Reliability and Uncertainty Quantification for Pipeline Systems Using Exponentiated- Exponentiated Distributions: A Case Study of Nigerian Oil and Gas Pipelines

Abstract

This paper introduces a comprehensive framework for modelling the reliability of pipeline systems
in Nigeria’s oil and gas industry, utilizing the Exponentiated-Exponentiated (E-E) distribution.
The E-E distribution, a generalization of the Exponentiated Weibull distribution, affords greater
flexibility to model failure rates that may increase, decrease, or remain constant over time — a
characteristic that renders it particularly suitable for capturing the non-monotonic degradation
patterns commonly observed in pipeline systems. To enhance predictive accuracy and uncertainty
quantification, the E-E distribution is integrated with hybrid neural-statistical models that combine
Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM) networks, and
Convolutional Neural Networks (CNN). The resulting model provides probabilistic predictions for
pipeline lifespan and Remaining Useful Life (RUL), offering appreciable improvements in
predictive maintenance capabilities. Real-world data from Nigeria’s oil and gas pipelines are used
to validate the model, demonstrating superior performance relative to traditional models such as
the Weibull distribution and ARIMA. The study highlights the potential of this framework to
enhance maintenance planning and decision-making, thereby improving pipeline operations across
the Nigerian oil and gas sector.
Keywords: Reliability Modelling, Exponentiated-Exponentiated (E-E) Distribution, Pipeline
Degradation, Hybrid Models, Predictive Maintenance, Oil and Gas Sector