Research On Typical Mechanical Structure Health And Life Prediction Based On SVM

This paper was supported by the National Natural Science Foundation of China and China Academy of Engineering Physics Mutual Funds under Grant(NSAF) “Research on the identificationof the mechanical structural characteristic parameters and its prediction based on kernel function”(NO.11176027). This project mainly aims to predict and monitor the health status of mechanical structures of large weapon equipments. The main research contents of this paper are as follows.(1) This paper dealed with machine learning, kernel function theory, support vector machine(SVM) and multi parameter optimization theory. It focuses on the small sample data processing capability of SVM.(2) SVM is a kind of machine learning algorithm with multiple parameters. The good regions of the optimal combinations of parameter is found by constraint relations between parameters of SVM model and the asymptotic behaviors of SVM with the radial basis function(RBF) kernel. Besides, improved artificial bee colony(ABC)optimization algorithm is adopted to search the optimal parameters in the good regions. Finally, a complete SVM model parameter optimization system, named as L-ABC-SVM method, is constructed.(3) In order to explore the influence of environmental factors on the connection structure of storage equipment, the accelerated life testing platform of three kinds of connection mode(riveting, welding,bolt connection) is built based on the accelerated life testing standard.The life prediction models of connection structures were predicted by L-ABC-SVM method based on the data preprocessing scheme of this paper. After that, the experimental results verified the feasibility and effectiveness of the research method in this paper, which were analyzed from three respects including the regression forecasting,different kernel functions and different parameters.(4) The prediction software was developed on VC6.0 platform.The L-ABC-SVM method is the core content of the software.