Constructed Of SVM Decision Tree Based On Particie Swarm Optimization Algorithm For Fault Diagnosis Of Gear Box

Gearbox is widely used in the development of modern industry as a general part which could join and pass power necessarily in mechanical instruments. In the industrial production process, there are about80%of the mechanical failure was caused by gearbox failure. The occurrence of these failures will directly affect the life of the machine, production safety and production efficiency. So it is of great factual significance to diagnose its failures in keeping instruments in health. This paper mainly studies the existing problem of process of Support Vector Machine in being establishing multiple-class classification by using the device of decision tree and brings up a model of Support Vector Machine based on decision tree of particle swarm optimization.First of all, this paper analyses theorem of vibration of gearbox and some ordinary types of failure and on this base experiment platform of gearbox is built. Experimental scheme of diagnosis of failure of gearbox is designed. Acceleration signals of vibration of gearbox under different failures are gotten after testing.Secondly, basic theory and development prospect of particle swarm optimization algorithm is introduced systematically. Influences of particle swarm optimization algorithm parameter on accuracy and efficiency. Commonly used parameters selection principle is introduced. To test the ability of algorithm of inertia weight PSO, this paper selects two commonly used standard functions to simulate. Proved by experiments, algorithm of inertia weight PSO could converges to the optimal solution quickly meeting the condition of the accuracy of convergence. On the basis of introduction of VC dimension theory and the principle of structural risk minimization, classification strategy of basic theory of Support Vector Machine and decision tree of Support Vector Machine is elaborated.Finally, decision tree of Support Vector Machine is optimized by using the fine performance of global searching of algorithm of inertia weight PSO, then the optimal decision tree is generated and the multiple classifier is built up which are used for experimental data of failure diagnosis of gearbox. So failure diagnosis could be accomplished effectively, and further the ratio of accuracy of failure diagnosis is raised. These are going to provide a more accurate and more effective method for failure diagnosis of machinery.