Applied Predictive Analytics to Evaluate Centrifugal Pumps Reliability
Based on Hydraulic Operating Region
- End-User 2 -
Description:
This study explores the application of predictive analytics and machine learning techniques to evaluate pump reliability in relation to hydraulic operating regions, particularly the influence of operating conditions around the Best Efficiency Point (BEP). By leveraging time-series data from high-energy multistage centrifugal pumps, the research employs classification models, primarily Random Forest and Neural Networks, to systematically assess the relationships among a diverse set of operational variables. These models allow for the identification of patterns in pump behavior and help uncover the key factors that contribute to mechanical degradation, such as vibration in critical components. The integration of machine learning enables a data-driven framework to support early fault detection and reliability-centered maintenance strategies. This work contributes to the growing field of industrial analytics by demonstrating the value of predictive modeling in optimizing pump operation and extending equipment life.
Presenter:
Ernesto Primera
Doctoral Candidate, Tickle College of Engineering - University of Tennessee
Ernesto Primera is a Mechanical and Maintenance Engineer with over 28 years of industry experience, specializing in the reliability of rotating equipment across the oil & gas, power generation, and OEM sectors. His technical background spans maintenance engineering, condition monitoring, performance diagnostics, and failure analysis, firmly rooted in data-driven decision-making and predictive strategies.
Throughout his career, Ernesto has held critical roles with major companies such as Chevron, ConocoPhillips, HF Sinclar, SKF, and Flowserve. For the past decade, he has been a key contributor to the Maintenance & Reliability teams at Chevron's Refineries and HF Sinclair Refineries, where he focused on enhancing equipment reliability through advanced monitoring techniques and structured root cause investigations.
Ernesto’s core expertise lies in the reliability of rotating machinery, including the deployment of wireless condition monitoring systems and the integration of AI and machine learning tools to enable early anomaly detection and proactive maintenance strategies. His innovative approach bridges traditional reliability engineering with cutting-edge digital technologies, helping industrial organizations shift from reactive to predictive maintenance.
In addition to his technical practice, Ernesto is a global instructor and industry consultant for institutions such as the American Society of Mechanical Engineers (ASME), the Hydraulic Institute, ASTM, and SRE. He holds multiple industry-recognized certifications including CMRP, VA CAT-III, CQRM, CSSGB, and GStat, underlining his commitment to continuous learning and professional excellence. Academically, Ernesto earned a bachelor’s in maintenance engineering (Venezuela: Polytechnic University Institute), a Master’s in Predictive Maintenance (Spain: Sevilla University), a Master’s in Reliability & Data Analytics (USA: University of Delaware), and is currently pursuing a Doctorate in Engineering at the University of Tennessee.
A former ASME Chair of the Fluid Engineering Division Technical Committee, Ernesto’s mission is to empower engineers with the tools, knowledge, and technologies necessary to drive higher equipment reliability, operational efficiency, and process safety across complex industrial systems.