Engine health monitoring in an aircraft by using Levenberg-Marquardt Feedforward Neural Network and Radial Basis Function Network

In maintenance of aircrafts, engine health monitoring (EHM) method has crucial importance. If an aircraft engine is in good condition, aircraft accidents could be prevented. On the other hand, healthy engines enable reduced fuel consumption and maintenance costs. Most of the EHM methods track the exshaust gas temperature (EGT) value in evaluating the health level of an engine. In this study, both Levenberg-Marquardt Feedforward Neural Network and Radial Basis Function Network methods are used to estimate the EGT values. 15 inputs are utilized as aircraft flight performance parameters for estimating the output (EGT parameter) in the two models. Obtained results show us Levenberg-Marquardt Feedforward Neural Network is more efficient compared to Radial Basis Function Network in estimating the EGT. Using this method, engine deteriorations could be caught by pilots without any expert knowledge before more serious damages in aircraft engines.

I. Introduction

Aviation industry has a fast technological development nowadays. Aircrafts were known as “flying coffin” during Second World War, while they are called engineering marvels today. Up to date, many problems have been solved but there are still issues that need to be solved for maintaining the flight safety. Problems noticed by pilots during the flights are noted down to the aircraft loogbook; so that mechanics can check the loogbook and have an idea on the flight history of the aircraft. In this study, we propose a new fault predictive system to enhance flight safety. The proposed system also enables time saving and reduced maintenance costs.