Flight data validation of an icing accretion estimation scheme using super-twisting observers

This paper develops a generalised multivariable super-twisting observer for a class of nonlinear systems in which the unmeasured variables linked to the known state dependent matrix function appear multiplicatively. A sufficient condition is given to guarantee that the reconstruction errors associated with the unmeasurable variables converge to zero in finite time. This approach is then used to address the aircraft icing accretion estimation problem despite unreliable sensor measurement. The efficacy of the approach has been evaluated via real flight data recorded under natural icing conditions. Results show that the observer has the capability to estimate the change of the drag coefficient induced by icing accretion and to reconstruct the unreliable pitch rate sensor measurement simultaneously.