Predictive assessment of corrosion growth in pipelines using probabilistic modeling and Monte Carlo simulation. (RIP2026-00110)

Luan Santos, Luan Carrera Santos, Carlos Eduardo Vieira Masalla, RHUAN SOUZA, Jose Antonio Da Cunha Ponciano Gomes , Alysson Helton Santos Bueno

Universidade Federal de São João del-Rei, Universidade Federal de Sao Joao del-Rei, Universidade Federal de Sao Joao del-Rei, Labcorr – Universidade Federal do Rio de Janeiro, Universidade Federal de São João del-Rei

Pipelines are critical infrastructures for the reliable and efficient transport of oil
and gas over long distances. However, steel pipelines are inherently vulnerable
to degradation mechanisms such as corrosion, cracking, and mechanical
damage, all of which can progressively compromise structural integrity. Without
proper monitoring and timely maintenance interventions, these defects may
evolve into severe safety risks and lead to significant financial losses. In this
context, quantitative methods capable of characterizing the severity of
anomalies and predicting their evolution over time are essential for modern
pipeline integrity management.
This study proposes an integrated framework that combines exploratory data
analysis, statistical modeling, and predictive assessment to evaluate the
progression of corrosion anomalies detected by In-line Inspection (ILI) tools.
The methodology is applied to an ethylene transmission pipeline with a nominal
diameter of 6 inches, constructed in 1994 from API 5L X52 steel. The dataset
comprises inspection results from PIG runs, including detailed geometric
information for each anomaly: depth of metal loss, affected area, longitudinal
and transverse dimensions, and axial position along the pipeline.
The exploratory stage includes descriptive statistics, boxplots, and outlier
identification to detect geometric patterns and differences between inspection
years. Subsequently, the anomaly data are fitted to several candidate
probability distributions, such as Normal, Lognormal, Gamma, Weibull, and
Generalized Pareto (GPD). The distribution parameters are modeled as
functions of time, enabling the representation of anomaly characteristics
through a closed-form probability density function dependent on t:

Where is the variable of interest (for example, the circumferential position of the
anomalies), and represents the time-dependent parameters. The adequacy of
the fitted distributions is evaluated using the Akaike Information Criterion (AIC),
which quantifies relative information loss, and Q–Q plots, which visually
compare empirical and theoretical quantiles.
Based on the selected distributions for corrosion depth, failure probability is
estimated following the most recent ASME B31G procedure. Monte Carlo
simulation is employed to understand uncertainties associated with
measurement errors, geometric variability, and model parameters, yielding a
probabilistic estimate of remaining strength and burst pressure.
This enhanced modeling fidelity enables more consistent application of the
ASME B31G methodology and results in more reliable burst pressure
estimations. Overall, the integration of statistical modeling and structural
assessment aims to establish a robust decision-support tool for pipeline integrity
management, reinforcing the value of data-driven analysis of ILI results within
modern risk-based maintenance strategies.