Fault Diagnosis of a Centrifugal Pump Using Electrical Signature Analysis and Support Vector Machine

Purpose Early detection of impending faults in centrifugal pumps could lower the repair and maintenance costs. Electrical Signature Analysis (ESA) is a powerful tool which uses the voltage and current signals of the motor driving the pump to infer the health status. In this study, we strive to develop an ESA-based algorithm to perform fault diagnosis for a centrifugal pump. Methods Support vector machine (SVM), derived from the statistical learning theory (SLT), is a relatively new technique based on the structural risk minimization principle. This paper presents the application of SVM algorithm in junction with ESA for fault diagnosis of a motor-driven centrifugal pump. To do this, SVM is formulated as a multi-class classification problem and suitable features from electrical signals are devised as the inputs to the SVM classifier. Results Experimental results show the effectiveness of the proposed method in detecting and diagnosing main types of pump faults with high accuracy. Conclusion Fault diagnosis of motor-driven centrifugal pumps can be carried out using only the electrical signals of the motor, i.e. voltage and current. This alleviates the need to install accelerometers for vibration sensing and allows for continuous monitoring of the target equipment.