基于改进BP神经网络的柴油机故障诊断研究

Fault Diagnosis in Diesel Engine Based on Improved BP Artificial Neural Network

根据柴油发动机故障与征兆之间关系来建立一种采用BP算法前馈型神经网络结构,然而采用标准BP算法对神经网络训练进行训练,但存在收敛速度慢等问题。因此,又采用添加动量项和自适应学习速率两种方法对标准BP算法进行改进,并将改进的BP算法运用于神经网络训练,结果表明改进的BP神经网络能够改善收敛速度慢的缺点,而且预测故障效果较好。

This paper introduces a forward neural network with BP learning algorithm based on the relation between faults and symptoms in diesel engine ,however,the convergence speed during training neural network with standard BP learning algorithm will be slow sometimes. Therefore, momentum term and self-correcting studying rate should also he adopted to improve BP learning algorithm. The result shows improved BP artificial neural network can accelerate convergence speed in training process, and predict the faults in diesel engine effectively.