遗传BP网络在滚动轴承故障诊断中的应用.pdfVIP

21 5 Vol. 21No. 5 1 9 9 9 10 JOURNAL OF TH CHINA RAILWAY SOCI TY October 1 9 9 9 : 1001-8360( 1999) 05-0029-04 遗传 网络在滚动轴承故障诊断中的应用 BP 骆志明,　冯庚斌 ( , 100081) : BP , BP , 3 , BP BP BP , : ; ; ; : T P18; V279: A BP Neural Network Based on Genetic Algorithm for Fault Diagnosis of Rolling Bearing LUO Zhi-ming, F NG Geng-bin ( Locomotive and Car Research Institute China Academy of Railway Sciences Beijing, , 100081, China) Abstract: T here are two main weak-points of the classical BP algorithm which is easily stacked into local minima and takes much more times. In this paper, the bearing fault intelligent diagnosis method has been studied based on BP neural network combined with genetic algorithm In genetic algorithm training of BP. network weights, the real number encoding was applied. For GA optimal of locomotive and car bearing fault diagnosis BP network weights, optimal population was selected as third individual, with the use of fitness- proportionate selection and discrete crossover as well as average mutation Comparing GA and traditional. gradient algorithm for weights optimal of the BP neural network GA takes much little optimal time T he, . efficiency of research work mentioned above has been shown by numerical simulation and practical bearing fault diagnosis. Keywords neural network genetic algorithm fault diagnosis rolling bearing: ; ; ; ( Artificial Neural Network, RBF (BAM) (BSB) [ 1] ANN) Boltzmann (ART) , BP , , , Rumehart Mcclelland 1985 , ( rror Ba