New energy automobile motor bearing online dynamic monitoring and fault positioning method and device based on multi-sensor fusion, equipment and medium

Abstract

The invention relates to a new energy automobile motor bearing online dynamic monitoring and fault positioning method and device based on multi-sensor fusion, equipment and a medium. The method comprises the following steps: collecting multi-source detection signals of a motor bearing, carrying out time sequence alignment on the multi-source detection signals to obtain a synchronous multi-source signal set, extracting working condition characteristics based on the synchronous multi-source signal set, carrying out bearing state analysis to obtain a bearing state analysis result, and determining the motor bearing state based on the working condition characteristics and the bearing state analysis result. Positioning a bearing potential fault to obtain a bearing potential fault positioning result, and predicting the residual life of the motor bearing based on the working condition characteristics, the bearing state analysis result and the bearing potential fault positioning result to obtain a residual life prediction result. And generating a motor bearing online monitoring report based on a bearing state analysis result, a potential fault positioning result and a residual life prediction result. By adopting the method, real-time monitoring, fault positioning and residual life prediction of the motor bearing can be realized.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring and fault diagnosis, and in particular relates to a method, device, equipment and medium for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion. Background Technology

[0002] With the development of intelligent monitoring and fault diagnosis technologies for key components of new energy vehicles, multi-sensor fusion fault diagnosis technology has emerged. This technology can achieve comprehensive perception and intelligent analysis of the condition of motor bearings by integrating data from multiple sensors such as vibration, current, temperature, and acoustics. Traditional fault diagnosis methods include manual experience-based diagnosis, fault code diagnosis, and single-parameter offline monitoring. These methods rely primarily on the subjective judgment of maintenance technicians or the reading of fault codes using specialized diagnostic instruments. However, these methods lack sensitivity for mechanical and gradual faults in motor bearings. Furthermore, single-parameter monitoring methods, derived from industrial scenarios, do not fully consider the complex automotive-grade environment. These current diagnostic methods or traditional approaches suffer from strong subjectivity, insensitivity to early fault detection, high false alarm rates, insufficient real-time performance, and a disconnect between monitoring and diagnosis, making it difficult to meet the real-time, accurate, and reliable health management needs of new energy vehicle motor bearings. Summary of the Invention

[0003] Therefore, it is necessary to provide a method, device, equipment, and medium for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion that can solve the above problems.

[0004] Firstly, this application provides a method for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion, including:

[0005] Multi-source detection signals of motor bearings are collected, and the multi-source detection signals are time-aligned to obtain a synchronous multi-source signal set;

[0006] Based on synchronous multi-source signal sets, operating condition features are extracted;

[0007] Based on the operating conditions, statistical methods are used to analyze the bearing condition and obtain the bearing condition analysis results.

[0008] Based on the operating condition characteristics and bearing condition analysis results, a machine learning model is used to locate potential bearing faults, and the bearing potential fault location results are obtained.

[0009] Based on the operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the remaining life of the motor bearing is predicted according to the physical degradation mechanism of the motor bearing, and the remaining life prediction results are obtained.

[0010] Based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, an online monitoring report for motor bearings is generated.

[0011] In one embodiment, the synchronous multi-source signal set includes vibration signals, current signals, temperature signals, and acoustic signals;

[0012] Based on a synchronous multi-source signal set, operating condition features are extracted, including:

[0013] Based on the vibration signal, time-domain composite features and frequency-domain modulation features are extracted to obtain the vibration characteristics;

[0014] Based on the current signal, electrical parameter distortion features and vibration-current coupling features are extracted to obtain current features. Among them, vibration-current coupling features are calculated based on electrical parameter distortion features and vibration features.

[0015] Based on the temperature signal, temperature change characteristic parameters are extracted to obtain temperature characteristics;

[0016] Based on the acoustic signal, acoustic amplitude features and acoustic frequency domain modulation features are extracted to obtain acoustic features;

[0017] By integrating vibration characteristics, current characteristics, temperature characteristics, and acoustic characteristics, the operating condition characteristics are obtained.

[0018] In one embodiment, a statistical method is used to perform bearing condition analysis, and the bearing condition analysis results are obtained, including:

[0019] Based on a synchronous multi-source signal set, motor speed and load parameters are obtained;

[0020] Based on motor speed and load parameters, and according to preset weighting rules, the vibration characteristics, current characteristics, temperature characteristics and acoustic characteristics of the operating conditions are weighted and fused to obtain a fused feature set.

[0021] Principal component analysis was used to reduce the dimensionality of the fused feature set, with a cumulative contribution rate of no less than 90% as the retention criterion.

[0022] Based on the dimensionality reduction feature set and combined with the pre-set dimensionality reduction feature-bearing state sample database, the Fisher discriminant analysis method is used to perform bearing state analysis and obtain the bearing state analysis results.

[0023] In one embodiment, based on operating condition characteristics and bearing condition analysis results, a machine learning model is used to locate potential bearing faults, obtaining the bearing potential fault location results, including:

[0024] For the bearing condition in the bearing condition analysis results, the working condition features are input into a convolutional neural network pre-trained based on different bearing conditions to extract local spatial correlation features of multi-dimensional features and obtain a spatial feature set.

[0025] The core fault features are obtained by filtering the spatial feature set through an attention mechanism.

[0026] Based on core fault characteristics, fault type matching is performed according to a preset fault characteristic-fault type database to obtain suspected fault categories.

[0027] Obtain the material and structural parameters of the motor bearing, and based on these parameters, determine the location of potential faults according to the suspected fault categories.

[0028] By integrating suspected fault categories and potential fault locations, the results of potential bearing fault location are obtained.

[0029] In one embodiment, based on operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the remaining life of the motor bearing is predicted according to the physical degradation mechanism of the motor bearing, resulting in a remaining life prediction result, including:

[0030] Based on the operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the bearing degradation failure threshold is determined according to the physical degradation mechanism of motor bearings.

[0031] The operating condition characteristics, the suspected fault category and the location of potential faults in the bearing are input into the pre-trained long short-term memory model to simulate the bearing degradation time series and generate the bearing degradation time series prediction curve.

[0032] Based on the bearing degradation failure threshold and the bearing degradation time sequence prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point.

[0033] In one embodiment, based on the bearing degradation failure threshold and the bearing degradation time prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point, and is achieved through the following mathematical formula:

[0034]

[0035] Where L represents the predicted remaining life of the motor bearing. For the current monitoring time point, The bearing degradation failure time point and satisfy , This is the threshold for bearing degradation failure. This is a function for predicting bearing degradation time series curves. This is a fault category correction factor. This is a correction factor for the faulty part. This is a comprehensive correction factor for operating conditions. For dynamic error compensation term and , is the time series prediction fluctuation compensation coefficient, and k is the degradation rate decay factor.

[0036] In one embodiment, based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, an online monitoring report for the motor bearing is generated, including:

[0037] Based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, coupled visualization results of bearing condition, fault location, and remaining life are generated through data visualization processing.

[0038] Based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, maintenance and protection suggestions are extracted according to the pre-set motor bearing manual.

[0039] By integrating and coupling visualization results and maintenance and protection recommendations, an online monitoring report for motor bearings is generated.

[0040] Secondly, this application also provides an online dynamic monitoring and fault location device for new energy vehicle motor bearings based on multi-sensor fusion, comprising:

[0041] The multi-source signal acquisition module is used to acquire multi-source detection signals of the motor bearing and to perform time-series alignment of the multi-source detection signals to obtain a synchronous multi-source signal set;

[0042] The operating condition feature extraction module is used to extract operating condition features based on a synchronous multi-source signal set;

[0043] The bearing condition analysis module is used to perform bearing condition analysis based on operating condition characteristics and statistical methods to obtain bearing condition analysis results.

[0044] The potential fault location module is used to locate potential bearing faults based on operating condition characteristics and bearing condition analysis results, and to obtain the bearing potential fault location results.

[0045] The remaining life prediction module is used to predict the remaining life of motor bearings based on operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, according to the physical degradation mechanism of motor bearings, and obtain the remaining life prediction results.

[0046] The monitoring report generation module is used to generate online monitoring reports for motor bearings based on bearing condition analysis results, potential fault location results, and remaining life prediction results.

[0047] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned method for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion.

[0048] Fourthly, this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-mentioned method for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion.

[0049] The aforementioned method, device, equipment, and medium for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion obtain a synchronous multi-source signal set by collecting multi-source detection signals of the motor bearing and aligning them in time sequence. This overcomes the limitations of single-parameter monitoring and achieves comprehensive perception of the bearing status. Based on this synchronous multi-source signal set, operating condition features are extracted, and statistical methods are used for bearing status analysis, avoiding the subjectivity of manual experience-based diagnosis and improving the objectivity and accuracy of bearing status judgment. Combining operating condition features and bearing status analysis results, a machine learning model is used to locate potential bearing faults, solving the problem of traditional methods being insensitive to early fault detection and achieving accurate identification and location of potential faults. Based on operating condition features, bearing status analysis results, and potential fault location results, the remaining life of the motor bearing is predicted in conjunction with the physical degradation mechanism of the motor bearing, breaking the limitation of the disconnect between monitoring and diagnosis. An online monitoring report for the motor bearing is generated, making up for the lack of real-time performance and meeting the real-time, accurate, and reliable health management needs of new energy vehicle motor bearings. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 The flowchart of the online dynamic monitoring and fault location method for bearings of new energy vehicle motors based on multi-sensor fusion is shown in the present invention.

[0052] Figure 2 This is a structural diagram of the online dynamic monitoring and fault location device for new energy vehicle motor bearings based on multi-sensor fusion according to the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] In one embodiment, such as Figure 1 As shown, a method for online dynamic monitoring and fault location of motor bearings in new energy vehicles based on multi-sensor fusion is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, or to a system including both a terminal and a server, and can be implemented through interaction between the terminal and the server. In the implementation environment, the terminal device can be a portable multi-parameter intelligent diagnostic instrument integrating vibration, current, temperature, and acoustic sensors. It has a built-in data acquisition card and preprocessing unit, responsible for real-time acquisition of multi-source detection signals of the motor bearing and completion of timing alignment. The server side is a cloud-based intelligent analysis platform used to deploy statistical analysis methods and machine learning models. Application scenarios include: when new energy vehicles require online dynamic monitoring and fault early warning of motor bearings under complex operating conditions, the terminal device continuously collects vibration, current, temperature, and acoustic signals through a sensor array, and uploads them to the server after signal synchronization and compression. The server receives the synchronized multi-source signal set, acquires multi-dimensional operating condition characteristics, and performs bearing condition assessment by combining principal component analysis and Fisher discriminant analysis. It then locates potential faults using convolutional neural networks and attention mechanisms, and predicts the remaining lifespan based on a long short-term memory model to simulate degradation curves. This generates a visualized report integrating status, fault, and lifespan information, which is returned to the terminal for display or directly pushed to the mobile terminal of maintenance personnel via an API interface, forming a closed-loop interaction of data collection, cloud analysis, and decision feedback. In this embodiment, the method includes the following steps:

[0055] S01: Collect multi-source detection signals of the motor bearing and perform time-series alignment on the multi-source detection signals to obtain a synchronous multi-source signal set.

[0056] The multi-source detection signals are various physical quantity signals collected from the motor bearing through a sensor array, including vibration signals, current signals, temperature signals, and acoustic signals. These signals can be collected based on the principles of piezoelectric, Hall effect, or thermocouple sensors to comprehensively perceive the bearing's operating status. Time alignment refers to eliminating time offsets between multi-source signals through synchronization mechanisms, such as hardware triggering, clock synchronization, or software interpolation algorithms (e.g., linear interpolation or dynamic time warping) to align the signals on a unified time axis, ensuring data consistency. The "synchronized multi-source signal set" is the aligned multi-dimensional signal set, which can be stored through data compression or caching techniques for real-time processing. In implementation, portable devices or vehicle systems with integrated multi-channel acquisition cards can be used to synchronously acquire signals at preset fixed or dynamic sampling rates, and alignment can be performed through timestamp matching or reference signals (e.g., speed pulses) to adapt to complex automotive-grade operating conditions.

[0057] S02, based on a synchronous multi-source signal set, extracts operating condition features.

[0058] Among them, the operating condition features are multi-dimensional parameters reflecting the health status of the bearing extracted from the signal, including time-domain features (such as amplitude and variance), frequency-domain features (such as spectral components and modulation features), and cross-signal coupling features (such as vibration-current correlation). The extraction process can be implemented through signal processing algorithms, such as applying short-time Fourier transform to extract frequency-domain modulation features from vibration signals, analyzing electrical parameter distortion to obtain harmonic components from current signals, calculating temperature gradient parameters from temperature signals, and performing envelope analysis on acoustic signals to capture abnormal frequency bands. Multi-source feature sets are integrated through feature weighted fusion or principal component analysis to form an operating condition feature vector. By extracting features from multi-source signals, the comprehensiveness and representativeness of features are achieved, adapting to the real-time monitoring needs under complex automotive-grade operating conditions.

[0059] S03. Based on the operating conditions, statistical methods are used to analyze the bearing condition and obtain the bearing condition analysis results.

[0060] The statistical methods refer to data analysis techniques based on multivariate statistical theory, including Principal Component Analysis (PCA) for feature dimensionality reduction (eliminating redundancy by using a cumulative contribution rate of no less than 90%), Fisher discriminant analysis (FDA) for state classification, and partial least squares (PLS) or multivariate control chart methods. These methods achieve state pattern recognition by calculating the statistical distribution of feature vectors (such as the covariance matrix and discriminant function). The bearing state analysis results are the bearing health status assessment conclusions output through statistical analysis, such as normal, slightly degraded, or faulty states. These can be determined based on a preset dimensionality reduction feature-bearing state sample database using distance metrics (such as Mahalanobis distance) or probabilistic models. In implementation, multi-source features can be weighted and fused according to motor speed and load parameters. PCA is used for dimensionality reduction to extract principal components, and FDA or other classifiers (such as K-nearest neighbors) are used for state discrimination to output the analysis results.

[0061] S04. Based on the operating condition characteristics and bearing condition analysis results, a machine learning model is used to locate potential bearing faults, and the bearing potential fault location results are obtained.

[0062] The machine learning model refers to data-driven intelligent classification and localization technologies, including Convolutional Neural Networks (CNNs) for extracting local spatial correlation features of multi-dimensional features, attention mechanisms for focusing on key fault features, and models such as Support Vector Machines (SVMs) or Random Forests. These models achieve pattern recognition by training on historical fault data. The bearing potential fault localization result is the fault type and location information output by the model analysis, such as suspected fault categories like bearing wear and rotor eccentricity, and their specific locations within the bearing assembly. In implementation, operating condition features are input into a pre-trained CNN model to extract spatial feature sets. Core features are filtered through the attention mechanism, and matched with a fault feature-fault type database to obtain suspected fault categories. Based on the motor bearing material and structural parameters (such as ball bearing size, seal type, etc., corresponding to different fault characteristics), the potential fault location is determined, and the potential fault localization result is integrated and output. Through various machine learning architectures (such as deep learning or traditional classifiers) and feature optimization strategies (such as adaptive thresholding or dynamic weighting), a fault identification foundation is provided for subsequent remaining life prediction and maintenance decisions.

[0063] S05. Based on the operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the remaining life of the motor bearing is predicted according to the physical degradation mechanism of the motor bearing, and the remaining life prediction result is obtained.

[0064] Among them, the physical degradation mechanism of motor bearings refers to the performance degradation law caused by physical processes such as wear, fatigue, and corrosion during operation. The degradation model parameters can be determined by experimental data or industry standards (such as ISO 281). The remaining life prediction is to estimate the failure time point by simulating the bearing degradation process. This can be achieved by relying on the fusion method of physical model and data-driven approach. For example, the degradation failure threshold can be determined by combining the fault type and location correction coefficient with the comprehensive correction coefficient of the working condition characteristics. The bearing degradation time series curve can be generated by the time series prediction model (such as long short-term memory network or other recurrent neural network). The failure time point can be located and the remaining life can be calculated based on integral operation. The remaining life of motor bearings can be predicted by using a variety of prediction architectures (such as physical model driven, deep learning or hybrid methods) and threshold adjustment strategies (such as adaptive weighting or multi-source fusion).

[0065] S06 generates an online monitoring report for motor bearings based on bearing condition analysis results, potential fault location results, and remaining life prediction results.

[0066] The motor bearing online monitoring report is a comprehensive document integrating bearing condition analysis results, potential fault location results, and remaining life prediction results. It visually displays bearing condition, fault information, and life prediction. In implementation, data visualization processing (such as generating coupled visualizations of bearing condition, fault location, and remaining life, including trend charts, heatmaps, or dashboards) can be used. Maintenance and protection suggestions (such as replacement cycles or maintenance measures) can be extracted from a pre-set motor bearing manual or real-time database. The report generation module integrates the visualization results and suggestions to output standardized or customized online monitoring reports, providing a comprehensive output basis for subsequent maintenance decisions and life management.

[0067] In one embodiment, the synchronous multi-source signal set includes vibration signals, current signals, temperature signals, and acoustic signals;

[0068] S11, based on a synchronous multi-source signal set, extracts operating condition features, including:

[0069] S12, based on the vibration signal, extract the time-domain composite features and frequency-domain modulation features to obtain the vibration features;

[0070] S13, Based on the current signal, extract the electrical parameter distortion features and vibration-current coupling features to obtain the current features. Among them, the vibration-current coupling features are calculated based on the electrical parameter distortion features and vibration features.

[0071] S14, Based on the temperature signal, extract the temperature change characteristic parameters to obtain the temperature characteristics;

[0072] S15, based on the acoustic signal, extract acoustic amplitude features and acoustic frequency domain modulation features to obtain acoustic features;

[0073] S16 integrates vibration characteristics, current characteristics, temperature characteristics, and acoustic characteristics to obtain the operating condition characteristics.

[0074] For example, the synchronous multi-source signal set includes vibration signals, current signals, temperature signals, and acoustic signals. Operating condition feature extraction based on this set can be achieved in the following ways: For vibration signals, time-domain analysis methods are used to calculate time-domain composite features such as peak value, RMS value, and kurtosis. Simultaneously, short-time Fourier transform is used to perform frequency-domain decomposition of the signal, extracting sideband components, modulation frequencies, and other frequency-domain modulation features, which are then integrated to form vibration features. For current signals, electrical parameter distortion features are obtained by analyzing the harmonic content and phase shift of the three-phase current. By calculating the cross-correlation coefficient between this electrical parameter distortion feature and the aforementioned vibration features, vibration-current coupling features reflecting the correlation between the two are obtained, and the two together constitute current features. For temperature signals, temperature data within a preset time window is continuously collected, and temperature change characteristic parameters such as the average temperature, temperature rise rate, and temperature gradient are calculated to form temperature features. For acoustic signals, acoustic amplitude features such as peak value and root mean square are calculated from signals collected by acoustic sensors. Wavelet packet decomposition technology is used to extract the energy distribution of a specific frequency band as acoustic frequency-domain modulation features, forming acoustic features. According to the preset feature dimension order, vibration features, current features, temperature features and acoustic features are connected in series and integrated to form a unified dimension of working condition feature vector, providing comprehensive data support for subsequent bearing condition analysis and fault location.

[0075] In one embodiment, a statistical method is used to perform bearing condition analysis, and the bearing condition analysis results are obtained, including:

[0076] S21, based on a synchronous multi-source signal set, obtains motor speed and load parameters;

[0077] S22, based on motor speed and load parameters, according to preset weighting rules, the vibration characteristics, current characteristics, temperature characteristics and acoustic characteristics of the operating condition are weighted and fused to obtain a fused feature set;

[0078] S23. Using principal component analysis, with a cumulative contribution rate of not less than 90% as the retention criterion, the fusion feature set is dimensionality reduced to obtain a dimensionality-reduced feature set.

[0079] S24. Based on the dimensionality reduction feature set and combined with the preset dimensionality reduction feature-bearing state sample database, the Fisher discriminant analysis method is used to perform bearing state analysis and obtain the bearing state analysis results.

[0080] Specifically, motor speed and load parameters can be extracted from a set of synchronous multi-source signals. The speed is calculated based on the fundamental frequency of the current signal combined with the number of motor pole pairs, while the load parameters are obtained by converting the output torque signal fed back from the motor controller. Feature weighting and fusion are performed according to a preset weighting rule, which is dynamically adjusted based on real-time speed and load. For example, under high-speed, high-load conditions, the weights for vibration features are set to 0.4, current features 0.3, temperature features 0.2, and acoustic features 0.1; under low-speed, low-load conditions, the weights are adjusted to 0.3 for vibration features, 0.2 for current features, 0.3 for temperature features, and 0.2 for acoustic features. The fused feature set is obtained by weighted summation. Principal component analysis can be used to construct a covariance matrix for the fused feature set and solve for the eigenvalues ​​and eigenvectors. After sorting the eigenvalues ​​in descending order, the filtering stops when the cumulative contribution rate reaches 90%. The corresponding eigenvectors are selected to form a dimensionality reduction matrix. The fused feature set is projected onto this matrix to obtain a dimensionality-reduced feature set. Based on a pre-defined dimensionality reduction feature-bearing condition sample database (containing feature samples of various states such as normal, slight degradation, and failure), Fisher discriminant analysis is used to calculate the discriminant function values ​​between the dimensionality reduction feature set and various samples in the database. According to the category corresponding to the maximum discriminant function value, the bearing condition analysis results are output, characterizing the bearing's current state as normal, slightly degraded, or a precursor to failure.

[0081] In one embodiment, based on operating condition characteristics and bearing condition analysis results, a machine learning model is used to locate potential bearing faults, obtaining the bearing potential fault location results, including:

[0082] S31, For the bearing condition in the bearing condition analysis results, the working condition features are input into the convolutional neural network pre-trained based on different bearing conditions to extract the local spatial correlation features of multi-dimensional features and obtain the spatial feature set;

[0083] S32 uses an attention mechanism to filter features from the spatial feature set to obtain core fault features;

[0084] S33, based on core fault characteristics, performs fault type matching according to a preset fault characteristic-fault type database to obtain suspected fault categories;

[0085] S34, obtain the material and structural parameters of the motor bearing, and based on the motor bearing structural parameters, determine the location of the potential fault according to the suspected fault category;

[0086] S35 integrates suspected fault categories and potential fault locations to obtain bearing potential fault location results.

[0087] For example, the current bearing status (e.g., normal, slight degradation) can be determined based on the bearing status analysis results. The operating condition features are then input into a pre-trained convolutional neural network (containing 3 convolutional layers and 2 pooling layers, with a 3×3 kernel size and ReLU activation to enhance nonlinear expression) under the corresponding status. A sliding window is used to calculate the local spatial correlation of multi-dimensional features, outputting a spatial feature set with a dimension of 128×64. A scaling dot product attention mechanism can be used to filter the spatial feature set, calculating the attention weights of each feature vector and retaining the top 30% of features by weight as core fault features, highlighting key fault-related information. A pre-defined fault feature-fault type database is then invoked, storing feature templates corresponding to 12 common faults such as bearing wear and rotor eccentricity. The cosine similarity between the core fault features and the templates is calculated, and the entry with the highest similarity (threshold ≥ 0.85) is selected as the suspected fault category. Obtain the structural parameters of the motor bearing material, such as ball diameter, raceway curvature radius, and cage thickness. Combine these parameters with pre-defined association rules based on suspected fault categories (e.g., wear faults correspond to raceways or balls, eccentric faults correspond to the rotor-stator mating area) to determine the location of potential faults. Integrate the suspected fault categories (e.g., ball wear) with the location (e.g., bearing inner ring raceway) according to a pre-defined format to generate a bearing potential fault location result that includes fault type, location, and confidence level.

[0088] In one embodiment, based on operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the remaining life of the motor bearing is predicted according to the physical degradation mechanism of the motor bearing, resulting in a remaining life prediction result, including:

[0089] S41. Based on the operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, determine the bearing degradation failure threshold according to the physical degradation mechanism of motor bearings.

[0090] S42, input the operating condition characteristics, the suspected fault category and the location of the potential fault of the bearing into the pre-trained long short-term memory model, perform bearing degradation time series simulation, and generate bearing degradation time series prediction curve;

[0091] S43. Based on the bearing degradation failure threshold and the bearing degradation time sequence prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point.

[0092] Specifically, the physical degradation mechanism of motor bearings (such as wear fatigue accumulation and material corrosion attenuation law, which can be referenced in ISO281 bearing life standard) can be combined with key parameters such as load intensity and temperature change rate in the working condition characteristics, bearing condition analysis results (such as slight degradation or failure precursor state) and potential fault location results (such as fault type and location), and the bearing degradation failure threshold can be determined by a quantitative degradation critical value algorithm. For example, for ball wear type faults, the vibration amplitude critical value can be set to 3 times the average value under normal conditions, and for raceway corrosion faults, the temperature stability threshold can be set to 85℃. The operating condition feature vector, suspected fault category codes (e.g., wear is set to 1, eccentricity to 2), and potential fault location codes (e.g., inner ring is set to A, outer ring to B) are input into a pre-trained long short-term memory model (this model is trained on 100,000 sets of degradation time-series data under different fault types, locations, and operating conditions, containing 3 hidden layers, and using ReLU activation function). The model simulates the degradation process of the bearing from its current state to failure, outputting degradation indicators (e.g., vibration amplitude, temperature rise) at different time points, and generating a bearing degradation time-series prediction curve. Linear interpolation is used to locate the degradation failure time point. When the degradation indicator in the degradation time-series prediction curve first reaches the preset bearing degradation failure threshold, this time point is recorded as the bearing degradation failure time point. The difference between this time point and the current monitoring time point can be used to calculate the predicted remaining life of the motor bearing.

[0093] In one embodiment, S51, based on the bearing degradation failure threshold and the bearing degradation time prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point, and is achieved through the following mathematical formula:

[0094]

[0095] Where L represents the predicted remaining life of the motor bearing. For the current monitoring time point, The bearing degradation failure time point and satisfy , This is the threshold for bearing degradation failure. This is a function for predicting bearing degradation time series curves. This is a fault category correction factor. This is a correction factor for the faulty part. This is a comprehensive correction factor for operating conditions. For dynamic error compensation term and , is the time series prediction fluctuation compensation coefficient, and k is the degradation rate decay factor.

[0096] For example, the fault category correction factor The system can retrieve coefficients from a preset coefficient library based on the suspected fault category (e.g., bearing wear is set to 1.2, rotor eccentricity to 1.5, and winding short circuit to 1.3), and the fault location correction coefficient can be applied. Based on the potential location of the failure (e.g., inner raceway set to 1.1, balls set to 1.3, cage set to 1.2), the comprehensive correction factor for the operating condition characteristics is determined. The time-series prediction fluctuation compensation coefficient is obtained by weighted summation of key parameters such as load intensity and temperature change rate in the operating condition characteristics (weights are allocated according to preset rules, and the sum is 1). The value is 0.05, and the degradation rate decay factor k is 0.02; both are fixed values ​​optimized through training with real vehicle data. Current monitoring time point. Directly read the system's real-time timestamp to determine the bearing degradation failure threshold. Given the critical value determined earlier, the bearing degradation time series prediction curve function This is a function representing the time-varying degradation index output by the Long Short-Term Memory (LSTM) model. The equation is solved using the bisection method. Determine the bearing degradation failure time point The integral term is calculated using Simpson's numerical integration method, and the integration interval is... The integrand expands as follows: and The ratio, multiplied by Superimposed dynamic error compensation term The integral result is the predicted remaining life L of the motor bearing.

[0097] In one embodiment, based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, an online monitoring report for the motor bearing is generated, including:

[0098] S61, based on the bearing condition analysis results, potential fault location results and remaining life prediction results, generates coupled visualization results of bearing condition-fault location-remaining life through data visualization processing;

[0099] S62, based on the bearing condition analysis results, potential fault location results and remaining life prediction results, extract maintenance and protection suggestions according to the preset motor bearing manual;

[0100] S63 integrates and couples visualization results and maintenance and protection suggestions to generate an online monitoring report for motor bearings.

[0101] Specifically, data visualization processing is performed on the bearing condition analysis results, potential fault location results, and remaining life prediction results. Using Matplotlib or ECharts visualization libraries, coupled visualization results of bearing condition, fault location, and remaining life are generated. Bearing condition is presented intuitively with a three-color dashboard (green represents normal, yellow represents slight degradation, and red represents early signs of failure). Fault location is indicated by marking the faulty parts (such as inner raceway and balls) using a 3D structural diagram of the bearing and overlaying fault type labels. Remaining life is displayed as a time-series curve showing the degradation trend and the current remaining time. All three are linked together via a time axis to create a synchronized visualization effect. Based on the above analysis results, the system's pre-set standardized motor bearing manual (containing a maintenance solution library corresponding to different conditions, fault types, and lifespans) is retrieved. Maintenance and protection recommendations are extracted according to matching rules. For example, under normal conditions, it is recommended to clean and lubricate the bearing every 3 months; for slight degradation and remaining lifespan exceeding 6 months, it is recommended to shorten the monitoring cycle to once a week; for a faulty condition and remaining lifespan less than 1 month, it is recommended to immediately stop the machine and replace the corresponding components. Specific recommendations are supplemented based on the fault location (e.g., for ball wear, the bearing fit tolerance needs to be checked). Following a fixed structure of "basic information - status analysis - fault details - life prediction - maintenance recommendations", the coupled visualization results are integrated with the text description of maintenance and protection recommendations in the form of embedded images to generate an online monitoring report for motor bearings. The report includes basic identifiers such as motor model, monitoring time, and report number, and is pushed to the mobile terminal or terminal device of maintenance personnel through a cloud platform to achieve intuitive and practical information transmission.

[0102] The aforementioned method for online dynamic monitoring and fault location of motor bearings in new energy vehicles based on multi-sensor fusion collects and aligns vibration, current, temperature, and acoustic signals from multiple sources in the motor bearing to construct a synchronous multi-source signal set, overcoming the limitations of single-parameter monitoring. Based on this signal set, multi-dimensional operating condition features in the time domain, frequency domain, and across signal coupling are extracted. After weighted fusion with motor speed and load parameters, principal component analysis for dimensionality reduction and Fisher discriminant analysis are used to objectively determine the bearing condition, avoiding the subjectivity of manual experience-based diagnosis. A pre-trained convolutional neural network is used to extract feature spatial correlations, and an attention mechanism is employed to filter core fault features and match them. By combining a fault type database with bearing structural parameters to locate fault locations, the sensitivity and accuracy of early fault detection are improved, and the false alarm rate is reduced. The failure threshold is determined based on the physical degradation mechanism of motor bearings, and the degradation time series curve is simulated through a long short-term memory model to accurately predict the remaining life, thus connecting the links of monitoring, diagnosis, and life assessment. The results of condition analysis, fault location, and life prediction are integrated to generate an online monitoring report with visualization results and maintenance suggestions, realizing real-time, accurate, and reliable health management of motor bearings. This effectively solves the problems of traditional methods, such as strong subjectivity, insensitivity to early faults, high false alarm rate, insufficient real-time performance, and disconnect between monitoring and diagnosis.

[0103] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0104] Based on the same inventive concept, this application also provides a device for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion, used to implement the above-mentioned method for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the device for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion provided below can be found in the limitations of the method for online dynamic monitoring and fault location of new energy vehicle motor bearings based on multi-sensor fusion described above, and will not be repeated here.

[0105] In one exemplary embodiment, such as Figure 2 As shown, a device for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion is provided, comprising:

[0106] The multi-source signal acquisition module 101 is used to acquire multi-source detection signals of the motor bearing and perform time-series alignment on the multi-source detection signals to obtain a synchronous multi-source signal set;

[0107] Operating condition feature extraction module 102 is used to extract operating condition features based on a synchronous multi-source signal set;

[0108] The bearing condition analysis module 103 is used to perform bearing condition analysis based on operating condition characteristics and statistical methods to obtain bearing condition analysis results.

[0109] The potential fault analysis module 104 is used to locate potential bearing faults based on operating condition characteristics and bearing condition analysis results, and to obtain the bearing potential fault location results.

[0110] The remaining life prediction module 105 is used to predict the remaining life of the motor bearing based on the operating condition characteristics, bearing condition analysis results and bearing potential fault location results, according to the physical degradation mechanism of the motor bearing, and obtain the remaining life prediction result.

[0111] The monitoring report generation module 106 is used to generate an online monitoring report for motor bearings based on the bearing condition analysis results, potential fault location results, and remaining life prediction results.

[0112] In one embodiment, the synchronous multi-source signal set in the working condition feature extraction module 102 includes vibration signals, current signals, temperature signals and acoustic signals;

[0113] Based on a synchronous multi-source signal set, operating condition features are extracted, including:

[0114] Based on the vibration signal, time-domain composite features and frequency-domain modulation features are extracted to obtain the vibration characteristics;

[0115] Based on the current signal, electrical parameter distortion features and vibration-current coupling features are extracted to obtain current features. Among them, vibration-current coupling features are calculated based on electrical parameter distortion features and vibration features.

[0116] Based on the temperature signal, temperature change characteristic parameters are extracted to obtain temperature characteristics;

[0117] Based on the acoustic signal, acoustic amplitude features and acoustic frequency domain modulation features are extracted to obtain acoustic features;

[0118] By integrating vibration characteristics, current characteristics, temperature characteristics, and acoustic characteristics, the operating condition characteristics are obtained.

[0119] In one embodiment, the bearing condition analysis module 103 is further configured to:

[0120] Based on a synchronous multi-source signal set, motor speed and load parameters are obtained;

[0121] Based on motor speed and load parameters, and according to preset weighting rules, the vibration characteristics, current characteristics, temperature characteristics and acoustic characteristics of the operating conditions are weighted and fused to obtain a fused feature set.

[0122] Principal component analysis was used to reduce the dimensionality of the fused feature set, with a cumulative contribution rate of no less than 90% as the retention criterion.

[0123] Based on the dimensionality reduction feature set and combined with the pre-set dimensionality reduction feature-bearing state sample database, the Fisher discriminant analysis method is used to perform bearing state analysis and obtain the bearing state analysis results.

[0124] In one embodiment, the potential fault analysis module 104 is further configured to:

[0125] For the bearing condition in the bearing condition analysis results, the working condition features are input into a convolutional neural network pre-trained based on different bearing conditions to extract local spatial correlation features of multi-dimensional features and obtain a spatial feature set.

[0126] The core fault features are obtained by filtering the spatial feature set through an attention mechanism.

[0127] Based on core fault characteristics, fault type matching is performed according to a preset fault characteristic-fault type database to obtain suspected fault categories.

[0128] Obtain the material and structural parameters of the motor bearing, and based on these parameters, determine the location of potential faults according to the suspected fault categories.

[0129] By integrating suspected fault categories and potential fault locations, the results of potential bearing fault location are obtained.

[0130] In one embodiment, the remaining lifetime prediction module 105 is further configured to:

[0131] Based on the operating condition characteristics, bearing condition analysis results, and bearing potential fault location results, the bearing degradation failure threshold is determined according to the physical degradation mechanism of motor bearings.

[0132] The operating condition characteristics, the suspected fault category and the location of potential faults in the bearing are input into the pre-trained long short-term memory model to simulate the bearing degradation time series and generate the bearing degradation time series prediction curve.

[0133] Based on the bearing degradation failure threshold and the bearing degradation time sequence prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point.

[0134] In one embodiment, the remaining life prediction module 105 is further configured to calculate the remaining life prediction result of the motor bearing by locating the degradation failure time point based on the bearing degradation failure threshold and the bearing degradation time sequence prediction curve using the following mathematical formula:

[0135]

[0136] Where L represents the predicted remaining life of the motor bearing. For the current monitoring time point, The bearing degradation failure time point and satisfy , This is the threshold for bearing degradation failure. This is a function for predicting bearing degradation time series curves. This is a fault category correction factor. This is a correction factor for the faulty part. This is a comprehensive correction factor for operating conditions. For dynamic error compensation term and , is the time series prediction fluctuation compensation coefficient, and k is the degradation rate decay factor.

[0137] In one embodiment, the monitoring report generation module 106 is further configured to:

[0138] Based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, coupled visualization results of bearing condition, fault location, and remaining life are generated through data visualization processing.

[0139] Based on the bearing condition analysis results, potential fault location results, and remaining life prediction results, maintenance and protection suggestions are extracted according to the pre-set motor bearing manual.

[0140] By integrating and coupling visualization results and maintenance and protection recommendations, an online monitoring report for motor bearings is generated.

[0141] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion.

[0142] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0143] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0144] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion, characterized in that, The method includes: Multi-source detection signals of the motor bearing are acquired, and the multi-source detection signals are time-aligned to obtain a synchronous multi-source signal set; Based on the synchronous multi-source signal set, operating condition features are extracted; Based on the aforementioned operating conditions, a statistical method is used to analyze the bearing condition, and the bearing condition analysis results are obtained. Based on the operating condition characteristics and the bearing condition analysis results, a machine learning model is used to locate potential bearing faults, and the bearing potential fault location results are obtained. Based on the operating condition characteristics, the bearing condition analysis results, and the bearing potential fault location results, the remaining life of the motor bearing is predicted according to the physical degradation mechanism of the motor bearing, and the remaining life prediction result is obtained. Based on the bearing condition analysis results, the potential fault location results, and the remaining life prediction results, an online monitoring report for the motor bearing is generated.

2. The method according to claim 1, characterized in that, The synchronous multi-source signal set includes vibration signals, current signals, temperature signals, and acoustic signals; The extraction of operating condition features based on the synchronous multi-source signal set includes: Based on the vibration signal, time-domain composite features and frequency-domain modulation features are extracted to obtain the vibration features; Based on the current signal, electrical parameter distortion features and vibration-current coupling features are extracted to obtain current features, wherein the vibration-current coupling features are calculated based on the electrical parameter distortion features and the vibration features; Based on the temperature signal, temperature change characteristic parameters are extracted to obtain temperature characteristics; Based on the acoustic signal, acoustic amplitude features and acoustic frequency domain modulation features are extracted to obtain acoustic features; The operating condition characteristics are obtained by integrating the vibration characteristics, the current characteristics, the temperature characteristics, and the acoustic characteristics.

3. The method according to claim 2, characterized in that, Based on the aforementioned operating condition characteristics, a statistical method is used to perform bearing condition analysis, yielding bearing condition analysis results, including: Based on the synchronous multi-source signal set, the motor speed and load parameters are obtained; Based on the motor speed and the load parameters, the vibration characteristics, current characteristics, temperature characteristics and acoustic characteristics of the operating condition are weighted and fused according to the preset weighting rules to obtain a fused feature set; Principal component analysis was used to reduce the dimensionality of the fused feature set, with a cumulative contribution rate of not less than 90% as the retention criterion, to obtain a dimensionality-reduced feature set. Based on the dimensionality reduction feature set, combined with the preset dimensionality reduction feature-bearing state sample database, the Fisher discriminant analysis method is used to perform bearing state analysis, and the bearing state analysis results are obtained.

4. The method according to claim 1, characterized in that, Based on the operating condition characteristics and the bearing condition analysis results, a machine learning model is used to locate potential bearing faults, resulting in bearing potential fault location results, including: For the bearing condition in the bearing condition analysis results, the working condition features are input into a convolutional neural network pre-trained based on different bearing conditions to extract the local spatial correlation features of multi-dimensional features and obtain a spatial feature set. The spatial feature set is filtered using an attention mechanism to obtain core fault features; Based on the core fault characteristics, fault type matching is performed according to the preset fault characteristic-fault type database to obtain the suspected fault category. Obtain the material and structural parameters of the motor bearing, and based on the motor bearing structural parameters, determine the location of the potential fault according to the suspected fault category; By integrating the suspected fault categories and the potential fault locations, the bearing potential fault location results are obtained.

5. The method according to claim 4, characterized in that, Based on the operating condition characteristics, the bearing condition analysis results, and the bearing potential fault location results, and according to the physical degradation mechanism of the motor bearing, the remaining life of the motor bearing is predicted, resulting in the remaining life prediction result, including: Based on the operating condition characteristics, the bearing condition analysis results, and the bearing potential fault location results, the bearing degradation failure threshold is determined according to the physical degradation mechanism of the motor bearing. The operating condition characteristics, the suspected fault category and the location of the potential fault in the bearing are input into a pre-trained long short-term memory model to simulate the bearing degradation time series and generate a bearing degradation time series prediction curve. Based on the bearing degradation failure threshold and the bearing degradation time sequence prediction curve, the remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point.

6. The method according to claim 5, characterized in that, The remaining life prediction result of the motor bearing is calculated by locating the degradation failure time point based on the bearing degradation failure threshold and the bearing degradation time prediction curve. This calculation is achieved through the following mathematical formula: Where L represents the predicted remaining life of the motor bearing. For the current monitoring time point, The bearing degradation failure time point and satisfy , This is the threshold for bearing degradation failure. This is a function for predicting bearing degradation time series curves. This is a fault category correction factor. This is a correction factor for the faulty part. This is a comprehensive correction factor for operating conditions. For dynamic error compensation term and , is the time series prediction fluctuation compensation coefficient, and k is the degradation rate decay factor.

7. The method according to claim 1, characterized in that, Based on the bearing condition analysis results, the potential fault location results, and the remaining life prediction results, an online monitoring report for the motor bearing is generated, including: Based on the bearing condition analysis results, the potential fault location results, and the remaining life prediction results, coupled visualization results of bearing condition, fault location, and remaining life are generated through data visualization processing. Based on the bearing condition analysis results, the potential fault location results, and the remaining life prediction results, maintenance and protection suggestions are extracted according to the preset motor bearing manual. By integrating the coupled visualization results and the maintenance and protection recommendations, an online monitoring report for the motor bearing is generated.

8. A device for online dynamic monitoring and fault location of bearings in new energy vehicle motors based on multi-sensor fusion, characterized in that, The device includes: A multi-source signal acquisition module is used to acquire multi-source detection signals of the motor bearing and to perform time-series alignment on the multi-source detection signals to obtain a synchronous multi-source signal set; The operating condition feature extraction module is used to extract operating condition features based on the synchronous multi-source signal set; The bearing condition analysis module is used to perform bearing condition analysis using statistical methods based on the operating condition characteristics, and obtain the bearing condition analysis results. The potential fault location module is used to locate potential bearing faults based on the operating condition characteristics and the bearing condition analysis results, and to obtain the bearing potential fault location results. The remaining life prediction module is used to predict the remaining life of the motor bearing based on the operating condition characteristics, the bearing condition analysis results, and the bearing potential fault location results, according to the physical degradation mechanism of the motor bearing, and obtain the remaining life prediction result. The monitoring report generation module is used to generate an online monitoring report for motor bearings based on the bearing condition analysis results, the potential fault location results, and the remaining life prediction results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images