An alternative neural airspeed computation method for aircrafts

Abstract

Purpose This paper aims to present an alternative airspeed computation method based on artificial neural networks (ANN) without requiring pitot-static system measurements. Design/methodology/approach The data set used to train proposed neural model is obtained from the Digital Flight Data Acquisition Unit records of a Boeing 737 type commercial aircraft for real flight routes. The proposed method uses the flight parameters as inputs of the ANN. The Levenberg–Marquardt training algorithm was used to train the neural model. Findings The predicted airspeed values obtained with ANN are in good agreement with the measured airspeed values. The proposed neural model can be used as an alternative airspeed computation method. Practical implications The proposed alternative airspeed computation method can be used when the air data computer or pitot-static system has failed. Originality/value The proposed method uses flight parameters as inputs for the ANN. As such, airspeed is calculated using flight parameters instead of the pitot-static system measurements.