Weight and longitudinal center-of-gravity position of an aircraft are valuable information for flight monitoring, automatic flight control, and fault detection and identification algorithms. In trimmed symmetric flight (as cruise, climb or descent), the aircraft weight is counteracted mainly by the aerodynamic lift, whereas the associated couple is counterbalanced by a moment generated by an appropriate elevator deflection. Since, for given flight conditions, the aerodynamic lift and pitching moment are functions of the angle of attack and elevator deflection only, the last two variables carry the information about the aircraft weight and its longitudinal center-of-gravity position. It is demonstrated that this information can be effectively recovered in flight using non-parametric artificial neural networks (NNs), pre-trained on flight test data. To enhance the accuracy and speed-up the training process, basic flight mechanics relations are incorporated into the NN. It is also demonstrated that the NN-based lift and pitching moment functions can be used a posteriori to estimate the aircraft longitudinal neutral point.
[1]
David J. Haas,et al.
Feasibility of Aircraft Gross Weight Estimation Using Artificial Neural Networks
,
2001
.
[2]
T. Teichmann,et al.
Dynamics of Flight: Stability and Control
,
1959
.
[3]
Robert J. Mack.
A Rapid Empirical Method for Estimating the Gross Takeoff Weight of a High Speed Civil Transport
,
1999
.
[4]
Ken-ichi Funahashi,et al.
On the approximate realization of continuous mappings by neural networks
,
1989,
Neural Networks.
[5]
George Cybenko,et al.
Approximation by superpositions of a sigmoidal function
,
1992,
Math. Control. Signals Syst..
[6]
Daniel P. Raymer,et al.
Aircraft Design: A Conceptual Approach
,
1989
.
[7]
Kurt Hornik,et al.
Multilayer feedforward networks are universal approximators
,
1989,
Neural Networks.