Prediction of soil-water characteristic curves using two artificial intelligence (AI) models and AI aid design method for sands

In this paper, two artificial intelligence (AI) models (i.e., neural networks (ANN) and multivariate adaptive regression splines (MARS)) were developed using cumulative percentiles from the grain-size distribution (GSD) curve as input information to predict soil-water characteristic (SWCC). The importance of each variable was testified different sensitivity analyses. results show a strong correlation between SWCC GSD curves based on large volume datasets. ANN provides higher accuracy due its unique structure; however, MARS model facilitates in developing equation that contributes stable performance. can be reliably predicted with one data point bulk density information. Sensitivity analysis suggests prediction is also possible reasonable degree by single an variable. Finally, novel AI aid design method proposed combining along physico-empirical fitting rapid reliable technique for predicting sands.