Dissertation > Aviation, aerospace > Aviation > Aeronautical Manufacturing Technology > Aero-engine manufacturing > Fault Analysis and

Applying the Artificial Neural Network to the Failure-diagnosis of Aero Engines

Author LiuYuJie
Tutor HeZuoLian
School Tianjin University
Course Applied Computer Technology
Keywords Artificial Neural Networks BP algorithm Aircraft engines Fault Diagnosis
CLC V263.6
Type Master's thesis
Year 2003
Downloads 402
Quotes 3
Download Dissertation

The issue is the application of neural network technology for the first time in our aircraft engine fault detection . This study combines the experience and knowledge of the some aircraft experts in the the maintenance of Boeing 747 engine failure over time to adapt to the needs of China's civil aviation modernization and teaching modern , and combined with the contents of the the engine fault isolation manual . Its main purpose is to help the crew , and especially help maintenance personnel in the repair , quickly , accurately locate and troubleshoot , improve work efficiency to achieve better economic efficiency and higher safety factor for the civil aviation industry . This study also can be applied to teaching , and help students learn aircraft maintenance , more convenient access to reliable and detailed knowledge to prepare as soon as possible to become experienced crewmen . In-depth study of artificial neural network (ANN) , especially in an attempt to improve the algorithm has done a lot , data collected from the full aircraft engine fault information , combined with the experience of experts , apply to civil aircraft engine , especially Boeing 767 and Boeing 747 assembly the JT9D engine and PW4000 engine fault diagnosis system . Disintegration of the situation so that the engine is not detected early hidden faults . The system uses a three-layer neural network structure , four input nodes , the intermediate layer 10 , corresponding to the output layer fault model , 12 . After several experiments, algorithms proposed an efficient modified FBP algorithm using gradient descent method iterative method constantly revised weights , network training, the number of iterations of 100,000 times , learning accuracy requirements to take the system total error E = 0.001. The system diagnostics success rate hit 91.6% , the system will eventually reach the design purpose .

Related Dissertations
More Dissertations