Active Fault Identification by Optimization of Test Designs

Model-based fault detection and isolation (FDI) methods are used to determine faults by examining the deviation of sensed information from anticipated system trajectories. In this paper, a comprehensive model-based FDI framework is proposed to improve fault identifiability and reduce false alarms during maintenance testing. In this framework, robust maintenance tests are designed and conducted, followed by false alarm analysis. The optimal tests designed improve the identifiability of faults by manipulating system inputs to maximize information with respect to faults, in the form of sensitivities of the system outputs. Each test design is evaluated a posteriori using the system model to explore whether false alarms are plausible, given system uncertainty and measurement noise. The proposed framework is applied on two case studies that compare the identifiability of faults at nominal and optimal system test conditions. The first case study focuses on a plate fin heat exchanger with various levels of particulate fouling at steady-state and transient conditions. The second case study deals with the same type of fault but in an aircraft environmental control system with multiple sources of uncertainty.