摘要
论述了神经网络应用于电机故障诊断的方法,采用非线性最小二乘法中的LM(Levenberg-M arquardt)算法,能够有效地减少训练时间,并且提高诊断识别的精度,明显优于其他改进型BP算法.针对神经网络易陷入局部最小点的固有缺点,采用遗传算法(Genetic A lgorithm,简称GA)对神经网络初始权值进行全局优化,再采用LM算法进行训练学习,同时避免了遗传算法局部搜索能力不强的缺点.应用遗传-神经网络方法对电机故障诊断进行了仿真实验研究,证实了此方法的正确性与有效性.
This paper proposes the motor fault diagnosis methods based on neural network. The Levenberg-Marquardt（LM） algorithm, which is one of the nonlinear least squares algorithms, can reduce the training time and increase the diagnose precision efficiently. It＇s better than other modified BP algorithm obviously. Genetic Algorithm（GA） is used to optimize the initial weights in global view to avoid the inherent defect of local minimal points. Then the LM algorithm is used to further train the neural network, which can avoid the defect of weak local searching ability. The motor diagnosis simulation verifies the correctness and effectiveness.
出处
《天津理工大学学报》
2006年第5期41-43,72,共4页
Journal of Tianjin University of Technology
基金
天津市自然科学基金(043601511)
天津市科技创新基金(2004BA08)
关键词
神经网络
遗传算法
故障诊断
neural network
genetic algorithm
fault diagnose
