CN-121994487-A - Bearing state evaluation method, system and medium based on vibration signals

CN121994487ACN 121994487 ACN121994487 ACN 121994487ACN-121994487-A

Abstract

The application provides a bearing state evaluation method, a system and a medium based on vibration signals, wherein the method comprises the steps of collecting original vibration signals of a bearing under different operation conditions through a vibration sensor, and preprocessing the original vibration signals; the method comprises the steps of respectively extracting time domain features, frequency domain features and wavelet domain features, adopting a combined feature screening algorithm to perform feature screening, removing redundant features and invalid features to obtain an optimal feature set, performing feature fusion processing on the optimal feature set to generate fusion feature vectors, constructing a bearing state evaluation model based on deep learning, inputting the fusion feature vectors into the bearing state evaluation model, outputting a bearing state evaluation result, synchronously extracting the multi-dimensional features of the time domain, the frequency domain and the wavelet domain, comprehensively capturing the feature information of different types of bearing faults, adopting the combined feature screening algorithm to remove the redundant and invalid features, reducing the calculation load of a subsequent model, and improving the evaluation efficiency.

Inventors

WANG YONG
Ye Linglin
ZHAO MING

Assignees

瑞湖智科数据(苏州)有限公司

Dates

Publication Date: 20260508
Application Date: 20260127

Claims (10)

1. A method of evaluating a state of a bearing based on a vibration signal, comprising: Collecting original vibration signals of the bearing under different operation conditions through the vibration sensor, and preprocessing the original vibration signals; respectively extracting time domain features, frequency domain features and wavelet domain features based on the preprocessed original vibration information; screening the time domain features, the frequency domain features and the wavelet domain features by adopting a combined feature screening algorithm, and removing redundant features and invalid features to obtain an optimal feature set; Performing feature fusion processing on the optimal feature set to generate a fusion feature vector; Building a bearing state evaluation model based on deep learning, inputting the fusion feature vector into the bearing state evaluation model, and outputting a bearing state evaluation result; And transmitting the bearing state evaluation result to the terminal in real time.
2. The method for evaluating the state of a bearing based on vibration signals according to claim 1, wherein the preprocessing of the original vibration signals specifically comprises: Acquiring an original vibration signal, analyzing the original vibration signal based on abnormal mutation condition information, removing a mutation abnormal signal section caused by sensor abnormality and electromagnetic interference in the original vibration signal, and reserving an effective signal section; denoising the effective signal segment by adopting a self-adaptive filtering algorithm, filtering environmental noise and equipment operation interference noise, and obtaining a denoised signal; carrying out amplitude standardization processing on the denoised signal, mapping the signal amplitude to a preset interval, eliminating the influence of the signal amplitude difference under different working conditions, and obtaining a standardized signal; calculating the signal-to-noise ratio of the standardized signal, and comparing the signal-to-noise ratio with a preset signal-to-noise ratio threshold; And if the signal-to-noise ratio is lower than the preset threshold, the original vibration signal under the corresponding working condition is collected again until an effective signal is obtained.
3. The vibration signal-based bearing state assessment method according to claim 2, wherein the time domain features include peak value, peak value factor, kurtosis, skewness, variance, standard deviation, mean value, pulse factor, margin factor and waveform factor; the frequency domain features include center of gravity frequency, mean square frequency, frequency variance, frequency standard deviation, spectral peak energy and spectral peak frequency.
4. The method for evaluating the bearing state based on the vibration signal according to claim 3, wherein the method for evaluating the bearing state based on the vibration signal is characterized by adopting a combined feature screening algorithm to screen time domain features, frequency domain features and wavelet domain features, eliminating redundant features and invalid features, and obtaining an optimal feature set, and specifically comprises the following steps: The combined characteristic screening algorithm is a combination of ReliefF algorithm and L1 regularization algorithm; calculating corresponding weight values of the time domain features, the frequency domain features and the wavelet domain features by adopting ReliefF algorithm, and eliminating features with weight values smaller than a preset weight threshold; And (3) carrying out secondary screening on the characteristics with the weight value larger than or equal to a preset weight threshold by adopting an L1 regularization algorithm, and eliminating the characteristics with the coefficient of 0 to obtain an optimal characteristic set.
5. The vibration signal-based bearing state evaluation method according to claim 4, wherein the feature fusion process uses a principal component analysis PCA algorithm to map an optimal feature set to a low-dimensional space, generating a fused feature vector having dimensions of 10-20 dimensions.
6. The vibration signal-based bearing state evaluation method according to claim 5, wherein the deep learning-based bearing state evaluation model is a CNN-LSTM hybrid model, and the CNN-LSTM hybrid model includes a convolution layer, a pooling layer, an LSTM layer, and a full connection layer connected in sequence; The convolution layer is used for extracting local features in the fusion feature vector, the pooling layer is used for carrying out dimension reduction processing on the features output by the convolution layer, the LSTM layer is used for capturing time sequence dependency relations in the feature sequence, and the full-connection layer is used for outputting classification results of bearing states; the loss function of the bearing state evaluation model adopts a cross entropy loss function, and the optimizer adopts an Adam optimizer.
7. The bearing state evaluation system based on the vibration signal is characterized by comprising a memory and a processor, wherein the memory comprises a program of a bearing state evaluation method based on the vibration signal, and the program of the bearing state evaluation method based on the vibration signal realizes the following steps when being executed by the processor: Collecting original vibration signals of the bearing under different operation conditions through the vibration sensor, and preprocessing the original vibration signals; respectively extracting time domain features, frequency domain features and wavelet domain features based on the preprocessed original vibration information; screening the time domain features, the frequency domain features and the wavelet domain features by adopting a combined feature screening algorithm, and removing redundant features and invalid features to obtain an optimal feature set; Performing feature fusion processing on the optimal feature set to generate a fusion feature vector; Building a bearing state evaluation model based on deep learning, inputting the fusion feature vector into the bearing state evaluation model, and outputting a bearing state evaluation result; And transmitting the bearing state evaluation result to the terminal in real time.
8. The vibration signal-based bearing condition assessment system according to claim 7, wherein the preprocessing of the raw vibration signal comprises: Acquiring an original vibration signal, analyzing the original vibration signal based on abnormal mutation condition information, removing a mutation abnormal signal section caused by sensor abnormality and electromagnetic interference in the original vibration signal, and reserving an effective signal section; denoising the effective signal segment by adopting a self-adaptive filtering algorithm, filtering environmental noise and equipment operation interference noise, and obtaining a denoised signal; carrying out amplitude standardization processing on the denoised signal, mapping the signal amplitude to a preset interval, eliminating the influence of the signal amplitude difference under different working conditions, and obtaining a standardized signal; calculating the signal-to-noise ratio of the standardized signal, and comparing the signal-to-noise ratio with a preset signal-to-noise ratio threshold; And if the signal-to-noise ratio is lower than the preset threshold, the original vibration signal under the corresponding working condition is collected again until an effective signal is obtained.
9. The vibration signal based bearing condition assessment system according to claim 8, wherein the time domain features include peak, peak factor, kurtosis, skewness, variance, standard deviation, mean, pulse factor, margin factor, and waveform factor; the frequency domain features include center of gravity frequency, mean square frequency, frequency variance, frequency standard deviation, spectral peak energy and spectral peak frequency.
10. A computer-readable storage medium, characterized in that it comprises a vibration signal-based bearing state evaluation method program, which, when executed by a processor, implements the steps of a vibration signal-based bearing state evaluation method according to any one of claims 1 to 6.

Description

Bearing state evaluation method, system and medium based on vibration signals Technical Field The application relates to the field of bearing state evaluation, in particular to a bearing state evaluation method, system and medium based on vibration signals. Background The bearing is used as a core transmission component of the rotary machine, is widely applied to key fields of industrial manufacture, rail transit, aerospace, wind power generation and the like, and the running state of the bearing directly determines the reliability, safety and service life of equipment. According to the statistics of industrial equipment faults, more than 40% of rotating machinery faults are caused by bearing failure, the bearing failure not only can cause equipment to stop and stop production, but also can cause huge economic loss, and in extreme cases, safety accidents can be caused, so that the life safety of personnel is threatened. Therefore, the method realizes accurate and real-time evaluation of the running state of the bearing, early warns potential faults in advance and judges the fault degree, and has important practical significance for guaranteeing stable running of equipment, reducing maintenance cost and improving industrial production safety. The existing bearing state evaluation method has the following defects: Firstly, the effect of preprocessing the vibration signal is poor. The industrial field environment is complex, the collected original vibration signal is easily influenced by factors such as equipment operation noise, electromagnetic interference, environmental vibration and the like, and the signal-to-noise ratio is low; Secondly, feature extraction is not comprehensive and effective enough. The existing method only extracts the characteristics of single dimension of the time domain or the frequency domain, is difficult to comprehensively characterize the complex running state of the bearing, and meanwhile, the extracted characteristics are not effectively screened, and the redundant characteristics and the invalid characteristics can increase the calculation load of a subsequent model and reduce the evaluation efficiency; thirdly, the state evaluation model has poor adaptability and generalization capability. The existing evaluation model is constructed based on a single algorithm, so that the local characterization capability and the time sequence dependency capture capability of the features are difficult to be considered. Disclosure of Invention The embodiment of the application aims to provide a bearing state evaluation method, a system and a medium based on vibration signals, which can comprehensively capture the characteristic information of different types of bearing faults by synchronously extracting multidimensional characteristics of time domain, frequency domain and wavelet domain, eliminate redundant and invalid characteristics by adopting a combined characteristic screening algorithm, generate a low-dimensional and efficient fusion characteristic vector by combining a characteristic fusion technology, and reduce the calculation load of a follow-up model and improve the evaluation efficiency while improving the characteristic distinction. The embodiment of the application also provides a bearing state evaluation method based on the vibration signal, which comprises the following steps: Collecting original vibration signals of the bearing under different operation conditions through the vibration sensor, and preprocessing the original vibration signals; respectively extracting time domain features, frequency domain features and wavelet domain features based on the preprocessed original vibration information; screening the time domain features, the frequency domain features and the wavelet domain features by adopting a combined feature screening algorithm, and removing redundant features and invalid features to obtain an optimal feature set; Performing feature fusion processing on the optimal feature set to generate a fusion feature vector; Building a bearing state evaluation model based on deep learning, inputting the fusion feature vector into the bearing state evaluation model, and outputting a bearing state evaluation result; And transmitting the bearing state evaluation result to the terminal in real time. Optionally, in the method for evaluating a bearing state based on a vibration signal according to the embodiment of the present application, preprocessing an original vibration signal specifically includes: Acquiring an original vibration signal, analyzing the original vibration signal based on abnormal mutation condition information, removing a mutation abnormal signal section caused by sensor abnormality and electromagnetic interference in the original vibration signal, and reserving an effective signal section; denoising the effective signal segment by adopting a self-adaptive filtering algorithm, filtering environmental noise and equipment operation interference noise, and ob