Abstract:

Accurate prediction of slope stability is critical for reducing the devastating impacts of slope failures and landslides. However, conventional slope stability analysis methods are complex and resource-intensive, requiring extensive soil property and field investigation data for wide-area assessments. To address these challenges, this ongoing research focuses on developing simplified and efficient slope stability prediction models using machine learning (ML) techniques. Specifically, multiple linear regression (MLR) and artificial neural networks (ANN) are utilized for modeling slope stability, while random forest (RF) and support vector machine (SVM) methods are applied for classifying slopes as safe or unsafe.

A dataset of 4,208 slope cases, generated using limit equilibrium-based Slide software, serves as the basis for model training and testing. The effectiveness of the models is assessed using statistical metrics and validated through roadside slope case studies from Nepal, India, Canada, and the UK. Preliminary results indicate that the Spencer’s method-based ANN model shows the highest reliability in slope stability prediction.

These findings are expected to offer valuable contributions to simplifying slope stability assessment, enhancing slope safety, and improving sustainability in engineering projects involving soil slopes. The ongoing study aims to further refine these models and provide more accurate and reliable tools for slope stability decision-making in diverse geographical regions.

Status:

Ongoing

Team:

Rajan KC, Milan Aryal, Keshab Sharma, Richa Pokhrel, Netra Prakash Bhandary, Indra Prasad Acharya