Efficient qualitative risk assessment of pipelines using relative risk score based on machine learning

Area of Science:

• Engineering
• Risk Management
• Data Science

Background:

• Pipelines are crucial for transporting oil, gas, and water globally.
• Pipeline integrity is threatened by damages like corrosion, dents, and ruptures.
• Regular evaluation is essential to ensure pipeline safety and prevent failures.

Purpose of the Study:

• To predict pipeline failures using the relative risk scoring (RRS) technique.
• To assess the effectiveness of RRS in classifying pipeline risks.
• To compare RRS performance against other machine learning models.

Main Methods:

• Utilized machine learning, a probabilistic technique, for risk forecasting.
• Applied the relative risk scoring (RRS) method considering parameters like corrosion, leakage, and environmental factors.
• Compared RRS with Naive Bayes, decision tree, support vector machine, and graph convolutional network models.

Main Results:

• The RRS technique achieved 97.5% accuracy in identifying pipeline risks.
• RRS provided precise and immediate classification of risks as high, medium, or low.
• RRS outperformed other tested machine learning models in accuracy and speed.

Conclusions:

• The RRS technique is a highly accurate and efficient method for predicting pipeline failures.
• RRS enhances pipeline integrity management through reliable risk assessment.
• Machine learning, specifically RRS, offers a robust approach to proactive pipeline safety.