A note on earthquake-induced liquefaction

Abstract

This technical note demonstrates that earthquake-induced liquefaction consists of a cyclic instability dominated by heterogeneities in the soil, and as such is a phenomenon that is fundamentally intractable analytically. Consequently it is shown that the liquefaction analyses proposed by the Berkeley school are fundamentally incorrect, not just after zero effective stress has been reached in the soil, but well before that - that is as soon as any pore pressure is generated and any softening of the soil takes place. The method is fundamentally in error in assuming that dynamic stressing of the soil takes place independently of soil softening. A distinction is also made between liquefaction failure and flowslides, which is a phenomenon originally addressed by the Harvard school in its studies of soil behaviour at densities looser than critical. It is argued that the only way in which reliable design procedures against liquefaction can be developed is by the creation of empirical design criteria based on direct experimental data. Where possible this should involve experimentation at full scale, which is by definition the most accurate method. However, such data is necessarily very limited and very expensive, and so it is suggested that these design rules should be developed from the second most accurate experimental method, namely centrifuge modelling, and then calibrated wherever possible against measured full-scale results. It is also suggested that the most accurate form of in situ index testing of soil liquefaction potential will incorporate the self-boring pressuremeter, and some preliminatry recommendations are given in this note. (Author/TRRL)