Research On Engineering Applications Of Suppor Vector Machine

As a new technology developing in the middle 1990's, the principle of Support Vector Machine based on statistics has made big progress and showed tremendous practical value in engineering application. Therefore, it is of great significance to study the algorithms and relevant engineering application of SVM.Firstly, the basic knowledge and typical algorithms of SVM are introduced. With the analysis of existing algorithms, the present problems and developing direction of SVM are indicated. By analyzing the principle of a typical SVM, guidance on sorting SVM is provided.Secondly, directing at the problem that calculating burden of choosing the parameters of traditional SVM is too large and parameters choosing always becomes unsuccessful when there are too many parameters, a new algorithm based on chaotic and particle swarm optimization is proposed. On the basis of the proposed algorithm, abnormity detection based on one-class SVM is established. The result of abnormity detection simulation upon moving robot sensor is satisfying.Consequently, directing at the arousing problem of garbage message , a garbage message classifier based on support vector classification is proposed. the technique of short message real-time monitoring and distinguishing is realized by this classifier. Simulation results show that the classifier has high distinguishing precision and can advisably meet the precision and real-time demand of distinguishing garbage message.Finally, Directing at the objective circumstance that friction coefficient is hard to confirm during metal plastic forming process, a friction coefficient prediction method based on SVM is proposed and application of SVM on the prediction field of friction coefficient during metal plastic forming process is inducted. A SVM prediction model in relation to lubricant,coarse,slide rate and friction coefficient is established,which can perfectly solve the realistic problem that friction coefficient during metal plastic forming process is hard to predict.