Abstract:Pipelines are critical arteries for the oil and gas industry, requiring massive capital investment to safely construct networks transporting hydrocarbons across diverse environments. However, these pipeline systems are prone to integrity failure, resulting in significant economic losses and environmental damage. The accurate prediction of pipeline failure events using historical oil pipeline accident data enables asset managers to plan sufficient maintenance, rehabilitation, and repair (MR&R) activities to prevent catastrophic failures. However, learning the complex interdependencies between pipeline attributes and rare failure events poses several analytical challenges. This study proposes a novel machine learning framework to accurately predict pipeline failure causes on highly class-imbalanced data compiled by the United States Pipeline and Hazardous Materials Safety Administration (PHMSA). Natural language processing techniques were leveraged to extract informative features from unstructured text data. Furthermore, class imbalance in the dataset was addressed via oversampling and intrinsic cost-sensitive learning strategies adapted for the multi-class case. Nine shallow and deep machine and deep learning architectures were benchmarked, with LightGBM demonstrating superior performance. The integration of cost-sensitive learning yielded an 86% F1 score and a 0.82 Cohen kappa score, significantly advancing prior research efforts. This study leveraged comprehensive Shapley Additive explanation (SHAP) analysis to interpret the predictions from the LightGBM algorithm, revealing the key factors driving failure probabilities. Leveraging sentiment analysis allowed the models to capture a richer, more multifaceted representation of the textual data. The paper developed a novel cost-sensitive learning approach that integrates domain knowledge about the varying cost impacts of misclassifying different failure types into machine learning models. The research demonstrated an effective fusion of text insights from inspection reports with structured pipeline data that enhances model interpretability. The resulting AI modeling framework generated data-driven predictions of failure causes, empowering transportation agencies with actionable insights. These insights enable tailored preventative maintenance decisions that proactively mitigate emerging pipeline failures.

Machine Learning, Linear and Bayesian Models for Logistic Regression in Failure Detection Problems

A Novel Bayesian Robust Model and Its Application for Fault Detection and Automatic Supervision of Nonlinear Process

Predicting machine failures from multivariate time series: an industrial case study

Predicting Future Machine Failure from Machine State Using Logistic Regression

Analyzing Risk of Service Failures in Heavy Haul Rail Lines: A Hybrid Approach for Imbalanced Data

Modeling failures in smart grids by a bilinear logistic regression approach

Using Big Data to Enhance the Bosch Production Line Performance: A Kaggle Challenge

Adoption of machine learning technology for failure prediction in industrial maintenance: A systematic review

Using machine learning and deep learning algorithms for downtime minimization in manufacturing systems: an early failure detection diagnostic service

Machine Learning in High Volume Media Manufacturing

Adaptable and Explainable Predictive Maintenance: Semi-Supervised Deep Learning for Anomaly Detection and Diagnosis in Press Machine Data

A two-level machine learning framework for predictive maintenance: comparison of learning formulations

Root cause prediction for failures in semiconductor industry, a genetic algorithm-machine learning approach

Strategies for overcoming data scarcity, imbalance, and feature selection challenges in machine learning models for predictive maintenance

Failure prediction in production line based on federated learning: an empirical study

Machine Learning Application Using Cost-Effective Components for Predictive Maintenance in Industry: A Tube Filling Machine Case Study

Using machine learning prediction models for quality control: a case study from the automotive industry

Data models for service failure prediction in supply-chain networks

Model-Driven Bayesian Network Learning for Factory-Level Fault Diagnostics and Resilience

Interpretable Machine Learning Models for Failure Cause Prediction in Imbalanced Oil Pipeline Data