Discussion of “Statistics of model factors in reliability-based design of axially loaded driven piles in sand”

Reply to the discussion by Flynn and McCabe on “Statistics of model factors in reliability-based design of axially loaded driven piles in sand”

“…It can be seen that (i) when the parameter b > 0.6, the difference within the hyperbolic parameters (a, b) is not very discernible and (ii) when the parameter b < 0.6, the parameter a for CEP piles is usually greater than that for DCIS piles. This is unlike the observation of Flynn and McCabe (2019), where the DCIS pile data are in close proximity to the CEP pile dataset. It should be noted that the size of the calibration database was 18 for DCIS piles and 27 for CEP piles in Flynn and McCabe (2019), which are lower than that of the present analysis with N = 41 for DCIS piles and N = 70 for CEP piles.…”

“…This is unlike the observation of Flynn and McCabe (2019), where the DCIS pile data are in close proximity to the CEP pile dataset. It should be noted that the size of the calibration database was 18 for DCIS piles and 27 for CEP piles in Flynn and McCabe (2019), which are lower than that of the present analysis with N = 41 for DCIS piles and N = 70 for CEP piles. The statistics of the hyperbolic parameters are summarized in Table R4.…”

“…The objective of this reply is to further investigate the observations made by the discussers (Flynn and McCabe 2019): (1) the hyperbolic parameters for DCIS piles are strongly dependent on the failure criterion employed to define R um , (2) when the slope tangent method was applied, a discernible difference in the hyperbolic parameters is apparent for the CEP and DCIS piles, and (3) the behaviour of DCIS piles at the serviceability limit state (SLS) may not be similar to that of CEP piles, due to the in situ concreting that eliminates the driving-induced residual loads and leads to enhanced interface friction properties along the pile shaft. On this basis, Flynn and McCabe (2019) suggested that DCIS pile test data should not be included to derive overall SLS model factors.…”

“…(1) Q R um ϭ s a ϩ bs where Q is applied load, R um is measured resistance, and s is settlement. In the paper under discussion (Tang and Phoon 2018a), R um was interpreted from documented load-settlement data using the L 1 -L 2 method (Hirany and Kulhawy 1989), while the slope tangent method (O'Rourke and Kulhawy 1985) was adopted by Flynn and McCabe (2019). Bauduin (2002) discussed the following three philosophies for using databases to characterize the model uncertainty in ultimate limit state design of pile foundations, which resulted in different quality levels in the derived model factors:…”

“…This is can be further verified with the statistics in Table R3. With regard to observation 2 in Flynn and McCabe (2019), the hyperbolic parameters based on the slope tangent resistances for DCIS and CEP piles are illustrated in Fig. R4.…”