基于神经网络信息融合的柴油机故障诊断

本文论述了神经网络信息融合的原理与方法，首先就BP网络训练速度慢，易陷入局部极小点的问题，提出将附加动量项与自适应学习速率相结合的改进BP算法，有效地抑制了网络陷于局部极小并提高了收敛速度。最后，将振动信号与血管压力信号作为特征参数，分别采用传统BP算法，改进BP算法对供油系统的三种故障进行信息融合诊断分析。实践表明，神经网络信息融合方法非常适用于多征兆机械系统的故障诊断。

23-Fault diagnosis of diesel engines based on neural network information fusion

The paper introduces the theory and method of neural network, and at first, an improved BP algorithm that connects additional momentum with self-adaptive learning rate is proposed, the improved algorithm can prevent network to stack into the minimal value locally and improve the convergent velocity effectively. Vibration signals and fuel pressure signals are used as characteristics to diagnose and analyze three faults of fuel supply system by using traditional BP and improving BP algorithm, finally, the practice shows that the method of neural network information fusion is applicable to fault diagnosis of multi- signs engines.