Ensemble feature selection for diagnosing multiple faults in rotating machinery

Recent research in fault classification has shown that one of the benefits of using ensembles of classifiers is that they achieve higher accuracy than single ones. For an ensemble to be effective, it should consist of base classifiers that give diverse predictions. One method for constructing an ensemble is to have the base classifiers work on different feature sets. In the current paper, the problem of selecting the feature sets for the base classifiers is handled by means of genetic algorithms aimed at maximizing the fault classification performance and at minimizing the number of features. A voting technique is then used to combine effectively the outputs of the base classifiers to construct the ensemble output. For verification, some numerical experiments are conducted on multiple fault classification in rotating machinery and the results are compared with those obtained using an optimal single classifier.