calculation processing unit of the state diagnosis device for the electric motor according to Embodiment 1. [ Fig. 4] Fig. 4 shows a connection relationship between the electric motor, a belt transmission portion, and a load equipment with respect to the state diagnosis device for the electric motor according to Embodiment 1. [ Fig. 5] Fig. 5 is a flowchart showing a processing procedure in learning the state diagnosis device for the electric motor according to Embodiment 1. [ Fig. 6] Fig. 6 is a flowchart showing a processing procedure in diagnosis by the state diagnosis device for the electric motor according to Embodiment 1. [ Fig. 7] Fig. 7 is a flowchart showing a processing procedure of frequency analysis in the state diagnosis device for the electric motor according to Embodiment 1. [ Fig. 8] Fig. Figure 8 shows the comparison of analysis results in the normal case and the abnormal case of the electric motor. [ Fig. 9] Fig. 9 shows a configuration of a calculation processing unit of a state diagnosis device for an electric motor according to Embodiment 2. [ Fig. 10] Fig. 10 is a flowchart showing a processing procedure in learning the state diagnosis device for the electric motor according to Embodiment 2. [ Fig. 11] Fig. 11 is a flowchart showing a processing procedure in diagnosis by the state diagnosis device for the electric motor according to Embodiment 2. [ Fig. 12] Fig. 12 shows an overall system configuration of a condition diagnosis device for an electric motor according to Embodiment 3. [ Fig. 13] Fig. 13 shows an overall system configuration of an abnormality sign inference device for an electric motor according to Embodiment 4. [ Fig. 14] Fig. 14 is a flowchart showing a processing procedure in learning an abnormality sign inference device for the electric motor according to Embodiment 4. [ Fig. 15] Fig. 15 is a flowchart showing a processing procedure when using the abnormality sign inference device for the electric motor according to Embodiment 4. DESCRIPTION OF EMBODIMENTS

[0015] Embodiments will be described in detail below with reference to the drawings. The embodiments described below serve only as examples. Any of the embodiments can be combined with one another as needed. Embodiment 1

[0016] Fig. 1 shows an overall system configuration of a condition diagnosis device for an electric motor according to Embodiment 1. Fig. 2 shows a hardware configuration of the condition diagnosis device for the electric motor according to Embodiment 1. Fig. 3 shows a configuration of a calculation processing unit of the state diagnosis device for the electric motor according to Embodiment 1. Fig. 4 shows a connection relationship between the electric motor, a belt transmission portion, and a load equipment with respect to the state diagnosis device for the electric motor according to Embodiment 1. Fig. 5 is a flowchart showing a processing procedure in learning the state diagnosis device for the electric motor according to Embodiment 1. Fig. 6 is a flowchart showing a processing procedure in diagnosis by the state diagnosis device for the electric motor according to Embodiment 1. Fig. 7 is a flowchart showing a processing procedure of frequency analysis in the state diagnosis device for the electric motor according to Embodiment 1. Fig. Figure 8 shows the comparison of analysis results in the normal case and the abnormal case of the electric motor.

[0017] As in Fig. 1, a circuit breaker 2, an electromagnetic contactor 3, a current detector 4, and an electric motor 5 are connected to the main circuit 1. The circuit breaker 2 prevents an overcurrent from flowing into the electric motor 5 when an abnormality has occurred in the grid. The electromagnetic contactor 3 serves as a switch for turning the electric motor 5 on and off. The current detector 4 detects the load current of the main circuit 1 and outputs the detected current to a current input unit 7 of a condition diagnosis device 100. The electric motor 5 is connected to mechanical equipment 6 constituting a production line or the like, and is operated and driven by a power conversion device or the like (not shown). The electric motor 5 is, for example, a three-phase asynchronous motor, but is not limited thereto.

[0018] The condition diagnosis device 100 includes a current input unit 7, a calculation processing unit 8, an evaluation information storage unit 9, an evaluation information setting unit 10, a display unit 11, a contactor driving unit 12, an output unit 13, and a communication unit 14. The current input unit 7 receives current data detected by the current detector 4 and outputs the current data to the calculation processing unit 8. The calculation processing unit 8 analyzes the current data input from the current input unit 7 and calculates a parameter used for diagnosing the electric motor 5 to perform abnormality diagnosis for the electric motor 5. The details of the calculation processing unit 8 will be described later.

[0019] Fig. 2 shows a hardware configuration of the condition diagnostic device 100 according to the present disclosure. The condition diagnostic device 100 is constituted by a processor 20, such as a central processing unit (CPU), a memory 30, such as a random access memory (RAM), a display 40, and an input / output interface (I / F) 50. The calculation processing unit 8 is realized by the processor 20, which executes a program stored in the memory 30. This can be implemented, for example, through the cooperation of multiple processors 20.

[0020] Evaluation information such as a power supply frequency, a rated power, a rated current, the number of poles, and a rated rotational speed of the electric motor 5 are inputted in advance to the evaluation information setting unit 10. The evaluation information is basic information listed in a manufacturer's catalog, manual, or the like for the electric motor 5. In the present embodiment, a case is shown where the number of electric motors 5 to be diagnosed is one. However, if a plurality of electric motors 5 are diagnosed, the evaluation information of each electric motor 5 is inputted in advance to the evaluation information setting unit 10.

[0021] The evaluation information storage unit 9 stores the evaluation information inputted into the evaluation information setting unit 10 and outputs the evaluation information to the calculation processing unit 8 as needed. Fig. 1 illustrates a case where the evaluation information storage unit 9 and the evaluation information setting unit 10 are provided separately, but the configuration is not limited thereto. The evaluation information storage unit 9 and the evaluation information setting unit 10 may be provided together and configured to store the evaluation information of the electric motor 5 and output it to the calculation processing unit 8 at the time of analyzing the current data input from the current input unit 7.

[0022] The evaluation information storage unit 9 and the evaluation information setting unit 10 do not necessarily have to be included in the condition diagnosis device 100 and may be provided externally. In this case, the evaluation information of the electric motor 5 may be stored, for example, in an external server and transmitted to the calculation processing unit 8 at the time of diagnosing the electric motor 5.

[0023] The display unit 11, the contactor driver unit 12, the output unit 13, and the communication unit 14 are connected to the calculation processing unit 8. When the calculation processing unit 8 detects an abnormality of the electric motor 5, the display unit 11 displays the detected current data and outputs the abnormal condition, a warning, or the like. The warning may also be provided by an alarm or the like instead of a display.

[0024] When the calculation processing unit 8 detects an abnormality of the electric motor 5, the contactor driver unit 12 outputs a control signal to open or close the electromagnetic contactor 3. The output unit 13 outputs the abnormal condition, a warning, or the like to a production line management department. The communication unit 14 transmits the abnormal condition of the electric motor 5 or the like via a network to a monitoring device 200, such as a PC or a tablet terminal, which serves as an external master device for the entire condition diagnosis device 100. Data transmission from the communication unit 14 to the monitoring device 200 can be performed using a wired or wireless method.

[0025] Next, the configuration of the calculation processing unit 8 will be described with reference to Fig. 3. As described in Fig. As shown in Figure 3, the calculation processing unit 8 includes a current fluctuation calculation unit 110, an analysis range determination unit 111, a calculation unit 112, a reference value storage unit 113, a diagnosis result storage unit 114, an analysis unit 120, and a diagnosis unit 130. The analysis unit 120 includes a frequency analysis unit 121, a peak detection calculation unit 122, a rotation frequency band extraction unit 123, a frequency axis conversion unit 124, and an averaging processing unit 125. The diagnosis unit 130 includes a characteristic peak extraction unit 131 and a judgment unit 132.

[0026] The current fluctuation calculation unit 110 calculates the presence or absence of fluctuations in the current data input from the current input unit 7 and determines whether the current data is in a stable state. Specifically, an analysis of the statistical fluctuations in the current data is performed. For example, an analysis method such as the Mahalanobis distance is used to determine whether the electric motor 5 is in a stable state.

[0027] The current fluctuation calculation unit 110 stores a range determination threshold, which is obtained in advance from the normally operating electric motor 5, as a threshold for determining whether the current data is in a stable state. For example, for the range determination threshold, current data from a plurality of electric motors 5 are acquired in advance, and the range determination threshold is set within a range smaller than one standard deviation of the acquired current data. Alternatively, current data from the electric motor 5 to be diagnosed may be acquired over a certain period of time, and the range determination threshold may be set using the standard deviation or the like of the acquired current data as a reference.The range determination threshold may also be stored in the evaluation information storage unit 9 together with the evaluation information of the electric motor 5 instead of being stored in the current fluctuation calculation unit 110.

[0028] The analysis range determination unit 111 compares a calculation result of the current fluctuation calculation unit 110 with the range determination threshold calculated in advance from the normally operating electric motor to determine an analysis range. That is, from the current data input from the current input unit 7, the analysis range determination unit 111 extracts a range where the current data is in a stable state to determine an analysis range within which the analysis unit 120 performs analysis. Specifically, a range in which the statistical fluctuation calculated by the current fluctuation calculation unit 110 does not exceed a predetermined threshold is determined as the range in which the current data is in a stable state, that is, as the analysis range.

[0029] The analysis unit 120 includes the frequency analysis unit 121, the peak detection calculation unit 122, the rotation frequency band extraction unit 123, the frequency axis conversion unit 124, and the averaging processing unit 125.

[0030] The frequency analysis unit 121 performs frequency analysis of the current data input from the current input unit 7 within the analysis range extracted by the analysis range determination unit 111 to calculate a spectral waveform as the analysis result. The current data is analyzed by, for example, a fast Fourier transform (FFT) analysis, a discrete Fourier transform, or the like.

[0031] The peak detection calculation unit 122 detects a plurality of sideband waves with respect to the power supply frequency from the spectrum waveform obtained by the frequency analysis unit 121 and calculates the peak intensities of the detected sideband waves. Specifically, the peak detection calculation unit 122 extracts regions where steep gradients are reversed by calculating the first, second, and third derivatives from the detected sideband waves with respect to the power supply frequency to calculate the peak intensities. In this way, a differential calculation up to the third order is performed when calculating the peak intensities, allowing even small peak intensities to be detected.

[0032] The rotation frequency band extraction unit 123 calculates the rotation frequency from the evaluation information of the electric motor 5 and extracts peak intensities of sideband waves with respect to the power supply frequency, which occur at positions spaced by a distance of the rotation frequency from the power supply frequency of the electric motor 5, from the peak intensities of the plurality of sideband waves detected by the peak detection calculation unit 122.Specifically, the rotation frequency band extraction unit 123 calculates the rotation frequency from the rated rotation speed in the evaluation information stored in the evaluation information storage unit 9, and, using the calculated rotation frequency, extracts sideband waves having approximately equal peak intensities that occur at positions spaced by a distance of the rotation frequency toward both the higher and lower frequency sides from the power supply frequency.

[0033] Regarding the calculation of the rotation frequency, the rotation frequency band extraction unit 123 calculates the rotation frequency using the evaluation information stored in the evaluation information storage unit 9 according to the following formula (1). Here, fn denotes the rotation frequency, f the power supply frequency, P the number of poles, and s the slip of the electric motor 5. (Mathematical Formula 1) fn=(2×f) / P×(1−s)

[0034] As described above, the rotation frequency band extraction unit 123 calculates the rotation frequency using Formula (1) and extracts peak intensities of sideband waves related to the power supply frequency that occur in rotation frequency bands from the peak intensities of the sideband waves calculated by the peak detection calculation unit 122. Here, the rotation frequency bands refer to frequencies spaced by a distance of the rotation frequency from the power supply frequency of the electric motor 5. This refers to two positions: a position spaced by the rotation frequency toward the lower frequency side from the power supply frequency and a position spaced by the rotation frequency toward the higher frequency side.

[0035] The frequency axis conversion unit 124 calculates a correction value for aligning the frequency axes of the spectral waveforms obtained by frequency analysis and corrects the spectral waveforms calculated by the frequency analysis unit 121. In a spectral waveform obtained by frequency analysis of the current data, the rotation frequency shifts depending on the load torque of the electric motor 5. Therefore, the frequency axis conversion unit 124 calculates a correction value for spectral waveforms and performs correction to align the frequency axes of the spectral waveforms. This ensures that the averaging processing unit 125 described later can perform averaging processing with a plurality of spectral waveforms accurately overlapping without shifting.

[0036] The correction value calculated by the frequency axis conversion unit 124 is described below. According to the load torque state of the electric motor 5, the slip of the electric motor 5 changes, and a rotation frequency fn calculated by the rotation frequency band extraction unit 123 using Formula (1) also changes according to the slip of the electric motor 5.

[0037] For example, in a case where a power supply frequency f is 60 Hz and the number of poles P is 4, the rotation frequency fn under the condition that there is no slip (s = 0) is calculated by the following formula (2). (Mathematical Formula 2) fn=(2×60) / 4×(1−0)=30 Hz

[0038] On the other hand, in a case where the slip is 2% depending on the state of the load torque of the electric motor 5, the rotation frequency fn is calculated by the following formula (3). (Mathematical Formula 3) fn=(2×60) / 4×(1−0.02)=30.6 Hz

[0039] Thus, the value of the rotation frequency fn calculated using formula (1) changes depending on the state of the load torque of the electric motor 5. Therefore, the frequency axis conversion unit 124, taking into account the shift due to slip described above, performs a correction for the shift from a reference corresponding to a state in which there is no slip, that is, an idling state. Specifically, in a case where the slip is 2%, according to formula (3), a shift of 0.06 Hz occurs compared to the rotation frequency in the idling state. In such a case, the frequency axis conversion unit 124 corrects the spectral waveforms with a correction value of 0.06 Hz. Thus, the frequency axis conversion unit 124 uses the rotation frequency in a state in which there is no slip in the electric motor 5, that is,in the idle state, as a reference and corrects each of the spectral waveforms by a correction value corresponding to a difference from the reference if the rotation frequency calculated by the rotation frequency band extraction unit 123 deviates from the reference.

[0040] The averaging processing unit 125 performs processing for averaging a plurality of spectral waveforms corrected by the frequency axis conversion unit 124. Thus, after the frequency axis conversion unit 124 calculates a correction value for aligning the rotation frequency and corrects each spectral waveform, the averaging processing is performed in a state where the frequency axes of the plurality of spectral waveforms are aligned and overlapped, whereby the plurality of spectral waveforms can be averaged in a precisely superimposed state.

[0041] The calculation unit 112 acquires a plurality of peak intensities of sideband waves with respect to the power supply frequency from the analysis result of the analysis unit 120. Specifically, the calculation unit 112 acquires a plurality of peak intensities of sideband waves with respect to the power supply frequency from the spectral waveform that has undergone averaging processing in the averaging processing unit 125. The output result of the calculation unit 112 is output to the characteristic peak extraction unit 131 of the diagnosis unit 130.

[0042] The diagnosis unit 130 includes the characteristic peak extraction unit 131 and the judgment unit 132.

[0043] Fig. 4 shows the transmission of vibrations to a belt transmission section 60 when an abnormality has occurred in the electric motor 5. Here, the belt transmission section 60 includes a belt and gears connecting the electric motor 5 to the load equipment 70.

[0044] The characteristic peak extraction unit 131 extracts peak intensities of sideband waves due to the natural frequency (hereinafter referred to as the belt transmission frequency) that occur when an abnormality exists in the belt transmission section 60 that connects the electric motor 5 to the load equipment 70. That is, the characteristic peak extraction unit 131 extracts peak intensities of sideband waves due to the belt transmission frequency from the plurality of peak intensities of the sideband waves acquired by the calculation unit 112. The belt transmission frequency, which is the natural frequency that occurs when an abnormality exists in the belt transmission section 60, may be the natural frequency that occurs when an abnormality exists in the belt or the natural frequency that occurs when an abnormality exists in the gears.

[0045] The extraction method is described below. The characteristic peak extraction unit 131 calculates the belt transmission frequency based on the evaluation information stored in the evaluation information storage unit 9, and extracts sideband waves with approximately equal peak intensities that occur at positions spaced apart from the power supply frequency by a distance of the belt transmission frequency in both the higher and lower frequency directions.

[0046] The belt transmission frequency is calculated using the evaluation information of the electric motor 5 stored in the evaluation information storage unit 9 according to the following formula (4). Here, F denotes the belt transmission frequency, f the power supply frequency, and P the number of poles. (Mathematical Formula 4) F=(2×f) / P×(reduction ratio)

[0047] For example, in a case where the evaluation information of the electric motor 5 specifies that the power supply frequency f is 60 Hz, the number of poles P is 4, and the reduction ratio is 1:3, the belt transmission frequency F can be calculated as (2 × 60) / 4 × (1 / 3) = 10 Hz according to formula (1).

[0048] Thus, in the above example, the characteristic peak extraction unit 131 extracts sideband waves with approximately equal peak intensities occurring at positions 10 Hz apart from the power supply frequency in both the higher and lower frequency directions. This means that sideband waves with approximately equal peak intensities are extracted on both the higher and lower frequency sides of the power supply frequency.

[0049] Fig. 4 shows a connection relationship between the electric motor 5, the belt transmission section 60 and the load equipment 70. As in Fig. 4, when an abnormality has occurred in the electric motor 5, the vibration of the electric motor 5 is transmitted to the belt transmission section 60 connected to the electric motor 5. Also, when an abnormality has occurred in the load equipment 70, the vibration of the load equipment 70 is transmitted to the belt transmission section 60. Therefore, by extracting sideband waves due to the belt transmission frequency by the characteristic peak extraction unit 131, it is possible to detect not only an abnormality of the belt transmission section 60, but also an abnormality of the electric motor 5 and the load equipment 70 connected to the belt transmission section 60.

[0050] The judgment unit 132 compares the peak intensities of the sideband waves due to the belt transmission frequency extracted by the characteristic peak extraction unit 131 with a peak intensity threshold, which is a peak intensity of a sideband wave calculated using the current data of a normally operated electric motor 5 and stored in advance in the reference value storage unit 113, to diagnose whether an abnormality exists in the electric motor 5. A specific processing method for diagnosis will be described later with reference to Fig. 7 described.

[0051] In a case where the electric motor 5 is diagnosed as abnormal by the judgment unit 132 of the diagnosis unit 130, the calculation processing unit 8 outputs information to notify the display unit 11, the contactor driving unit 12, the output unit 13, and the communication unit 14 of the abnormality, and the diagnosis result is stored in the diagnosis result storage unit 114. The diagnosis result of the electric motor 5 may be stored in the diagnosis result storage unit 114 in such a way as to be associated with the current data that has not been subjected to frequency analysis by the frequency analysis unit 121.

[0052] The reference value storage unit 113 stores a peak intensity of a sideband wave due to the belt transmission frequency detected in advance from the current data of a normally operated electric motor 5 as a reference value used for diagnosing the electric motor 5, that is, the peak intensity threshold. The peak intensity threshold may be stored together with the evaluation information in the evaluation information storage unit 9. A specific processing method therefor will be described below with reference to Fig. 5 described.

[0053] Next, the processing methods for learning and diagnosis are described with reference to Fig. 5 to Fig. 7 described.

[0054] First, the processing procedure for learning is described with reference to Fig. 5 described. Fig. Figure 5 is a flowchart showing a calculation method for the peak intensity threshold used in performing the comparison with peak intensities of sideband waves.

[0055] In step S101, the evaluation information of the normally operating electric motor 5 is input into the evaluation information setting unit 10 and stored in the evaluation information storage unit 9. The evaluation information of the electric motor 5 can be input into and stored in the evaluation information storage unit 9. If a plurality of electric motors 5 are to be diagnosed, the evaluation information of each electric motor 5 is input and stored.

[0056] In step S102, the current of the electric motor 5 to be diagnosed detected by the current detector 4 is input to the current input unit 7.

[0057] In step S103, the current fluctuation calculation unit 110 calculates the fluctuation in the current data input to the current input unit 7, that is, whether the current data is in a stable state or not. The analysis range determination unit 111 extracts a range in which the current data is in a stable state and determines this range as the analysis range in which the frequency analysis is to be performed. The current fluctuation calculation unit 110 performs calculations on the current data and compares them with the range determination threshold. If the range determination threshold is not met, the current data is determined to be unstable, and the process returns to step S102. If the current data is determined to be stable, the process proceeds to step S104.

[0058] In step S104, the frequency analysis unit 121 performs frequency analysis of the current data in the analysis range judged to be stable to calculate a spectral waveform representing an analysis result. The spectral waveform obtained by the frequency analysis is output to the calculation unit 112. The details of the frequency analysis in step S104 will be described later with reference to Fig. 7 described.

[0059] In step S105, the characteristic peak extraction unit 131 calculates the belt transmission frequency due to the belt transmission portion 60 corresponding to the electric motor 5 from the information of the electric motor 5 stored in the evaluation information storage unit 9.

[0060] In step S106, the characteristic peak extraction unit 131 extracts the peak intensities of sideband waves due to the belt transmission frequency from the peak intensities of the sideband waves input from the calculation unit 112. That is, using the belt transmission frequency calculated in step S105, the characteristic peak extraction unit 131 extracts sideband waves with approximately equal peak intensities occurring at positions spaced apart by a distance of the belt transmission frequency in both the higher and lower frequency sides from the power supply frequency.

[0061] In step S107, the reference value storage unit 113 stores the peak intensity of the sideband waves extracted by the characteristic peak extraction unit 131 in step S106 as a threshold value for diagnosis, that is, the peak intensity threshold.

[0062] Next, the processing procedure for diagnosis is described with reference to Fig. 6 described. Fig. Figure 6 is a flowchart showing the processing procedure in diagnosis.

[0063] In step S201, the current of the electric motor 5 to be diagnosed detected by the current detector 4 is input to the current input unit 7.

[0064] In step S202, the current fluctuation calculation unit 110 calculates the fluctuation in the current data input to the current input unit 7, that is, whether the current data is in a stable state or not. The analysis range determination unit 111 extracts a range in which the current data is in a stable state and determines this range as the analysis range in which the frequency analysis is to be performed. The current fluctuation calculation unit 110 performs calculations on the current data and compares them with the range determination threshold. If the range determination threshold is not met, the current data is determined to be unstable, and the process returns to step S201. If the current data is determined to be stable, the process proceeds to step S203.

[0065] In step S203, the frequency analysis unit 121 performs frequency analysis of the current data in the analysis range judged to be stable in step S202 to calculate a spectral waveform representing an analysis result. The spectral waveform obtained by the frequency analysis is input to the calculation unit 112. The details of the frequency analysis will be described later with reference to Fig. 7 described.

[0066] In step S204, the characteristic peak extraction unit 131 calculates the belt transmission frequency due to the belt transmission portion 60 of the electric motor 5 to be diagnosed from the evaluation information of the electric motor 5 stored in the evaluation information storage unit 9.

[0067] In step S205, the characteristic peak extraction unit 131 extracts the peak intensities of sideband waves due to the belt transmission frequency from the peak intensities of the sideband waves input from the calculation unit 112. That is, using the belt transmission frequency calculated in step S105, the characteristic peak extraction unit 131 extracts sideband waves with approximately equal peak intensities occurring at positions spaced apart by a distance of the belt transmission frequency in both the higher and lower frequency sides from the power supply frequency.

[0068] In step S206, the judging unit 132 compares the peak intensities of the sideband waves due to the belt transmission frequency extracted in step S205 with the peak intensity threshold stored in the reference value storage unit 113.

[0069] In step S207, the judgment unit 132 determines whether an abnormality exists in the electric motor 5 based on the comparison between the peak intensities of the sideband waves due to the belt transmission frequency and the peak intensity threshold stored in the reference value storage unit 113. If the peak intensities of the sideband waves satisfy the peak intensity threshold, the judgment unit 132 determines that an abnormality exists in the electric motor 5, and the process proceeds to step S208. Otherwise, the judgment unit 132 determines that the electric motor 5 is normal, and the process returns to step S201.

[0070] In step S208, the judgment unit 132 outputs the diagnosis result to the display unit 11, the contactor driver unit 12, the output unit 13, and the communication unit 14. Furthermore, the diagnosis result is stored in the diagnosis result storage unit 114.

[0071] Next, the frequency analysis is performed in steps S104 and S203 with reference to Fig. 7 described in detail. Fig. 7 shows the process of frequency analysis. The Fig. 7 The process shown in steps S301 to S308 corresponds to the process in steps S101 to S103 in Fig. 5 and the process in steps S201 to S203 in Fig. 6. In Fig. 7, step S301 corresponds to steps S102 and S201, and step S302 corresponds to steps S103 and S202, so their description is omitted.

[0072] In step S303, the frequency analysis unit 121 performs frequency analysis of the current data in the analysis range judged to be stable to calculate a spectral waveform representing an analysis result. The frequency analysis unit 121 outputs the calculated spectral waveform to the peak detection calculation unit 122.

[0073] In step S304, the peak detection calculation unit 122 detects a plurality of peak intensities of sideband waves with respect to the power supply frequency from the spectrum waveform which is a result of frequency analysis and which is input from the frequency analysis unit 121.

[0074] In step S305, the rotation frequency band extraction unit 123 extracts the peak intensities of sideband waves occurring in the rotation frequency bands from the plurality of peak intensities of the sideband waves detected by the peak detection calculation unit 122. The rotation frequency band extraction unit 123 extracts the peak intensities of the sideband waves occurring in the rotation frequency bands on both the higher and lower frequency sides of the power supply frequency. The extraction of the peak intensities is performed by calculating the rotation frequency from the rated rotation speed in the evaluation information stored in the evaluation information storage unit 9 and by extracting sideband waves occurring at positions spaced by a distance of the rotation frequency, with the power supply frequency as the center.

[0075] In step S306, the frequency axis conversion unit 124 calculates a correction value for the spectrum waveform and performs correction to align the frequency axis of the spectrum waveform using the calculated correction value to reduce the influence of the rotation frequency shift due to the load torque. This can eliminate the rotation frequency shift depending on the load torque state.

[0076] In step S307, the averaging processing unit 125 repeats steps S301 to S306 multiple times to collect a plurality of spectral waveforms that have undergone steps S301 to S306. The number of repetitions is not particularly limited. For example, the number of repetitions may be set in the evaluation information storage unit 9, and the above steps may be repeated the set number of times. Once the number of repetitions is reached, the process proceeds to step S318.

[0077] In step S308, the frequency-aligned spectral waveforms collected in step S307 are superimposed and subjected to averaging processing. This allows the influence of noise or the like to be removed.

[0078] As described above, the condition diagnosis method for the electric motor 5 includes the steps of: receiving current data detected by the current detector 4 from the electric motor 5; performing frequency analysis of the received current data to calculate an analysis result; detecting a plurality of peak intensities of sideband waves related to the power supply frequency of the electric motor 5 from the analysis result; calculating the natural frequency of the belt transmission section 60 from the evaluation information of the electric motor, and extracting, from the plurality of detected peak intensities, a peak intensity of a sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor 5 as a reference;and comparing the peak intensity with the peak intensity threshold calculated from the normally operating electric motor 5 to perform a diagnosis for the electric motor 5. This makes it possible to perform a condition diagnosis for the electric motor 5 using a single parameter, namely the peak intensity of the sideband wave due to the belt transmission frequency, without having to calculate a plurality of parameters.

[0079] Fig. 8 shows the comparison of frequency analysis results in the normal case and the abnormal case of the electric motor 5. A solid line shows the analysis result in a case where the electric motor 5 is normal, and a dashed line shows the analysis result in a case where the electric motor 5 is abnormal. As in Fig. As shown in Figure 8, when an abnormality has occurred in the electric motor 5, the peak intensities of the sideband waves due to the belt transmission frequency increase compared to the normal case. Therefore, in the condition diagnosis device 100 for the electric motor 5 according to the present embodiment, the peak intensity threshold, which is a threshold value for a peak intensity of a sideband wave due to the belt transmission frequency obtained in advance from the normally operating electric motor 5, is used as a threshold value for judging whether the electric motor 5 is abnormal. If the peak intensity threshold is exceeded, an abnormality in the electric motor 5 is diagnosed.

[0080] As described above, the condition diagnosis device 100 for the electric motor 5 according to the present embodiment performs frequency analysis of the current data acquired from the electric motor 5 and performs condition diagnosis for the electric motor 5 by using a peak intensity of a sideband wave due to the belt transmission frequency extracted by the characteristic peak extraction unit 131 from a spectrum waveform that is an analysis result. This makes it possible to perform condition diagnosis for the electric motor 5 using a single parameter, namely the peak intensity of the sideband wave due to the belt transmission frequency, as a parameter in the diagnosis. This reduces the number of parameters required for diagnosis and shortens the time required for diagnosis.

[0081] In the present embodiment, the configuration in which the analysis unit 120 includes the peak detection calculation unit 122, the rotation frequency band extraction unit 123, the frequency axis conversion unit 124, and the averaging processing unit 125 has been described. However, the present disclosure is not limited to this. The frequency analysis unit 121 may perform frequency analysis on the current data input from the current input unit 7 to calculate a spectral waveform as the analysis result, and output the calculated spectral waveform to the calculation unit 112. In this case, the analysis unit 120 only needs to include the frequency analysis unit 121.

[0082] In the above description, it has been illustrated that the condition diagnosis device 100 according to the present embodiment diagnoses an abnormality when a peak intensity of a sideband wave extracted by the characteristic peak extraction unit 131 satisfies the peak intensity threshold. However, the method for judging whether the peak intensity threshold is satisfied is not limited to this. For example, it may be determined that an abnormality exists when the peak intensity threshold has been satisfied for a certain period of time. Alternatively, a certain number of occurrences may be set, and it may be decided that an abnormality exists when the number of times the peak intensity threshold is exceeded exceeds the set number. Example 2

[0083] The present embodiment will be described with reference to Fig. 9 to Fig. 11 described. Fig. 9 shows a configuration of a calculation processing unit 80 of a state diagnosis device for an electric motor according to the present embodiment. Fig. 10 is a flowchart showing a processing method for learning the state diagnosis device for the electric motor according to the present embodiment. Fig. 11 is a flowchart showing a processing procedure for diagnosis by the state diagnosis device for the electric motor according to the present embodiment.

[0084] In Embodiment 1, a condition diagnosis device 100 was described in which, from a spectral waveform representing an analysis result of the analysis unit 120, the characteristic peak extraction unit 131 extracts peak intensities of sideband waves occurring at positions spaced apart from the power supply frequency by a distance of the belt transmission frequency in both the lower and higher frequency directions. Subsequently, the diagnosis unit 130 compares the extracted peak intensities of the sideband waves with the peak intensity threshold to perform condition diagnosis for the electric motor 5.The present embodiment shows a condition diagnosis device 101 for the electric motor 5, in which a calculation unit 140 includes an OA value calculation unit 141 that calculates a total value (hereinafter referred to as "OA value") used in the condition diagnosis of the electric motor 5. The other components are the same as those in Embodiment 1. The same components as in Embodiment 1 are denoted by the same reference numerals, and their description will be omitted.

[0085] As in Fig. As shown in FIG. 9, the calculation processing unit 80 of the condition diagnosis device 101 includes a current fluctuation calculation unit 110, an analysis range determination unit 111, a reference value storage unit 113, a diagnosis result storage unit 114, an analysis unit 120, a diagnosis unit 130, and a calculation unit 140. With this configuration, the diagnosis unit 130 can perform condition diagnosis for the electric motor 5 by comparing a peak intensity of a sideband wave with the peak intensity threshold. This reduces the number of parameters required for condition diagnosis, and the diagnosis time can be shortened.

[0086] Furthermore, the calculation unit 140 of the calculation processing unit 80 according to the present embodiment includes the OA value calculation unit 141.

[0087] The OA value calculation unit 141 calculates an OA value, which is an average value of the amplitudes as information for determining an abnormality of the electric motor 5, from a spectral waveform that is an analysis result calculated by the analysis unit 120. The OA value calculated in the learning process is stored in the reference value storage unit 113, and the OA value calculated in the diagnosis process is output to a judgment unit 133.

[0088] The OA value refers to the average magnitude of the amplitudes across the entire spectral waveform obtained as an analysis result by frequency analysis of the power data. Here, the signal intensities at the power supply frequency and their harmonic components account for a large proportion of the calculated OA value. Therefore, frequency components peaking at the power supply frequency and higher-order frequency components that are integer multiples of the power supply frequency—that is, harmonic components of the power supply frequency—can be excluded from the values used for calculating the OA value by the OA value calculation unit 141. By excluding values that account for a large proportion of the OA value, misdiagnosis can be prevented.

[0089] Next, the processing methods for learning and diagnosis are described with reference to Fig. 10 and Fig. 11. In Fig. 10 are the same processing steps as in Fig. 5 are provided with the same reference numerals. Likewise, Fig. 11 the same processing steps as in Fig. 6 are provided with the same reference numerals. Therefore, step S107 and step S108 are subsequently Fig. 10 and step S207, step S209 and step S210 in Fig. 11, while the description of the other steps is omitted.

[0090] First, the processing method for learning will be described. In step S108, the OA value calculation unit 141 calculates an OA value from the spectral waveform output as the analysis result from the analysis unit 120. The OA value calculated by the OA value calculation unit 141 is output to the reference value storage unit 113.

[0091] In step S107, the reference value storage unit 113 stores the OA value output from the OA value calculation unit 141 as a reference value used in diagnosis, that is, the OA threshold. Furthermore, the reference value storage unit 113 stores the peak intensity of the sideband wave due to the belt transmission frequency extracted by the characteristic peak extraction unit 131 in step S106 as a reference value used in diagnosis, that is, the peak intensity threshold. As described above, in step S107, the OA threshold and the peak intensity threshold are stored as learned data in the reference value storage unit 113. The OA threshold and the peak intensity threshold may be stored in the evaluation information storage unit 9.

[0092] Next, the processing method for diagnosis will be described. In step S209, the OA value calculation unit 141 calculates an OA value from the spectral waveform output as the analysis result from the analysis unit 120. The calculated OA value is output from the OA value calculation unit 141 to the judgment unit 133 of the diagnosis unit 130.

[0093] In step S210, the judging unit 133 compares the OA value calculated by the OA value calculating unit 141 in step S209 with the OA threshold stored in the reference value storing unit 113.

[0094] In step S207, the judgment unit 133 determines that an abnormality exists in the electric motor 5 if it is determined in step S210 that the OA value satisfies the OA threshold and it is determined in step S206 that the peak intensity satisfies the peak intensity threshold. In this case, the process proceeds to step S208. If either the OA value or the peak intensity does not satisfy the respective threshold, the judgment unit 133 determines that the electric motor 5 is normal, and the process returns to step S201.

[0095] As described above, the condition diagnosis method for the electric motor 5 includes the steps of: receiving current data detected by the current detector 4 from the electric motor 5; performing frequency analysis of the received current data to calculate an analysis result; detecting a plurality of peak intensities of sideband waves related to the power supply frequency of the electric motor 5 from the analysis result; calculating the natural frequency of the belt transmission portion 60 from the evaluation information of the electric motor 5 and extracting, from the plurality of detected peak intensities, a peak intensity of a sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor 5 as a reference; calculating an OA value from the analysis result;Comparing the peak intensity with the peak intensity threshold calculated from the normally operating electric motor 5, and comparing the OA value with the OA threshold calculated from the normally operating electric motor 5 to perform a diagnosis for the electric motor 5. This can improve the accuracy of the diagnosis result compared to the state diagnosis method for the electric motor 5 according to Embodiment 1.

[0096] As described above, the condition diagnosis device 101 for the electric motor 5 according to the present embodiment includes a calculation unit 140 including the OA value calculation unit 141, and the judgment unit 133 performs the condition diagnosis for the electric motor 5 using the OA value calculated by the OA value calculation unit 141 and the peak intensity of the sideband wave due to the belt transmission frequency extracted by the characteristic peak extraction unit 131. This reduces the number of parameters required for diagnosis and shortens the time required for diagnosis. In addition, the accuracy of the diagnosis result can be improved compared to the condition diagnosis method for the electric motor 5 according to Embodiment 1. Embodiment 3

[0097] The present embodiment will be described with reference to Fig. 12 described. Fig. 12 shows an overall system configuration of a condition diagnosis device for an electric motor according to the present embodiment.

[0098] The present embodiment shows a state diagnosis device 102 for the electric motor 5, which is configured such that a learning unit 15, a model storage unit 16, and an inference unit 17 are additionally provided in the state diagnosis device 100 for the electric motor 5 according to Embodiment 1. The other components are the same as those of Embodiment 1. The same components are denoted by the same reference numerals, and their description is omitted.

[0099] As in Fig. 12, the state diagnosis device 102 includes, for example, the current input unit 7, the calculation processing unit 81, the evaluation information storage unit 9, the evaluation information setting unit 10, the display unit 11, the contactor driving unit 12, the output unit 13, the communication unit 14, the learning unit 15, the model storage unit 16, and the inference unit 17.

[0100] In the condition diagnosis device 102, the characteristic peak extraction unit 131 extracts a peak intensity of a sideband wave due to the belt transmission frequency from the peak intensities of the sideband waves of a spectrum waveform acquired by the calculation unit 112, and the extracted peak intensity is compared with the peak intensity threshold to perform condition diagnosis for the electric motor 5. This makes it possible to perform abnormality diagnosis for the electric motor 5 using the peak intensity of the sideband wave due to the belt transmission frequency as a parameter in the diagnosis, thereby reducing the number of parameters required for diagnosis and shortening the diagnosis time.

[0101] The learning unit 15 acquires the current data acquired by the current input unit 7 and a diagnosis result for the electric motor 5 stored in the diagnosis result storage unit 114, and performs learning to associate the diagnosis result and the current data to generate a trained model. That is, the diagnosis result stored in the diagnosis result storage unit 114 is stored in such a way that it is associated with current data that has not been subjected to frequency analysis. The learning unit 15 learns a temporal pattern indicative of an abnormality based on the diagnosis result from temporal current data acquired by the current input unit 7 to generate a trained model.

[0102] When learning to create a trained model by the learning unit 15, the following method is used: A temporal pattern that frequently occurs in an abnormality case and includes current data that is not included in a normal temporal pattern is derived by deep learning from current data of a certain period before an abnormality is detected in the electric motor 5. Alternatively, known machine learning methods such as genetic programming, inductive logic programming, or a support vector machine can be used.

[0103] The model storage unit 16 stores the trained model generated by the learning unit 15. The stored trained model is not limited to the model generated by the learning unit 15. For example, pre-learned data may be read externally and stored as a trained model. The model storage unit 16 may also be provided in an external server or the like instead of being implemented in the state diagnosis device 102.

[0104] The inference unit 17 derives an abnormality sign of the electric motor 5 from current data newly acquired by the current input unit 7 using the trained model stored in the model storage unit 16, and outputs an inference result to the monitoring device 200. That is, the inference unit 17 acquires current data from the current input unit 7 and, using the trained model stored in the model storage unit 16, infers whether the current data corresponds to a temporal pattern detected as a sign of an abnormality of the electric motor 5. If an abnormality sign is present, the inference unit 17 determines that an abnormality exists in the electric motor 5 and outputs an inference result to the monitoring device 200.

[0105] Instead of outputting to the monitoring device 200, the inference result may be output to the display unit 11, the contactor driving unit 12, the output unit 13, and the communication unit 14 via the calculation processing unit 8. By detecting an abnormality sign of the electric motor 5 from the current data as described above, it is possible to plan the maintenance of the electric motor 5 in a targeted manner, thereby reducing the downtime of the mechanical equipment 6 connected to the electric motor 5 to a minimum.

[0106] As described above, the condition diagnosis device 102 for the electric motor 5 according to the present embodiment additionally includes the learning unit 15, the model storage unit 16, and the inference unit 17. The learning unit 15 generates a trained model from the current data acquired based on an analysis result, and the inference unit 17 infers an abnormality of the electric motor 5 from the current data using the trained model. With this configuration, an abnormality of the electric motor 5 can be predicted from the current data, thereby enabling smooth operation of the electric motor 5 and performing maintenance planning without unnecessary equipment downtime.

[0107] The present embodiment has shown a state diagnosis device 102 configured such that the learning unit 15, the model storage unit 16, and the inference unit 17 are additionally included in the state diagnosis device 100 of Embodiment 1. However, a configuration is also possible in which the learning unit 15, the model storage unit 16, and the inference unit 17 are additionally included in the state diagnosis device 101 of Embodiment 2. Embodiment 4

[0108] The present embodiment will be described with reference to Fig. 13 to Fig. 15 described. Fig. 13 shows an overall system configuration of an abnormality sign inference device for an electric motor according to the present embodiment. Fig. 14 is a flowchart showing a processing method for learning the abnormality sign inference means for the electric motor according to the present embodiment. Fig. 15 is a flowchart showing a processing method for using the abnormality sign inference device for the electric motor according to the present embodiment.

[0109] The present embodiment shows an abnormality sign inference device 300 for the electric motor 5, which is provided externally to the state diagnosis device 100 for the electric motor 5. The other components are the same as those in Embodiment 1. The same components are denoted by the same reference numerals, and their description is omitted.

[0110] As in Fig. 13, the condition diagnosis device 100 includes, for example, the current input unit 7, the calculation processing unit 8, the evaluation information storage unit 9, the evaluation information setting unit 10, the display unit 11, the contactor driving unit 12, the output unit 13, and the communication unit 14. In the condition diagnosis device 100, the characteristic peak extraction unit 131 extracts a peak intensity of a sideband wave due to the belt transmission frequency from the peak intensities of the sideband waves of a spectrum waveform acquired by the calculation unit 112, and the extracted peak intensity is compared with the peak intensity threshold to perform condition diagnosis for the electric motor 5.This makes it possible to perform abnormality diagnosis for the electric motor 5 using the peak intensity of the sideband wave due to the belt transmission frequency as a parameter in the diagnosis, thereby reducing the number of parameters required for diagnosis and shortening the diagnosis time.

[0111] The abnormality sign inference device 300 for the electric motor 5 includes a data acquisition unit 310, a trained model generation unit 320, a trained model storage unit 330, and a current data inference unit 340.

[0112] The data acquisition unit 310 acquires current data from the current input unit 7 of the condition diagnosis device 100 and acquires a diagnosis result for the electric motor 5 associated with current data that has not been subjected to frequency analysis from the diagnosis result storage unit 114 of the calculation processing unit 8.

[0113] The trained model generation unit 320 learns the stream data acquired by the data acquisition unit 310 based on the diagnosis result to generate a trained model. Specifically, the trained model generation unit 320 learns a temporal pattern indicative of an abnormality from temporal stream data acquired by the stream input unit 7 based on a diagnosis result to generate a trained model. The number of data to be learned for a trained model is not particularly limited. For example, the number of data to be learned may be set in advance in the trained model generation unit 320 or a similar device, and learning may be terminated when the set number of data is learned.

[0114] When learning to create a trained model, the following method is used: A temporal pattern that frequently occurs in an abnormality case and includes current data that is not included in a normal temporal pattern is derived by deep learning from current data of a certain period before an abnormality is detected in the electric motor 5. Instead of deep learning, well-known machine learning methods such as genetic programming, inductive logic programming, or a support vector machine can also be used.

[0115] The trained model storage unit 330 stores the trained model generated by the trained model generation unit 320. The stored trained model is not limited to the model generated by the trained model generation unit 320. For example, pre-learned data may be read from external sources and stored as a trained model. The trained model storage unit 330 may be provided in an external server or similar device instead of being implemented in the abnormality sign inference device 300.

[0116] The current data inference unit 340 derives an abnormality sign of the electric motor 5 from current data newly acquired by the data acquisition unit 310 using the trained model and outputs an inference result to the monitoring device 200. That is, the data acquisition unit 310 acquires current data from the current input unit 7, and the current data inference unit 340 determines whether the acquired current data corresponds to a temporal pattern detected as a sign of abnormality of the electric motor 5 using the trained model stored in the trained model storage unit 330. If an abnormality sign is present, the current data inference unit 340 determines that an abnormality exists in the electric motor 5 and outputs an inference result to the monitoring device 200.

[0117] Instead of outputting to the monitoring device 200, the inference result may be output to the display unit 11, the contactor driving unit 12, the output unit 13, and the communication unit 14 via the calculation processing unit 8. By detecting an abnormality sign of the electric motor 5 from the current data as described above, it is possible to plan the maintenance of the electric motor 5 in a targeted manner, thereby reducing the downtime of the mechanical equipment 6 connected to the electric motor 5 to a minimum.

[0118] Next, the learning procedure is described with reference to Fig. 14 described. Fig. Figure 14 is a flowchart showing the learning phase.

[0119] In step S301, the data acquisition unit 310 acquires current data from the current input unit 7 and acquires a diagnosis result associated with current data that has not been subjected to frequency analysis from the diagnosis result storage unit 114 of the calculation processing unit 8.

[0120] In step S302, the trained model generation unit 320 learns an abnormality sign of the electric motor 5 based on the current data output from the data acquisition unit 310 and the diagnosis result. That is, from current data acquired by the current input unit 7 in a certain period before the diagnosis of an abnormality of the electric motor 5, the trained model generation unit 320 learns a temporal pattern of the current data that commonly occurs in cases where an abnormality is diagnosed, to generate a trained model.

[0121] In step S303, the trained model storage unit 330 stores the trained model generated in step S302.

[0122] Next, the usage procedure will be explained with reference to Fig. 15 described. Fig. Figure 15 is a flowchart showing the utilization phase.

[0123] In step S401, the data acquisition unit 310 acquires current data from the current input unit 7.

[0124] In step S402, the current data acquired by the data acquisition unit 310 in step S401 is input into the trained model stored in the trained model storage unit 330.

[0125] In step S403, the current data inference unit 340 outputs an inference result based on the current data input to the trained model, that is, a result that determines by inference whether or not there is an abnormality sign in the electric motor 5 from the current data acquired by the data acquisition unit 310.

[0126] In step S404, the inference result output by the current data inference unit 340 is output to the monitoring device 200. Instead of outputting to the monitoring device 200, the inference result may be output to the display unit 11, the contactor driving unit 12, the output unit 13, and the communication unit 14 via the calculation processing unit 8.

[0127] As described above, the abnormality sign inference device 300 for the electric motor 5 according to the present embodiment learns current data acquired by the condition diagnosis device 100 based on a diagnosis result to generate a trained model, and inputs newly acquired current data into the trained model to derive an abnormality sign of the electric motor 5. Therefore, before the condition diagnosis device 100 for the electric motor 5 diagnoses an abnormality, the abnormality sign inference device 300 can derive an abnormality sign from the current data and output the presence or absence of an abnormality sign to the monitoring device 200. This enables targeted maintenance of the electric motor 5. Furthermore, by scheduling maintenance in advance, the unnecessary downtime of the mechanical equipment 6 connected to the electric motor 5 can be reduced.

[0128] In the present embodiment, the configuration is such that a trained model is generated from current data and a diagnosis result acquired by the condition diagnosis device 100 of Embodiment 1, and newly acquired current data is input to the trained model to derive an abnormality sign of the electric motor 5. However, a configuration is also possible in which a trained model is generated from current data and a diagnosis result acquired by the condition diagnosis device 101 of Embodiment 2, and newly acquired current data is input to the trained model to derive an abnormality sign of the electric motor 5. DESCRIPTION OF REFERENCE SYMBOLS 1 main circuit 2 circuit breakers 3 Electromagnetic contactor 4 Current detector 5 Electric motor 6 Mechanical equipment 7 Power input unit 8, 80 Calculation processing unit 9 Rating information storage unit 10 Rating information setting unit 11 Display unit 12 Contactor driver unit 13 Output unit 14 Communication unit 15 learning units 16 Model storage unit 17 Inference Unit 20 processor 30 storage 40 ad 50 Input interface 60 Belt drive section 70 load equipment 100, 101 Condition diagnostic device 110 Power fluctuation calculation unit 111 Analysis area determination unit 112, 140 Calculation unit 113 Reference value storage unit 114 Diagnostic result storage unit 120 analysis units 121 Frequency analysis unit 122 Peak detection calculation unit 123 Rotational frequency band extraction unit 124 Frequency axis conversion unit 125 averaging processing unit 130 diagnostic unit 131 Characteristic peak extraction unit 132, 133 Assessment unit 141 OA value calculation unit 200 monitoring device 300 Abnormality Sign Inference Facility 310 Data acquisition unit 320 Trained Model Generation Unit 330 Trained model storage unit 340 Power data acquisition unit QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] JP 2017-181437

[0005]

Claims

A condition diagnosis device for an electric motor, comprising: a current input unit that receives current data of the electric motor detected by a current detector; an analysis unit that performs frequency analysis of the current data received by the current input unit to calculate an analysis result; a calculation unit that detects a plurality of peak intensities of sideband waves with respect to a power supply frequency of the electric motor from the analysis result; a characteristic peak extraction unit that calculates a natural frequency of a belt transmission section based on evaluation information of the electric motor and extracts, from the plurality of peak intensities detected by the calculation unit, the peak intensity of the sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor as a reference;anda judgment unit that compares the peak intensity extracted by the characteristic peak extraction unit with a peak intensity threshold calculated from the normally operated electric motor to perform a diagnosis for the electric motor.; The condition diagnosis device for the electric motor according to claim 1, wherein the analysis unit comprises a frequency analysis unit that performs frequency analysis on the current data, a peak detection calculation unit that detects a plurality of peak intensities of sideband waves related to the power supply frequency of the electric motor from a spectrum waveform that is a frequency analysis result of the frequency analysis unit, a rotation frequency band extraction unit that calculates a rotation frequency based on evaluation information of the electric motor and extracts, from the plurality of peak intensities detected by the peak detection calculation unit, the peak intensity of the sideband wave appearing at a position spaced by a distance of the rotation frequency from the power supply frequency of the electric motor as a reference, a frequency axis conversion unit,which calculates a correction value for aligning the frequency axis of each of the spectral waveforms and corrects each of the spectral waveforms, and an averaging processing unit that performs averaging processing on a plurality of the spectral waveforms corrected by the frequency axis conversion unit, thereby calculating the analysis result. A condition diagnosis device for the electric motor according to claim 1 or 2, wherein the analysis unit performs the frequency analysis in an analysis range in which it has been determined that the current data is in a stable state. A state diagnosis device for the electric motor according to claim 3, wherein the analysis unit performs the frequency analysis of the current data in the analysis range extracted by a current fluctuation calculation unit that calculates whether or not the current data is in the stable state, and an analysis range determination unit that performs a determination by comparing a calculation result of the current fluctuation calculation unit with a range determination threshold calculated in advance from the normally operated electric motor. A condition diagnosis device for the electric motor according to any one of claims 1 to 4, wherein the calculation unit has an OA value calculation unit that calculates an OA value, which is an average value of amplitudes, from the analysis result calculated by the analysis unit. The condition diagnosis device for the electric motor according to claim 5, wherein the judging unit compares the OA value calculated by the OA value calculation unit with an OA threshold calculated from the normally operated electric motor to perform a diagnosis for the electric motor. The condition diagnosis device for the electric motor according to any one of claims 1 to 6, further comprising: a learning unit that acquires the current data acquired by the current input unit and a diagnosis result for the electric motor, and performs learning to combine the diagnosis result and the current data to generate a trained model; and an inference unit that derives an abnormality sign of the electric motor from the current data newly acquired by the current input unit using the trained model and outputs an inference result. An abnormality sign inference device for an electric motor, comprising: a data acquisition unit that acquires current data and a diagnosis result for the electric motor associated with current data that has not been subjected to frequency analysis from the condition diagnosis device for the electric motor according to any one of claims 1 to 6; a trained model generation unit that learns the current data acquired by the data acquisition unit based on the diagnosis result to generate a trained model; and a current data inference unit that derives an abnormality sign of the electric motor from the current data newly acquired by the data acquisition unit using the trained model and outputs an inference result. A condition diagnosis method for an electric motor, comprising the steps of: receiving current data detected by a current detector from an electric motor; performing frequency analysis of the received current data to calculate an analysis result; detecting a plurality of peak intensities of sideband waves with respect to a power supply frequency of the electric motor from the analysis result; calculating a natural frequency of a belt transmission portion based on evaluation information of the electric motor, and extracting, from the plurality of detected peak intensities, the peak intensity of the sideband wave appearing at a position spaced by a distance of the natural frequency from the power supply frequency of the electric motor as a reference; and comparing the peak intensity with a peak intensity threshold calculated from the normally operated electric motor to perform diagnosis for the electric motor.

Images