Application Of Extreme Learning Machine In Aero-engine Gas Path Fault Diagnosis

Aero-engine is the heart of the aircraft, the stability of aero-engine is directly related to the safety of flight. If there is a big fault happen to the aero-engine, it will cause great losses. Aero-engine need regular periodic maintenance before, it will use lots of manpower and material resources. Thus it is necessary to have a fault diagnosis to aero-engine. Found faulty parts in the early stages of the aero-engine failure to have a repair or replacement in time. Gas path component failure is the main fault of aero-engine, we mainly study the aero-engine gas path component failure in this paper.In this paper, we have done the works bellow:1) We have designed the aero-engine gas path component faultiagnosis system and the sensor fault diagnosis system.2) Use the aero-engine gas path single component failure data, combinethe kernel function with extreme learning machine,we proposed aaero-engine fault diagnosis algorithm based on Kernel ExtremeLearning Machine(KELM) after comparing multiple fault diagnosisalgorithm.3) Through the use of aero-engine gas path composite componentsfailure data, add the interference signal and real test data ofaero-engine, it further confirmed the validity of aero-engine faultdiagnosis algorithm based on KELM.4) Simulated the aero engine sensor faults, using algorithm of KPCA +KELM to do fault diagnosis of sensor, applied KLEM to the sensorfault diagnosis.5) Designed a distinction program of sensor failure and aero-engine gaspath fault by KELM.The experiment results show that, applied fault diagnosis algorithm based on the KELM to aero-engine gas path components fault diagnosis and sensor fault diagnosis could have a high fault diagnosis accuracy and the program could run very fast, it can satisfy the requirement of engineering practice very well.