Contributions of maintenance policies for centrifugal pumps used in an iron ore concentrate plant

Abstract

Implementing an effective maintenance policy on industrial equipment is crucial to ensuring optimal performance. In particular, maintaining the optimal condition of centrifugal pumps, which play a critical role in various industrial processes such as the transportation of slurries and fluids, is of utmost importance. Hence, this thesis presents two studies related to the maintenance of centrifugal pumps used in the production of iron ore concentrates. The first study proposes a reinforcement learning approach for condition-based maintenance of the pumps, utilizing a variance gamma process degradation model to simulate the degradation process and recommend actions to minimize long-term maintenance costs. The study filled the gap in the literature by considering three performance features, namely pressure, temperature, and vibration, in modelling the state environment. The second study develops an integrated opportunistic maintenance policy for a cold-standby system consisting of a principal set of pumps and a standby set that is activated in the event of failure, utilizing the delay-time concept and defining two maintenance policies for combined optimization. The study’s major contribution to knowledge is in the integrated policy that considers actions for the principal system and for the standby system as well. A sensitivity analysis was performed on the two models. The results obtained justified the robustness of the policies, and demonstrated potential for improving maintenance strategies in the iron ore industry and other industries utilizing similar equipment. The two works therefore provide valuable insights into improving maintenance policies for centrifugal pumps and highlight the importance of developing maintenance strategies at reduced costs.