Engine Start Identification Based on Parameter Optimization of Improved PSO-SVM

For kernel function and relative parameters of affecting SVM identification performance,the best result has been found.Relative parameters are optimized by combining improved PSO with two methods.Parameters affecting identification performance and the best parameter selection are defined.Contrasting engine start identification results of BP neural network and SVM,identification precision and converage time of SVM which are consistent with starting data are superior to BP neural network.And generalized performance of identification model is good.Based on the model,engine start performance is calculated,and the results are good agreement with test-drive data.This method is instructional for start performance calculation.