差分进化小波神经网络在微动齿轮故障诊断中的应用  

Application of Differential Evolution Wavelet Neural Networks in Micro-Gears Fault Diagnosis

作　　者：姚宁[1] 王武[1] 张飞云[1] 

机构地区：[1]许昌学院电气信息工程学院,许昌461000

出　　处：《机械设计与制造》2012年第12期89-91,共3页Machinery Design & Manufacture

基　　金：河南省教育厅自然科学研究资助项目(2011B470005);河南省科技厅科技攻关计划资助项目(112102210339)

摘　　要：构建立了三层小波神经网络,给出了小波神经网络结构及算法,给出了差分进化的原理和实施步骤。给出了微动齿轮的机械振动机理和故障特征,将微动齿轮故障分为无故障、齿轮断层、齿轮面磨损脱落、齿轮面损伤,齿轮面裂痕等五种故障,通过振动试验测试故障信息,将其作为小波神经网络的训练样本,将差分进化应用于小波神经网络结构和参数的优化中。仿真结果表明差分进化小波神经网络能够有效避免神经网络不收敛的缺点,提高学习速度,采用差分进化小波神经网络进行微动齿轮故障诊断,具有较高的诊断精度和效率,可以有效应用于其他系统的故障诊断工程中。The micro-vibration mechanism and fault characteristics of micro-gears were described, and the three layer wavelet neural network was constructed and the algorithm was proposed,and also the principles and steps of differential evolution algorithm were presented.The micro-vibration mechanism and fault characteristics of micro-gears were described and the faults were classified with no fault,gear crack, gear face wear,tooth face attrition, and tooth face crack.The diagnosis information acquired with vibration experiment and designed as training samples of wavelet neural networks ,and the network was optimized with differential evolution algorithm^Simulation result shows the new method can avoid normal neural net- works with convergence quality and enhance learning speed,and also the higher diagnosis precision and efficiency was got,and this proposed method can be effectively used in engineering diagnosis system.

关 键 词：小波神经网络 差分进化算法 故障诊断 微动齿轮 仿真 

分 类 号：TH16[机械工程—机械制造及自动化]