Centrifugal pumps (CPs) fail due to anomalies in fluid flow patterns and/or due to failure of mechanical subsystems in them. In this work, a technique built on the multiclass support vector machine (MSVM) is developed to identify multiple faults in the CP. In addition, the complex problem of fault combinations and their classification is dealt with in this work. The combination of features from motor line current sensors and accelerometers is used to train the algorithm. To take into account the transient as well as harmonic components of fault signatures, continuous wavelet transform (CWT) analysis is used. Thereafter, the most important information from the CWT coefficients is selected using the two proposed novel methods CWT-based on energy (BE)-MSVM and CWT-principal component analysis (PCA)-MSVM, which are BE as well as PCA, respectively. It is experimentally observed that faults in the CPs have a very strong association with its operating speed. Thus, in order to make the CP versatile in operation, it is important that the fault diagnosis methodology is also efficient at large speed range of CP operation. This work attempts to develop a fault classification methodology, which is independent of the CP operating speed.