Experimental-Modeling Evaluation Between Hydraulic and Electrical Variables Using Copulas and Spectral Analysis for a Centrifugal Pump

Abstract

Centrifugal pumps are turbomachines that have wide industrial applications and could perform in different ways such as pump and turbine mode. The maintenance of this equipment is mostly carried out using invasive methods that are expensive, time-consuming, and even complicated. The application of non-invasive methods is sought since they offer the advantage of real-time monitoring without stopping the process, reducing component assembly and disassembly times and providing a faster response. The aim of this work is done an experimental investigation that shows evidence about how the information on the hydraulic variables can be obtained if the electrical variables are monitored for the modes of operation such as pump and turbine. This work is divided into two parts, the first part is based on a statistical analysis to perform a multivariate adjustment through copulas and probability distributions. The second part focuses on the graphical analysis of the power density spectra for the hydraulic variables, the torque, and the defined electrical variables. The amplitude peaks of each variable and which peaks are common between them are determined. A statistically significant fit for Tawn type 2 copula is obtained with the indicator variable of pressure fluctuation and a multivariate transformation of the three-phase network currents. In the spectra analysis, common amplitude peaks are observed between the spectra that indicate the information flow on the phenomena between the hydraulic variables and the electrical variables.