|
The main objective of this study was to redevelop a method that allows the determination of the landing and takeoff distance of an aircraft using an on-board data processing unit. The developed method is based on the indications of the inertial measurement unit (acceleration in 3-axis and Euler angles), which are processed by a developed architecture artificial neural network. The model is based on a segmentation task to indicate the position in the multivariate time series where the change of state occurred, that is, the transition from flare to ground roll (the moment of touchdown), and from ground roll to climb (the takeoff moment). The indications are correlated with data from the Global Positioning System module to determine the geographic location of both the position of the state change and where the aircraft remained stationary. Both geographic locations allow the determination of the length of the landing distance and the takeoff distance. The model, along with auxiliary algorithms designed to determine if the aircraft is stationary, allows the model to infer only at times when required, which translates into reduced energy requirements.
The developed method was compared with a reference method specially developed for these purposes, which is also based on determining the position of a change of state in a multivariate time series. The reference method is based on the indications of an observer. The two methods were compared with each other, and the obtained differences between the indications of the model and the reference method for the length of the touchdown and takeoff distance were 1.7725% and 2.7712%, respectively. The tests were conducted on one type of aircraft under varying weather conditions.
|
