Predicting and Optimizing the Soil–Water Characteristic Curve Parameters with Limited Data using the Performance Guided Jaya Algorithm

Abstract

The soil water characteristic curve (SWCC) is a fundamental tool for studying the intrinsic mechanical properties of unsaturated soils. The data for SWCCs must be obtained via laboratory tests, which require considerable time and labour. Therefore, in practice, empirical models are often developed with a limited number of laboratory data samples to produce SWCC curves. However, empirical models fitted with limited data often lead to inaccurate modelling of SWCCs. This research aims to propose a data-driven optimization framework to accurately and effectively fit empirical SWCC models with limited amounts of laboratory data. First, the van Genuchten model is adopted as the empirical model to construct the SWCCs. Second, the performance-guided Jaya algorithm (PGJAYA) is introduced to optimize the parameters of the van Genuchten model and increase the fitting accuracy. In the experiments, two cases are investigated, one with limited data and another with abundant data, for four types of soils, i.e., sand, silt, clay and loess. To demonstrate the accuracy and effectiveness of the proposed approach, a comparison with other benchmark optimization algorithms is performed. With limited data, the PGJAYA algorithm outperforms the other optimization algorithms tested. Furthermore, the PGJAYA algorithm produces the smallest fitting error for the SWCCs in the case with full data. Overall, the PGJAYA algorithm is an effective and accurate optimizer with respect to the parameter fitting tasks based on empirical SWCC models.