System health management for safe automatic take-off

This paper presents a model-based monitor architecture for automatic aircraft take-off. Only a limited runway length is available for aircraft to become airborne, or to reject the take-off and come to a standstill. Automatic take-off systems, which emerge in unmanned applications, require monitoring to ensure safety during this critical flight phase. We propose a model-based monitor architecture, which continuously predicts probability distributions of take-off and stopping distances. This is accomplished with particle filter and Monte-Carlo based algorithms. In this paper, we use flight test data to show that the performance of an automatically controlled take-off has better repeatability than when flown manually. This enables a prediction of the complete take-off distance until the aircraft reaches the screen height. We evaluate the prediction accuracy of a proposed model with low computational requirements for a nominal take-off with flight test data. The simple model is able to predict take-off distances with satisfactory accuracy before reaching rotation speed.