Remaining useful life prediction of a piping system using artificial neural networks: A case study

Abstract

Oil producers or operators such as Shell, Petronas, Petron, Chevron, and Lukoil have always placed their equipment as the highest priority for operations. Still, the study shows that many failures in the facility associated with piping systems lead to billions of dollars’ loss. In the oil and gas industry, these piping systems are subjected to various failure mechanisms since it has been operated in various processes and harsh geographical environment. Most of the piping systems are susceptible to corrosion caused by several factors, as reported in the literature. Corrosions of the piping system weakened the piping strength as well as its fittings, thus reducing its ability to withstand the fluctuation of temperature and pressure generated towards the piping system. This work focussed on the factors that contribute to the life of the piping system based on the real-time risk inspection data that were obtained from PETRONAS facilities. The parameters considered were pressure, corrosion, wall thinning, age, nominal thickness, outer radius, and product type. The neural network model has been developed to predict the remaining useful life of piping based on the selected parameters. The proposed model showed promising results of R2 value 0.99, which is close to 1.0, and the validation accuracy of a model was found 97.51% when compared with the actual data. The deterioration trends of individual factors considered in this study are generated to know the effect on pipe life conditions. This work may help oil and gas companies in determining the Fitness For service (FFS) of the piping system by estimating the life of the piping system affected by various corrosion phenomena.

Keywords

• Artificial neural networks
• Corrosion
• Deterioration
• Piping
• Prediction
• RUL