BSP-BDT classification technique: Application to rolling elements bearing

To maintain an efficient operating unit and avoid failure of mineral critical equipment, fault detection and diagnosis are the solution for the critical parts of these equipments. This paper presents a non parametric classification technique using Best Selective Parameters (BSP) embedded in Binary Decision Tree (BDT). The method arises from the question: how can we choose suitable types of parameters in order to achieve accurate classification? DSP-BDT technique answers this question and provides the solution. BSP-BDT method is based on multi-class Support Vector Machines (SVM). BSP-BDT takes advantage of both: the efficient computation of tree by selecting appropriate parameters and the high classification accuracy of SVM. BSP-BDT method is applied on rolling elements bearing not only to detect and diagnose faults (inner race, ball, outer race) but also to identify the magnitudes of these faults. Results are compared to Artificial Neural Network (ANN) technique to improve the DSP-BDT classification precision.