基于集成法的汽轮机组智能故障诊断仿真研究

Simulation Research on Intelligent Fault Diagnosis for Turbine Generator Unit Based on Integrated Method

针对BP神经网络在故障诊断中容易陷入局部极小点及收敛速度慢等问题,引入了小波神经网络(WNN)和概率神经网络(PNN)两种故障诊断算法。为了避免采用单一的故障诊断方法进行故障诊断可能会存在误诊或漏诊的现象,提高故障诊断的可靠性,又在决策层引入D-S理论进行融合故障诊断。但是由于在处理冲突证据合成时,D-S合成公式完全没有考虑冲突证据信息而造成故障信息的浪费,因此为了充分利用所有故障信息,采用了改进的D-S合成算法,提出了一种基于WNN-PNN和改进的D-S集成的汽轮机组故障诊断方法。方法首先通过WNN和PNN对故障信息进行诊断并构造证据体,然后根据改进的D-S证据理论进行融合诊断。通过仿真验证了集成诊断方法的有效性。

Considering that BP neural network has slow convergence speed and easy to fall into local minima,the wavelet neural network（ WNN） and probabilistic neural network（ PNN） are employed to fault diagnosis for turbine generator unit. In order to avoid misdiagnosis and missed diagnosis of a single fault diagnosis method and to improve the reliability of fault diagnosis,the D-S evidential theory was introduced into the fusion fault diagnosis in decisionmaking. However,D-S combination formula is likely to waste the fault information because it abandons the conflict evidence information when the conflict evidence is being dealt with. In order to take full advantage of all the evidence,the improved D-S combination formula was employed to fusion diagnosis in this paper. An integrate fault diagnosis method using WNN-PNN and improved D-S evidential theory was presented for turbine generator unit. In the method,first,the results of preliminary diagnosis of fault information with WNN and PNN network constituted evidences; and the evidences were fused under the improved D-S evidential theory,then the diagnostic results were obtained. The simulation results show that the integrated fault diagnosis method is an efficient diagnosis approach.