Abstract

In this thesis a new approach for aircraft jet engine condition monitoring
is proposed based on system identification and by using Genetic Programming (GP). This approach consists of two fault detection and isolation parts. In the detection part, the relationship between the engine Exhaust Gas Temperature (EGT), as a major indicator of the engine health condition, and other engine parameters and operating conditions corresponding to different phases of the flight is modelled using the GP technique. Towards this end, flight characteristics are divided into several phases such as the take-off and the cruise. The GP scheme is then used to discover the structure of the interrelations among engine variables. The constructed model is then used to detect abrupt faults in the engine performance.

For the isolation purpose, a hierarchical approach is proposed which narrows down the number of possible faults toward the target fault. The GP algorithm is then exploited to extract a series of nonlinear functions of the engine variables called fault indices. These indices attempt to magnify the signature of a fault in the engine by combining the effects of a fault on the engine parameters. These indices subsequently provide the necessary residuals for classifying the faults.

The approaches developed in this thesis provide an effective strategy for inspecting the aircraft jet engine health condition without requiring any specific information on the engine internal characteristics. The main advantage of the proposed approaches over other data driven methods such as neural networks is that our approaches provide a simple and tangible mathematical model of the engine rather than a black box model. The performance of the proposed algorithms are demonstrated and illustrated by implementing them on a double spool jet engine data that is generated by using the Gas turbine Simulation Program (GSP) software.