Method Study On Ensemble Learning Based For Fault Diagnosis

Machine learning is the most intelligent artificial intelligence features and one of theforefront research areas. Support Vector Machine (SVM) is a very typical machine learningmethod, it can well solve many small sample, nonlinear, high dimension and can overcomeother problems that neural network can not. It has good performance in mall sample problemsand has strong generalization ability. But the good performance of SVM in sample problems ishard to play in massive sample data. Therefore, Ensemble learning give a new route for us.The purpose of this paper is to study fault diagnosis method on ensemble learning, usingsupport vector machine (SVM) as a learner, through the ensemble of multiple learners to solvethe problem of mechanical fault diagnosis. The paper is analyze the ensemble learning and itsrelated algorithm, and studied the composition of support vector machine learner, using theAdaBoost algorithm constructed the method of support vector machine on ensemble learning,,and applies this method to the mechanical fault diagnosis. The paper is aim at two differentmechanical equipment of the diesel engine and gear boxes, monitoring vibration signals underdifferent running condition, applies different eigenvalue extraction (dynamic index method,the wavelet packet energy method) method, then to verify the fault diagnosis method onensemble learning. The experimental validation results indicate the proposed method isfeasible, and confirm the imagination that the classification result of strong classifier onensemble learning is better than single classifier, and meanwhile the results shows thatensemble learning method can improved the generalization ability of the system significantly.The paper made some stretch on ensemble learning, using the experimental confirm theimagination about the "selective integration", that means the select part learner to ensembemay be better than all.