Simulation of mechanical behavior of carbonate gravel with hybrid PSO-SVR algorithm

Large-scale testing is one of the common methods used to study the mechanical behavior of carbonate gravels. However, due to the huge workload of larger-scale tests and a dimensional limitation of laboratory specimens, there are only a limited number of large-scale tests, and the mechanical behavior of carbonate gravels is not fully understood. In this study, a hybrid algorithm combining two techniques of artificial intelligence - particle swarm optimization and the support vector regression technique - are used to simulate soil behavior using on a dataset obtained from large-scale direct tests of carbonate gravels. Three types of single-output models have been developed, results show that the models developed by the proposed hybrid algorithm are highly accurate. In addition, one of the models used is to predict the shear strength of carbonate gravel under different normal stress. The results show that when the normal stress is larger than 600 kPa, the nonlinear shear strength model is more suitable for carbonate gravel with high particle breakage. Modeling the mechanical behavior of soil with artificial intelligence is thus a useful method to evaluate the mechanical properties of gravel economically and conveniently.

Automated summary

In this study, artificial intelligence techniques were used to simulate the mechanical behavior of carbonate gravels, and highly accurate models were developed. The shear strength of the gravel was also predicted, showing that the nonlinear model is more suitable for heavily broken gravel with high normal stress. Modeling the mechanical behavior of soil using artificial intelligence is an economical and convenient method to evaluate the mechanical properties of gravel.