基于LabVIEW FPGA的音圈电机神经网络控制方法

音圈电机是一种常用的高动态性能驱动器。为了优化、简化不同音圈电机的控制算法调试过程,提出了一种基于LabVIEW FPGA的改进RBF神经网络控制算法。该网络采用3-输入的3层网络结构,通过梯度下降法对隐含层权值系数更新,实现对音圈电机的电流环控制。仿真结果表明,随着隐含层节点数增加,RBF神经网络控制算法性能将优于单神经元自适应控制算法。当隐含层节点数为6时,系统电流上升时间为14μs,相对于单神经元自适应控制缩短了26.3%。将该算法用于激光精密指向系统,实验结果表明,隐含层4节点时,电流上升时间为25μs。实验结果与仿真结果较吻合,验证了RBF神经网络算法的有效性。

Voice Coil Motor Neural Network Control Method Based on LabVIEW FPGA

Voice coil motor is a common driver with high dynamic performance.In order to optimize and simplify the debugging process of control algorithms for different voice coil motors,an improved RBF neural network control algorithm based on LabVIEW FPGA was proposed.The network adopted a 3-input three-layer network structure,and updated the weight coefficient of hidden layer by gradient descent method to realize the current loop control of voice coil motor.Simulation result shows that the performance of RBF neural network control algorithm is better than the performance of single neuron adaptive control algorithm when the number of hidden layer nodes increases.When the number of hidden layer nodes is 6,the system current rise time is 14μs,which is 26.3%quicker than the single neuron adaptive control.The proposed algorithm was applied to the laser precise pointing system and the experimental result shows that the current rise time is 25μs when there are 4 nodes in the hidden layer.The result is consistent with the simulation results and verifies the effectiveness of the RBF neural network algorithm.