CN-122016312-A - Bearing fault detection method, device and train

CN122016312ACN 122016312 ACN122016312 ACN 122016312ACN-122016312-A

Abstract

The disclosure provides a bearing fault detection method, which relates to the technical field of rail transit and comprises the steps of acquiring original vibration time sequence data, original strain time sequence data and original displacement time sequence data of a bearing to be detected; the method comprises the steps of performing enhancement processing on fault time sequence data in original vibration data to obtain enhanced vibration time sequence data, determining filtering parameters based on reference noise data, filtering the enhanced vibration time sequence data according to the filtering parameters to obtain initial vibration time sequence data, performing multi-mode data fusion and feature extraction processing on the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data to obtain fusion time sequence features, inputting the fusion time sequence features into a bearing fault detection model, and outputting fault detection results aiming at a bearing to be detected. The disclosure also provides a bearing fault detection device and a train.

Inventors

CUI YULONG
XIN LIANG
YANG MUCHEN
ZHU HUILONG
YIN JILEI

Assignees

中车青岛四方机车车辆股份有限公司

Dates

Publication Date: 20260512
Application Date: 20260324

Claims (10)

1. A bearing failure detection method comprising: Acquiring original vibration time sequence data, original strain time sequence data and original displacement time sequence data of a bearing to be detected, wherein the original vibration time sequence data comprises fault time sequence data representing that the bearing to be detected has faults; Performing enhancement processing on the fault time sequence data in the original vibration time sequence data to obtain enhanced vibration time sequence data; Determining a filtering parameter based on reference noise data, and filtering the enhanced vibration time sequence data according to the filtering parameter to obtain initial vibration time sequence data, wherein the reference noise data comprises noise data collected at a preset position around the bearing to be detected; Performing multi-mode data fusion and feature extraction processing on the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data to obtain fusion time sequence features; And inputting the fusion time sequence characteristics into a bearing fault detection model, and outputting a fault detection result aiming at the bearing to be detected.
2. The method of claim 1, wherein the method further comprises: acquiring a reference part training sample set corresponding to a reference part, wherein the reference part comprises a rotating part except the bearing to be detected; Training an initial reference component fault detection model by using the reference component training sample set to obtain a target reference component fault detection model; and updating the weight of the target reference component fault detection model to obtain the bearing fault detection model.
3. The method of claim 2, wherein the weight updating the target reference component fault detection model to obtain the bearing fault detection model comprises: Acquiring a bearing training sample set, wherein the bearing training sample set comprises fault marking data; inputting the bearing training sample set into the target reference component fault detection model, and outputting a prediction result of the bearing training sample set; And performing iterative updating operation on the initial network weight of a preset high-level network in the target reference component fault detection model based on the error between the fault labeling data and the prediction result, and determining the updating weight of the preset high-level network to obtain the bearing fault detection model.
4. The method of claim 1, the inputting the fused timing features into a bearing failure detection model, outputting a failure detection result for the bearing to be detected comprising: Dividing the fusion time sequence characteristics into a plurality of groups of fusion characteristics according to a plurality of time periods; Inputting the multiple groups of fusion features into the bearing fault detection model, and outputting multiple fault states; and determining the fault detection result according to the plurality of time periods corresponding to the plurality of groups of fusion features and the plurality of fault states.
5. The method of claim 1, wherein the method further comprises: and determining the value of the preset parameter based on the vibration frequency range of the original vibration time sequence data.
6. The method of claim 1, wherein the performing multi-modal data fusion and feature extraction processing on the original strain time series data, the original displacement time series data, and the initial vibration time series data to obtain a fused time series feature comprises: preprocessing the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data to obtain target strain time sequence data, target displacement time sequence data and target vibration time sequence data; and utilizing a target countermeasure network to perform multi-mode data fusion and feature extraction on the target strain time sequence data, the target displacement time sequence data and the target vibration time sequence data to obtain the fusion time sequence feature.
7. The method of claim 6, the method further comprising: acquiring a multi-mode data training sample set; inputting the multimodal data training sample set into an initial countermeasure network; performing iterative updating operation on the initial weights of the generator and the discriminator in the initial countermeasure network to obtain target weights of the generator and the discriminator; Wherein performing a round of prediction operations includes: Determining respective loss functions of the generator and the arbiter; And updating the initial weight according to the loss function.
8. The method of claim 6, wherein the preprocessing the raw strain timing data, the raw displacement timing data, and the initial vibration timing data comprises: Performing time calibration processing on the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data; And normalizing the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data after time calibration processing to obtain a target strain signal, a target displacement signal and a target vibration signal.
9. A bearing failure detection apparatus comprising: The acquisition module is used for acquiring original vibration time sequence data, original strain time sequence data and original displacement time sequence data of the bearing to be detected, wherein the original vibration time sequence data comprises fault time sequence data representing that the bearing to be detected has faults; The enhancement module is used for enhancing the fault time sequence data in the original vibration time sequence data to obtain enhanced vibration time sequence data; the filtering module is used for determining filtering parameters based on reference noise data, filtering the enhanced vibration time sequence data according to the filtering parameters to obtain initial vibration time sequence data, wherein the reference noise data comprises noise data collected at preset positions around the bearing to be detected; the data fusion and feature extraction module is used for carrying out multi-mode data fusion and feature extraction processing on the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data to obtain fusion time sequence features; And the input module is used for inputting the fusion time sequence characteristics into a bearing fault detection model and outputting a fault detection result aiming at the bearing to be detected.
10. A train, comprising: A bearing to be detected; The sensor is used for acquiring original vibration time sequence data, original strain time sequence data and original displacement time sequence data of the bearing to be detected; fault detection device for performing the steps of the method according to any of claims 1-8.

Description

Bearing fault detection method, device and train Technical Field The disclosure relates to the technical field of rail transit, and more particularly, to a bearing fault detection method, a bearing fault detection device and a train. Background The bearing is used as a key component of the train, and the running state of the bearing directly influences the safety and reliability of the train. The related bearing fault diagnosis method is usually manual inspection and manual periodic maintenance, but the manual detection method is difficult to accurately capture weak fault signals of the bearing in early fault stages in real time, and problems can be found only when the faults develop to a relatively serious degree, so that the maintenance is not timely, and the train shutdown risk and the maintenance cost are increased. In addition, the manual detection workload is large, and the detection efficiency is low. Disclosure of Invention In view of the above, the disclosure provides a method, a device and a train for detecting bearing faults. An aspect of the disclosure provides a bearing fault detection method, which includes obtaining original vibration time sequence data, original strain time sequence data and original displacement time sequence data of a bearing to be detected, wherein the original vibration time sequence data comprises fault time sequence data representing that the bearing to be detected has failed, performing enhancement processing on the fault time sequence data in the original time sequence vibration data to obtain enhanced vibration time sequence data, determining filtering parameters based on reference noise data, filtering the enhanced vibration time sequence data according to the filtering parameters to obtain initial vibration time sequence data, wherein the reference noise data comprises noise data collected at preset positions around the bearing to be detected, performing multi-mode data fusion and feature extraction processing on the original strain time sequence data, the original displacement time sequence data and the initial vibration time sequence data to obtain fusion time sequence features, inputting the fusion time sequence features into a bearing fault detection model, and outputting fault detection results aiming at the bearing to be detected. According to the embodiment of the disclosure, the bearing fault detection method further comprises the steps of obtaining a reference part training sample set corresponding to a reference part, wherein the reference part comprises a rotating part except a bearing to be detected, training an initial reference part fault detection model by using the reference part training sample set to obtain a target reference part fault detection model, and carrying out weight updating on the target reference part fault detection model to obtain the bearing fault detection model. According to the embodiment of the disclosure, the method for updating the weight of the target reference component fault detection model to obtain the bearing fault detection model comprises the steps of obtaining a bearing training sample set, inputting the bearing training sample set into the target reference component fault detection model, outputting a prediction result of the bearing training sample set, executing iterative updating operation on initial network weights of a preset high-level network in the target reference component fault detection model based on errors between the fault marking data and the prediction result, and determining updating weights of the preset high-level network to obtain the bearing fault detection model. According to the embodiment of the disclosure, inputting the fusion time sequence characteristics into a bearing fault detection model and outputting a fault detection result aiming at a bearing to be detected comprises dividing the fusion time sequence characteristics into a plurality of groups of fusion characteristics according to a plurality of time periods, inputting the plurality of groups of fusion characteristics into the bearing fault detection model and outputting a plurality of fault states, and determining the fault detection result according to the plurality of time periods corresponding to the plurality of groups of fusion characteristics and the plurality of fault states. According to the embodiment of the disclosure, the fault bearing detection method further comprises the step of determining the value of the preset parameter based on the vibration frequency range of the original vibration time sequence data. According to the embodiment of the disclosure, multimode data fusion and feature extraction processing are performed on original strain time sequence data, original displacement time sequence data and initial vibration time sequence data to obtain fusion time sequence features, wherein preprocessing is performed on the original strain time sequence data, the original displacement time sequence data and the initia