Statistical Analysis to Determine Appropriate Design Methodologies of Drilled Shaft Foundations

Detailed geotechnical investigations were performed at two sites within the state of Arkansas (Monticello and Turrell). The soil parameters, predicted axial capacity, and predicted load-movement response values varied depending on (1) which geotechnical investigation methods and/or (2) which predictive software programs (FB-Deep, SHAFT) were utilized. The uncertainty associated with the different soil properties and the discrepancies between the different software programs are discussed. Parametric and nonparametric statistical testing methods, including the: T-test, F-Test, and Wilcoxon test were utilized to evaluate the soil parameters (corrected blow count, total unit weight, and undrained shear strength) and the predicted axial capacity data. No statistical differences (95 % confidence interval) were observed for the respective undrained shear strength, total unit weight (clay), and correlated corrected blow count parameters as determined from University of Arkansas (UofA) method and from Missouri Department of Transportation (MODOT) method. However, differences were observed for the predicted axial capacity and load-movement values that were predicted using the aforementioned soil parameters (percent differences ranging from 0.5 to 29.2 % for load-movement values). Because an inverse relationship was observed between the percent difference in the load-movement values and the number of statistically similar soil properties, it was determined that the predicted axial capacity and predicted load-movement response were dependent upon the soil sampling and testing methods and the utilized software program.