Nonlinear Filter and Neural Modeling for Calibration of Aircraft Airdata System

Abstract

Calibration of aircraft airdata system deals with the reconstruction of flight path trajectories from the noisy flight data using six-degree-of-freedom equations of aircraft. Aircraft system dynamics are highly nonlinear in rapid variations of the aircraft motion and require the use of a nonlinear filtering algorithm. In this paper, a methodology based on data-driven decision making is introduced to obtain the accurate values of aircraft flow angles (angle of attack and angle of sideslip) and static pressure from its noisy measurements. For this, the integration of fault detection and isolation approach to the adaptive nonlinear filter is applied to dynamic maneuvers, and a neural model of calibration function is established over a flight envelope using the filter estimates. A deterministic airdata calibration function is derived by estimating its coefficients from the established neural model using the neural partial differentiation method. The cascading impact of adaptive estimation and neural modeling of airdata calibration function reduces the development cost of an aircraft. The investigations are initially made on simulated flight data under various conditions of wind and turbulence and later extended to the flight data of aircraft to identify the calibration function valid over a flight envelope. The complementary flight data are used to validate the calibration function and are compared with online estimation results of the robust filter. The experimental results show that the proposed algorithm can isolate and rectify the fault and exhibits more accurate estimates directly with neural modeling than the nonadaptive version of the filter.