中文摘要：

针对无刷同步发电机旋转整流器常见故障特点,提出一种基于小波包分解和BP神经网络的旋转整流器故障识别方法。运用小波包分析（WAP）提取故障特征信号,建立故障识别的神经网络模型,采取粒子群-人工蜂群（PSO-ABC）算法优化神经网络的最优初始连接权值和阈值等结构参数。仿真结果表明,该方法具有识别速度快、准确性高等优点。

英文摘要：

In terms of some common faults in rotating rectifier of brushless synchronous generator, a way based on wavelet packet decomposition and BP nerve net to identify faults is discussed. Generally, using WAP to pick up fault signals and to make nerve net model to identify faults. Besides, PSO-ABC algorithm could be used to optimize such structural parameters as the optimum initial connection weights and threshold value in nerve net. The simulation results reveal that this method is characterized with its advantages of being quick and accurate in detect-ing faults.