Altitude Estimation From Flight Data Recorder By Using Artificial Neural Network

Abstract

Aircrafts fly safely at high altitudes with the help of certain proven reliable systems. Among the basic instruments, the altimeter is one of the most essential elements to maintain a safe altitude. To obtain altitude information in aircraft, barometric pressure data from static ports is calculated with certain correction parameters in the air data computer and then sent to the pilots and the flight management computer. The purpose of using correction parameters is to enable the air data computer to calculate the updated parameters in order to determine the aircraft’s current altitude accurately, accounting for unpredictable factors such as the Earth’s geographical effects, day-to-day temperature variations, and variable weather resistance. The necessity of keeping altimeter calculation data up-to-date has prompted the scientific community to explore alternative calculation methods. The aim of this study is to create a primary parameter using acceleration data The second parameter is the time domain. Acceleration and time data were defined as input data in artificial neural networks. In this study, the proposed method uses a feedforward backpropagation algorithm within a multilayer artificial neural network architecture (FFBP). In each flight phase, learning algorithms were applied to each network architecture. Levenberg-Marquardt (LM) algorithm provided more successful results compared to other algorithms. In the proposed method, a safe alternative calculation is developed using artificial neural networks with data obtained from flight data recorders, without the need for correction parameters applied in air data computers, and resilient against issues such as the exposure and degradation of static systems due to disruptive atmospheric factors.