基于神经网络预测的网络化控制系统故障检测

Fault Detection of Networked Control Systems Based on Neural Network Prediction

  • 摘要: 针对存在传感器数据丢失的网络化控制系统,研究其故障检测问题.首先,利用基于迭代算法的线性矩阵不等式得到使系统故障观测器稳定的充分条件;然后,为补偿丢包给系统造成的不良影响,采用改进的神经网络预测控制器对系统输出进行预测,在利用反馈校正对预测输出值进行修正的基础上通过在线调整学习效率改变误差大小,从而得到更好的收敛性;并给出故障判定准则,结合观测器对带故障系统进行检测.最后通过仿真分析验证所提方法的有效性.

     

    Abstract: The fault detection of the problem affecting the networked control systems was investigated by analyzing the data packet dropout of the sensor. Firstly, the linear matrix inequality based on the iterative algorithm was used to obtain sufficient conditions for the stability of the fault observer. Then, the output of system was determined by applying the predictive controller through the improved neural network to offset the negative impact which has been caused by the packet dropout. The overall error was reduced by adjusting the learning efficiency based on the corrected predictive output and by using the feedback correction to get the higher convergence. Finally, several examples were given to demonstrate the effectiveness of the proposed method.

     

/

返回文章
返回