Indonesia Conference Directory

The cavitation phenomena increase noise and vibration level in a centrifugal pump, which, if not properly maintain leads to catastrophic failure and total stop for the whole production process. It is important to develop a method that can detect cavitation as early as possible. Support vector machine (SVM) is one of pattern recognition techniques which requires statistical features as input for classifier modelling. However, the selection of statistical features is arbitrary, hence further investigation is needed. In this study ten statistical features are extracted from time domain vibration signal and selected using Relief Feature Selection. The vibration signal is taken from cavitation test-rig under four different pump conditions i.e., normal condition, early, medium and full cavitation. The selected features are used as input for two types of SVM, binary and multi class, to classify the new vibration data. Feature selection process reveals that variance, RMS, and SD are the best feature to use for SVM classification. The binary SVM method shows the best plot on early cavitation with accuracy 99% where Bayesian Optimization algorithm with multi class SVM is the best combination to classify all pump conditions with overall accuracy 100%.