基于RBF神经网络的无刷直流电机无位置传感器控制

通过分析无刷直流电机间接位置检测原理 ,提出了基于径向基函数 (RBF)神经网络的无位置传感器控制方法。该方法建立动态的RBF网络模型 ,采用k 均值聚类法和递推最小二乘法 (RLS)离线训练得到网络初始参数 ,采用基于梯度下降纠正误差法在线训练更新网络参数 ,通过对电机端电压和相电流的映射 ,得到电机换相信号 ,取代了传统的位置传感器。最后实验结果表明了该控制方法的有效性

Based on RBF Neural Network Position Sensorless Control for Brushless DC Motors

In this paper,the principle of position sensorless control for brushless DC Motors (BLDCM) is analyzed,and a new control method for BLDCM is proposed,which is based on radial basis function (RBF) neural network.The network gets the initial centers through a k means algorithm,and gets the weights through recursive least squares (RLS) in off-line training.In on-line training,the network updates the network parameters using a gradient descending error algorithm.The outputs of the network,rectified by the logical module,are regarded as the control signals as well as the teacher for network training.By mapping the terminal voltages and phase currents to communication signals,the network can replace the traditional position sensors.The theory in this paper is verified by the experimental results.