Developing Predictive Models of Collapse Settlement and Coefficient of Stress Release of Sandy-Gravel Soil via Evolutionary Polynomial Regression

The collapse settlement of granular soil, which brings about considerable deformations, is an important issue in geotechnical engineering. Several factors are involved this phenomenon, makes it difficult to predict. present study aimed develop a model predict the and coefficient stress release sandy gravel soil through evolutionary polynomial regression (EPR). To achieve this, dataset containing 180 records obtained from large-scale direct shear test was used. In study, five models were developed with secant hyperbolic, tangent natural logarithm, exponential, sinusoidal inner functions. Using sand content (SC), normal (σn), level (SL), relative density (Dr) values, can (∆H) (CSR). results indicated that exponential functions best models. With these models, values R2 for training, testing, all data prediction 0.9759, 0.9757, respectively, predicting 0.9833, 0.9820, respectively. sensitivity analysis also revealed (SC) parameters had highest lowest degrees importance settlement. contrast, Dr SC showed release. Finally, conducted parametric line previous studies.