Residual Life Estimation of Corroded Critical Piping Using Probabilistic and Statistical Methods

Oil and gas piping systems are subject to corrosion that can reduce wall thickness. Given the importance of the piping system for connecting components within a plant or reservoir and transporting fluids, the failure of such a sensitive system will result in a high-hazard consequence. Hence, adequate inspection, repair, and maintenance of the piping system are required. However, not all inspection, repair, and maintenance processes are efficient and cost-effective; therefore, it would be helpful to estimate the residual life of the piping systems using several methods and compare the results to assess each technique's conservatism. This study aims to evaluate the residual life of a corroded piping system using different approaches with similar wall-thickness data, compare solutions from the various methods, and identify the advantages and disadvantages of each technique. Evaluate the minimum data required to comfortably predict wall thickness reduction in a piping system by investigating and reviewing related studies. The United Kingdom Health and Safety Executive (HSE) highlighted the need for an efficient, cost-saving process for estimating the remaining wall thickness of a corroded metal surface, recommending statistical methods and computer-aided tools. This study applied statistical analysis of sixteen (16) year wall thickness measurements from an oil and gas piping system using Reliasoft's Weibull++ tool and compared the results with conventional industry methods. The application of the computer-aided tool makes the statistical extreme value analysis easier and efficient. The study revealed that extreme-value statistical analysis addresses the uncertainties associated with the data. It is a practical and conservative means of determining the remaining life of corroded piping. The extreme value statistical method provides a probabilistic perspective on the current state of the piping system, accounting for the entire system/line. However, all the techniques can produce reliable results if a large amount of data with lower uncertainty is available.