Remained Useful Life Prediction Research Of The Rolling Bearing Based On Dynamic State Space Model

Rolling is a key component of rotating machinery and predicting its remaining life is particularly important for the normal operation and maintenance of equipment.Accurately predicting the life of rolling bearings can rationalize its performance inspection and maintenance or replacement work, which can avoid excessive maintenance and resulting in unnecessary economic losses and ensure production safety.Rolling bearing life prediction method is divided as the following three types:the rolling bearing life prediction method based on the statistical model, the rolling bearing life prediction method based on fracture mechanics model, the rolling bearing life prediction method based on data driven.Synthesize the advantages of the three methods, this paper based on the degradation mechanism of rolling bearing degradation made use of characteristic values to improve the common fatigue degradation formula and established the physical dynamic state space model, then used the dynamic state space model life prediction method to predict the life, the study substance as follows:(1) Based on the fatigue degradation mechanism of rolling bearing, improved crack growth formula-Paris formula, proposed a physical dynamic state space model, and used the physical dynamic state space model method to predict the life of rolling bearing.(2) On the basis of the physical dynamic state space model, improved crack growth formula-Foreman formula, proposed multi-stage physical dynamic state space model, and used the multi-stage physical dynamic state space model method to predict the life of rolling bearing.(3) Using the time domain characteristic value and the frequency domain characteristic value of the rolling bearing to construct the multi-dimensional characteristic matrix. The multidimensional feature matrix is processed by manifold learning to establish a new distance index. Finally, the distance evaluation algorithm is used to select the most sensitive characteristic value.(4) Compare and analyze resampling algorithm of particle filter algorithm, Select the best resampling algorithm, Improve the accuracy of life prediction results.