A Cost-Efficient MFCC-Based Fault Detection and Isolation Technology for Electromagnetic Pumps

Fluid pumps serve critical purposes in hydraulic systems so their failure affects productivity, profitability, safety, etc. The need for proper condition monitoring and health assessment of these cannot be overemphasized this has resulted extensive research studies on standard techniques ensuring optimum fault detection isolation (FDI) results pumps. Interestingly, mechanical vibrational signals reflect operating conditions by exploring the robust time–frequency-domain feature extraction techniques, underlying nonlinear characteristics can captured reliable diagnosis/condition assessment. This study is based use electromagnetic From signals, Mel frequency cepstral coefficients (MFCCs), first-order second-order differentials were extracted salient features selected a rank-based recursive elimination (RFE) uncorrelated features. proposed framework was tested validated five VSC63A5 at various isolated/classified using Gaussian kernel SVM (SVM-RBF-RFE). Results show that selection approach computationally cheaper significantly improves diagnostics performance. In addition, yields comparatively better comparison with other methods, hence more tool