摘要
针对电机保护只在被测参数达到或者超过设定动作阈值才动作,缺乏预测控制能力,设计了一种基于粒子群的径向神经网络。利用小波变换的时频分解能力、优异的奇异检测能力进行故障特征分量的提取;用粒子群算法和径向神经网络配合优化权重,从而使网络收敛快,训练时间短。通过电动机故障进行仿真实验,结果表明PSO-RBF神经网络实现了对故障的识别。
Because the motor can’t take action until the measured parameters meet or exceed the threshold, which lack of prediction control, a Radial Basis Function（RBF） Neural Network is designed based on Particle Swarm Optimization（PSO）. Using strong time-frequency decomposition capabilities and outstanding singularity detection capability of wavelet transform, the eigenvector of fault can be gained；the connection weight is optimized by RBF Neural Network with PSO, which makes the Neural Network convergence faster and training shorter. According to simulation of fault by motor, fault is recognized by PSO-RBF Neural Network.
出处
《计算机工程与应用》
CSCD
2012年第31期216-219,共4页
Computer Engineering and Applications
基金
浙江省重大科技专项(No.2007C11072)
国家重大专项"极大规模集成电路制造装备及成套工艺"子课题(No.2009ZX02011-02)
关键词
径向神经网络
小波变换
故障诊断
粒子群算法
Radial Basis Function（RBF） Neural Network wavelet transform fault diagnosis Particle Swarm Optimization（PSO）
