The Study On The Improved Particle Swarm Optimization Algorithm

Particle Swarm Optimization algorithm developed rapidly in recent years, due to its simple concept, easy to implement, few parameters, etc. And as an important mechanical component, rotating machinery will directly affect the whole mechanical systems, so timely failure diagnosis is very important. So PSO has been successfully applied to the area of fault diagnosis. In this paper, two improved algorithms are proposed based on PSO-BP neural networks, PSO clustering algorithm.Firstly, on the basic of Mulit-Species Cooperative PSO, the improved algorithm which increases the population diversity is proposed, through the simulation of the test functions, to determine the parameters which can make the algorithm performance best. Training the neural network instead of BP algorithm, then can prove the excellent performance of the new algorithm.Then, the improved PSO-K means clustering algorithm is proposed, through the simulation of the datas from the UCI ,new algorithm performans well on the classification of low and high dimensional.Finally, point out the the types of bearing fault and common methods of fault diagnosis, Training the neural network instead of BP algorithm, then can prove the excellent performance of the new algorithm.Make two new fault diagnosis algorithm applied in the field respectively,comparing with other methods, shows the advantages of the new algorithm on failure diagnosis and diagnostic efficiency.