The introduction of Fly-By-Wire (FBW) and the increasing level of automation contribute to improve the safety of civil aircraft significantly. These technological steps permit the development of advanced capabilities for detecting, protecting and optimizing A/C guidance and control. Accordingly, this higher complexity requires extending the availability of aircraft states, some flight parameters becoming key parameters to ensure a good behaviour of the flight control systems. Consequently, the monitoring and consolidation of these signals appear as major issues to achieve the expected autonomy.Two different alternatives occur to get this result. The usual solution consists in introducing many functionally redundant elements (sensors) to enlarge the way the key parameters are measured. This solution corresponds to the classical hardware redundancy, but penalizes the overall system performance in terms of weight, power consumption, space requirements, and extra maintenance needs. The second alternative consists in estimating the parameters thanks to signal processing or model-based techniques. The only drawbacks concern the implementation on aboard computers and especially the computational time. This solution relies on analytical redundancy and resorts to the notion of virtual sensor.Within the framework of model-based approaches, this paper describes an estimation scheme involving a nonlinear Kalman Filtering (Extended KF) to estimate some of the longitudinal flight parameters of a civil aircraft during usual flight conditions. In order to ease onboard implementation, the main aerodynamic components of the A/C modelling are approximated by a set of grey-box neural networks. The method used to accurately and automatically derive these surrogate models is explained, and some details are given to illustrate its performance when dealing with look-up tabulated coefficients issued from wind tunnel or flight tests.On the other part, trends are explored to define a self-adaptive strategy for switching between different EKF tunings when some measurements become unavailable after sensor failures. They exploit statistical tests on the residual signals provided by the filter. Besides, some simulation results are displayed to evaluate the performances of this approach in different realistic flight conditions, including external disturbances (noises and wind) and robustness issues (modelling errors). They correspond to the flight mechanics of a generic commercial aircraft and are simulated by means of Airbus simulation environment.
On-line Estimation of Longitudinal Flight Parameters
Sae Technical Papers
Aerospace Technology Conference and Exposition ; 2011
2011-10-18
Aufsatz (Konferenz)
Englisch
Estimation of Longitudinal Parameters from Real-Flight Data Using Neural Networks
Online Contents | 2002
|Wiley | 2024
|