Bibliographic citations
Landeo, J., (2023). Modelos explicables basados en datos para la predicción de la resistencia a corte y modos de falla de membranas de concreto reforzado [Tesis, Universidad de Ingeniería y Tecnología]. https://hdl.handle.net/20.500.12815/347
Landeo, J., Modelos explicables basados en datos para la predicción de la resistencia a corte y modos de falla de membranas de concreto reforzado [Tesis]. PE: Universidad de Ingeniería y Tecnología; 2023. https://hdl.handle.net/20.500.12815/347
@misc{renati/230644,
title = "Modelos explicables basados en datos para la predicción de la resistencia a corte y modos de falla de membranas de concreto reforzado",
author = "Landeo Diego, Jostin Gabriel",
publisher = "Universidad de Ingeniería y Tecnología",
year = "2023"
}
Currently, the shear strength evaluation of complex structures can be based on the grouping of membrane elements, which are commonly studied by reinforced concrete (RC) panels. The way to describe the shear strength of membrane elements is based on empirical and analytical equations, so these models are not generalizable, have high variability and dispersion, even more in predicting the failure modes of these elements. This study proposes a framework to develop two explainable data-driven models for the shear strength and failure mode prediction of RC membranes over a wide range of configurations. This is based on a comprehensive database of 195 experimental tests of RC panels for the predictive model of the shear strength and 89 for the classification model of the failure mode. For the classification problem, three data augmentation techniques were implemented, thus the initial data was increased to 180 experimental tests with two of these techniques. Subsequently, six ML models, including ensemble models (Random Forest, AdaBoost, and XGBoost), were evaluated with four performance metrics for the predictive model of regression and two for the classification model. The proposed models of regression and classification (based on XGBoost) achieve high accuracy and generalization, with a value of R 2=0.98 for the predictive model of shear strength and F1-score = 0.93 for the classification model. Moreover, the SHAP (Shapley Additive exPlanations) technique was used to evaluate the importance of the parameters over the outputs of the models in a global and local way. In addition, the proposed models were compared with the Modified Compression Field Theory (MCFT), SMCS method (Simple Model for Combined Stress-resultants) and the finite element analysis, showing that the proposed models have comparable and even better predictions. Finally, a graphical interface was developed for practical and educational applications to perform calculations of shear strength and failure modes of RC membranes.
This item is licensed under a Creative Commons License