Neural network model to predict strength parameters of dune sand at Jodhpur City. (2022)

• Record Type: Journal Article
• Authors: Sharma, Yagya; Dave, Mayank; Gehlot, Tarun; Solanki, Deepanshu
• Is Part Of: Materials today. Volume 62:Part 6(2022)
• Page Start: 4498
• Page End: 4503
• Publication Date: 2022
• Subjects: Soft computing -- Soil strength -- Neural Network Model -- Dune sand -- Jodhpur City; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2022.04.945
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: The shear strength of the soil is a significant factor in determining bearing value and determining stability. This study investigates a neural network model for determining the strength parameters (cohesion & friction angle) of locally accessible dune sand in Jodhpur city in India. Model has been developed considering grain sizes, dry density, relative density, void ratio, liquid limit, plastic limit etc as independent variables and strength of dune sand as dependent variable. The findings show that the model is capable of forecasting dune sand shear strength parameters . R programming is used to generate an ANN model, which is then compared to MS-Excel. A response analysis was used to establish the best algorithm and neuronal numbers for improving the model design. The use of a neural network to forecast dune sand strength variables was shown to be very accurate. For the experimental results and model outputs, the relative mean squared error, coefficient of decision (R 2 ), and mean absolute relative error are determined. For performance analysis, Levenberg Marquardt algorithm is used. In the ANN model, the quantity and least value of R 2 for training data sets are 98.2 percent and 96.2 percent, respectively. The suggested ANN models for dune sand strength are accurate and highly close to the experimental values, according to statistical results. As a result, the ANN model appears to be a useful tool for predicting dune sand strength parameters.