Font Size: a A A

Roller Bearing Performance Degradation Assessment Based On Anomaly Detection Algorithm

Posted on:2019-10-21Degree:MasterType:Thesis
Country:ChinaCandidate:H J GuoFull Text:PDF
GTID:2382330566959351Subject:Instrument Science and Technology
Abstract/Summary:PDF Full Text Request
The health of rotating machine during industrial processes is the key to ensuring its reliability.Among them,bearings are commonly used components in rotating machinery and are the main source of breakdown failure.Bearings usually run from normal condition to failure.If the real-time bearing operating conditions can be monitored,the occurrence of bearing failure can be prevented,so as to avoid unnecessary economic losses.In this paper,the wavelet packet decomposition and autoregressive model(AR)theory are used to extract features of signal.And the anomaly detection algorithm is exploited to evaluate the performance of the rolling bearings,the specific work is as follows:(1)For the problems of traditional time-domain features such as limited excavation of feature information and insufficient monitoring effect,AR model is established for failure-free data of rolling bearing and failure data.The coefficient and residuals of AR model and the energy values of nodes of wavelet packet decomposition are input into the subsequent degradation assessment model as feature sequence?(2)The statistical distance-based anomaly detection algorithm based on fuzzy C-means has no definite upper bound,while the algorithm of anomaly detection based on Hidden Markov Model often appears early saturation.Performance degradation assessment model based on FCM-HMM is presented.After extracting the features,the measured datasets are input into the established HMM and FCM model getting the degradation indexes P and DI.Then P and DI are input into the established FCM model.This fusion algorithm integrates the advantages of both the spatial statistical distance and the similarity method.And this model can realize the performance degradation assessment.Finally,the demodulation analysis method combining EEMD and Hilbert envelope demodulation is adopted to verify the correctness of the assessment results by comparing envelope spectrum with bearing fault feature frequency.(3)In order to prove the applicability of FCM-HMM model,this paper validates the model with two sets of experiments of IEEE PHM2012 and WTG high-speed bearing.The results show that FCM-HMM model has obvious advantages in evaluating the performance degradation of rolling bearing.Finally,the FCM-HMM model is compared with the anomaly detection algorithm based on model reconstruction,and the results show the superiority of fusion anomaly detection algorithm.
Keywords/Search Tags:Rolling bearing, Fault feature extraction, Anomaly detection algorithm, Performance degradation assessment
PDF Full Text Request
Related items