Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
ANALYSIS AND CLASSIFICATION OF THE CAVITATION LEVEL IN CENTRIFUGAL PUMPS USING ARTIFICIAL NEURAL NETWORKS OF VIBRATION SIGNALS
Submission Author:
Tobias Anderson Guimarães , MG
Co-Authors:
Tobias Anderson Guimarães, Paulo Balduino Flabes Neto, José Gustavo Coelho, Matheus Fraga Teixeira Lara, Túlio Benez Ornellas Graciano
Presenter: Tobias Anderson Guimarães
doi://10.26678/ABCM.COBEM2021.COB2021-0742
Abstract
The cavitation phenomena may produce wear and erosion in the rotor blades from centrifugal pump. Thus, it is interesting to detect the cavitation level in pumps in order to optimize its operation conditions. In this sense, the vibration signal processing measured in the centrifugal pump in operating may be used to identify the cavitation level. In this work, it was used the Power Cepstrum in order to extract the amplitude modulation components generated by the cavitation. The vibration signals were measured in the centrifugal pump by operating in the rotation range of 2500 to 3500 rpm. Subsequently, an artificial neural network was designed and trained to detect the presence of the cavitation in the centrifugal pump. During the learning process from neural network, it was used several samples of the vibration signal have processed by the Power Cepstrum and measured in the directions x, y, z with respect to the pump rotor. The results proved that the Power Cepstrum and neural network may be successfully used in the classification procedure of the cavitation phenomena in centrifugal pumps.
Keywords
Cavitation, centrifugal pumps, neural networks, vibration signal
