A Hybrid Multimodel-Based Condition Monitoring and Sensor Fault Detection Method for Aero Gas Turbine

Diagnosing gas path faults in aero gas turbines is vital for maintaining reliability. This article presents a hybrid architecture for fault diagnosis, integrating condition monitoring with sensor fault detection. For the engine condition monitoring, the methodology uses a data-driven approach for fault classification and a multimodel approach for condition monitoring. Initially, the data-driven method identifies the specific gas path failure mode in the engine. Once identified, the corresponding adaptive model provides detailed condition monitoring results. This approach provides a more accurate state assessment than a single-model approach by determining in advance where the fault occurs. For sensor fault diagnosis, a distributed multimodel framework analyzes discrepancies among various measurements using the state consistency strategy. Multiple thresholds are then applied to detect and isolate sensor faults effectively. Implemented in MATLAB/Simulink and validated through a gas turbine simulation model, the proposed multithreshold method shows promise in improving the engine sensor system’s self-diagnostic quantitative capabilities over traditional single-threshold methods.