Synthetic Intelligence Framework for Soil Geotechnical Property Prediction

Abstract

This study uses Artificial Intelligence (AI) technologies in geotechnical properties, such as Relevance Vector Machines (RVM) models built with various settings. Microsoft Excel 2019's Data Analysis Tool creates the RA models. The content of clay, silt, gravel, and sand are the input parameters used by AI models. Two soil datasets are used to compute the correlation coefficient. The correlation demonstrates how important the contents of geotechnical parameters sand, silt, and clay are in determining. The best-performing AI models are identified by comparing their respective performances. This study makes use of the restricted soil datasets. Relevance vector machines and artificial neural networks could not function effectively. They fared better than the ANNs, SVM, RA, and RVM AI technologies based on the performance of the models. As a result, the GPR AI technique can forecast soil's geotechnical characteristics based on its clay, sand, gravel, and silt content. Additionally, a Monte-Carlo global sensitivity analysis is conducted, and the results show that the amount of sand and clay in the soil impacts the prediction of its geotechnical qualities.