A Comparative Multivariate Reliability Analysis Approaches Accelerated Failure Time (AFT) and Cox PH regression: Case in Oil \& Gas Well

Deep-offshore operations subject well components to extreme conditions where physical maintenance is often impossible. Given the limited number of wells per platform, high reliability and real-time risk monitoring are essential. Because retrieving components for inspection is difficult, failure analysis relies heavily on operational variables, requiring robust multivariate reliability assessments. This paper evaluates two primary methods for monitoring well integrity: the Cox Proportional Hazards (Cox PH) model and the Accelerated Failure Time (AFT) model. While Cox PH is a staple of survival analysis, AFT is more frequently utilized in engineering reliability. The study compares these approaches using a dataset of Down Hole Safety Valve (DHSV) lifetimes, comprising 550 observations from 448 wells. The analysis focuses on the impact of two critical stressors-temperature and pressure-on component longevity. By implementing goodness-of-fit tests and graphical reliability curve comparisons, the research highlights the advantages and disadvantages of each method. The results demonstrate that an integrated approach-combining the AFT method with a 2-parameter Weibull model-effectively generates reliability curves that capture the dynamic effects of temperature and pressure variations. This multivariate framework provides a superior tool for predicting well degradation and ensuring operational safety in challenging offshore environments.

Keywords: Multivariate Reliability Analysis, Cox regression, Proportional Hazard, Accelerated Failure Time, Well, Oil and Gas.