Scholarlink

To ensure the stability and safety of the structure against the risks of sliding along the sliding surface of civil engineering foundations, it is necessary to predict the soil's bearing capacity. Conventional analytical techniques are costly, time-consuming, and difficult. A relative comparison of software computational models is presented. Ultimate bearing capacity of soil was predicted in this study using Three different machine learning techniques namely: feed forward neural network (FFNN), support vector machine (SVM), and multilinear regression (MLR). Cohesion (C), internal friction angle (Ø), and ultimate bearing capacity are examples of input variables. 200 datasets were acquired from secondary sources, of which 175 were used for model training and 25 for model validation. The findings show that the proposed method for estimating the final bearing capacity has good predictive reliability. The findings demonstrate the effectiveness of the three models (FFNN, SVM, and MLR) in forecasting the ultimate bearing capacity of the three different foundation types, however, FFNN outperforms SVM and MLR by a little margin. Consequently, in order to compare the performance of the models, a number of error criteria have been taken into consideration, including correlation coefficient (R), mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). Likewise, the corresponding R2 values of ANN-strip (0.9501), ANN-square (0.9846), ANN-circular (0.9680, SVM-Strip (0.9627), SVM-Square (0.9915), SVM-Circular (0.9517), MLRA-Strip (0.9501), MLRA-Square (0.9549) and MLRA-Circular footing (0.9578) respectively. Utilizing soft computing can offer fresh perspectives and techniques to reduce the risk of correlation problems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementation of Nonlinear Computing Models and Classical Regression for Predicting of Soil Bearing Capacity

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Awaisu Shafiu Ibrahim,
  • Ahmad Idris

摘要

To ensure the stability and safety of the structure against the risks of sliding along the sliding surface of civil engineering foundations, it is necessary to predict the soil's bearing capacity. Conventional analytical techniques are costly, time-consuming, and difficult. A relative comparison of software computational models is presented. Ultimate bearing capacity of soil was predicted in this study using Three different machine learning techniques namely: feed forward neural network (FFNN), support vector machine (SVM), and multilinear regression (MLR). Cohesion (C), internal friction angle (Ø), and ultimate bearing capacity are examples of input variables. 200 datasets were acquired from secondary sources, of which 175 were used for model training and 25 for model validation. The findings show that the proposed method for estimating the final bearing capacity has good predictive reliability. The findings demonstrate the effectiveness of the three models (FFNN, SVM, and MLR) in forecasting the ultimate bearing capacity of the three different foundation types, however, FFNN outperforms SVM and MLR by a little margin. Consequently, in order to compare the performance of the models, a number of error criteria have been taken into consideration, including correlation coefficient (R), mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R2). Likewise, the corresponding R2 values of ANN-strip (0.9501), ANN-square (0.9846), ANN-circular (0.9680, SVM-Strip (0.9627), SVM-Square (0.9915), SVM-Circular (0.9517), MLRA-Strip (0.9501), MLRA-Square (0.9549) and MLRA-Circular footing (0.9578) respectively. Utilizing soft computing can offer fresh perspectives and techniques to reduce the risk of correlation problems.