Reply to the discussion by Fellenius on “Load tests on full-scale bored pile groups”

We thank Dr. Fellenius for his discussion of our paper. The discussion suggests that we have left out four references on load tests on pile groups (O'Neill et al. 1982a, 1982b; Phung 1993; O'Neill and Reese 1999). Our paper focused on instrumented field load tests on bored piles, with a clear focus on pile groups. The references mentioned in the discussion, while of interest, are not directly applicable in the present case. The Ph.D. thesis of Phung (1993) focuses on model, steel pipe piles driven in sand. O'Neill and Reese (1999) discuss load tests on instrumented single bored piles, but not on pile groups. The papers by O'Neill et al. (1982a, 1982b) focus on instrumented driven pile groups. The discussion appears to argue that the position of the pile within the group (i.e., whether a corner, side or central pile) would have no effect on load response. Our tests show otherwise. The pile spacing in the data reported by Phung (1993), as shown in Fig. 1 of the discussion, is of the order of 10 times the equivalent diameter of the piles, while the pile spacing is 2.5 and 3 times the pile diameter in our tests. The focus of Phung (1993) was on piles as settlement-reducing elements (i.e, piled mats or rafts); our work, in contrast, focused on spacings that would bemore typical in a traditional pile group, for which spacings are kept relatively small so that pile cap costs do not increase. The different spacing obviously has an impact on pile interaction and on the influence of pile position on pile response. The extent of the effect of pile position also depends on other factors, chiefly the relative stiffnesses of the various components of the foundation system (pile cap, individual piles, and soil), and could be obscured by random variation of soil properties around the pile group or, particularly in the case of driven piles, by variability related to installation procedure and sequence. Another issue with model pile testing is the scale effects, which become quite significant as shear strain begins to localize next to the pile (see, for example, Foray et al. (1998); Lehane et al. (2005); Loukidis and Salgado (2008)). These scale effects may distort results significantly. The next point raised refers to the use of the term settlement by the geotechnical engineering profession. It is useful in issues related to notation to refer back to the underlying science: mechanics in this case. In mechanics, the pertinent quantity is displacement. In the context of the present paper, we are dealing with vertical displacements, routinely referred to as settlement in foundation engineering. The term has of course been used universally (e.g., Poulos 1989; Briaud et al. 2000; Lehane and Randolph 2002; Randolph 2003; de Sanctis andMandolini 2006; McCabe and Lehane 2006; Xu and Zhang 2007; Salgado 2008) and is appropriate to describe the results of our tests. Another question raised regards the use of the traditional 10% relative settlement criterion for estimation of the ultimate load of an individual pile. It is a criterion that is indeed widely used internationally both in design and in the interpretation of load tests (e.g., Skempton 1959; Lee and Salgado 1999; Briaud et al. 2000; Paik and Salgado 2003; Paik et al. 2003; Randolph 2003; Jardine et al. 2005; de Sanctis and Mandolini 2006; McCabe and Lehane 2006; Salgado 2008; Xu et al. 2008; Fleming et al. 2009; Kim et al. 2009; Seo et al. 2009). The reason for its widespread use is that it does work well for a range of design situations. As pointed out by Salgado et al. (2011), deviations would occur, for example, for piles bearing in rock (when a smaller relative settlement would be appropriate) or for piles in very weak clay (when a limit or plunging load would be reached before a relative settlement of 10% could develop). The notion of a relative settlement (in contrast to a set value of settlement) is useful in designing foundations for frame structures in that the pile diameter has some correlation with span, which, in turn, correlates with tolerable differential settlement (Salgado 2008). The correlation weakens as we move from one pile per column to multiple piles per column, but it still exists. In any case, relative settlement is typically used to define an ultimate load for an individual pile. Many authors (e.g., Lee and Salgado 1999; Lehane and Randolph 2002; Fleming et al. 2009) also explicitly use or refer to use of the relative settlement of the base, not of the top of the pile, which would specifically separate out the contribution of pile compression to total settlement at the pile head. In summary, if the range of applicability of the criterion is properly considered, the 10% relative settlement criterion has a role in pile design. It is noworse than anymethod in the literature and definitely superior to the so-called graphical methods that attempt to arrive at an ultimate load based on the appearance of the load–settlement curve, without any reference to the fact that the foundation element is supporting a structural load and has as its primary functionminimization of the chances of the structure reaching a limit state. Ideally, an engineer would be able to use much more specific criteria, applicable to a given structure and circumstances, and definition of an “ultimate load” for an individual pile, much less for a pile group, would not even be required in such a case. Now we turn to some specific clarifications regarding the test results. With respect to the last applied load (equal to 1680 kN) on pile DZ1L, this was a plunging load. As it could not reach equilibrium, we must assume that plunging would have occurred between the 1540 and 1680 kN loads. Details were provided only for the 1540 kN load. With respect to load transfer plot estimation, taking the location of layer interfaces, if accurately known, into account would lead to an improved estimation of these plots and therefore of the

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.