This paper presents an “ad hoc” methodology for the design of diagnostic software for the detection and isolation of faults on sensors and actuators of a remotely controlled semi scale YF-22 research aircraft. Starting from the structural analysis of the non-linear dynamic equations of the aircraft, an algorithm, based on the “variables elimination method”, is proposed to compute a set of residual equations having all the possible fault signatures. The quality of each residual equation has been ranked according to a cost function chosen to represent implementation issues such as the sensitivity to measurement noise in the numerical computation of high order derivatives. An algorithm is then proposed for selecting a subset of residual equations with maximum “failure isolability” and minimum cost, according to the selected performance criteria. The issue of robustification of the residual equations to modeling errors and measurement noise has been addressed through nonlinear uncertainty mapping using Neural Networks in conjunction to FIR filters. The Fault Detection and Isolation method has been applied by injecting simulated faults to flight data collected by a semi-scale YF-22 research aircraft model.
Design of robust redundancy relations with application to a semi-scale yf-22 aircraft model
FRAVOLINI, Mario Luca;
2009
Abstract
This paper presents an “ad hoc” methodology for the design of diagnostic software for the detection and isolation of faults on sensors and actuators of a remotely controlled semi scale YF-22 research aircraft. Starting from the structural analysis of the non-linear dynamic equations of the aircraft, an algorithm, based on the “variables elimination method”, is proposed to compute a set of residual equations having all the possible fault signatures. The quality of each residual equation has been ranked according to a cost function chosen to represent implementation issues such as the sensitivity to measurement noise in the numerical computation of high order derivatives. An algorithm is then proposed for selecting a subset of residual equations with maximum “failure isolability” and minimum cost, according to the selected performance criteria. The issue of robustification of the residual equations to modeling errors and measurement noise has been addressed through nonlinear uncertainty mapping using Neural Networks in conjunction to FIR filters. The Fault Detection and Isolation method has been applied by injecting simulated faults to flight data collected by a semi-scale YF-22 research aircraft model.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
