Cálculo de capacidade de carga em estacas utilizando métodos semi-empíricos: abordagem probabilística dos resultados

Abstract

Studying pile load estimation has historically been a major challenge for foundation engineering, as most methods are based on empirical correlations with the SPT test, and there is high variability in soil characteristics. culminate in inaccurate results. In that way, this work wants to make comparative analyses between methods of Aoki and Velloso (1975), Décourt and Quaresma (1978), Teixeira (1996), Antunes and Cabral (1996), Brasfond (1991), Lizzi (1982) and Vorcaro and Velloso (2000), through probabilistic analyzes of the load capacity results calculated using data from 8 (eight) SPT tests and a static load test performed on a root pile in the city of Fortaleza, Ceará. Results were compared using statistical parameters such as Percentage Variation (PV), Mean Square Error Root (REQM) and Vies. Besides, probability distributions were generated based on SPT tests results and by means of random variable generation. Starting the distributions, it was estimated the probability of each method presenting values close to the load test Therefore, it was possible to conclude that the Brasfond (1991) method presented results close to the PCE, since the average load capacities showed a good approximation with 110 tf (rupture value found in the load test) and the deviation-relatively low standard. Besides, this method had the highest probability of find values within the interval of ±20 tf, besides showing acceptable results for Vies and the REQM, being -8 and 34.76, respectively. The methodology proposed by Vorcaro and Velloso (2000) also stood out for revealing good results, reaching low REQM and high probability (34.24%) of find values close to the load test. Lizzi’s (1982) formulation, however, showed up exacerbated results in all tests, resulting in an average of 142.59 tf and, therefore, a probability distribution outside the range analysed. According to the result of REQM, Aoki and Velloso (1975), Decourt and Quaresma (1978) and Antunes and Cabral (1996) exhibited the greatest divergences in the load test values, being 114.05 tf, 62.70 tf and 72, 27 tf, respectively. Therefore, it is concluded that probability distributions allow the analysis of results to be closer to reality since one does not analyze single values, but rather a distribution of them.