For the nonlinear vibration signals caused by vehicle gearbox,a state recognition method combined of multi-fractal and support vector machine is proposed.The vibration signal characteristics are described by using the singular spectrum and generalized dimension which as the input into characteristics of SVM.The intelligent optimization selected of penalty function c and radial basis functionσis realized are through introducing the improved chaotic particle swarm optimization algorithm into support vector machine parameter optimization.The results show that the different operational status of the vehicle gearbox can be classified by using SVM classification model established by this method,and the accuracy is increased.