Comparison of regression analysis for estimation of initial and total fracture energy of concrete

• Published in: Multiscale and Multidisciplinary Modeling, Experiments and Design Vol. 7; no. 1; pp. 173 - 190
• Main Author: Peng, Jia
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.03.2024
• Subjects: Characterization and Evaluation of Materials; Engineering; Mathematical Applications in the Physical Sciences; Mechanical Engineering; Numerical and Computational Physics; Original Paper; Simulation; Solid Mechanics; Concrete; Random tree; Multivariate adaptive regression spline; Estimation model; Fracture energy
• ISSN: 2520-8160; 2520-8179
• DOI: 10.1007/s41939-023-00190-9

The investigation into the energy requirements for crack propagation in concrete structures has been a captivating subject ever since the application of fracture mechanics to concrete. Based on previous experimental observations, various regression and classification techniques were utilized to estimate the initial ( G f ) and total ( G F ) fracture energy of concrete, considering factors, such as compressive strength, largest aggregate size, curing age, and water-to-cement ratio, as independent variables while treating them as output parameters. In the present study, to this aim, it was aimed to develop some equations to estimate G f and G F of concrete using multivariate adaptive regression spline ( MARS ) and Random Tree ( RT ) analysis. The results reveal that both analyses perform well in forecasting the G f and G F , which represent the allowable association between measured and simulated values. RT procedure had a higher level of competency than the MARS in both the learning and testing phases, considering all indices in forecasting the G f with the coefficient of determination ( R 2 ) value at 0.9763 and 0.9984, and for G F with the R 2 value at 0.9419 and 0.9381 in the train and test portion. All in all, from descriptions and outcomes of models from metrics, it could have resulted that the visualization tree from RT classifier is really dependable to use for estimating G f and G F of concrete.