Flight Control System improvement based on a software sensor derived from Partial Least Squares Algorithm

Abstract

Global aircraft optimization is a main concern for future and upcoming programs. In particular, great research efforts are dedicated to Electrical Flight Control Systems (EFCS). Obviously, their reliability increases with the redundancy of the flight parameter sensors. However, physical redundancy, obtained by increasing the number of sensors, penalizes the aircraft weight and cost. This paper proposes a sensor failure detection method based on analytic redundancy. The flight parameter of interest is modelled as a linear function of independent sensor measurements on a sliding observation window. The Partial Least Squares (PLS) algorithm is used to estimate regression coefficients on this window. The PLS computes the solution via an iterative processing, and thus can be implemented in the flight control computer for a real time use. Two different failure detection strategies based on the behaviour of the regression coefficients are proposed. Simulation results show that the proposed method leads to robust detections.