Estudo de confiabilidade de um extenso estaqueamento em estacas pré-moldadas na Baixada de Jacarepaguá

Abstract

In view of the innumerable uncertainties inherent in a foundation design it is the engineer's the task to conceive and optimize it without compromising safety. ABNT's NBR6122 rule adopts the deterministic approach of global or partial safety factors. On the other hand, Eurocode already requires the verification of the probability of foundations' failure, together with the adoption of partial safety factors in each country. In Brazil, the probabilistic approach is already used in large projects, under the requirement of insurance companies to define insurance premiums for large constructions. The present research investigates the reliability index of a piling and the influence of the Bayesian update procedure in reduction uncertainties, using as the likelihood function dynamic formulas based on records obtained during pile driving. An initial analysis of the piles that were submitted to dynamic loading test led to two procedures and two likelihood functions: in the first procedure only the data obtained during driving (set and rebound) were considered; in the second, the dynamic load tests results and soil load transfer mechanism at failure were also taken into account. The a priori estimates were calculated using the semi-empirical methods of Aoki-Velloso and Decourt-Quaresma. The updating Bayesian procedure using two distinct likelihood functions resulted in a posteriori estimates with a lower coefficient of variation. Both dynamic formulas adopted as a likelihood function, Chellis (1951) or Sorensen-Hansen (1957), were compared to verify the one that best fitted to the dynamic loading test. With the adoption of Sorensen-Hansen (1957) formula and the first procedure, a much larger number of piles was analyzed forming an extensive database. In this analysis of a larger database, the reliability index was quantified, initially for a priori estimates and, in sequence, after the uncertainty reduction by the Bayesian update. A significant increase in the reliability index was observed by updating, with a consequent reduction in the probability of failure. It was concluded on the relevance of the driving records and its ability to reveal the natural soil variability, as well as the capability of the Bayesian update procedure in reducing the uncertainties and increasing the piling reliability. This is a significative aspect to be considered during the piling control, such as the set and rebound registered in driven piles. The lack of accurate alternative registers for other pile types is the reason for much high uncertainty compared to driven piles.