تجزیه و تحلیل حساسیت و نقش آن در طراحی شبه استاتیکی شالوده شمعی

Lateral spreading, a phenomenon associated with earthquake-induced soil liquefaction, can impose large lateral demands on piles and has been responsible for the failure of many foundations. One of the key issues in the simplified analysis of piles using the pseudo-static approach is how to deal with the uncertainties associated with soil liquefaction and lateral spreading, an understanding of which is necessary for the consistent and reliable use of these methods in practice. In this paper a comprehensive series of analyses is used to examine the parametric sensitivity of the pile response for a broad range of soil-pile systems and magnitudes of lateral spreading displacement. The parametric sensitivity results clearly demonstrate the fundamental link between the relative importance of the various model parameters and the mechanism of soil-pile interaction, with soil strength-related parameters most important in the case of ‘stiff’ pile behaviour, and soil stiffness-related parameters in the case of ‘flexible’ pile behaviour. The SPT blowcount, N, in particular was arguably the most influential parameter given the magnitude of its uncertainty, and its use in determination of the stiffness and strength of soil.

Widespread damage to pile foundations has been observed after many strong earthquakes in areas where extensive soil liquefaction and lateral spreading have occurred [1], [2] and [3]. In response to this, numerous methods for pile design and analysis have been developed, ranging from simplified pseudo-static procedures to advanced dynamic effective stress techniques. The most commonly adopted simplified procedure is the pseudo-static method, in which kinematic and inertial loads on the pile are imposed via free-field ground displacements and forces, respectively. Despite its commonality, there are numerous different implementations of the pseudo-static method, which primarily reflect different parameterisations for the ground displacement, inertial forces, soil force-displacement and pile moment-curvature relationships (e.g., [4], [5] and [6]). The interaction between piles and liquefying soils is a complex and intense dynamic process, and it is very difficult to predict the magnitude and distribution of lateral spreading displacements and the mechanical properties of the liquefied soil. When undertaking a simplified pseudo-static analysis for a particular scenario, it is important to recognise that key parameters affecting the predicted pile response may take values over a very wide range, due to the limitations of a simplified static analysis for representation of a complex dynamic problem. In this paper, a simplified, pseudo-static analysis method developed by Cubrinovski and others [7] and [8] has been used to explore the effects of variation of the model parameters on the predicted pile response by means of sensitivity analyses. The analyses focus solely on the response of single piles in order to first identify key mechanism-driven sensitivities in the pile response for the simpler single-pile case, while sensitivity analyses covering more complex pile-group effects are addressed in subsequent studies [9].We first introduce the adopted simplified analysis method, before outlining a procedure for undertaking a deterministic parametric sensitivity study. The interpretation of parametric sensitivity results is then explored by means of a large series of systematic pseudo-static analyses.

A comprehensive study has been conducted to investigate the parametric sensitivity of the pseudo-static analysis for piles in liquefying soils subjected to lateral spreading. The key findings from the study can be summarised as follows: • There is a fundamental link between the parametric sensitivity and response mechanism that allows us to summarise all results for various soil-pile systems and spreading demands in the form of normalised charts, as shown in Fig. 15. In this study, the yield ratio and curvature sensitivity have been used as parameters depicting the response ‘mechanism’ and ‘sensitivity’, respectively. • Model parameters related to the stiffness of liquefied soils (particularly the stiffness degradation ratio) are the most important prior to yielding of the soil. Conversely, strength parameters for the liquefied soil dominate post-(soil) yielding behaviour. • The sensitivity of the pile response to variation of liquefied soil parameters decreases with the thickness of the crust layer. For a lateral spreading soil with a thickness of 10 m, a 1.5 m crust reduced the sensitivity of the pile response to the properties of the liquefied layer to 30–50% of the no-crust case. • The pile response is relatively less sensitive to the properties of the base layer. • Uncertainties in the geotechnical characterisation parameter (SPT blow count in this study) are significant and consistently affect both stiffness and strength related model parameters. Note that here, unlike for most geotechnical problems, a higher blowcount may correspond to a greater demand (as in the case when the soil is driving the movement of the pile rather than resisting it), illustrating the potential of deterministic parametric analyses for revealing counterintuitive mechanical interactions. • The study demonstrates that for any given soil-pile system (baseline model) the pile response may change significantly when varying a given model parameter within the acceptable (realistic) range of values. This clearly points out to the need for parametric studies when evaluating the pile response by means of a pseudo-static analysis. The proposed preliminary assessment of the pile response can be summarised as: 1. Define reference (best guess) values for all model parameters for the given soil-pile system and lateral spreading demand considered. 2. Identify potential sources of uncertainty and their effects on model parameters and specify lower and upper bound values for relevant model parameters. 3. Specify the range and magnitude of relevant levels/combinations of demand (ground displacements). 4. Perform 2n pseudo-static analyses for each demand combination, setting each uncertain model parameter in turn at its lower and upper bound values. 5. Conduct a baseline (reference) analysis for each demand combination. 6. Note that a separate (additional) pseudo-static analysis will be required to investigate the pile response during the cyclic phase when inertial and kinematic demands concurrently occur. Finally, it should be recognised that the considerable uncertainties affecting the performance of piles mean that ‘unique’ or ‘exact’ values for the various parameters in the model cannot be defined and the precise response of the pile cannot be ‘predicted’. Ultimately, these simplified methods provide the engineer with a tool that should enhance, but not replace, their engineering judgement.