Distributed fault detection with correlated decision fusion

Quick and accurate fault detection is critical to the operation of modern dynamic systems. In this paper, the fault detection problem when using multiple sensors is investigated. At each time step, local sensors transmit binary data to the fusion center, where decision fusion is performed to detect the potential occurrence of a fault. Since the sensors observe a common dynamic process, their measurements, and thus the local decisions, are correlated. Under a likelihood-ratio-based local decision rule constraint, we propose efficient suboptimal system designs involving local sensor rules and fusion rule that include the correlation consideration. Two correlation models are proposed to approximate the complicated correlation between sensor measurements for general systems. Experimental results show that the designs with correlation consideration outperform the design under the independence assumption significantly when the correlation between sensor measurements is strong.