基于小波概率神经网络的旋转机械振动故障诊断技术  

Research on Fault Diagnosis Method for Rotating Machinery Vibration Based on Wavelet Transformation and Probabilistic Neural Network

作　　者：吴文杰[1] 黄大贵[1] 

机构地区：[1]电子科技大学机械电子工程学院,成都611731

出　　处：《核动力工程》2012年第6期105-109,共5页Nuclear Power Engineering

摘　　要：利用小波分解提取故障特征,应用概率神经网络(PNN)诊断故障,提出一种基于小波PNN的信息融合故障诊断技术,并用MATLAB进行仿真验证。仿真验证表明:应用小波分解提取故障能量向量特征,具有很强的泛化能力和抗噪声干扰能力,适应转速频率结构的动态变化范围宽,所需样本容量小;构建的PNN具有适应性好、抗噪声干扰能力强、分类诊断准确率高的特点。将两者融合构成小波PNN应用,可获得更佳的分类诊断效果,大大提高其故障诊断的泛化性、可靠性和准确率。Based on wavelet transformation and Neural Network Data Fusion, a Fault Diagnosis Technology is proposed. Fault feature extraction is carried out using wavelet decomposition, probabilistic neural network fault diagnosis technologies by optimizing the selection, and by the MATLAB Simulation. The simulation and results verify that using wavelet decomposition extract fault characteristics of the energy vector, which has strong generalization ability and anti-noise ability to adapt to Wide dynamic range and small sample, and building the adaptive probabilistic neural network is a good anti-noise capability, classification advantage of the high rate of diagnostic accuracy. Integration of the wavele and neural network application will provide a better classification of diagnosis results, and better reliability and accuracy.

关 键 词：概率神经网络 旋转机械 振动 故障诊断 

分 类 号：V263.6[航空宇航科学与技术—航空宇航制造工程]