A Model of Fuel Consumption Estimation and Abnormality Detection based on Airplane Flight Data Analysis

Due to the importance of fuel consumption to cost reduction of air transport, the major airlines in the world pay close attention to research on fuel saving. How to improve the ability of fuel consumption estimation and abnormality detection of fuel system becomes an important topic of airlines. To enhance the ability of fuel consumption estimation and abnormality detection, this paper analyzes the flight data of fuel system and establishes a model of fuel consumption estimation and abnormality detection, which can upgrade the precision of flight plan's fuel estimation and abnormality detection of fuel system to reduce the fuel consumption cost and find faults or faults once them occur in the fuel system. The fuel estimation and abnormality detection model is closely related to airplane performance, engine performance, flight track and meteorology. Most traditional fuel consumption models are based on energy balance principle, which need to inquire performance graphs. However, it is difficult to obtain the performance data from performance graphs. To deal with this problem, this paper builds an airplane fuel estimation model by training a BP neural network based on the airplane flight data. What's more, using the abnormal data when faults or faults occur in the fuel system, the model can detect the abnormal condition of fuel system and warn the pilots. Because it is not enough accurate to estimate fuel consumption of the entire flight course once, based on flight data, the flight course is divided automatically into five routes by the model in this paper, which is takeoff, climb, cruise and decline. Models of different routes are trained in different BP neural networks to estimate the fuel consumption of each route. Adding all the fuel consumption of different routes, the fuel consumption of entire flight course will be calculated. Considering the factors of meteorology, the speed and acceleration of airplane, including (but not limited to), wind direction, wind speed, tilt, longitudinal acceleration and transverse acceleration, and so on, are taken as inputs when the model is established. These factors make the fuel consumption estimation much closer to the real flight fuel consumption data and abnormal condition be easily detected. The experiment indicates that the model of airplane fuel consumption estimation and abnormality detection which is proposed performs with high precision.