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.