Skip to Main Content
Fault detection of networked control systems subject to long delay and packet dropouts is discussed in this paper. No assumption about the statistics of long delay and packet dropouts is required for the proposed approach. Long delay and packet dropouts in sensor-to-controller link, which are known to the fault detection system, are handled with non-uniform sampling technique. Due to the unknown delay and packet dropouts in controller-to-actuator link, the true control commands implemented by the actuator are unknown to the fault detection system. However, given the upper bounds of delay and the number of successive packet dropouts, in a time window the number of possibilities of packet transmission in controller-to-actuator link is limited; then the limited possibilities of the control commands implemented by the actuator can be determined from the control commands sent by the controller. This fact is useful but has not been utilized by existing methods for fault detection of networked control systems to the knowledge of the authors. Then, consistency of parity relations is tested by evaluating residuals generated with the received outputs and each possibility of the corresponding control commands implemented by the actuator. The effectiveness of the proposed method is illustrated by a simulation example.