Vibration fault diagnosis of hydro-turbine generating unit based on rough 1-v-1 multiclass support vector machine

The traditional vibrant fault diagnosis classifier of hydro-turbine generating unit (HGU) can´t reflect the uncertain information in fault pattern recognition. To overcome the above problem a novel classifier based on rough set (RS) and 1-v-1 multiclass support vector machine (SVM) is introduced. In this method, the basic ideas of RS: upper approximation, lower approximation and boundary region are used to describe the positive region, negative region and margin of SVM. By using 1-v-1 method, the multiclass classification of SVM is realized. Then the description of upper approximation, lower approximation and boundary region of multiclass are determined. At last, the rules of classifier are acquired. The results show that the proposed classifier has high classification reliability, more concise rule, and lower requirement of memory space in operation stage, and can reflect the uncertain information of fault diagnosis.