Title :
Application and Research of the Train Fault Diagnosis Based on Improved BP Neural Network Algorithm
Author :
Yingwei Qu ; Yinnan Yan ; Guanghai Zheng
Author_Institution :
Software Technol. Inst., Dalian Jiaotong Univ., Dalian, China
Abstract :
Traditional BP model of neural network is easy to get a local minimum rather than the global optimal solution. As the training times increases, the learning efficiency is falling low, so as the convergence rate. Improvement on the traditional model of BP neural network algorithm improves the convergence rate of the neural network, and reduces the training times, so that the output of the neural network can not only determine the type of the train failure occurred, to improve the accuracy of diagnostic results, but also to diagnose within a certain range even the fault does not appear, to make the fault of train intelligent and simple. The simulation results show that the improved algorithm is effective.
Keywords :
backpropagation; fault diagnosis; neural nets; parallel processing; railways; associative memory; convergence rate; distributed parallel processing; global optimal solution; improved BP neural network algorithm; learning efficiency; train fault diagnosis; Accuracy; Biological neural networks; Convergence; Fault diagnosis; Neurons; Training; Transfer functions; Fault diagnosis; Neural network; global optimal solution; step length;
Conference_Titel :
Control Engineering and Communication Technology (ICCECT), 2012 International Conference on
Conference_Location :
Liaoning
Print_ISBN :
978-1-4673-4499-9
DOI :
10.1109/ICCECT.2012.26