In-flight fault accommodation of safety-critical faults requires rapid detection and remediation. Indeed, for a class of safety-critical faults, detection within a millisecond range is imperative to allow accommodation in time to avert undesired engine behavior. We address these issues with an integrated detection and accommodation scheme. This scheme comprises model-based detection, a bank of binary classifiers, and an accommodation module. The latter biases control signals with pre-defined adjustments to regain operability while staying within established safety limits. The adjustments were developed using evolutionary algorithms to identify optimal biases off-line for multiple faults and points within the flight envelope. These biases are interpolated online for the current flight conditions. High-fidelity simulation results are presented showing accommodation applied to a high-pressure compressor fault on a commercial, high-bypass, twin-spool, turbofan engine throughout the flight envelope.
Handling index-out-of-bounds in safety-critical embedded C code using model-based development