无刷直流电机的神经滑模变结构控制

Neural Sliding Mode Control for Brushless DC Motors

针对工业控制中对无刷直流电机位置控制的高精度要求,研究了滑模变结构控制和神经网络相结合的控制方法。为了消除滑模变结构控制方法中存在的抖振缺点,提出了一种神经滑模控制方法。方法首先设计了一个二阶时变滑模面,使系统的初始状态就在滑模面上,可以增强系统的鲁棒性。然后,通过径向基函数神经网络学习电机的负载、干扰等参数,使滑模控制的切换控制项能随着负载参数的变化而变化,削弱了滑模变结构控制的抖振。对上述方法进行仿真,结果证明了上述方法的有效性,为无刷直流电机优化控制提供了有效手段。

The approach which combined sliding mode control and neural networks is researched for the position controller of brushless DC motors in industry. A new neural sliding mode control scheme was proposed for reducing chattering of sliding mode control in the paper. A global sliding mode manifold was designed in this approach, which guarantees that the system states can be on the sliding mode manifold at initial time and the system robustness can be increased. A radial basis function neural network （RBFNN） was applied to learn the maximum of unknown loads and external disturbances. Based on the neural networks, the switching control parameters of sliding mode control can be adaptively adjusted with uncertain external disturbances and unknown loads. Therefore, the chattering of the sliding mode controller was reduced. The simulation results prove that this control scheme is valid by simulation experiments.