Centrifugal pump vibration analysis with the aim of early detection of cavitation phenomenon
Subject Areas :
Keywords: Machine Learning, Cavitation, vibrations, pump,
Abstract :
Khalid Naeem Hassan Al-Abidi 1, Ahmad Soheili Mehdizadeh 2*1-2- Department of Mechanical Engineering, Technical and Engineering Faculty, Kermanshah Branch, Islamic Azad University, Kermanshah, Iran* Coresponding author: Soheili2008@gmail.comCavitation is one of the most important issues that, in addition to causing mechanical damage to the pump, causes disturbances in the pumping process, and if it occurs, the pumping process will be stopped. In this research, the cavitation phenomenon was investigated using methods based on vibration analysis. The vibration signals were measured in three sensor installation positions in horizontal, vertical and lateral directions on the pump shaft carrier bearing. Two healthy states and cavitation state were investigated. 50 vibration signals were recorded for each mode. Each signal was processed with the methods of fast Fourier transform (FFT), wavelet transform (WT) and emperical mode decomposition (EMD) and was broken into signals with more detailed information. After each broken signal, 10 statistical characteristics such as mean, standard deviation, skewness, kurtosis, etc. were extracted. And these characteristics were used to create a cavitation prediction model. The results showed that the vibration behavior of the pump changed with the occurrence of cavitation. Also, changing the direction of the sensor caused a change in the amount of recorded vibrations. In general, the analysis of the results showed that the EMD method and the sensor installation position in the vertical direction have the highest accuracy in detecting cavitation. The best artificial neural network model had a sensitivity of 98% in cavitation detection.
_||_