A Comparative study between Support Vector Machine (SVM) and Extreme Learning Machine (ELM) for Fault Detection in Pumps

Background/Objectives: To apply Support Vector Machine and Extreme Learning Machine for fault diagnosis of centrifugal pump and to compare them with respect to classification accuracy and learning time. The overall best one is reported.

Methods/Statistical Analysis: The vibration signals are extracted from the experimental setup. Then signals are then filtered and trimmed for ease of processing. The wavelet features are extracted in its discrete form thus it forms the feature set. The set of features are then fed as an input and classified using Extreme Learning Machine and Support Vector Machine. These two algorithms are state of the art techniques for classification of different conditions of the setup. The results are compared with respect to classification accuracy and learning rate. Finally, it is concluded that the ELM could achieve 99.2% classification accuracy at a very faster rate than SVM.

Application/Improvements: Extreme Learning Machine has been used for the first time for fault diagnosis applications. Discrete wavelet transform features in combination with ELM have been attempted for the first time. To conclude, ELM could act as a better alternate for machine learning approach.