Optimal Hydrocarbon Piping Replacement Decisions Using the Markov Decision Process

Abstract

Piping is an essential component of processing plants that transfers process fluids from one piece of equipment to another. To detect the unpredictable deterioration of piping systems, periodic or continuous pipe inspection for wall loss is conducted, which determines when preventative or corrective maintenance activities, usually replacements, should be done. Therefore, it is necessary to develop an optimal maintenance strategy to replace piping systems to reduce the overall maintenance expenditure and avert the effects of pipe failure. Most of the literature in this field concentrates on maintenance activities executed during decision-making. This paper investigates creating a system and tool to assist decision-makers in evaluating and selecting the optimal moment to replace a deteriorating pipe based on the current state of the piping system. Considering that preventative replacement of the pipeline is done during a scheduled shutdown, also referred to as a turnaround, this paper investigates making a preventative replacement ruling two turnaround cycles ahead. We account for the varying thickness at different monitoring locations of two piping systems in a refinery by applying a nonstationary gamma process with random effects. We employ the Markov decision process (MDP) and the value iteration algorithm to determine the most effective maintenance replacement policy. The outcome of our analysis is an optimal maintenance policy that minimizes the cost of maintenance. This paper’s practical use is to aid in maintenance planning in refineries, midstream or pipelines, chemical plants, and gas facilities by planning piping replacement in two shut-down cycles ahead of time. The pipe wall is classified into states ranging from the present thickness reading to the retirement thickness. The likelihood of transiting from one state to another is calculated based on the corrosion rate. Given that the pipe wall thickness is classified into states, the current wall thickness, transitional probabilities, and preventative and corrective maintenance costs, including lost profit opportunity, can be inputted into the model to help decision-makers decide if it is optimal to replace the piping in the subsequent two shut-down cycles. This policy will assist in reducing the risk of a pipe failure, which has financial, safety, and environmental consequences. It is also critical not to replace a still-in-good-shape pipe. This model may be customized to pipe systems in the various petrochemical sectors with known retirement thicknesses to estimate the risk of operating near the retirement thickness.