TAUNO TOIKKA FEASIBILITY OF SELECTED MACHINE LEARNING METH- ODS FOR FAILURE FORECASTING OF AEROPLANE FLIGHT CONTROL SURFACES