Rough Set and Neural Network Based Fault Diagnosis for Aeroengine Gas Path

Aeroengine is a very complex nonlinear object. Traditional methods for its fault diagnosis are proved time-consuming and low efficient. A new system based on rough sets and neural networks for the fault diagnosis of aeroengine gas path is presented in this paper. At first, the rough set theory is used to determine qualitatively fault and isolate the fault. It consists of three steps: discretizing sensed data, reducing the decision table and generating rules. After that, feed-forward neural networks are added into the system to construct several sub-systems, which take the engine sensible data pretreated by rough sets as inputs and compute damage degrees of the aeroengine fault state. Last, the noise rejection abilities of the engine fault diagnosis system are analyzed. The test results show that the system can quantitatively diagnose the faults of aeroengine gas path with precision and efficiency, while it is robust for noise rejection.