Avaliação de modelos de regressão na estimativa da força de ruptura de blocos sobre quatro estacas sob pilares quadrados e retangulares em diferentes escalas com auxílio de simulação numérica

Abstract

The prediction of the load-bearing capacity of four-pile concrete caps lacks normative consensus, as does the truss idealization in strut-and-tie models available in the literature, which are typically evaluated using small-scale models and square columns. A regression model results in an equation with terms calibrated from a database, a methodology already used in the literature for other structural elements to obtain load-bearing capacity values. Therefore, the objective of this study was to develop predictive equations, using regression models, as an alternative for determining the load-bearing capacity of four-pile caps based on their geometry. To this end, the study focused on compiling a database associated with numerical simulations. These numerical models were extrapolated from a calibration based on literature records, encompassing square and rectangular columns, as well as effective depths greater than those of the original experimental tests. From these data, nine equations with different variable interactions were developed, all showing R² values greater than 90%. It was observed that, for columns with a cross-sectional area of less than 50% of the total pile area, a dispersion in the predictions occurred across all equations due to the complex stress flow at the contact interface between the column and the pile cap. It was also found that the increase in load-bearing capacity associated with the elongation of one column dimension followed a quadratic function. Thus, despite the higher dispersion for area ratios below 50%, the regression models proved to be an effective and viable alternative. They bypass the truss idealization difficulties debated in literature and can be continuously refined by incorporating new experimental tests