Operational Reliability Assessment Of Rolling Bearing Based On EEMD And FSVDD

Mechanical equipment is indispensable in social production and people’s life,and its operational reliability is required to be higher and higher.Rotating machinery occupies a large proportion in mechanical equipment,and rolling bearing is a necessary part of it,and it is prone to failure.In addition,the limitations and disadvantages of the traditional regular maintenance and post maintenance strategy make it difficult to meet the actual engineering needs.Therefore,according to the principle of barrel effect,the running reliability of rolling bearing is evaluated to reflect the running state of the whole rotating machinery.Based on EEMD and FSVDD,the evaluation method of rolling bearing operation reliability is proposed.Firstly,in the aspect of data acquisition,operational reliability assessment needs rolling bearing life test data.At present,the authoritative life-cycle test data at home and abroad are published by IMS of University of Cincinnati,FEMTO-ST Research Institute of France and Xi’an Jiaotong University.By comparing the experimental data obtained by Xi’an Jiaotong University for two years,the method is verified to ensure the reliability of the data source.Secondly,the vibration signal of rolling bearing collected by sensors contains noise and other useless information,the original vibration signal is decomposed by using ensemble empirical mode decomposition,and then the modal components with correlation coefficient greater than 0.3 are selected by using correlation coefficient criterion to achieve the purpose of noise reduction.Thirdly,feature extraction is the key step of rolling bearing reliability evaluation.In order to extract the reliable features in line with the operational reliability evaluation,firstly,multi-dimensional feature extraction is performed on the de-noising signal,including the traditional time-domain features and the two complexity analysis methods of fuzzy approximate entropy and permutation entropy.Then,two improved feature selection indexes proposed in this paper: monotonic ability and robust ability are used for feature selection.Finally,the most suitable features are selected The feature vector is constructed from the extracted three-dimensional features and used as the input of the subsequent reliability evaluation model.Finally,after the fault features are extracted,the fuzzy support vector data model is used for reliability modeling.The fuzzy membership degree determination method of cosine Euclidean distance is proposed,and the operation reliability is calculated by using the proposed normalized operation state parameter method.The proposed method is applied to the rolling bearing accelerated life test data released to the world by Xi’an Jiaotong University,and the effectiveness of the proposed method is verified.