Adopting Support Vector Machine Algorithm for Fault Diagnosis in Polymer Impeller Centrifugal Pump

With rapid advancement and application of polymer materials in current pumping industry, it is important to continuously monitor the health of the system. In this research work, an experimental study is conducted on a centrifugal pump with semi-open polymer impeller installed on the rotor shaft of the pump. Pump data with healthy and faulty features are extracted using vibration and temperature signals from the pump. Correlation matrix shows a good correlation among selected variables and classifier SVM with three kernel functions viz. RBF, linear and polynomial are used on extracted data for diagnosing fault accuracy. Result depicts high accuracy of classification of faults by polynomial kernel function as compare to RBF and linear kernel function.