Prediction of the Bearing Capacity Values of Course and Fine Soils (CBR) Using the Artificial Neural Network Method

Abstract

Soft computing-based forecasting is a vital tool in engineering and is used successfully to make informed decisions. Therefore, it is crucial for engineers to predict the behavior of geotechnical engineering elements used in infrastructure, such as earth dams, road fills, road pavements, and airports. The bearing capacity of soil in road foundations and pavement layers is usually expressed by the California Bearing Capacity (CBR). This index evaluates the soil's ability to withstand loads under certain conditions. Engineers use various techniques to estimate and determine soil bearing capacity, but they have difficulty in arriving at accurate estimates due to the wide variation in soil properties and the multitude of factors affecting them. Locally, determining soil bearing capacity is difficult, so in this research, we attempted to develop valid models to determine the CBR value using information and data on soil properties (Atterberg limits, sand equivalent, density, and CBR). Predictive models were built to estimate soil bearing capacity values ​​(CBR) using artificial neural networks (ANN). These models were applied to two real-world projects: the northern ring road of Latakia and latakia-Hamma road. Both models demonstrated good flexibility in calculating the design bearing capacity values ​​(CBR) using ANN.