Improved ensemble learning in fault diagnosis system

• Published in: 2009 International Conference on Machine Learning and Cybernetics Vol. 1; pp. 54 - 60
• Main Authors: Chao Ren, Jian-Feng Yan, Zhan-Huai Li
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2009
• Subjects: Aerospace industry; Artificial intelligence; Artificial neural networks; D-S theory; Defense industry; Ensemble learning; Fault diagnosis; Fault diagnosis system; Intelligent networks; Machine learning; Machine learning algorithms; Maintenance; Patter classification; Space technology
• ISBN: 9781424437023; 1424437024
• ISSN: 2160-133X
• DOI: 10.1109/ICMLC.2009.5212527

To improve the performance of diagnosis system, the ensemble learning mechanism using Dempster-Shafer evidence theory (D-S) in pattern classification problem is introduced, which allows multiple diagnosis agents to work together. However, the one-vote veto problem existing in D-S theory affects the performance of the ensemble learning algorithm using D-S theory. To solve this problem, a new improved ensemble learning algorithm is put forth in this paper. Simulations and experiments show that our algorithm holds high performance. The diagnosis system based on the improve ensemble learning algorithm proves effective in an aero-engine automatic diagnosis system.