Authors: V. Muralidharan; V. Sugumaran
Addresses: Department of Mechatronics Engineering, School of Mechanical Engineering, SRM University, Kattankulathur – 603203, Kanchipuram Dt, Tamil Nadu, India ' School of Mechanical and Building Sciences, VIT University, Chennai Campus, Chennai – 600 048, Tamil Nadu, India
Abstract: Centrifugal pumps play a significant role in many critical engineering applications. Continuous monitoring of such machine components becomes essential. Continuous monitoring means to monitor the condition of the machine whether the machine is in good condition or affected by some faults. This fault diagnosis problem is conceived as a pattern recognition problem. Generally, pattern recognition problems are approached by following three steps such as feature extraction, feature selection and feature classification. In this paper, fault diagnosis of monoblock centrifugal pump is carried out using vibration signals. Among the number of available feature extraction techniques, wavelet features are found to be good and encouraging for such critical applications and hence it is chosen. The extracted features from the vibration signal are given as input to the decision tree to frame a set of rules and to feed them as an input to the fuzzy classifier. The fuzzy classifier is built and tested with the representative data. The results show that it is good for real time applications.
Keywords: mono-block centrifugal pumps; decision tree; fault diagnosis; discrete wavelet transforms; DWT; fuzzy logic; fuzzy classifiers; pattern recognition; vibration signals; condition monitoring.
International Journal of Computational Systems Engineering, 2013 Vol.1 No.3, pp.175 - 183
Published online: 26 Jul 2014 *Full-text access for editors Access for subscribers Purchase this article Comment on this article