Predictive Analysis of Debris Flow Impact on Pipelines in Mountainous Regions: An Experimental and Machine Learning Approach
Abstract ID: 3.10774 | Accepted as Talk | Talk | TBA | TBA
Afnan Ahmad (1)
Mudassir Ali Khan (2), Muhammad Sumair (3), Manoj Kumar (4), Vivi Anggraini (5)
(2) Universiti Teknologi PETRONAS, Persiaran UTP, 32610 Seri Iskandar, Perak, MY
(3) Universiti Malaysia Sarawak, Jalan Datuk Mohammad Musa, 94300, Kota Samarahan, Sarawak, MY
(4) Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, 47500, Subang Jaya, Selangor, MY
(5) Monash University Malaysia, Jalan Lagoon Selatan, Bandar Sunway, 47500, Subang Jaya, Selangor, MY
Debris flows are a significant geohazard in mountainous regions, posing substantial risks to critical infrastructure, including water, oil, and gas pipelines. Understanding and predicting the impact forces exerted by debris flows on pipelines is crucial for enhancing the resilience and sustainability of these infrastructures. This study investigates the relationship between the rheological properties of debris flows such as shear stress, shear strain, viscosity, and normal stress and the resulting impact forces on pipelines. The significance of this study lies in its potential to improve the design and protection of pipeline infrastructure, thereby reducing the risk of catastrophic failures and ensuring the safe and efficient operation of critical energy systems. The study utilizes a comprehensive dataset obtained from experimental measurements of debris flow samples with varying volume fractions (S0 to S7). Each sample was characterized by its rheological properties, including shear stress, shear strain, viscosity, and normal stress. These properties were measured using a digital hybrid rheometer. Utilizing experimental data from seven samples with varying volume fractions, various machine learning techniques (ML) were employed to predict impact forces. The ML techniques include Random Forest and Gradient Boosting Machines (GBM) models which are developed and validated using 5-fold cross-validation, demonstrating high predictive accuracy. The analysis revealed that rheological properties significantly influence the magnitude and distribution of impact forces on pipelines. The Random Forest model demonstrated high predictive accuracy, with an RMSE of 0.12 and an R² of 0.95, indicating strong model performance. The GBM model also showed robust predictive capabilities, with an RMSE of 0.15 and an R² of 0.93. These findings underscore the necessity for incorporating rheological properties into the design and protection of pipeline infrastructure. By integrating experimental data with advanced predictive modeling, this research provides practical guidance for enhancing the resilience and longevity of pipeline systems in geohazard-prone areas.
N/A | ||||||||
|