基于小波神经网络控制的无刷直流电机调速系统  

THE SPEED-ADJUSTMENT SYSTEM OF BRUSH DC MOTOR BASED ON WAVELET ANN

作　　者：李先祥[1] 徐小增[1] 肖红军[2] 

机构地区：[1]佛山科学技术学院工学院,广东省佛山市528000 [2]华南理工大学自动化科学与工程学院,广东省广州市510640

出　　处：《中国电机工程学报》2005年第9期126-129,共4页Proceedings of the CSEE

基　　金：广东省自然科学基金项目(020118)

摘　　要：该文提出了基于小波神经网络控制的无刷电机调速系统新方法,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。A new approach of brush-less DC motor speed-adjustment system is presented in this paper based on artificial neural network (ANN) and wavelet transform. Three-layer forward artificial neural network is used in this paper. The network parameters can be in-line renewed by training of gradient descending error algorithm. The time-frequency characteristics of discrete wavelet transform (DWT) and local maximum of modulus of continuous wavelet transform (CWT) are used, so the working and fault states of brush-less DC motor speed-adjustment system can be detected. The simulation result shows that the system using the approach has very good dynamic and static performances, and it is sensitive to fault and robust to noise, therefore the system has a vast applying prospect.

关 键 词：神经网络控制 直流电机调速系统 人工神经网络 连续小波变换 离散小波变换 检测灵敏度 网络参数 在线训练 梯度下降 故障状态 运行状态 无刷电机 检测信号 时频特性 仿真结果 动态性能 控制效果 前馈式 鲁棒性 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程] TP273.4[自动化与计算机技术—控制科学与工程]