Modelo em aprendizagem de máquina para previsão de capacidade de carga de fundações diretas

Abstract

The problem of estimating load capacity is essential in civil engineering since adequate dimensioning of foundations is essential to guarantee the safety and stability of structures. Recent studies have been using machine learning models in foundation problems, demonstrating significant accuracy. The present study sought to develop a model for predicting the load capacity of direct foundations laid in different types of soil. The results obtained in the accurate literature review on the topic highlight the effectiveness of using machine learning techniques in direct foundations. Studies have shown that machine learning models are capable of providing accurate and reliable estimates, outperforming traditional approaches based on empirical methods. Furthermore, the literature also explored the influence of different geotechnical parameters and input variables such as: SPT, soil type, footing installation level, presence of water level, among others, on the performance of machine learning models. They showed that the use of relevant geotechnical variables, such as soil shear strength and compaction characteristics, can significantly improve the accuracy of the output data. The model created was calibrated based on existing literature and was determined using Machine Learning techniques, using the random forest algorithm, to predict the load capacity based on feeding the machine with input data.