基于多神经网络多参数综合的旋转机械故障诊断系统研究

STUDY ON THE SYSTEM OF FAULT DIAGNOSIS FOR ROTATING MACHINERY BASED ON THE SYNTHESIS OF MULTIPLE PARAMETERS' HYBRID NEURAL NETWORKS

本文提出并建立了基于多神经网络多参数综合的旋转机械故障诊断系统。在多层多输出前向神经网络的算法基础上,对多个征兆域分别建立相应的诊断网络,然后利用加权方法进行综合评判,并且该诊断系统具有自学习、自适应能力,以便能够适应大型旋转机械,特别是汽轮发电机组等实际产生故障的振动原因的复杂性及诱发的振动征兆的多元性等特点,从而提高了故障诊断的可靠性和诊断精度。本系统对工程应用具有较高的实用价值。

In the paper a kind of fault diagnostic system for rotating machinery based on the synthesis of multiple parameters' hybrid neural networks is introduced. With the help of the multi-layer-multi-output neural network diagnostic model, the diagnostic networks corresponding to multiple symptom domains are built up and the comprehensive judgement is carried out with weighted average method. Meanwhile, the system has the ability of self-leanring and self-adaptation in order to fit the complexity of practical vibration signals and the plurality of vibrationsymptoms induced really in large rotating machinery, especially in turbogenerators. The reliability and accuracy of diagnosis with this system is satisfactory. It seems that the system is of practical value in engineering applications.