Fault Diagnosis by Using Selective Ensemble Learning Based on Mutual Information

Fault diagnosis on diesel engine is a difficult problem due to the complex structure of the engines and the presence of multi-excite sources. There have been previous attempts to solve this problem by using artificial neural networks and others methods. In this paper, a novel algorithm named MISEN (Mutual Information based Selective Ensemble) is proposed to improve diagnosis accuracy and efficiency. MISEN is compared with the general case of bagging and GASEN, a baseline method, namely Genetic Algorithm Based Selective ENsemble, on UCI data sets. Then, MISEN is used to diagnose the diesel engine. Computational results show that MISEN obtains higher accuracy than other several methods like bagging of neural networks and GASEN.