基于神经网络混合建模的结构振动滑模控制

将神经网络和标称系统混合建模方法引入到离散滑模控制当中,得到神经网络滑模控制,然后对结构振动进行控制,振动结构为具有不确定性参数的柔性附件,并受到随机外扰作用。离散滑模控制的滑模面是以标称系统为基础,由最优二次型价值函数求解黎卡提方程确定。利用标称模型和神经网络混合建模方法来减小系统的不确定性,达到提高滑模控制在实际控制系统中的控制效果。其中利用前馈神经网络来对不确定部分进行建模。最后通过对滑模控制和神经网络滑模控制进行仿真,结果表明,本文所提出的神经网络滑模控制对具有不确定性参数和随机外扰的柔性结构系统振动的控制效果要优于滑模控制。

Sliding mode control for flexible structural vibration reduction based on hybrid neural network models

In this paper,the hybrid modelling based on neural network is introduced for discrete sliding mode control.This model is used to control the random vibration of the flexible structures with parametric uncertainty .The design objective of the discrete sliding mode surface is determinated by the quadratic optimal cost function and solved from the algebraic Riccati equation based on the nominal system.To reduce the chatter from the large uncertainty of the system,the hybrid model of neural network and nominal system is adopted.The neural network method with multilayer feed-forward neural network is used to model the uncertainty part.In the end,the system controlled by sliding mode control and that by neural network sliding mode control are simulated,respectively.The simulation results show that the neural network sliding mode control is more effective than the sliding mode control for the vibration of flexible structures with large parametric uncertainty and random external disturbances.