Driving motor fault diagnosis method and device based on artificial intelligence technology

Abstract

The invention discloses a driving motor fault diagnosis method and device based on an artificial intelligence technology, and relates to the technical field of fault diagnosis, and the method comprises the steps: arranging a multi-modal sensor at a preset part of a driving motor, and obtaining a multi-dimensional data signal collected by the multi-modal sensor; performing real-time filtering, amplification and noise reduction processing on the multi-dimensional data signal, and performing feature extraction on the multi-dimensional data signal by using a feature extraction algorithm to obtain a plurality of key feature parameters; based on a preset dynamic weighted fusion algorithm, fusing the plurality of key feature parameters to obtain fused key feature parameters; and establishing a hybrid deep network architecture consisting of a long-short term memory network and a convolutional neural network, and inputting the fusion key feature parameters into the pre-trained hybrid deep network architecture to obtain corresponding fault information of the driving motor. The problem of low fault diagnosis accuracy of the driving motor in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis, and particularly to a method and device for diagnosing faults of a drive motor based on artificial intelligence technology. Background Art

[0002] As a core component of the entire powertrain of a new energy vehicle, higher requirements are put forward for the safe, efficient and long-life operation of the drive motor. At present, in the field of motor fault diagnosis and prevention, many technologies have adopted ultrasonic detection technology, which collects parameters such as structural vibration and noise during the operation of the motor in real time and uses data modeling for fault prediction.

[0003] However, only limited ultrasonic waves or other single sensors are used for fault detection in the existing technology, resulting in difficulty in comprehensively capturing local subtle cracks or early abnormal signals in the complex structure of the motor, and thus leading to the problem of low accuracy in diagnosing faults of the drive motor. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method and device for diagnosing faults of a drive motor based on artificial intelligence technology, aiming to solve the problem of low accuracy in diagnosing faults of the drive motor in the existing technology.

[0005] On the one hand, the present invention proposes a method for diagnosing faults of a drive motor based on artificial intelligence technology, and the method includes: Multimodal sensors are respectively arranged at preset parts of the drive motor, and multi-dimensional data signals related to the fault diagnosis of the drive motor collected by the multimodal sensors are obtained; The multi-dimensional data signals are respectively subjected to real-time filtering, amplification and noise reduction processing, and then a feature extraction algorithm is used to extract features from the multi-dimensional data signals to obtain a plurality of key feature parameters; Based on a preset dynamic weighted fusion algorithm, according to the signal quality and current working condition information collected by the multimodal sensors, different weights are assigned to the key feature parameters obtained from different sensor channels among the multimodal sensors, so as to fuse a plurality of key feature parameters to obtain a fused key feature parameter; A hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network is established, and the fused key feature parameter is input into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor.

[0006] Further, in the above method for diagnosing faults of a drive motor based on artificial intelligence technology, the step of using a feature extraction algorithm to extract features from the multi-dimensional data signals to obtain a plurality of key feature parameters includes: Time domain features are extracted from the multi-dimensional data signals based on the statistics of the multi-dimensional data signals; Extract frequency-domain features from the multi-dimensional data signal based on the Fourier transform of the multi-dimensional data signal; Extract time-frequency domain features from the multi-dimensional data signal based on the wavelet transform of the multi-dimensional data signal; Extract transform-domain features from the multi-dimensional data signal based on the Hilbert-Huang transform of the multi-dimensional data signal.

[0007] Further, in the above drive motor fault diagnosis method based on artificial intelligence technology, wherein, the step of assigning different weights to the key feature parameters obtained from different sensor channels in the multi-modal sensors according to the signal quality and current working condition information collected by the multi-modal sensors based on a preset dynamic weighted fusion algorithm to fuse multiple key feature parameters to obtain fused key feature parameters includes; Respectively obtain the signal-to-noise ratio and signal variance of the data signals collected by the multi-modal sensors, linearly map the signal-to-noise ratio and signal variance to a preset interval, and then perform weighted fusion to obtain the corresponding signal quality score for a single sensor channel; Look up the basic weight coefficients of each sensor channel corresponding to the current working condition information in a preset basic weight mapping table according to the current working condition information; Calculate the initial weights of different sensor channels by combining the signal quality score of a single sensor channel and the basic weight coefficients under the corresponding working conditions; Perform weighted summation on the corresponding key feature parameters according to the initial weights of different sensor channels to obtain fused key feature parameters.

[0008] Further, in the above drive motor fault diagnosis method based on artificial intelligence technology, wherein, before the step of performing weighted summation on the corresponding key feature parameters according to the initial weights of different sensor channels to obtain fused key feature parameters, it further includes: Introduce a penalty factor to the initial weights of the sensor channels with transmission lag and perform smoothing processing using exponential moving average filtering; Among them, the expression of the initial weight processing process is: ; ; ; Among them, The weight after introducing the penalty factor to the initial weight of the th sensor channel, is the initial weight of the th sensor channel, is the penalty factor, is the natural logarithm, is the attenuation coefficient, which controls the influence intensity of the delay on the weight, is the transmission delay time of the th sensor channel, is the final smoothed weight of the th sensor channel, is the smoothing coefficient, is the final smoothed weight of the th sensor channel at the previous moment.

[0009] Further, in the above drive motor fault diagnosis method based on artificial intelligence technology, the steps of establishing a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and inputting the fused key feature parameters into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor include: Using the long short-term memory network to learn the long-term dependencies in the time-series data among the fused key feature parameters through the gating mechanism, and using the convolutional kernels in the convolutional neural network to slide and extract local structured features; Fusing the temporal global features output by the long short-term memory network and the local structured features output by the convolutional neural network, and mapping them to the classification / regression target through the fully connected layer to obtain the fault information of the drive motor.

[0010] Further, in the above drive motor fault diagnosis method based on artificial intelligence technology, the method further includes: Determining the corresponding abnormal thresholds under different working conditions according to the fault probability distributions of normal samples under different working conditions; In the inference stage of the hybrid deep network architecture, loading the corresponding abnormal thresholds according to the identified working conditions, and comparing the abnormal probability output by the hybrid deep network architecture with the abnormal thresholds to determine whether there is a fault in the drive motor.

[0011] Further, in the above drive motor fault diagnosis method based on artificial intelligence technology, the step of respectively arranging multi-modal sensors at preset parts of the drive motor includes: Arranging ultrasonic probes on the stator, rotor and bearings of the drive motor, and installing vibration and temperature sensors both outside and inside the drive motor.

[0012] Another object of the present invention is to provide a drive motor fault diagnosis device based on artificial intelligence technology, and the device includes: An acquisition module, configured to respectively arrange multi-modal sensors at preset parts of the drive motor, and acquire multi-dimensional data signals related to drive motor fault diagnosis collected by the multi-modal sensors; An extraction module for performing real-time filtering, amplification, and noise reduction processing on multi-dimensional data signals respectively, and then using a feature extraction algorithm to extract features from the multi-dimensional data signals to obtain multiple key feature parameters; A fusion module for, based on a preset dynamic weighted fusion algorithm, according to the signal quality and current working condition information collected by multi-modal sensors, assigning different weights to the key feature parameters obtained from different sensor channels among the multi-modal sensors, so as to fuse multiple key feature parameters to obtain fused key feature parameters; A diagnosis module for establishing a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and inputting the fused key feature parameters into the pre-trained hybrid deep network architecture to obtain the fault information of the corresponding drive motor.

[0013] Another object of the present invention is to provide a readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0014] Another object of the present invention is to provide an electronic device, including a memory, a processor, and a computer program stored on the memory and running on the processor, and when the processor executes the program, the steps of the above method are implemented.

[0015] In the present invention, multi-modal sensors are arranged at preset positions of the drive motor to collect multi-dimensional data signals. After real-time filtering, amplification, and noise reduction processing, key feature parameters are extracted. The dynamic weighted fusion algorithm is used to combine the signal quality and working condition information to assign weights to the parameters of different sensor channels and fuse them. Finally, the fused key feature parameters are input into a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network for fault diagnosis. The multi-modal sensors break through the information dimension limitation of a single sensor, can cover multiple types of fault features, improve data reliability and enhance working condition adaptability through dynamic weight optimization, and enhance the discrimination of complex fault modes by means of deep feature learning, ultimately effectively improving the accuracy of fault diagnosis. It solves the problem of low accuracy of drive motor fault diagnosis in the prior art. Brief Description of the Drawings

[0016] Figure 1 It is a flowchart of a drive motor fault diagnosis method based on artificial intelligence technology in the first embodiment of the present invention; Figure 2 It is a structural block diagram of a drive motor fault diagnosis device based on artificial intelligence technology in the third embodiment of the present invention.

[0017] The following specific embodiments will further illustrate the present invention in conjunction with the above drawings. Specific Embodiments

[0018] To facilitate the understanding of the present invention, the present invention will be described more comprehensively below with reference to the relevant drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, these embodiments are provided to make the disclosure of the present invention more thorough and comprehensive.

[0019] It should be noted that when an element is referred to as being "fixedly provided on" another element, it can be directly on the other element or there may also be an intermediate element. When an element is considered to be "connected to" another element, it can be directly connected to the other element or there may be an intermediate element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are only for the purpose of illustration.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0021] Embodiment 1 Please refer to Figure 1 , which shows a drive motor fault diagnosis method based on artificial intelligence technology in the first embodiment of the present invention. The method includes steps S10 to S12.

[0022] Step S10: Multimodal sensors are respectively arranged at preset parts of the drive motor, and multidimensional data signals related to drive motor fault diagnosis collected by the multimodal sensors are obtained.

[0023] Among them, during the operation of the drive motor, in order to obtain more comprehensive and accurate information required for fault diagnosis, various different types of sensors are respectively installed at multiple preset key parts of the drive motor to form a multimodal sensor array. Specifically, ultrasonic probes can be arranged on the stator, rotor and bearings of the drive motor, and vibration and temperature sensors can be installed outside and inside the drive motor at the same time.

[0024] Exemplarily, during the operation of the drive motor, the stator, rotor, and bearings, as core components, their operating states are directly related to the overall performance and stability of the motor. When these components malfunction, specific ultrasonic signals are generated. By installing ultrasonic probes at the stator position, the changes in ultrasonic signals caused by faults such as partial discharge and insulation aging during the energized operation of the stator winding can be monitored in real time. For example, when the insulation layer of the stator winding is damaged, the high-frequency ultrasonic signals generated by the discharge can be sensitively captured by the probe; by arranging ultrasonic probes at the rotor, the ultrasonic characteristics generated by abnormal friction or collision caused by component looseness and structural defects during the high-speed rotation of the rotor can be sensed. For instance, the ultrasonic fluctuations generated by the vibration transmission caused by broken rotor bars to the surface can be collected by the probe; by arranging ultrasonic probes on the bearings, the ultrasonic signals generated by the abnormal friction of the contact surface due to problems such as wear and poor lubrication between the bearing rolling elements and the inner and outer rings can be monitored. For example, when the wear of the bearing balls intensifies, the ultrasonic probe will record stronger and more complex-frequency ultrasonic signals.

[0025] Install vibration sensors at the bearings to capture the vibration signals generated during the operation of the bearings. These signals contain key information about mechanical faults such as bearing wear and looseness. Meanwhile, arrange temperature sensors on the surface of the motor housing to collect the temperature data of the entire motor in real time. An abnormal increase in temperature often indicates potential faults such as overload and poor heat dissipation inside the motor.

[0026] Step S11: Perform real-time filtering, amplification, and noise reduction processing on the multi-dimensional data signals respectively, and then use the feature extraction algorithm to extract features from the multi-dimensional data signals to obtain multiple key feature parameters.

[0027] Among them, after collecting multi-dimensional data signals of the drive motor using multi-modal sensors, due to the existence of complex interference factors such as electromagnetic interference and mechanical vibration in the actual collection environment, a large amount of noise is often mixed in the original data signals, and some useful signals may be difficult to identify due to too small amplitudes. Therefore, it is necessary to perform preprocessing and feature extraction operations on these multi-dimensional data signals. First, real-time filtering processing filters out high-frequency noise or low-frequency interference mixed in the data signals by designing appropriate filters, such as low-pass filters, band-pass filters, or adaptive filters, according to the frequency characteristics of the signals, and retains the effective signals in the frequency bands related to fault diagnosis. For example, a band-pass filter is used to extract the vibration components related to bearing faults within a specific frequency range in the vibration signal; then, amplification processing is aimed at the situation where the signal amplitude is weak, and the amplitude of the effective signal is increased through an amplifier to make the signal characteristics more obvious, facilitating subsequent processing and avoiding the key information being ignored due to the signal being too weak; then, noise reduction processing further uses algorithms such as wavelet denoising and mean filtering to suppress the random noise in the signal, optimize the signal quality, and ensure the accuracy and reliability of the data. After completing real-time filtering, amplification, and noise reduction processing, the multi-dimensional data signals already have high purity and clarity. At this time, using feature extraction algorithms such as Fourier transform, wavelet transform, and principal component analysis (PCA), key information that can characterize the operating state and fault characteristics of the drive motor is mined from the processed multi-dimensional data signals. For example, fault characteristic frequencies are extracted from the vibration signal, harmonic components are extracted from the current signal, and the temperature rise rate is extracted from the temperature signal. These information are converted into multiple key characteristic parameters. These parameters, as the "feature labels" of the operating state of the drive motor, can more accurately and concisely reflect the operating conditions and potential faults of the motor, providing an effective basis for subsequent fault diagnosis.

[0028] Specifically, in the embodiments of the present invention, after completing real-time filtering, amplification, and noise reduction processing on the multi-dimensional data signals, in order to deeply mine the drive motor fault information contained in the signals, a variety of feature extraction algorithms are used to extract key characteristic parameters from different dimensions. First, time-domain features are extracted based on the statistical characteristics of the multi-dimensional data signals. By calculating statistical quantities such as the mean, variance, peak value, kurtosis, and skewness of the signals, the change trend and fluctuation degree of the signals on the time axis are intuitively reflected. For example, the mean of the vibration signal during the normal operation of the drive motor is within a stable range. If the mean suddenly increases, it may indicate that the motor has an abnormality; the variance reflects the degree of dispersion of the signal, and the variance of the vibration signal will increase significantly when the bearing wear intensifies. These time-domain features can quickly capture the overall change trend of the signal, providing preliminary clues for fault diagnosis.

[0029] Secondly, the Fourier transform is used to extract the frequency-domain features of the multi-dimensional data signal. The Fourier transform converts the time-domain signal into a frequency-domain representation, which can reveal the distribution of different frequency components in the signal. By analyzing the amplitude and phase of each frequency component, information such as the natural frequency and fault characteristic frequency during the operation of the drive motor can be identified. For example, when a fault occurs in the motor bearing, a peak at a specific frequency will appear in the spectrum of the vibration signal. By comparing the spectrum in the normal state, it is possible to determine whether there is a fault in the bearing and the type of the fault.

[0030] Furthermore, the wavelet transform is used to extract the time-frequency domain features of the multi-dimensional data signal. The wavelet transform overcomes the deficiency of the Fourier transform that it cannot reflect the local characteristics of the signal in the time domain. It can analyze the signal in both the time and frequency dimensions simultaneously. By selecting an appropriate wavelet basis function, the signal is decomposed into wavelet coefficients of different scales and frequencies, clearly showing the change of the frequency components of the signal at different times. For non-stationary signals, such as the current signal during the start-up and braking processes of the motor, the wavelet transform can accurately capture the dynamic change of the signal frequency over time, which helps to detect transient and sudden faults.

[0031] Finally, based on the Hilbert-Huang transform, the transform-domain features of the multi-dimensional data signal are extracted. This method decomposes the complex signal into multiple intrinsic mode functions (IMFs) through empirical mode decomposition (EMD), and then performs the Hilbert transform on each IMF to obtain the instantaneous frequency and amplitude information of the signal, forming the Hilbert spectrum, so as to more accurately describe the non-linear and non-stationary characteristics of the signal. When dealing with the complex signals generated during motor faults, especially the signals caused by compound faults, the Hilbert-Huang transform can effectively separate the signal components generated by different fault sources, accurately extract the fault features, and provide a more comprehensive and detailed basis for the diagnosis of complex faults. Through the above feature extraction of the multi-dimensional data signal from multiple perspectives of the time domain, frequency domain, time-frequency domain, and transform domain, multiple key feature parameters that can comprehensively reflect the operating state and fault information of the drive motor are obtained, laying a solid foundation for subsequent fault diagnosis and analysis.

[0032] Step S12: Based on the preset dynamic weighted fusion algorithm, according to the signal quality and the current working condition information collected by the multi-modal sensors, different weights are assigned to the key feature parameters obtained from different sensor channels among the multi-modal sensors, so as to fuse multiple key feature parameters to obtain the fused key feature parameters.

[0033] Among them, after obtaining multiple key feature parameters collected by the multi-modal sensors of the drive motor and extracted through feature extraction, due to the differences in the reliability and sensitivity of different sensors under different working conditions, and the signal quality collected by each sensor will also change due to factors such as environmental interference and its own performance, it is necessary to use the preset dynamic weighted fusion algorithm to process these key feature parameters.

[0034] The core logic of this algorithm lies in comprehensively considering the signal quality collected by multi-modal sensors and the current working condition information, and assigning different weights to the key feature parameters obtained from different sensor channels. Among them, the signal quality evaluation is mainly based on indicators such as the signal-to-noise ratio and stability of the data collected by the sensors. For example, when the vibration sensor is affected by environmental electromagnetic interference and the signal noise is too large, its signal-to-noise ratio decreases, and the algorithm will correspondingly reduce the weight of the key feature parameters of this sensor channel; while for sensors with stable signals and low noise, a higher weight is assigned. At the same time, the current working condition information includes factors such as the speed, load, and running time of the drive motor. Because the sensitivity of each sensor to fault characteristics is different under different working conditions. For example, under high-speed operating conditions, the vibration sensor is more sensitive to bearing faults, and the algorithm will increase its weight; while under heavy-load conditions, the current sensor has more advantages in detecting faults such as winding overheating and short circuits, and at this time, the weight of the key feature parameters of the current sensor channel will be increased. By this way of dynamically adjusting the weights, the key feature parameters from different channels of multi-modal sensors are weighted and fused, and finally the fused key feature parameters are obtained. The fused key feature parameters integrate the effective information of each sensor under the current working condition, which not only avoids misjudgment caused by poor signal quality of a single sensor or insensitivity to specific working conditions, but also gives full play to the collaborative advantages of multi-modal sensors, making the fused parameters more accurately and comprehensively reflect the actual operating state and potential fault characteristics of the drive motor.

[0035] Step S13, establish a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and input the fused key feature parameters into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor.

[0036] Among them, in order to accurately extract the fault information of the drive motor from the fused key feature parameters, the embodiment of the present invention constructs a hybrid deep network architecture composed of a long short-term memory network (LSTM) and a convolutional neural network (CNN). In practical applications, first use a large amount of drive motor data containing normal states and different fault types to pre-train the hybrid deep network architecture, and continuously adjust the parameters in the network through the backpropagation algorithm, so that the network can accurately identify the feature patterns corresponding to different fault types. After the pre-training is completed, the fused key feature parameters obtained by the dynamic weighted fusion algorithm are input into this hybrid deep network architecture, and finally the corresponding drive motor fault information is output through the output layer of the network, including the fault type (such as bearing wear, winding short circuit, etc.), the degree of the fault, and the possibility of the fault occurring, etc., to achieve accurate diagnosis of the drive motor fault.

[0037] Specifically, the long short-term memory network (LSTM), with its unique gating mechanism including an input gate, a forget gate, and an output gate, can efficiently process the time-series data in the fused key feature parameters. During the operation of the drive motor, the trends of parameters such as temperature and vibration changing over time contain important clues about the occurrence and development of faults. The gating mechanism of LSTM can selectively retain and forget historical information, thereby learning the long-term dependencies in this time-series data. For example, the wear of the motor bearing is a gradual process. LSTM can capture the trend of the vibration signal gradually increasing over a long period of time, accurately memorize the historical information related to the fault development, avoid the problem that traditional recurrent neural networks are difficult to learn long-term dependencies due to the vanishing gradient, and then extract the global features reflecting the overall change law of the time-series data. The convolutional neural network (CNN), on the other hand, performs a sliding operation on the fused key feature parameters through a convolutional kernel to automatically extract the local structured features in the data. For the multi-dimensional data of the drive motor, such as the distortion characteristics of the current waveform and the specific frequency components of the vibration signal, the convolutional kernel of CNN can keenly capture the correlations and patterns between these local data, and continuously compress and abstract the features through convolutional operations and pooling operations, turning the complex raw data into more representative local structured features, such as extracting the local pattern of abnormal harmonic components from the current signal. After obtaining the time-series global features output by LSTM and the local structured features output by CNN respectively, the two are fused to fully combine the advantages of the time-series change trend and the local data features, enabling the network to more comprehensively understand the fault information in the data. The fused features are processed through a fully connected layer, and the fully connected layer maps the fused features to the classification or regression target. If it is used for fault type diagnosis, the probability of different fault types is output through the classification target to determine what kind of fault has occurred in the drive motor; if it is used for fault degree assessment, a specific numerical value is output through the regression target to quantify the severity of the fault, ultimately achieving the accurate acquisition of the fault information of the drive motor. In addition, in some alternative embodiments of the present invention, the method further includes: Determining corresponding abnormal thresholds according to the fault probability distributions of normal samples under different working conditions; In the inference stage of the hybrid deep network architecture, loading the corresponding abnormal threshold according to the identified working condition, and determining whether there is a fault in the drive motor by comparing the abnormal probability output by the hybrid deep network architecture with the abnormal threshold.

[0038] Among them, during the actual operation of the drive motor, there are significant differences in the normal operating states and data performances of the motor under different working conditions (such as no-load, light load, full load, and different rotational speeds, etc.). In order to more accurately judge whether the motor has a fault, an abnormal threshold determination mechanism based on the working condition is further introduced. First, a large number of normal operation sample data of the drive motor under different working conditions are collected. These data cover the fusion key feature parameters after processing and feature extraction of multi-modal sensor signals of the motor under various typical working conditions. Through in-depth analysis of these normal sample data and using statistical methods such as probability density estimation and confidence interval calculation, the fault probability distribution of normal samples under each working condition is determined. For example, under the no-load working condition, analyze the fusion feature parameters such as current, vibration, and temperature during a large number of normal operations, construct its corresponding fault probability distribution model, and set a reasonable abnormal threshold based on this distribution model. This threshold is used to define the boundary from the normal state to the abnormal state under the no-load working condition; similarly, for other working conditions such as light load and full load, establish corresponding fault probability distributions and determine their respective abnormal thresholds. When the hybrid deep network architecture enters the inference stage, that is, when diagnosing the fault of the drive motor during actual operation, first identify the current working condition by collecting the operating parameters (such as rotational speed, load, etc.) of the current motor through sensors, and then load the pre-determined corresponding abnormal threshold according to the identified working condition. At this time, input the current fusion key feature parameters of the drive motor into the pre-trained hybrid deep network architecture, and the network outputs a probability value indicating that the current motor state is abnormal after calculation. Finally, compare this abnormal probability with the abnormal threshold under the corresponding working condition loaded. If the abnormal probability is greater than the abnormal threshold, it indicates that the current operating state of the motor deviates from the normal range under this working condition, and it is determined that the drive motor has a fault. By setting abnormal thresholds according to different working conditions and making a comparison and judgment in combination with the working condition during diagnosis, it effectively avoids misjudgment caused by working condition differences, significantly improves the accuracy and reliability of drive motor fault diagnosis, and ensures that motor faults can be identified in a timely and accurate manner under various complex working conditions.

[0039] In summary, in the above embodiments of the present invention, the drive motor fault diagnosis method based on artificial intelligence technology collects multi-dimensional data signals by arranging multi-modal sensors at preset parts of the drive motor, extracts key feature parameters after real-time filtering, amplification and noise reduction processing, uses a dynamic weighted fusion algorithm to assign different weights to the parameters of different sensor channels in combination with signal quality and working condition information and fuses them. Finally, the fused key feature parameters are input into a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network for fault diagnosis. The multi-modal sensors break through the limitation of the information dimension of a single sensor, can cover various types of fault features, improve data reliability and enhance working condition adaptability through dynamic weight optimization, and enhance the discrimination of complex fault modes with the help of deep feature learning, ultimately effectively improving the accuracy of fault diagnosis. It solves the problem of low accuracy in drive motor fault diagnosis in the prior art.

[0040] Embodiment 2 This embodiment also proposes a drive motor fault diagnosis method based on artificial intelligence technology. The difference between the drive motor fault diagnosis method based on artificial intelligence technology in this embodiment and the drive motor fault diagnosis method based on artificial intelligence technology in Embodiment 1 is as follows: The step of, based on the preset dynamic weighted fusion algorithm, assigning different weights to the key feature parameters obtained from different sensor channels among the multi-modal sensors according to the signal quality collected by the multi-modal sensors and the current working condition information to fuse multiple key feature parameters to obtain fused key feature parameters includes: Respectively obtain the signal-to-noise ratio and signal variance of the data signals collected by the multi-modal sensors, linearly map the signal-to-noise ratio and signal variance to a preset interval and then perform weighted fusion to obtain the corresponding single-channel signal quality score of the sensor; According to the current working condition information, look up the basic weight coefficients of each sensor channel corresponding to the current working condition information in the preset basic weight mapping table; Calculate the initial weights of different sensor channels by combining the single-channel signal quality score of the sensor and the basic weight coefficients under the corresponding working conditions; Perform weighted summation on the corresponding key feature parameters according to the initial weights of different sensor channels to obtain fused key feature parameters.

[0041] Among them, in the fault diagnosis of the drive motor, to achieve the effective fusion of multi-modal sensor data, this technical solution adopts a dynamic weighted fusion algorithm, which differentially weights the key characteristic parameters of each sensor channel by comprehensively considering the signal quality and working condition information. First, for the data signals collected by each sensor, calculate their signal-to-noise ratio (SNR) and signal variance respectively. The SNR reflects the proportional relationship between the effective information and noise in the signal. The higher the SNR, the better the signal quality. The signal variance measures the degree of signal fluctuation. Excessive variance may mean that the signal is unstable or there are abnormal interferences. Map these two indicators to a preset interval (such as [0,1]) through a linear mapping function to make the indicators with different dimensions comparable, and then perform weighted fusion on the mapped results to obtain the signal quality score of each sensor's single channel. This score intuitively reflects the reliability of the data collected by each sensor at the current moment. At the same time, according to the current working condition information of the drive motor (such as speed, load, etc.), look up the corresponding basic weight coefficients of each sensor channel in the pre-established basic weight mapping table. This mapping table is constructed based on a large amount of historical data and expert knowledge, which clarifies the importance differences of each sensor for fault diagnosis under different working conditions. For example, the basic weight of the vibration sensor will be relatively high during high-speed operation, while the basic weight of the current sensor will increase under heavy load conditions. Subsequently, combine the single-channel signal quality score with the basic weight coefficient under the corresponding working condition to calculate the initial weight of each sensor channel, which can be specifically achieved by multiplying the two or more complex non-linear combination methods. In this way, it not only considers the influence of the working condition on the importance of the sensor, but also dynamically adjusts the reliability difference caused by the signal quality change. Finally, use the calculated initial weights to perform weighted summation on the key characteristic parameters of each sensor channel, and fuse the multi-source heterogeneous characteristic parameters into a comprehensive key characteristic parameter. This parameter fully integrates the advantageous information of each sensor under the current working condition, effectively suppresses the interference of low-quality signals, and provides a more accurate and reliable basis for subsequent fault diagnosis.

[0042] In addition, before the step of performing weighted summation on the corresponding key characteristic parameters according to the initial weights of different sensor channels to obtain the fused key characteristic parameter, the following steps are also included: Introduce a penalty factor to the initial weight of the sensor channel with transmission lag and perform smoothing processing using exponential moving average filtering; Among them, the expression of the initial weight processing process is: ; ; ; Among them, The weight of the initial weight of the th sensor channel after introducing the penalty factor, The initial weight of a sensor channel is the penalty factor is the natural logarithm is the decay coefficient, which controls the influence intensity of the delay on the weight is the transmission delay time of the th sensor channel The final smoothed weight of the th sensor channel is the smoothing coefficient The final smoothed weight of the

[0043] Among them, in the process of multi-modal sensor data fusion of the drive motor, due to factors such as the hardware performance and transmission lines of each sensor, there may be problems of data transmission lag in some sensor channels, which will cause the data collected by them to be out of sync with the data of other sensors in time. If the fusion is directly carried out according to the initial weight, errors may be introduced and even diagnostic errors may occur. To solve this problem, before obtaining the fusion key feature parameters by weighted summation of the key feature parameters according to the initial weight, special processing is performed on the initial weights of the sensor channels with transmission lag. First, a penalty factor is introduced to adjust the initial weights of the sensor channels with transmission lag. However, simple penalty adjustment may cause the weight change to be too drastic. Therefore, exponential moving average filtering is further used to smooth the adjusted weights, making the weight change smoother and more continuous, and avoiding affecting the stability of data fusion due to weight mutation. Effectively eliminate the negative impact brought by transmission lag, and then obtain more accurate and reliable fusion key feature parameters, improving the accuracy and stability of drive motor fault diagnosis.

[0044] In summary, the drive motor fault diagnosis method based on artificial intelligence technology in the above embodiments of the present invention collects multi-dimensional data signals by arranging multi-modal sensors at preset parts of the drive motor, extracts key feature parameters after real-time filtering, amplification and noise reduction processing, uses a dynamic weighted fusion algorithm to assign weights to different sensor channel parameters in combination with signal quality and working condition information and fuse them, and finally inputs the fusion key feature parameters into a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network for fault diagnosis. The multi-modal sensors break through the limitation of the information dimension of a single sensor, can cover multiple types of fault features, improve data reliability and enhance working condition adaptability through dynamic weight optimization, and enhance the discrimination of complex fault modes by means of deep feature learning, and finally effectively improve the accuracy of fault diagnosis. Solve the problem of low accuracy of drive motor fault diagnosis in the prior art.

[0045] Embodiment III Please refer to Figure 2, shown is a driving motor fault diagnosis device based on artificial intelligence technology proposed in the third embodiment of the present invention, and the device includes: An acquisition module 100, configured to respectively deploy multi-modal sensors at preset parts of the driving motor, and acquire multi-dimensional data signals related to driving motor fault diagnosis collected by the multi-modal sensors; An extraction module 200, configured to respectively perform real-time filtering, amplification, and noise reduction processing on the multi-dimensional data signals, and then use a feature extraction algorithm to extract features from the multi-dimensional data signals to obtain multiple key feature parameters; A fusion module 300, configured to, based on a preset dynamic weighted fusion algorithm, assign different weights to the key feature parameters obtained from different sensor channels among the multi-modal sensors according to the signal quality and current working condition information collected by the multi-modal sensors, so as to fuse the multiple key feature parameters to obtain a fused key feature parameter; A diagnosis module 400, configured to establish a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and input the fused key feature parameter into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the driving motor.

[0046] The functions or operation steps realized when the above-mentioned modules are executed are substantially the same as those in the above method embodiment, and will not be elaborated here.

[0047] Embodiment 4 On the other hand, the present invention also provides a readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the method in any one of the above-mentioned Embodiments 1 to 2 are realized.

[0048] Embodiment 5 On the other hand, the present invention also provides an electronic device, the electronic device includes a memory, a processor, and a computer program stored on the memory and running on the processor, and when the processor executes the program, the steps of the method in any one of the above-mentioned Embodiments 1 to 2 are realized.

[0049] The technical features of each of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0050] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable storage medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatuses, or devices. For the purposes of this specification, a "computer-readable storage medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

[0051] More specific examples (a non-exhaustive list) of computer-readable storage media include the following: an electrical connection part with one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, a computer-readable storage medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or other suitable processing as necessary, and then stored in a computer memory.

[0052] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application-specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0053] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples", etc., mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0054] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

Claims

1. A drive motor fault diagnosis method based on artificial intelligence technology, characterized in that, The method includes: Multimodal sensors are respectively arranged at preset parts of the drive motor, and multidimensional data signals related to the fault diagnosis of the drive motor collected by the multimodal sensors are obtained; The multidimensional data signals are respectively subjected to real-time filtering, amplification and noise reduction processing, and then a feature extraction algorithm is used to extract features from the multidimensional data signals to obtain a plurality of key feature parameters; Based on a preset dynamic weighted fusion algorithm, according to the signal quality and current working condition information collected by the multimodal sensors, different weights are assigned to the key feature parameters obtained from different sensor channels among the multimodal sensors, so as to fuse the plurality of key feature parameters to obtain a fused key feature parameter; A hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network is established, and the fused key feature parameter is input into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor; Among them, the step of, based on a preset dynamic weighted fusion algorithm, according to the signal quality and current working condition information collected by the multimodal sensors, assigning different weights to the key feature parameters obtained from different sensor channels among the multimodal sensors, so as to fuse the plurality of key feature parameters to obtain a fused key feature parameter includes; The signal-to-noise ratio and signal variance of the data signals collected by the multimodal sensors are respectively obtained, and the signal-to-noise ratio and signal variance are linearly mapped to a preset interval and then weighted and fused to obtain a corresponding sensor single-channel signal quality score; According to the current working condition information, the basic weight coefficients of each sensor channel corresponding to the current working condition information are found in a preset basic weight mapping table; Combining the sensor single-channel signal quality score and the basic weight coefficient under the corresponding working condition to calculate the initial weights of different sensor channels; Performing weighted summation on the corresponding key feature parameters according to the initial weights of different sensor channels to obtain a fused key feature parameter.

2. The drive motor fault diagnosis method based on artificial intelligence technology according to claim 1, characterized in that The step of using a feature extraction algorithm to extract features from the multidimensional data signals to obtain a plurality of key feature parameters includes: Based on the statistics of the multidimensional data signals, time domain features are extracted from the multidimensional data signals; Based on the Fourier transform of the multidimensional data signals, frequency domain features are extracted from the multidimensional data signals; Based on the wavelet transform of the multidimensional data signals, time-frequency domain features are extracted from the multidimensional data signals; Based on the Hilbert-Huang transform of the multidimensional data signals, transform domain features are extracted from the multidimensional data signals.

3. The drive motor fault diagnosis method based on artificial intelligence technology according to claim 1, wherein, Before the step of performing weighted summation on the corresponding key feature parameters according to the initial weights of different sensor channels to obtain a fused key feature parameter, it further includes: Introducing a penalty factor to the initial weight of the sensor channel with transmission lag and performing smoothing processing by using exponential moving average filtering; Among them, the expression of the initial weight processing process is: ； ； ； Among them, The weight after introducing the penalty factor adjustment for the initial weight of the th sensor channel, is the initial weight of the th sensor channel, is the penalty factor, is the natural logarithm, is the attenuation coefficient, which controls the influence intensity of the delay on the weight, is the transmission delay time of the th sensor channel, is the final smoothed weight of the 4. The drive motor fault diagnosis method based on artificial intelligence technology according to claim 3, wherein, The step of establishing a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and inputting the fused key feature parameter into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor includes: Utilize the long short-term memory network to learn and fuse the long-term dependencies in the time-series data among the key feature parameters through the gating mechanism, and utilize the convolutional kernels in the convolutional neural network to slide and extract local structured features; Fuse the time-series global features output by the long short-term memory network and the local structured features output by the convolutional neural network, and map them to the classification / regression target through the fully connected layer to realize the acquisition of the fault information of the drive motor.

5. The drive motor fault diagnosis method based on artificial intelligence technology according to claim 4, wherein The method further includes: Determine the corresponding anomaly thresholds under different working conditions according to the fault probability distributions of the normal samples under different working conditions; In the inference stage of the hybrid deep network architecture, load the corresponding anomaly thresholds according to the identified working conditions, and determine whether there is a fault in the drive motor by comparing the anomaly probability output by the hybrid deep network architecture with the anomaly thresholds.

6. The drive motor fault diagnosis method based on artificial intelligence technology according to claim 1, characterized in that The step of respectively arranging multi-modal sensors at the preset parts of the drive motor includes: Arrange ultrasonic probes on the stator, rotor and bearings of the drive motor, and install vibration and temperature sensors outside and inside the drive motor at the same time.

7. A driving motor fault diagnosis device based on artificial intelligence technology, characterized in that, For implementing the drive motor fault diagnosis method based on artificial intelligence technology according to any one of claims 1 to 6, the device includes: An acquisition module, configured to respectively arrange multi-modal sensors at the preset parts of the drive motor, and acquire multi-dimensional data signals related to the drive motor fault diagnosis collected by the multi-modal sensors; An extraction module, configured to respectively perform real-time filtering, amplification and noise reduction processing on the multi-dimensional data signals, and then use a feature extraction algorithm to extract features from the multi-dimensional data signals to obtain a plurality of key feature parameters; A fusion module, configured to, based on a preset dynamic weighted fusion algorithm, assign different weights to the key feature parameters obtained from different sensor channels among the multi-modal sensors according to the signal quality and the current working condition information collected by the multi-modal sensors, so as to fuse the plurality of key feature parameters to obtain fused key feature parameters; A diagnosis module, configured to establish a hybrid deep network architecture composed of a long short-term memory network and a convolutional neural network, and input the fused key feature parameters into the pre-trained hybrid deep network architecture to obtain the corresponding fault information of the drive motor.

8. A readable storage medium, on which a computer program is stored, characterized in that, When the program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 6.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored on the memory and running on the processor. When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 6.

Images