Investigating pump cavitation based on audio sound signature recognition using artificial neural network. (July 2020)

• Record Type: Journal Article
• Authors: Arendra, Anis; Akhmad, Sabarudin; Winarso, Kukuh; Herianto,
• Is Part Of: Journal of physics. Volume 1569:Number 3(2020)
• Publication Date: 2020-07
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1569/3/032044
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: How to investigate the occurrence of cavitation in the pump? Several studies have shown the sound characteristic that occurs during cavitation. This research attemps to build a pump cavitation detection system based on the audio signal of the operating pump. Audio signal is recorded using a microphone through a computer sound card. Then perform the frequency domain feature extraction and the correlation analysis for feature selection. From this process, 9 frequency domain features are selected as the artificial neural network classifier input. This artificial neural network classifier is trained with the Resilient backprogation algorithm The performance of this detection system is able to determine the existence of cavitation with an accuracy rate of 82.5%.