Evaluation of Sliding Mode Observers for Fault Reconstruction on the ADDSAFE Functional Engineering Simulator 2011-01-2802
This paper presents a preliminary evaluation of the results from using second order sliding mode observer schemes applied to an aircraft fault detection benchmark problem for a class of sensor faults. The scheme has been evaluated on the ADDSAFE Functional Engineering Simulator (FES). This is part of ongoing work on a European FP7 funded project entitled Advanced Fault Diagnosis for Sustainable Flight Guidance and Control (ADDSAFE) which aims to study advanced fault detection and isolation (FDI) methods for aircraft. The simulation and verification FES used in this evaluation incorporates a high fidelity nonlinear aircraft model from AIRBUS (which includes sensor and process noise). It is intended to provide a design tuning and assessment tool for the proposed FDI schemes developed within the ADDSAFE consortium and acts as a platform to test the robustness of the proposed schemes using a predetermined grid or sequence to cover a wide range of flight conditions and perturbations in the aircraft parameters. The FDI schemes will ultimately be evaluated in the Industrial Benchmarking and Verification FES which uses Monte Carlo methods to generate a statistical distribution and covers all of the desired flight conditions, perturbations and uncertainties in the aircraft parameters under investigation. This paper considers an LPV sliding mode observer for yaw rate measurement faults. A robust sliding mode reconstruction scheme for the aircraft lateral axis, based on an LPV system representation is presented. The sliding mode observers provide information about the unmeasurable signals (in this case the faults) through appropriate processing of the nonlinear injection signal required to maintain a sliding motion. The sliding mode LPV scheme utilizes LMIs to synthesize the gains of the observer to minimize the effect of uncertainty on the reconstruction signal. The equivalent output error injection signals are obtained from a second order sliding mode structure, which is less susceptible to noise compared to the typical signum function associated with first order sliding. Preliminary evaluation results from the FES are presented for the yaw rate sensor fault to highlight the potential of the proposed scheme for fault detection and reconstruction.