Gear state detection method and device based on motor current signal and support vector data description

Abstract

The invention provides a gear state detection method and device based on motor current signals and support vector data description, and relates to the technical field of detection.The method comprises the steps that the current signals of a motor in target equipment are collected, and the fault characteristic frequency of a to-be-detected gear is determined based on the tooth number of the to-be-detected gear and the rotating speed of the motor; determining a monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency; for each monitoring fault frequency, determining corresponding fault characteristic energy based on the current signal and the monitoring fault frequency; and inputting the fault characteristic energy corresponding to each monitoring fault frequency into SVDD for gear state detection to obtain a gear state detection result output by the SVDD. According to the method, the fault characteristic energy at the fault monitoring frequency is adopted as the input of SVDD, the method is more accordant with the gear fault signal characteristics, the interference of other factors except the gear fault on gear detection can be reduced, and therefore the accuracy of gear state detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of detection technologies, and particularly to a gear state detection method and device based on motor current signals and support vector data description. Background Art

[0002] Gears are one of the most widely used mechanical components in industrial systems. Gears have high load-carrying capacity, high transmission ratio, and high power-to-volume ratio, and have strong transmission stability. They are widely used in industries such as wind power, coal mines, and robots. However, due to the high-pair mating of gears, the working environment is harsh and the load is heavy, so gear wear and faults are inevitable. Common gear faults include broken teeth, wear, pitting, and cracks. Since hidden gear faults may lead to unreliable and unstable production operations, it means there are potential economic losses and safety hazards. Therefore, on-line state monitoring of gears is crucial.

[0003] In related technologies, gear states are usually detected based on machine learning methods. However, machine learning methods require fault data when the gears are in abnormal states as training data, and there is a running-in phenomenon in gear systems. Each new installation will introduce additional non-fault features, resulting in the training data itself having overfitting characteristics, thus reducing the accuracy of gear state detection. Summary of the Invention

[0004] The present invention provides a gear state detection method and device based on motor current signals and support vector data description to solve the defect of reducing the accuracy of gear state detection in the prior art.

[0005] The present invention provides a gear state detection method based on motor current signals and support vector data description, including the following steps.

[0006] Collect the current signal of the motor in the target device, where the motor is used to drive the gear to be detected; Based on the number of teeth of the gear to be detected and the rotational speed of the motor, determine the fault characteristic frequency of the gear to be detected, and based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, determine the monitoring fault frequency set; For each monitoring fault frequency in the monitoring fault frequency set, based on the current signal and the monitoring fault frequency, determine the fault characteristic energy corresponding to the monitoring fault frequency; Input the fault characteristic energy corresponding to each monitoring fault frequency into the support vector data description (SVDD) for gear state detection, and obtain the gear state detection result output by the SVDD.

[0007] A gear state detection method based on motor current signals and support vector data description provided by the present invention, wherein the gear to be detected includes a sun gear to be detected and a planet gear to be detected that mesh with each other; Based on the number of teeth of the gear to be detected and the rotational speed of the motor, determine the fault characteristic frequency of the gear to be detected, and based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, determine a set of monitored fault frequencies, including: Determine the set of monitored fault frequencies based on the following formula (1); (1) Wherein, represents the set of monitored fault frequencies, represents the monitored fault frequency, represents the frequency of the current signal, represents the order of the analyzed fault, represents the fault characteristic frequency of the sun gear to be detected or the fault characteristic frequency of the planet gear to be detected , , , represents the frequency components already existing in the motor current signal, the neighborhood of the existing frequency components, and the harmonic frequencies of the motor current signal, represents the reduction ratio of the motor, represents the rotational speed of the motor, represents the number of teeth of the sun gear to be detected, represents the number of teeth of the planet gear to be detected.

[0008] A gear state detection method based on motor current signals and support vector data description provided by the present invention, determining the fault characteristic energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency, including: Determine the fault characteristic energy corresponding to the monitored fault frequency based on the following formula (2); (2) Wherein, represents the monitored fault frequency corresponding to the fault characteristic energy, represents the 3dB bandwidth of the main lobe of the Hanning window, represents the spectral resolution, , represents the result of performing a Fourier transform on the current signal, represents the number of the frequency domain signal in the result of the Fourier transform, represents a piecewise function.

[0009] A gear state detection method based on motor current signals and support vector data description according to the present invention. Inputting the fault feature energies corresponding to each monitored fault frequency into a support vector data description (SVDD) for gear state detection to obtain the gear state detection result output by the SVDD, including: Inputting the fault feature energies corresponding to each monitored fault frequency into the SVDD, and performing kernel mapping on each fault feature energy by the SVDD based on a Gaussian kernel function to obtain a mapping vector, and determining the distance between the mapping vector and the center of the hypersphere corresponding to the SVDD; When the distance is less than or equal to the radius of the hypersphere, determining that the gear state detection result is a healthy state; When the distance is greater than the radius of the hypersphere, determining that the gear state detection result is an abnormal state.

[0010] A gear state detection method based on motor current signals and support vector data description according to the present invention. Determining the fault feature energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency, including: Based on the current signal and the monitored fault frequency, determining the initial fault feature energy corresponding to the monitored fault frequency; Determining the logarithm value of the initial fault feature energy; Based on the logarithm value range, linearly transforming the logarithm value to obtain the fault feature energy corresponding to the monitored fault frequency, where the logarithm value range is determined based on the maximum logarithm value and the minimum logarithm value among the logarithm values of all the initial fault feature energies.

[0011] A gear state detection method based on motor current signals and support vector data description according to the present invention. The SVDD is trained based on the following method: When the sample device is in a healthy state, collecting the sample current signal of the sample motor in the sample device, where the sample motor is used to drive the sample gear; Based on the frequency of the sample current signal, the sample fault feature frequency of the sample gear, and the number of teeth of the sample gear, determining a set of sample monitored fault frequencies; For each sample monitored fault frequency in the set of sample monitored fault frequencies, based on the sample current signal and the sample monitored fault frequency, determining the sample fault feature energy corresponding to the sample monitored fault frequency; Taking some of the sample fault feature energies among the multiple sample fault feature energies as training data; Construct an optimization problem based on the training data, and solve the optimization problem to obtain the support vectors of the hypersphere of the SVDD; Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies, to obtain the SVDD.

[0012] According to a gear state detection method based on motor current signals and support vector data description provided by the present invention, the constructing an optimization problem based on the training data includes: Construct an optimization problem based on the following formula (3): (3) where the constraint condition is , represents the total number of the training data, represents the th training data 's Lagrange multiplier, represents the th training data 's Lagrange multiplier, and represent the Gaussian kernel function; The solving the optimization problem to obtain the support vectors of the hypersphere of the SVDD includes: Determine the obtained when the constraint condition is satisfied as the support vectors; The determining the center of the hypersphere and the radius of the hypersphere based on the support vectors to obtain the SVDD includes: Determine the radius of the hypersphere based on the following formula (4) , and determine the center a of the hypersphere based on the following formula (5): (4) (5) where, represents the support vectors, represents the Gaussian kernel function, represents the th training data 's corresponding sample mapping vector.

[0013] According to a gear state detection method based on motor current signals and support vector data description provided by the present invention, the determining the radius of the hypersphere based on the support vectors, and determining the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies, to obtain the SVDD includes: Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies, to obtain the to-be-tested SVDD; Take another part of the sample fault feature energies among the multiple sample fault feature energies as test data and input them into the to-be-tested SVDD. Through the to-be-tested SVDD, perform kernel mapping on each of the test data based on the Gaussian kernel function to obtain test mapping vectors, and determine the test distances between each of the test mapping vectors and the center of the hypersphere; For each of the test data, determine the gear state test result corresponding to the test data based on the test distance corresponding to the test data and the radius of the hypersphere; Determine the accuracy rate of the to-be-tested SVDD based on the gear state test results corresponding to each of the test data; In the case where the accuracy rate is not within the preset accuracy rate range, fine-tune the parameters of the Gaussian kernel function of the to-be-tested SVDD until the accuracy rate is within the preset accuracy rate range, to obtain the SVDD.

[0014] The present invention further provides a gear state detection device based on motor current signals and support vector data description, including: An acquisition unit, configured to acquire the current signal of the motor in the target device, where the motor is used to drive the to-be-detected gear; A first determination unit, configured to determine the fault feature frequency of the to-be-detected gear based on the number of teeth of the to-be-detected gear and the rotational speed of the motor, and determine the monitoring fault frequency set based on the frequency of the current signal and the fault feature frequency of the to-be-detected gear; A second determination unit, configured to, for each monitoring fault frequency in the monitoring fault frequency set, determine the fault feature energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency; A detection unit, configured to input the fault feature energies corresponding to each of the monitoring fault frequencies into a support vector data description SVDD for gear state detection, to obtain the gear state detection result output by the SVDD.

[0015] The present invention further provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, it implements the gear state detection method based on motor current signals and support vector data description as described in any one of the above.

[0016] The present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the gear state detection method based on the motor current signal and support vector data description as described in any one of the above.

[0017] The present invention also provides a computer program product, including a computer program. When the computer program is executed by a processor, it implements the gear state detection method based on the motor current signal and support vector data description as described in any one of the above.

[0018] The gear state detection method and device based on the motor current signal and support vector data description provided by the present invention collect the current signal of the motor in the target device, determine the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, determine the monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, and determine the fault characteristic energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency. Finally, the fault characteristic energy corresponding to each monitoring fault frequency is input into the support vector data description (SVDD) for gear state detection, and the gear state detection result output by the SVDD is obtained. It can be seen that the present invention uses the fault characteristic energy at the monitoring fault frequency as the input of the SVDD, which is more in line with the characteristics of the gear fault signal, can reduce the interference of other factors other than the gear fault on the gear detection, and thus improves the accuracy of the gear state detection. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0020] Figure 1 It is one of the schematic flowcharts of the gear state detection method based on the motor current signal and support vector data description provided by the embodiment of the present invention.

[0021] Figure 2 It is the second of the schematic flowcharts of the gear state detection method based on the motor current signal and support vector data description provided by the embodiment of the present invention.

[0022] Figure 3 It is the schematic diagram of the training method of the SVDD provided by the embodiment of the present invention.

[0023] Figure 4 It is the schematic diagram of the test process of the SVDD provided by the embodiment of the present invention.

[0024] Figure 5It is a schematic diagram of the overall process of the gear state detection method based on motor current signals and support vector data description provided by an embodiment of the present invention.

[0025] Figure 6 It is a schematic structural diagram of the gear state detection device based on motor current signals and support vector data description provided by an embodiment of the present invention.

[0026] Figure 7 It is a schematic diagram of the physical structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0027] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below with reference to the accompanying drawings in the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] Existing gear state detection technologies can be classified into: detection based on vibration signals and detection based on current signals. According to the signal processing method, it can be divided into direct diagnosis based on fault characteristics and gear detection technology based on machine learning.

[0029] Since vibration signals are one of the most easily collected mechanical signals, vibration signals are used for gear fault detection. Gear faults usually cause changes in the meshing torque, which in turn causes changes in the supporting force, resulting in periodic vibrations of the system. The acquisition of vibration signals is mainly used to extract and analyze the periodic components in the vibration signals. The main analyses include spectral analysis and envelope spectral analysis. Other analyses include Empirical Mode Decomposition (EMD) and Ensemble Empirical Mode Decomposition (EEMD) to improve the fault detection effect. Continuous Wavelet Transform (CWT) is used for time-frequency domain analysis to calculate the impact envelope spectrum. To solve the problem of system resonance, resonance demodulation technology has also been proposed. Generally speaking, vibration signals require additional sensors, and the transmission and various signal modulations of vibration signals are relatively complex, so the cost is high and the technology migration is poor. At the same time, the installation of sensors introduces additional aging factors, reducing the reliability of the system.

[0030] In recent years, Machine Current Signature Analysis (MCSA) has also been used for gear fault detection. The advantage of MCSA lies in its non-invasiveness, as most motor systems are equipped with current sensors for motor drive by themselves. The basic principle of MCSA is to monitor torque changes through current signals. Since the propagation path of gear fault information in torque is different from that in vibration signals, MCSA does not have the disadvantages of the above vibration signal detection methods. Fast Fourier Transform (FFT), Wavelet Transform (WT) and Envelope Spectrum Analysis (ESA) are commonly used methods in gear fault detection because gear fault characteristics are closely related to gear rotation angle. The main idea of these methods is to extract periodic fault information in the frequency domain. However, most MCSA-based studies mainly focus on whether the faults have obvious characteristics. For minor faults, detection is difficult due to the lack of quantitative indicators. The definition of fault indicators and the setting of their thresholds remain unresolved. In addition, the research on planetary gear transmission systems is also insufficient.

[0031] In addition, various methods for detecting gear status based on machine learning have also been proposed. Multiple models have been applied to gear fault detection, including Deep Belief Network (DBN), Deep Convolutional Neural Networks (CNNs), Probabilistic Confidence Convolutional Neural Network (PCCNN), Deep Enhanced Fusion Network (DEFN), and Robust Auxiliary Classifier Generative Adversarial Network (RAC-GAN). Due to the complexity and inherent errors of the gear system, it is difficult to obtain fault data of the gear system. There are significant differences between simulated faults and actual working conditions, and they cannot completely replace real data, while the acquisition of real data is extremely challenging. Existing research usually uses input features such as Root Mean Square (RMS), Standard Deviation (STD), and signal kurtosis, or extracts features using wavelet transform WT and Fourier transform FFT. In addition, Ensemble Empirical Mode Decomposition (EEMD) is also used for noise reduction before machine learning. Factors such as faults and installation errors will affect the input data, so the interpretability of most deep learning methods is relatively low, resulting in overfitting of the input data.

[0032] Based on this, the present invention proposes a gear status detection method based on motor current signals and Support Vector Data Description (SVDD), which uses the fault feature energy at the monitored fault frequency as the input of SVDD, better conforms to the characteristics of gear fault signals, can reduce the interference of other factors except gear faults on gear detection, and thus improves the accuracy of gear status detection.

[0033] The following combines Figures 1-5 to describe the gear status detection method based on motor current signals and Support Vector Data Description of the present invention. The execution subject of the gear status detection method based on motor current signals and Support Vector Data Description can be an electronic device such as a terminal, a tablet computer, or a computer, or it can also be a gear status detection device based on motor current signals and Support Vector Data Description set in the electronic device. The gear status detection device based on motor current signals and Support Vector Data Description can be implemented through software, hardware, or a combination of both.

[0034] Figure 1It is one of the schematic flowcharts of the gear state detection method based on motor current signals and support vector data description provided by an embodiment of the present invention. As Figure 1 shown, the gear state detection method based on motor current signals and support vector data description includes the following steps: Step 101: Collect the current signal of the motor in the target device, where the motor is used to drive the gear to be detected.

[0035] Among them, the target device can be any device that needs to drive the gear to be detected through a motor, and the present invention does not limit this.

[0036] Exemplarily, the current signal of the motor can be collected by the current sensor equipped in the motor system itself. Of course, if the motor system itself does not have a current sensor, a current sensor can also be installed to collect the current signal of the motor.

[0037] It should be noted that the present invention does not limit the type of the motor, and it needs to be determined based on the type of the motor configured in the target device. For example, the motor configured in the target device is a permanent magnet synchronous motor.

[0038] Step 102: Determine the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, and determine the monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected.

[0039] Among them, the gear to be detected includes a sun gear to be detected and a planet gear to be detected that mesh with each other. The rotational speed of the motor can be collected by a speed sensor or determined based on the current signal of the motor and the number of teeth of the gear to be detected. For specific reference, please refer to the related technology, and the present invention will not elaborate here.

[0040] Optionally, the monitoring fault frequency set is determined based on the following formula (1).

[0041] (1) Among them, represents the monitoring fault frequency set, represents the monitoring fault frequency, represents the frequency of the current signal, represents the order of the fault to be analyzed, represents the fault characteristic frequency of the sun gear to be detected or the fault characteristic frequency of the planet gear to be detected , , , represents the frequency components already existing in the current signal of the motor, the neighborhood of the existing frequency components, and the harmonic frequencies of the current signal of the motor, represents the reduction ratio of the motor, represents the rotational speed of the motor, represents the number of teeth of the sun gear to be detected, represents the number of teeth of the planet gear to be detected.

[0042] Exemplarily, a series of monitored fault frequencies can be obtained through formula (1) , can be based on and ratio to determine, The setting of considers the control performance of the motor system, that is This formula calculates the rotational speed of the sun gear to be detected in the reference system of the fixed-axis gear train, that is, the meshing frequency of a specific tooth of the sun gear to be detected, which is the fault characteristic frequency of the sun gear to be detected, This formula calculates the rotational speed of the planet gear to be detected in the reference system of the fixed-axis gear train, that is, the meshing frequency of a specific tooth of the planet gear to be detected, which is the fault characteristic frequency of the planet gear to be detected.

[0043] It should be noted here that when calculating the monitored fault frequency , it is necessary to use as substitute into formula (1) to calculate multiple monitored fault frequencies , and it is also necessary to use as substitute into formula (1) to calculate multiple monitored fault frequencies , and then use all the calculated monitored fault frequencies as the monitored fault frequency set. Each monitored fault frequency included in the monitored fault frequency set is a frequency obtained based on the gear fault mechanism analysis. Therefore, it is necessary to observe the corresponding fault characteristic energy at these monitored fault frequencies to determine whether the gear to be detected is abnormal.

[0044] It should be noted that the gear to be detected can also be other types of two meshing gears. For example, the gear to be detected is two spur gears or helical gears that mesh with each other, etc. The present invention does not limit this.

[0045] Step 103: For each monitored fault frequency in the monitored fault frequency set, based on the current signal and the monitored fault frequency, determine the fault characteristic energy corresponding to the monitored fault frequency.

[0046] Optionally, determine the fault characteristic energy corresponding to the monitored fault frequency based on the following formula (2).

[0047] (2) Among them, represents the monitored fault frequency The corresponding fault characteristic energy represents the 3dB bandwidth of the main lobe of the Hanning window represents the spectral resolution , represents the result of performing a Fourier transform on the current signal represents the number of the frequency-domain signal in the result of the Fourier transform represents a piecewise function

[0048] Exemplarily, the fast Fourier transform FFT is used as the main spectral processing means, and the spectral energy is used to calculate each monitored fault frequency The corresponding fault characteristic energy, and windowing processing is used to suppress the spectral leakage error. The Hanning window is selected as the window function because the function form of the Hanning window is simple and has a relatively narrow main lobe width and a high sidelobe attenuation rate. The main energy of the frequency components is concentrated in the main lobe region, and the 3dB bandwidth of the main lobe of the Hanning window , therefore, given the monitored fault frequency The approximate fault characteristic energy at can be calculated by the above formula (2). Through the calculation of the integration interval, formula (2) can be conveniently transformed from the integral form to the summation form. The spectral leakage error is considered in the above formula (2), making the calculated fault characteristic energy more accurate.

[0049] Step 104, input the fault characteristic energy corresponding to each monitored fault frequency into the Support Vector Data Description (SVDD) for gear state detection, and obtain the gear state detection result output by the SVDD.

[0050] Among them, Support Vector Data Description (SVDD) is a one-class classification method based on statistical learning theory, mainly used for anomaly detection and fault diagnosis. The core idea of SVDD is to find a hypersphere with the smallest radius to enclose the normal data, so as to distinguish normal samples from abnormal samples.

[0051] Exemplarily, when obtaining the fault characteristic energy corresponding to each monitored fault frequency in the monitored fault frequency set, the fault characteristic energy corresponding to each monitored fault frequency is used as the input of the pre-trained SVDD. By using SVDD, a hypersphere with the smallest radius is found to enclose the normal data, so as to distinguish normal data from abnormal data, and finally obtain the gear state detection result output by the SVDD. The gear state detection result is a healthy state or an abnormal state.

[0052] The gear state detection method based on motor current signals and support vector data description provided by the present invention collects the current signals of the motor in the target device, determines the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, determines the monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, determines the fault characteristic energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency, and finally inputs the fault characteristic energy corresponding to each monitoring fault frequency into the support vector data description (SVDD) for gear state detection to obtain the gear state detection result output by the SVDD. It can be seen that the present invention uses the fault characteristic energy at the monitoring fault frequency as the input of the SVDD, which is more in line with the characteristics of gear fault signals, can reduce the interference of other factors other than gear faults on gear detection, and thus improves the accuracy of gear state detection.

[0053] In one embodiment, Figure 2 is the second schematic flow chart of the gear state detection method based on motor current signals and support vector data description provided by the embodiment of the present invention. As Figure 2 shown, in step 104 above, the fault characteristic energy corresponding to each of the monitoring fault frequencies is input into the support vector data description (SVDD) for gear state detection to obtain the gear state detection result output by the SVDD. Specifically, it can be implemented through the following steps: Step 201: Input the fault characteristic energy corresponding to each of the monitoring fault frequencies into the SVDD, and the SVDD performs kernel mapping on each of the fault characteristic energies based on the Gaussian kernel function to obtain a mapping vector, and determine the distance between the mapping vector and the center of the hypersphere corresponding to the SVDD.

[0054] Exemplarily, the fault characteristic energy corresponding to each monitoring fault frequency is input into the SVDD, and the SVDD performs kernel mapping on each fault characteristic energy based on the Gaussian kernel function to obtain the analyzed mapping vector , and calculate the distance between the mapping vector and the center a of the hypersphere corresponding to the SVDD. Specifically, it can be calculated using to represent the distance between the mapping vector and the center of the hypersphere corresponding to the SVDD, to represent the vector corresponding to the center of the sphere, to represent the mapping vector and the norm of the vector difference between the vector corresponding to the center of the sphere .

[0055] Step 202: When the distance is less than or equal to the radius of the hypersphere, determine that the gear state detection result is a healthy state.

[0056] For example, when obtaining the distance between the mapping vector and the center of the hypersphere corresponding to SVDD, compare this distance with the radius of the hypersphere. When it is determined that this distance is less than or equal to the radius of the hypersphere, it indicates that the input fault feature energy is within the hypersphere and belongs to normal data. Therefore, it is determined that the gear status monitoring result is in a healthy state.

[0057] Step 203: In the case where the distance is greater than the radius of the hypersphere, determine that the gear status detection result is in an abnormal state.

[0058] For example, when obtaining the distance between the mapping vector and the center of the hypersphere corresponding to SVDD, compare this distance with the radius of the hypersphere. When it is determined that this distance is greater than the radius of the hypersphere, it indicates that the input fault feature energy is outside the hypersphere and belongs to abnormal data. Therefore, it is determined that the gear status monitoring result is in an abnormal state, that is, a fault state.

[0059] In this embodiment, SVDD provides a threshold setting method, that is, determining the gear status based on the comparison result of the distance between the mapping vector and the center of the hypersphere corresponding to SVDD and the radius of the hypersphere, which intuitively enhances the interpretability of the method, thereby making the gear status detection result more credible.

[0060] In one embodiment, in the above step 103, based on the current signal and the monitored fault frequency, determining the fault feature energy corresponding to the monitored fault frequency can be specifically implemented in the following manner: Based on the current signal and the monitored fault frequency, determine the initial fault feature energy corresponding to the monitored fault frequency; determine the logarithm value of the initial fault feature energy; based on the logarithm value range, linearly transform the logarithm value to obtain the fault feature energy corresponding to the monitored fault frequency, and the logarithm value range is determined based on the maximum logarithm value and the minimum logarithm value among the logarithm values of all the initial fault feature energies.

[0061] Exemplarily, after calculating the initial fault feature energy corresponding to each monitored fault frequency based on the above formula (2), since the magnitudes of the initial fault feature energies are inconsistent, it is necessary to preprocess the initial fault feature energies. Specifically, the logarithmic regularization method is used to normalize the initial fault feature energies. That is, for each initial fault feature energy, first take the logarithm of each initial fault feature energy to obtain the logarithmic values, then sort all the obtained logarithmic values to determine the maximum logarithmic value and the minimum logarithmic value. Take the minimum logarithmic value as the minimum value of the logarithmic value range, and take the maximum logarithmic value as the maximum value of the logarithmic value range to obtain the logarithmic value range. Based on this logarithmic value range, perform a linear transformation on each logarithmic value to obtain the fault feature energy corresponding to each monitored fault frequency. The finally obtained fault feature energy is a value within the range of 0 to 1. For gear anomalies, the relationship between the characteristic quantities in the healthy state and the abnormal state is closer to a binary "yes / no" relationship rather than a quantitative relationship. Logarithmic regularization is precisely to adapt to this binary relationship.

[0062] It should be noted that considering that the current signal of the motor collected in real time may contain extreme situations, the finally obtained fault feature energy may also be outside the range of 0 to 1.

[0063] In this embodiment, the initial fault feature energies corresponding to each monitored fault frequency are preprocessed by logarithmic regularization, so that the obtained fault feature energy is more in line with the binary relationship, that is, more in line with the nature of the actual fault features of the gear, further improving the accuracy of gear state detection.

[0064] In one embodiment, Figure 3 is a schematic diagram of the training method of SVDD provided by an embodiment of the present invention. As Figure 3 shown, the SVDD is trained based on the following method: Step 301, when the sample device is in a healthy state, collect the sample current signal of the sample motor in the sample device, and the sample motor is used to drive the sample gear.

[0065] Exemplarily, at the end of the running-in period of the sample device when it is in a healthy state, collect the sample current signal of the sample motor in the sample device.

[0066] It should be noted that when collecting the sample current signal, various working conditions of the actual operation of the sample device need to be included. The specific working conditions include slow running conditions, fast running conditions, light load running conditions, heavy load running conditions, etc. Among them, slow and fast can be determined based on the comparison between the speed and the preset speed. When the speed is less than the preset speed, it is slow, and when the speed is greater than or equal to the preset speed, it is fast; light load and heavy load can also be determined based on preset values, which will not be elaborated in the present invention.

[0067] Step 302: Determine a set of sample monitoring fault frequencies based on the frequency of the sample current signal, the sample fault characteristic frequency of the sample gear, and the number of teeth of the sample gear.

[0068] Exemplarily, the specific method for determining the set of sample monitoring fault frequencies based on the frequency of the sample current signal, the sample fault characteristic frequency of the sample gear, and the number of teeth of the sample gear is similar to the specific method in the above Step 102 for determining the set of monitoring fault frequencies based on the frequency of the current signal, the fault characteristic frequency of the gear to be detected, and the number of teeth of the gear to be detected. Reference can be made to the specific description in the above Step 102, and the present invention will not elaborate herein.

[0069] Step 303: For each sample monitoring fault frequency in the set of sample monitoring fault frequencies, determine the sample fault characteristic energy corresponding to the sample monitoring fault frequency based on the sample current signal and the sample monitoring fault frequency.

[0070] Exemplarily, the specific method for determining the sample fault characteristic energy corresponding to the sample monitoring fault frequency based on the sample current signal and the sample monitoring fault frequency is similar to the specific method in the above Step 103 for determining the fault characteristic energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency. Reference can be made to the specific description in the above Step 103, and the present invention will not elaborate herein.

[0071] Step 304: Use some of the sample fault characteristic energies among the multiple sample fault characteristic energies as training data.

[0072] Exemplarily, when obtaining the multiple sample fault characteristic energies of the sample device, some of the sample fault characteristic energies are used as training data, and some of the sample fault characteristic energies are used as test data.

[0073] Step 305: Construct an optimization problem based on the training data and solve the optimization problem to obtain the support vectors of the hypersphere of the SVDD.

[0074] Optionally, construct the optimization problem based on the following formula (3): (3) where the constraint condition is , represents the total number of the training data, represents the Lagrange multiplier of the th training data , represents the Lagrange multiplier of the th training data , and denotes a Gaussian kernel function; when the constraint conditions are satisfied, the is determined as the support vector.

[0075] Exemplarily, when obtaining the training data, the Gaussian kernel function is used to perform kernel mapping on the training data to obtain sample mapping vectors. The sample mapping vector of the th training data is denoted by , and the problem is transformed into the original optimization problem shown in the following formula (6): (6) where the constraint condition of formula (6) is , denotes the radius of the hypersphere, denotes the penalty parameter, which is used to control the tolerance for outliers, denotes the th slack variable of the training data, denotes the total number of training data, denotes the vector of the center of the hypersphere. The objective of formula (6) is to minimize the radius of the hypersphere while allowing a small number of samples to fall outside the hypersphere, specifically controlled by the slack variable .

[0076] Furthermore, the original optimization problem of formula (6) is transformed into the optimization problem shown in the above formula (3) by the Lagrange multiplier method. The objective of this optimization problem is to transform the original optimization problem into a dual problem by the Lagrange multiplier method to simplify the solution. Solving the above formula (3), the Lagrange multipliers obtained when the constraint conditions are satisfied are determined as the support vectors , and the support vectors refer to the data points located on the boundary of the hypersphere.

[0077] Step 306: Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies to obtain the SVDD.

[0078] Optionally, determine the radius of the hypersphere based on the following formula (4) , and determine the center a of the hypersphere based on the following formula (5): (4) (5) where denotes the support vector, denotes the Gaussian kernel function, denotes the th training data The corresponding sample mapping vector.

[0079] Exemplarily, when the support vector is calculated Substituting the support vector into the above formula (4), the radius of the hypersphere can be obtained , and substituting the sample mapping vector into the above formula (5), the vector of the center of the hypersphere can be obtained. After obtaining the radius of the hypersphere and the vector of the center of the hypersphere, it is equivalent to training the SVDD.

[0080] In this embodiment, an optimization problem is constructed based on the training data, and the optimization problem is solved to obtain the support vectors of the hypersphere of the SVDD. The radius of the hypersphere is determined based on the support vectors, and the center of the hypersphere is determined based on the sample mapping vectors corresponding to the sample fault feature energies of each sample. Finally, the SVDD is trained, and the trained SVDD is used as a gear state detection model to facilitate subsequent detection of the gear state based on the classification results of the SVDD.

[0081] In one embodiment, Figure 4 is a schematic diagram of the test process of the SVDD provided by the embodiment of the present invention. As Figure 4 shown, in the above step 306, the radius of the hypersphere is determined based on the support vectors, and the center of the hypersphere is determined based on the sample mapping vectors corresponding to the sample fault feature energies of each sample to obtain the SVDD. Specifically, it can be implemented through the following steps: Step 401: Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies of each sample to obtain the to-be-tested SVDD.

[0082] Exemplarily, the radius of the hypersphere and the center of the hypersphere are calculated based on the above formula (4) and formula (5) to obtain the initial SVDD, which is also called the to-be-tested SVDD.

[0083] Step 402: Use another part of the sample fault feature energies among the multiple sample fault feature energies as test data and input them into the to-be-tested SVDD. The to-be-tested SVDD performs kernel mapping on each test data based on the Gaussian kernel function to obtain test mapping vectors, and determines the test distances between the test mapping vectors and the center of the hypersphere.

[0084] Exemplarily, another part of the sample fault feature energy among the multiple sample fault feature energies is used as test data and input into the SVDD to be tested. The SVDD to be tested uses a Gaussian kernel function to perform kernel mapping on each sample fault feature energy, obtaining the test mapping vectors to be analyzed, and calculating and determining the test distance between the test mapping vectors and the center of the hypersphere. Specifically, the test distance is represented by the norm of the vector difference between the test mapping vector and the center vector of the hypersphere.

[0085] Step 403: For each of the test data, based on the test distance corresponding to the test data and the radius of the hypersphere, determine the gear state test result corresponding to the test data.

[0086] Exemplarily, when obtaining the test distance between the test mapping vector and the center of the hypersphere, compare this test distance with the radius of the hypersphere. When it is determined that this test distance is less than or equal to the radius of the hypersphere, it indicates that the input sample fault feature energy is within the hypersphere and belongs to normal data. Therefore, the gear state test result corresponding to the test data is determined to be the healthy state; when it is determined that this test distance is greater than the radius of the hypersphere, it indicates that the input sample fault feature energy is outside the hypersphere and belongs to abnormal data. Therefore, the gear state test result corresponding to the test data is determined to be the abnormal state, that is, the fault state.

[0087] Step 404: Based on the gear state test results corresponding to each of the test data, determine the accuracy rate of the SVDD to be tested.

[0088] Exemplarily, when obtaining the gear state test results corresponding to each test data, count the gear state test results corresponding to each test data, count the target number of gear state test results that are in the healthy state, and determine the ratio of the target number to the total number of test data as the accuracy rate of the SVDD to be tested.

[0089] Step 405: In the case where the accuracy rate is not within the preset accuracy rate range, fine-tune the parameters of the Gaussian kernel function of the SVDD to be tested until the accuracy rate is within the preset accuracy rate range, obtaining the SVDD.

[0090] Among them, the preset accuracy rate range can be set based on requirements. For example, the preset accuracy rate range is from 96% to 98%, and the present invention does not make any limitations in this regard.

[0091] Exemplarily, when obtaining the accuracy rate of the SVDD to be tested, it is determined whether the accuracy rate of the SVDD to be tested is within the preset accuracy rate range. When the accuracy rate of the SVDD to be tested is not within the preset accuracy rate range, it indicates that the performance of the SVDD to be tested is poor, and it is necessary to iteratively fine-tune the parameters of the Gaussian kernel function of the SVDD to be tested. Specifically, the kernel width parameter of the Gaussian kernel function is adjusted until the accuracy rate of the SVDD to be tested is within the preset accuracy rate range, indicating that the performance of the SVDD to be tested meets the requirements, and finally the SVDD is obtained.

[0092] It should be noted that both the training data and the test data need to be preprocessed by the logarithmic regularization method before being used as input data, which will not be elaborated in this invention.

[0093] In this embodiment, the SVDD to be tested obtained by training is tested based on the test data. When the accuracy rate of the SVDD to be tested is not within the preset accuracy rate range, the parameters of the Gaussian kernel function of the SVDD to be tested are iteratively fine-tuned until the accuracy rate is within the preset accuracy rate range, so as to improve the accuracy rate of the finally obtained SVDD and further improve the accuracy of gear state detection.

[0094] Figure 5 is the overall flow schematic diagram of the gear state detection method based on the motor current signal and the support vector data description provided by the embodiment of the present invention. As Figure 5 shown, it is mainly divided into three stages, namely the training stage of SVDD, the testing stage of SVDD, and the application stage of SVDD; in the training stage and the testing stage, when the sample device is in a healthy state, the sample current signal of the sample motor in the sample device is collected. In the application stage, the current signal of the motor is collected when the target device is in an unknown state. The preprocessing processes of the sample current signal and the current signal are the same. Taking the application stage as an example below, the collected current signal is windowed and subjected to FFT transformation. Based on the rotational speed of the motor, the reduction ratio, and the number of teeth of the gear, the fault characteristic frequency of the gear to be detected is calculated. Based on the result of the FFT transformation, the frequency of the current signal, and the fault characteristic frequency of the gear to be detected, the initial fault characteristic energy corresponding to each monitored fault frequency is calculated. The logarithmic regularization preprocessing is performed on each initial fault characteristic energy to obtain each fault characteristic energy, and each fault characteristic energy is used as the input of the SVDD; for the training stage and the testing stage, each sample fault characteristic energy is used as the input of the SVDD. The fault characteristic energy and the rotational speed of the motor, etc. are called motor signals. It should be noted here that the rotational speed of the motor can be estimated through electrical signals or obtained through other means such as sensors. The present invention does not make any limitations in this regard.

[0095] In the training phase, part of the sample fault feature energies among multiple sample fault feature energies are used as training data. The training data is subjected to Gaussian kernel function mapping to construct an optimization problem, and the optimization problem is solved to obtain the support vectors of the hypersphere. The radius of the hypersphere is calculated based on the support vectors, and the center of the hypersphere is calculated based on the sample mapping vectors corresponding to the respective training data, and the SVDD to be tested is trained.

[0096] In the testing phase, another part of the sample fault feature energies among multiple sample fault feature energies are used as test data and input into the SVDD to be tested. The test data is subjected to kernel mapping by the SVDD to be tested based on the Gaussian kernel function to obtain test mapping vectors, and the test distances between the respective test mapping vectors and the center of the hypersphere are calculated. For each test data, based on the test distance corresponding to the test data and the radius of the hypersphere, the test data is classified, that is, the gear state test result corresponding to the test data is determined. Based on the gear state test results corresponding to the respective test data, the accuracy of the SVDD to be tested is determined, and it is determined whether the accuracy of the SVDD to be tested is within the preset accuracy range. In the case where the accuracy is not within the preset accuracy range, the parameters of the Gaussian kernel function of the SVDD to be tested are fine-tuned until the accuracy is within the preset accuracy range, and the training ends to obtain the SVDD.

[0097] In the application phase, the fault feature energies corresponding to the respective monitored fault frequencies are used as actual data and input into the SVDD. The SVDD performs kernel mapping on the respective fault feature energies based on the Gaussian kernel function to obtain mapping vectors, and the distance between the mapping vectors and the center of the hypersphere corresponding to the SVDD is determined. In the case where the distance is less than or equal to the radius of the hypersphere, the gear state detection result is determined to be the healthy state; in the case where the distance is greater than the radius of the hypersphere, the gear state detection result is determined to be the abnormal state, thereby ending the detection of the gear state.

[0098] The gear state detection method based on motor current signals and support vector data description provided by the present invention uses SVDD to achieve gear state detection. As a single-classification technique, SVDD only requires the sample current collected under healthy conditions to complete the training. This method can perform model training during the healthy operation stage of the equipment, thus avoiding the problem of obtaining comprehensive fault data. In addition, SVDD can be independently trained for different devices, has self-adaptability, and can adapt to the characteristic changes between devices caused by machining errors and running-in state differences. This method can also effectively solve the gear state detection under multi-condition conditions. In addition, the present invention uses the fault feature energy at the fault frequency to be monitored as the input of SVDD, which is more in line with the characteristics of gear fault signals, can reduce the interference of other factors other than gear faults on gear detection, and thus improve the accuracy of gear state detection. Generally speaking, using SVDD as a machine learning technique and using the Gaussian kernel function and logarithmic regularization method can simplify the internal logic of machine learning and enhance the interpretability of the model, thereby improving the trustworthiness of the gear state detection results.

[0099] The gear state detection device based on motor current signals and support vector data description provided by the present invention will be described below. The gear state detection device based on motor current signals and support vector data description described below can be correspondingly referred to the gear state detection method based on motor current signals and support vector data description described above.

[0100] Figure 6 is a schematic structural diagram of the gear state detection device based on motor current signals and support vector data description provided by an embodiment of the present invention. As Figure 6 shown, the gear state detection device 600 based on motor current signals and support vector data description includes a collection unit 601, a first determination unit 602, a second determination unit 603, and a detection unit 604; where: The collection unit 601 is configured to collect the current signal of the motor in the target device, and the motor is used to drive the gear to be detected; The first determination unit 602 is configured to determine the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotation speed of the motor, and determine the monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected; The second determination unit 603 is configured to, for each monitoring fault frequency in the monitoring fault frequency set, determine the fault characteristic energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency; The detection unit 604 is configured to input the fault characteristic energy corresponding to each monitoring fault frequency into the support vector data description SVDD for gear state detection, and obtain the gear state detection result output by the SVDD.

[0101] The gear state detection device based on the motor current signal and support vector data description provided by the present invention collects the current signal of the motor in the target device, determines the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, determines the monitoring fault frequency set based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, and determines the fault characteristic energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency. Finally, the fault characteristic energy corresponding to each monitoring fault frequency is input into the support vector data description (SVDD) for gear state detection, and the gear state detection result output by the SVDD is obtained. It can be seen that the present invention uses the fault characteristic energy at the monitoring fault frequency as the input of the SVDD, which is more in line with the characteristics of the gear fault signal, can reduce the interference of other factors other than gear faults on gear detection, and thus improves the accuracy of gear state detection.

[0102] Based on any of the above embodiments, the gear to be detected includes a to-be-detected sun gear and a to-be-detected planet gear that mesh with each other; the first determination unit 602 is specifically configured to: Determine the monitoring fault frequency set based on the following formula (1); (1) Wherein, represents the monitoring fault frequency set, represents the monitoring fault frequency, represents the frequency of the current signal, represents the order of the analyzed fault, represents the fault characteristic frequency of the to-be-detected sun gear or the fault characteristic frequency of the to-be-detected planet gear , , , represents the frequency components already existing in the current signal of the motor, the neighborhood of the existing frequency components, and the harmonic frequencies of the current signal of the motor, represents the reduction ratio of the motor, represents the rotational speed of the motor, represents the number of teeth of the to-be-detected sun gear, represents the number of teeth of the to-be-detected planet gear.

[0103] Based on any of the above embodiments, the second determination unit 603 is specifically configured to: Determine the fault characteristic energy corresponding to the monitoring fault frequency based on the following formula (2); (2) Wherein, represents the monitoring fault frequency The corresponding fault feature energy represents the 3dB bandwidth of the main lobe of the Hanning window represents the spectral resolution , represents the result of performing a Fourier transform on the current signal represents the number of the frequency-domain signal in the result of the Fourier transform represents a piecewise function

[0104] Based on any of the above embodiments, the detection unit 604 is specifically configured to: Input the fault feature energy corresponding to each monitored fault frequency into the SVDD, and perform kernel mapping on each fault feature energy by the SVDD based on the Gaussian kernel function to obtain a mapping vector, and determine the distance between the mapping vector and the center of the hypersphere corresponding to the SVDD; When the distance is less than or equal to the radius of the hypersphere, determine that the gear state detection result is a healthy state; When the distance is greater than the radius of the hypersphere, determine that the gear state detection result is an abnormal state.

[0105] Based on any of the above embodiments, the second determination unit 603 is further specifically configured to: Based on the current signal and the monitored fault frequency, determine the initial fault feature energy corresponding to the monitored fault frequency; Determine the logarithm value of the initial fault feature energy; Based on the logarithm value range, perform linear conversion on the logarithm value to obtain the fault feature energy corresponding to the monitored fault frequency, where the logarithm value range is determined based on the maximum logarithm value and the minimum logarithm value among the logarithm values of all the initial fault feature energies.

[0106] Based on any of the above embodiments, the SVDD is trained in the following manner: When the sample device is in a healthy state, collect the sample current signal of the sample motor in the sample device, where the sample motor is used to drive the sample gear; Based on the frequency of the sample current signal, the sample fault feature frequency of the sample gear, and the number of teeth of the sample gear, determine a set of sample monitored fault frequencies; For each sample monitored fault frequency in the set of sample monitored fault frequencies, based on the sample current signal and the sample monitored fault frequency, determine the sample fault feature energy corresponding to the sample monitored fault frequency; Use some of the sample fault feature energies among the multiple sample fault feature energies as training data; Construct an optimization problem based on the training data, and solve the optimization problem to obtain the support vectors of the hypersphere of the SVDD; Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies to obtain the SVDD.

[0107] Based on any of the above embodiments, the constructing an optimization problem based on the training data includes: Construct an optimization problem based on the following formula (3): (3) where the constraint condition is , represents the total number of the training data, represents the th training data 's Lagrange multiplier, represents the th training data 's Lagrange multiplier, and represent the Gaussian kernel function; The solving the optimization problem to obtain the support vectors of the hypersphere of the SVDD includes: Determine the obtained when the constraint condition is satisfied as the support vectors; The determining the center of the hypersphere and the radius of the hypersphere based on the support vectors to obtain the SVDD includes: Determine the radius of the hypersphere based on the following formula (4) , and determine the center a of the hypersphere based on the following formula (5): (4) (5) where, represents the support vectors, represents the Gaussian kernel function, represents the th training data 's corresponding sample mapping vector.

[0108] Based on any of the above embodiments, the determining the radius of the hypersphere based on the support vectors, and determining the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies to obtain the SVDD includes: Determine the radius of the hypersphere based on the support vectors, and determine the center of the hypersphere based on the sample mapping vectors corresponding to the sample fault feature energies, to obtain the to-be-tested SVDD; Use another part of the sample fault feature energies among the multiple sample fault feature energies as test data and input it into the to-be-tested SVDD. Through the to-be-tested SVDD, perform kernel mapping on each piece of test data based on the Gaussian kernel function to obtain test mapping vectors, and determine the test distances between each test mapping vector and the center of the hypersphere; For each piece of test data, determine the gear state test result corresponding to the test data based on the test distance corresponding to the test data and the radius of the hypersphere; Based on the gear state test results corresponding to each piece of test data, determine the accuracy rate of the to-be-tested SVDD; In the case that the accuracy rate is not within the preset accuracy range, fine-tune the parameters of the Gaussian kernel function of the to-be-tested SVDD until the accuracy rate is within the preset accuracy range, to obtain the SVDD.

[0109] Figure 7 It is a schematic physical structure diagram of an electronic device provided by an embodiment of the present invention. As Figure 7 shown, the electronic device may include: a processor 710, a communication interface 720, a memory 730, and a communication bus 740. Among them, the processor 710, the communication interface 720, and the memory 730 complete communication with each other through the communication bus 740. The processor 710 can call the logical instructions in the memory 730 to execute a gear state detection method based on a motor current signal and support vector data description. The method includes: collecting the current signal of a motor in a target device, where the motor is used to drive a to-be-detected gear; Based on the number of teeth of the to-be-detected gear and the rotational speed of the motor, determine the fault characteristic frequency of the to-be-detected gear, and based on the frequency of the current signal and the fault characteristic frequency of the to-be-detected gear, determine a monitored fault frequency set; For each monitored fault frequency in the monitored fault frequency set, based on the current signal and the monitored fault frequency, determine the fault characteristic energy corresponding to the monitored fault frequency; Input the fault characteristic energies corresponding to each monitored fault frequency into a support vector data description (SVDD) for gear state detection, to obtain the gear state detection result output by the SVDD.

[0110] In addition, when the logical instructions in the above-mentioned memory 730 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0111] On the other hand, the present invention also provides a computer program product. The computer program product includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the gear state detection method based on the motor current signal and support vector data description provided by the above-mentioned various methods. The method includes: collecting the current signal of the motor in the target device, where the motor is used to drive the gear to be detected; Based on the number of teeth of the gear to be detected and the rotational speed of the motor, determining the fault characteristic frequency of the gear to be detected, and based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, determining a set of monitoring fault frequencies; For each monitoring fault frequency in the set of monitoring fault frequencies, based on the current signal and the monitoring fault frequency, determining the fault characteristic energy corresponding to the monitoring fault frequency; Inputting the fault characteristic energy corresponding to each monitoring fault frequency into the support vector data description SVDD for gear state detection to obtain the gear state detection result output by the SVDD.

[0112] On another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it realizes the gear state detection method based on the motor current signal and support vector data description provided by the above-mentioned various methods. The method includes: collecting the current signal of the motor in the target device, where the motor is used to drive the gear to be detected; Based on the number of teeth of the gear to be detected and the rotational speed of the motor, determining the fault characteristic frequency of the gear to be detected, and based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected, determining a set of monitoring fault frequencies; For each monitoring fault frequency in the set of monitoring fault frequencies, based on the current signal and the monitoring fault frequency, determine the fault feature energy corresponding to the monitoring fault frequency; Input the fault feature energy corresponding to each monitoring fault frequency into the Support Vector Data Description (SVDD) for gear state detection, and obtain the gear state detection result output by the SVDD.

[0113] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative effort.

[0114] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A gear state detection method based on motor current signals and support vector data description, characterized in that including: collecting a current signal of a motor in a target device, where the motor is used to drive a gear to be detected; determining a fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, and determining a set of monitored fault frequencies based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected; for each monitored fault frequency in the set of monitored fault frequencies, determining a fault characteristic energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency; inputting the fault characteristic energies corresponding to the monitored fault frequencies into a support vector data description (SVDD) for gear status detection, and obtaining a gear status detection result output by the SVDD.

2. The gear state detection method based on motor current signals and support vector data description according to claim 1, wherein The gear to be detected includes a sun gear to be detected and a planet gear to be detected that mesh with each other; The determining the fault characteristic frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, and determining a set of monitored fault frequencies based on the frequency of the current signal and the fault characteristic frequency of the gear to be detected includes: determining the set of monitored fault frequencies based on the following formula (1); （1） Among them, represents the set of monitored fault frequencies, represents the monitored fault frequency, represents the frequency of the current signal, represents the order of the analyzed fault, represents the fault characteristic frequency of the sun gear to be detected or the fault characteristic frequency of the planet gear to be detected , , , represents the existing frequency components in the current signal of the motor, the neighborhood of the existing frequency components, and the harmonic frequencies of the current signal of the motor, represents the reduction ratio of the motor, represents the rotational speed of the motor, represents the number of teeth of the sun gear to be detected, represents the number of teeth of the planet gear to be detected.

3. The gear state detection method based on motor current signals and support vector data description according to claim 1, characterized in that The determining the fault characteristic energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency includes: determining the fault characteristic energy corresponding to the monitored fault frequency based on the following formula (2); （2） Among them, represents the monitored fault frequency The corresponding fault feature energy represents the 3dB bandwidth of the main lobe of the Hamming window represents the spectral resolution , represents the result of performing a Fourier transform on the current signal represents the number of the frequency-domain signal in the result of the Fourier transform represents a piecewise function 4. The gear state detection method based on motor current signals and support vector data description according to claim 1, wherein The inputting the fault characteristic energies corresponding to the monitored fault frequencies into a support vector data description (SVDD) for gear status detection and obtaining a gear status detection result output by the SVDD includes: inputting the fault characteristic energies corresponding to the monitored fault frequencies into the SVDD, and performing kernel mapping on each of the fault characteristic energies by the SVDD based on a Gaussian kernel function to obtain a mapping vector, and determining a distance between the mapping vector and the center of a hypersphere corresponding to the SVDD; when the distance is less than or equal to the radius of the hypersphere, determining that the gear status detection result is a healthy state; when the distance is greater than the radius of the hypersphere, determining that the gear status detection result is an abnormal state.

5. The gear state detection method based on motor current signals and support vector data description according to claim 1, characterized in that The determining the fault characteristic energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency includes: determining an initial fault characteristic energy corresponding to the monitored fault frequency based on the current signal and the monitored fault frequency; determining a logarithm value of the initial fault characteristic energy; performing linear transformation on the logarithm value based on a logarithm value range to obtain the fault characteristic energy corresponding to the monitored fault frequency, where the logarithm value range is determined based on the maximum logarithm value and the minimum logarithm value among the logarithm values of all the initial fault characteristic energies.

6. The gear state detection method based on motor current signals and support vector data description according to any one of claims 1-5, characterized in that The SVDD is trained based on the following method: when a sample device is in a healthy state, collecting a sample current signal of a sample motor in the sample device, where the sample motor is used to drive a sample gear; determining a set of sample monitored fault frequencies based on the frequency of the sample current signal, the sample fault characteristic frequency of the sample gear, and the number of teeth of the sample gear; For each sample monitoring fault frequency in the sample monitoring fault frequency set, based on the sample current signal and the sample monitoring fault frequency, determine the sample fault feature energy corresponding to the sample monitoring fault frequency; Use some of the sample fault feature energies among the multiple sample fault feature energies as training data; Construct an optimization problem based on the training data and solve the optimization problem to obtain the support vectors of the hypersphere of the SVDD; Determine the radius of the hypersphere based on the support vectors and determine the center of the hypersphere based on the sample mapping vectors corresponding to the respective sample fault feature energies to obtain the SVDD.

7. The gear state detection method based on motor current signals and support vector data description according to claim 6, characterized in that, The constructing an optimization problem based on the training data includes: Construct an optimization problem based on the following formula (3): （3） Among them, the constraint condition is , represents the total number of the training data, represents the th training data 's Lagrange multiplier, represents the th training data 's Lagrange multiplier, and represents the Gaussian kernel function; The solving the optimization problem to obtain the support vectors of the hypersphere of the SVDD includes: When the constraint conditions are satisfied, the is determined as the support vector; The determining the center of the hypersphere and the radius of the hypersphere based on the support vectors to obtain the SVDD includes: Determine the radius of the hypersphere based on the following formula (4) , and determine the center a of the hypersphere based on the following formula (5): （4） （5） Among them, represents the support vector, represents the Gaussian kernel function, represents the th training data corresponding sample mapping vector.

8. The gear state detection method based on motor current signals and support vector data description according to claim 6, wherein, The determining the radius of the hypersphere based on the support vectors and determining the center of the hypersphere based on the sample mapping vectors corresponding to the respective sample fault feature energies to obtain the SVDD includes: Determine the radius of the hypersphere based on the support vectors and determine the center of the hypersphere based on the sample mapping vectors corresponding to the respective sample fault feature energies to obtain the to-be-tested SVDD; Use another part of the sample fault feature energies among the multiple sample fault feature energies as test data and input the test data into the to-be-tested SVDD. Through the to-be-tested SVDD, perform kernel mapping on each test data based on the Gaussian kernel function to obtain test mapping vectors, and determine the test distances between the respective test mapping vectors and the center of the hypersphere; For each test data, based on the test distance corresponding to the test data and the radius of the hypersphere, determine the gear state test result corresponding to the test data; Based on the gear state test results corresponding to the respective test data, determine the accuracy rate of the to-be-tested SVDD; In the case where the accuracy rate is not within the preset accuracy rate range, fine-tune the parameters of the Gaussian kernel function of the to-be-tested SVDD until the accuracy rate is within the preset accuracy rate range to obtain the SVDD.

9. A gear state detection device based on motor current signals and support vector data description, characterized in that Includes: An acquisition unit for acquiring the current signal of the motor in the target device, where the motor is used to drive the gear to be detected; A first determination unit for determining the fault feature frequency of the gear to be detected based on the number of teeth of the gear to be detected and the rotational speed of the motor, and determining the monitoring fault frequency set based on the frequency of the current signal and the fault feature frequency of the gear to be detected; A second determination unit for, for each monitoring fault frequency in the monitoring fault frequency set, determining the fault feature energy corresponding to the monitoring fault frequency based on the current signal and the monitoring fault frequency; The detection unit is used to input the fault feature energy corresponding to each of the monitored fault frequencies into the Support Vector Data Description (SVDD) for gear state detection, and obtain the gear state detection result output by the SVDD.

10. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the gear state detection method based on the motor current signal and the Support Vector Data Description as described in any one of claims 1 to 8.

Images