摘要
提出了基于小波神经网络控制的无刷电机控制系统新方法,该方法使用三层前馈式人工神经网络,采用基于梯度下降纠正误差法在线训练更新网络参数,使用离散小波变换的时频特性和连续小波变换检测信号边沿的原理进行无刷电机运行状态和故障状态的检测以便能实时保护。仿真结果表明该方法能大大改善调速系统的静态、动态性能,具有优良的控制效果,小波检测灵敏度高,对噪声有较高的鲁棒性,具有广阔的应用前景。
Based on artificial neural network (ANN) and wavelet transform,a new approach of brushless DC motor servo_system is presented with three_layer forward artificial neural network.The approach is designed to train and replace network parameters in_line using a gradient descending error algorithm. The work and fault states of brush_less DC motor are detected, in which the time_frequency characteristics of discrete wavelet transform (DWT) and the principle of signal edge_detecting of continuous wavelet transform (CWT) are used. The simulation result shows that the system using the approach has very good dynamic and static performances, it is sensitive to fault and robust to noise, and it has a vast applying prospect.
出处
《中山大学学报（自然科学版）》
CAS
CSCD
北大核心
2004年第6期102-105,共4页
Acta Scientiarum Naturalium Universitatis Sunyatseni
基金
广东省自然科学基金资助项目(020118)
广东省高教厅自然科学基金资助项目(1999267)
佛山市科技发展资金资助项目(0003012B)
关键词
小波变换
无刷电机
神经网络
wave transform
neural network
brushless DC motor
