Abstract: Aiming at the problem of complex vibration signal and the difficulty to predict the fault type of gearboxes, this paper proposes a wavelet neural network optimization algorithm (WCPSO) for fault diagnosis of the gearboxes. This algorithm is a particle swarm optimization based on the dynamic acceleration constant coordinating with inertia weight. The diagnosis result of the WCPSO optimizing wavelet neural network is compared with that of the traditional wavelet neural network. It is concluded that this method can obviously improve the accuracy and raise the convergence speed, and has high recognition rate for multi-fault symptoms. Thus, it is an effective method for fault diagnosis.

Key words: vibration and wave, WCPSO, wavelet neural network, swarm intelligence, fault diagnosis

摘要： 针对传动箱振动信号复杂及故障类型难以预知的问题，提出一种基于动态加速常数协同惯性权重的粒子群优化算法(WCPSO)优化的小波神经网络进行传动箱的故障诊断，并比较经WCPSO优化的小波神经网络和传统小波神经网络诊断的结果。结论是该方法能明显提高收敛精度，对多故障征兆有较好的故障识别率，是解决故障诊断问题的有效途径。

关键词: 振动与波, WCPSO, 小波神经网络, 群体智能, 故障诊断