Support vector machine for diagnosis of inductioi motors: A comparative analysis in terms of the quantity and the signal processing tool used to build the feature space

The use of advanced diagnosis techniques for induction motor (IM) faults relies on the use of automated classifiers, such as those based on support vector machines (SVMs), which are able to assess the condition of the machine using a set of relevant features extracted either from the time domain or from the frequency domain machines signals. But the performance of such systems depends on two main factors: the quantity that is used to obtain the machine's condition, and the signal processing tool used for extract the features set. In this paper, a combination of the most used quantities and signal processing tools is used for diagnosis a set of machines with broken bars, fed from the mains and from variable speed drives, using the same SVM. In this way, the most efficient combination can be chosen, from the point of view of the performance of the automatic classifier system.