• angle of attack
• angle of sideslip
• virtual sensor

The aim of this thesis is to design a Virtual Sensor to estimate the aerodynamic angles of an aircraft (that is, the angle of attack and the angle of sideslip). The Virtual Sensor is a software module that reconstructs the signal of interest from other measurements. Virtual Sensors can be used to achieve the required reliability of an Air Data System, introducing analytical redundancies instead of hardware redundancies. This leads to savings in weight and space.
The thesis tries to find an algorithm, that is able to reconstruct the aerodynamic angles from the aerodynamic forces acting on the aircraft. Such forces are estimated from the measures of the Inertial Measurement Unit.
The key idea is to exploit the relationship between aerodynamic angles and inertial measurement.
Three different algorithms are evaluated: Stepwise regression, Neural Network and Genetic Algorithm.
The algorithms developed in this work try to be simple and fully meet the requirements for the reconstruction of aerodynamic angles.
For Stepwise regression and the Genetic Algorithm, a linear connection is sought between the regressors and the outputs, significantly reducing the memory requirement.
The Neural Network, on the other hand, is an attempt to move away from the linear regression to evaluate a possible solution that further improves the previously developed results.
These algorithms are developed and tuned using data measured during the flight test of the Aermacchi M-346 MASTER.