Kilic, Ugur

Airspeed data is so important for an aircraft operation. This study is focused on the estimation of the airspeed data without any additional measurement source such as hardware redundancy. The flight data obtained from a commercial aircraft is processed with a deep learning algorithm, particularly LSTM recurrent neural networks that are developed based on Matrix Laboratory (MATLAB). Correlation analysis is carried out for related data according to a 95% confidence interval for each coefficient in the study to show strong predictor candidates. Data related to the airspeed are processed using Holdout Cross-Validation Technique. According to the results, the designed model achieved an R-squared of 0.9999, a root-mean-squared error of 0.8303 knots, and a standard error of 0.0092 knots. These results show that it is possible to accurately estimate aircraft airspeed data using LSTM recurrent neural network in case of the airspeed data cannot be provided to the flight crew.

Kurt, Bülent; Yıldırım, Mustagime Tülin

Faults in aircraft performance can be identified by the experts via analysis of the recorded flight information in today`s aircraft technology. The parameters used in identification of the faults include exhaust gas temperature, fuel flow, engine fan speed, vibration, oil pressure and oil temperature. In this study, a model that predicts the vibration parameters of the low pressure turbine using real time data of a Boeing 737-500 is developed. Using the developed model, it is aimed to determine a possible deterioration in performance by predicting vibration parameters of low pressure turbine and allowed vibration limits. Multiple regression analysis technique was used in the developed model. In our study, very highly significant relationships between vibration parameters of the low-pressure turbine and air speed, thrust lever angle right, N2 speed left and exhaust gas temperature were explored.