In the realm of geotechnical engineering, predicting slope stability has long been a complex and costly endeavor. Traditional on-site tests require sophisticated equipment and extensive logistics, making them both time-consuming and expensive. However, a groundbreaking study led by Kennedy C. Onyelowe from the Department of Civil Engineering at Michael Okpara University of Agriculture, has introduced a novel approach that could revolutionize how we ensure safety in geotechnical projects, with significant implications for the energy sector.
The research, published in ‘Scientific Reports’ (Scientific Reports), delves into the use of advanced machine learning techniques to predict the factor of safety (FOS) of slopes. By leveraging the learning abilities of Class Noise Two (CN2), Stochastic Gradient Descent (SGD), Group Method of Data Handling (GMDH), and artificial neural networks (ANN), Onyelowe and his team have developed a more efficient and accurate method for assessing slope stability.
The study began with a comprehensive literature search, curating and sorting a dataset of 349 entries on the FOS of slopes. After removing unrealistic data, the team was left with 296 realistic data points. These data points included parameters such as unit weight, cohesion, angle of internal friction, slope angle, slope height, and pore water pressure ratio. The initial analysis showed poor correlation with individual factors, leading the researchers to group these parameters into three non-dimensional inputs based on the physics of flows. These inputs were then used to predict the FOS.
The results were striking. The ANN model outperformed other techniques with a significant reduction in error metrics, achieving an SSE of 62%, MAE of 0.27, MSE of 0.21, RMSE of 0.46, average total error of 24%, and an impressive R2 of 0.946. This makes ANN the decisive intelligent model in this exercise. However, the GMDH model, which came in second, offers a unique advantage: it develops a closed-form equation that can be applied manually in the design of slope stability problems.
“This research not only improves the accuracy of slope stability predictions but also provides a practical tool for engineers to use in the field,” Onyelowe said. “The ability to manually apply the GMDH model’s closed-form equation is a game-changer for geotechnical engineering.”
The implications for the energy sector are profound. Slope stability is crucial for the construction and maintenance of infrastructure such as pipelines, power plants, and renewable energy sites. By providing a more accurate and efficient method for predicting slope stability, this research could significantly reduce the risk of geohazards, lower costs, and enhance the safety of energy projects.
The study’s success is attributed to the sorting and elimination of unrealistic data entries, the application of dimensionless combinations of slope stability parameters, and the superiority of the selected machine learning techniques. This approach outperformed eleven previous models, setting a new benchmark in the field.
As we look to the future, this research paves the way for further advancements in geotechnical engineering. The integration of advanced machine learning techniques with geophysical data could lead to even more sophisticated models, enhancing our ability to predict and mitigate geohazards. For the energy sector, this means safer, more reliable infrastructure and a more robust approach to managing geotechnical risks. The potential for commercial impact is vast, as companies can leverage these models to optimize their operations and reduce the likelihood of costly and dangerous slope failures.