Applying The Artificial Neural Network To The Failure-diagnosis Of Aero Engines

This research is the first time to use the ANN technical in the failure-diagnostic of aero engines in China. To meet the needs of the developing civil aviation and the education modernizing of our country, this research combines lots of experience of the experts of maintenance of airplane with the contents of the broke-down isolated manual. The main goal of the research is to assist aircraft crew, especially the maintenance men, to eliminate failures rapidly and accurately when they maintain the airplane, to enhance the work efficiency and gain more economic benefits and higher safety quotiety. While this system also can apply to teaching, to help the students to get more credible and detailed knowledge when learning the maintenance of airplane, to help them prepare to be seasoned aircraft crews. Through a deep exploration of artificial neural network, including experiments concerning algorithms-improvement in particular, we collect sufficient information on aero-engine breakdown, combining it with data provided by experienced specialists, then we develop this failure-diagnostic system which apply to the JT9D and PW4000 engines of Boeing767 and Boeing747. This system can detect the possible failure in the first time without dismantling the engines. This system adopts a three-layer net structure, and an improved BP algorithm. Using gradient iteration Method, constantly correct the authority, after iteration 100 thousand times, the precision of study (the total error of this system) is E=0.001. The successful rate is as high as 91.6%, so the system achieves the aim of design.