DETECTION AND ISOLATION OF A SENSOR DRIFT FAULT

DETECTION AND ISOLATION OF A SENSOR DRIFT FAULT Yilun Zhou Prof. Thomas Parisini Imperial College London Univ logo UKACC Ph. D Presentation Showcase

Introduction Ø Background • • Complex control systems and demand for fault tolerance Process fault and sensor fault in nonlinear systems Model-based fault detection and isolation schemes (FDI) Time evolution of faults: abrupt and incipient faults Ø Objectives • A class of nonlinear systems with sensor drift faults • Adaptive and less conservative threshold Univ logo UKACC Ph. D Presentation Showcase Slide 2

Problem Formulation and Methodology Ø A class of nonlinear MIMO system with sensor drift faults Ø Detection Estimators and Isolation Estimators Ø Residual and Threshold > > Univ logo crossing s Drift faults, s Isolators. d. Any crossing excludes faulty assumption. UKACC Ph. D Presentation Showcase Slide 3

Simulation Result Ø Detection At t = 13 s, a fault occurs. Then the fault is detected at t = 16. 64 s Univ logo UKACC Ph. D Presentation Showcase Slide 4

Simulation Result Ø Isolation There is a crossing in isolator 2. The possibility of fault occurrence in Sensor 2 is excluded. Drift fault occurred in Sensor 3 Univ logo UKACC Ph. D Presentation Showcase Slide 5

Conclusion and Future work Ø Conclusion: • Adaptive threshold increases robustness and stability • Fault compensation in isolators improves performance Ø Future Work: • Design a FDI approach for a hybrid sensor fault • Design a distributed fault diagnosis approach • Design a benchmark test using the FDI scheme Univ logo UKACC Ph. D Presentation Showcase Slide 6