This research endeavours to enhance the efficiency and accuracy of predicting pile-bearing capacity by implementing machine learning algorithms based on historical Standard Penetration Test (SPT) data. The purpose stems from the critical importance of accurately assessing the behaviour of pile foundations for ensuring the stability and safety of structures which is a resource-intensive process. To achieve this, three machine learning algorithms were selected; Multiple Linear Regression (MLR), Random Forest Regression (RFR), XG Boost (XGB) Algorithm, and five models were produced for each one. The study leveraged a diverse dataset to train and test machine learning models, showcasing their efficacy in reliably estimating pile-bearing capacity. Notably, ensemble learning algorithms, such as Random Forest Regression (RFR) and Extreme Gradient Boosting (XGB), consistently outperformed Multiple Linear Regression (MLR). MLR models exhibited higher Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) values, with average RMSE and MAE of 1440.866 and 1809.860, respectively, compared to 708.088 and 1023.989 for RFR, and 770.721 and 1147.407 for XGB. The Coefficient of Determination (R2) further highlighted the superiority of RFR, achieving the highest value of 0.983 using Principal Component Analysis (PCA) with 10 components. In contrast, MLR models with PCA (3) recorded the lowest R2 values (0.659) along with elevated error metrics (2505.883 and 3006.418). The limited components in PCA (3) hindered its ability to capture essential dataset traits. The superior predictive capabilities of ensemble learning, attributed to its adeptness in handling non-linear relationships, were underscored through comparative analyses with traditional geotechnical methods. This research signifies the algorithms’ potential to streamline the assessment of pile foundations. The implications extend to advancements in geotechnical engineering, offering a cost-effective and time-efficient solution for designing and evaluating pile foundations. By integrating machine learning techniques, this study contributes to the evolving landscape of civil engineering, fostering safer and more reliable structural design and construction practices.
Key words: Pile foundation, Standard Penetration Test, Multiple Linear Regression, Random Forest Regression, Extreme Gradient Boosting, Mean Absolute Error, Root Mean Squared Error
|
