California bearing ratio of black cotton soil using soft computing techniques: Fid-Bau Portal

California bearing ratio of black cotton soil using soft computing techniques

Shukla, Dharmendra Kumar / Iyer Murthy, Yogesh

Precise evaluation of the index and engineering properties of black cotton soil in the laboratory is a difficult task. Therefore, in this research, California Bearing Ratio (CBR) is predicted using typical soft computing tools for black cotton soil. Adaptive Neuro Fuzzy Interference System (ANFIS), Multi Layer Perceptron (MLR), Radial Basis Function (RBF), and 10 commonly used Artificial Neural Network (ANN) algorithms were tested to predict CBR values utilizing particle size, Optimum Moisture Content (OMC), Maximum Dry Density (MDD), Plasticity Index (PI), Shrinkage Limit (SL), and Swell pressure as input parameters. The dataset included 211 experimentally obtained values of these parameters. Many statistical measures assessed the effectiveness of all soft computing methods. Most soft computing algorithms accurately predicted CBR values; however RBF was the least accurate, followed by MLR, ANFIS, and ANN. The optimum neural architecture for the dataset was found to be 7-5-1. LMA is the most effective ANN algorithm, with excellent statistical parameters. The gradient error is highest for GDA, while BFG has the most epochs. According to MLR, OMC is the most critical element impacting CBR, followed by MDD, SL, swell pressure, PI, and particle size below and above 0.075 mm. In conclusion, the research advances geotechnical engineering by improving CBR forecast accuracy, directing computational tool selection, and boosting understanding of black cotton soil CBR variables.