1. Identification of faults for centrifugal pump using Wavelet analysis with machine learning classifiers.
- Author
-
Shah, Mit, Bhatt, Pina, Bhavsar, Keval, and Parmar, Umang
- Subjects
- *
ROLLER bearings , *CENTRIFUGAL pumps , *WAVELETS (Mathematics) , *SYSTEM downtime , *DISCRETE wavelet transforms , *MACHINE learning , *WAVELET transforms - Abstract
Rotary machines is a very significant part of the industry and one of the majorly used mechanical device is the centrifugal pump. Bearings are the key element in the pumps and hence, early identification of the defects of bearing can help in lowering the maintenance cost as well as downtime. In the field of machine fault detection, time and frequency-based methods are well known for their capability to handle complex signals. In this research rolling element bearing with defects in the roller, outer race, and inner race, and defects in a impeller of a pump are considered that are compared against the healthy condition of the pump. Maximum Overlap Discrete Wavelet Transform (MODWT) and Wavelet Packet Transform (WPT) are applied as signal processing techniques for the dataset. Generated final results are then taken to the mother wavelet selection and both techniques are compared for the same. After extraction of coefficients, Statistical time domain features are calculated from it and fed to Support Vector Machine, Ensemble Bagging Tree and K-Nearest Neighbour algorithms for training. A comparison of all these three algorithms is presented in this study and a conclusion has been made based on the 10-fold cross-validation. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF