Fault Diagnosis of Gearbox by Selective Ensemble Learning Based on Artificial Immune Algorithm

Because faults of gearbox are one of the most important factors of wind turbine, it is very important to research gearbox fault diagnosis technology. The problem of class unbalance is encountered in fault diagnosis of gearbox, which has a serious negative impact on the performance of the classifier of the hypothetical class. While it is critical that previous work rarely pays attention to the problem of gearbox fault diagnosis in this class imbalance. In the problem of imbalance, some features of dataset are superfluous, even irrelevant. These characteristics will affect the generalization performance of the learning machine. Here, it is propose AIEE (based on artificial immune algorithm function selection Easy Ensemble) to improve the gearbox fault diagnosis class imbalance problem. The experimental results of UCI datasets and gearbox fault data sets shows that AIEE algorithm can improve the classification performance and prediction accuracy of unbalanced datasets.