Applications of Artificial Neural Networks for the Prediction of Subgrade CBR Values

Abstract

This study delves into the development of Artificial Neural Networks for predicting subgrade strength properties based on experimental data. Based on the validity and quality of soil tests, using ANN to predict CBR value may offer a suitable replacement for lengthy and expensive laboratory testing based on validated data for materials supplied from all over Pakistan. Key soil parameters like liquid limit (LL), plasticity index (PI), coarse content, fines content, optimum moisture content (OMC), and maximum dry density (MDD) serve as model inputs. The trained ANN models predict California bearing ratio (CBR), a crucial parameter for flexible pavement design. An optimized ANN model was built and evaluated in MATLAB using performance indicators like correlation coefficient, mean square error, and mean absolute error. The predicted CBR values are then compared to actual values, with the coefficient of determination (R2) of 0.92 serving as the final validation metric. The close agreement between predicted and actual CBR values demonstrates the model’s effectiveness. Furthermore, the study highlights the flexibility of ANNs by suggesting that altering input and output parameters can enable prediction of various other soil engineering properties