A comparative study of machine learning approaches for modeling concrete failure surfaces

摘要: A general shape function, which is able to generate types of surfaces that cover most the hypothesis concrete failure criterion, introduced.Realizations this function with different noise levels emulating experimental errors are used for verification purposes machine learning approaches.Artificial neural networks, support vector machines, and regression adapted model surface C25 starting from 88 tests.These model-free approaches independent any predefined models eliminate need new types. This study introduces an enhanced approach strongly needed a realistic simulation behavior, by employing instead traditional surfaces. Since not exactly known, introduced. Artificial realizations levels. After successful fitting these surfaces, algorithms employed tests. The three fit data low error compared one another. DruckerPrager BreslerPister solved same surface. main advantage they