基于D-S证据理论的直接空冷凝汽器故障诊断方法研究

Research of Fault Diagnosis Method for Air-cooled Condensers Based on the D-S Evidence Theory

提出一种基于Elman神经网络和RBF神经网络通过D-S证据理论融合的故障诊断方法,把该方法应用在直接空冷凝汽器的故障诊断中。首先对故障进行神经网络初步诊断,得到属于不同故障状态的隶属度,然后采用D-S证据理论融合的方法进行决策诊断,得到最终结果。研究了直接空冷凝汽器的故障特征提取、样本选择、诊断系统结构和学习算法,并通过诊断实例阐述了该方法的具体实现过程,验证了所提方法的可行性,结果表明:该方法适用于直接空冷凝汽器故障诊断,故障定位准确率高。

A fault diagnosis method based on Elman network and RBF network fused by the D-S evidence theory was proposed and applied in the fault diagnosis of air-cooled condensers. In which, having the neural networks adopted for preliminary diagnosis of faults to get the membership degree in relation to different fault status, and then having D-S evidence adopted for decision-making diagnosis to get final result. In addition, the air-cooled condenser＇ s fault feature extraction, sample selection, diagnosis system structure and the learning algorithm concerned were discussed and this method＇ s feasibility and implementation were expounded. The diagnosis examples verify this method＇ s fault tolerance and effectiveness in dealing with complicated fault conditions and it has high accuracy in the fault location.