Engine Fault Diagnosis Based on PSO-BP Network

Abstract:

BP network could deal with some nonlinear problems such as fault diagnosis, but the gradient-based method tends to get stuck in local optima or premature convergence. As a population-based heuristic optimization algorithm, PSO method could increase training accuracy of BP network so as to improve discriminant precision of fault diagnosis of equipment such as engine. Ladder diminishing inertia weights tend to improve network’s convergence precision and efficiency during the late stage of PSO iterations, in addition, LPSO or RegPSO contributes to confining premature convergence. The PSO-BP fault diagnosis model of 190A diesel engine has verified the above opinions.

• Periodical: Advanced Materials Research (Volumes 299-300)
• Pages: 1307-1311
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.299-300.1307
• Online since: July 2011
• Authors: Li Xin Nie, Tian Xia Zhang, Shu Ju Wang, Li Ping Zhang
• Keywords: BP Network, Fault Diagnosis, Inertia Weight, Particle Swarm Optimization Algorithm (PSO), Premature Convergence

References

[1] Fu Huixuan and Zhao Hong: Application Designs of Neural Network Based on MATLAB (China Machine Press, China, 2010).

[2] Liu Yingji: Research on New Method of Condition Monitoring and Fault Diagnosis for Vehicle (Ph.D., Northeastern University, China, 2008).

[3] Wei Xiuye and Pan Hongxia: Particle Swarm Optimization and Intelligent Fault Diagnosis (National Defence Industry Press, China, 2010).

[4] Evers, George: An Automatic Regrouping Mechanism to Deal with Stagnation in Particle Swarm Optimization (M.S., the University of Texas – Pan American, America, 2009).