Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 Liquefaction and dynamic properties of gravelly soils M.D. Evans Civil Engineering Department, Northeastern University, ^20 Snell Engineering Center, CASL4 ABSTRACT Liquefaction of gravelly soil has received much attention in recent years as several case histories have been reported where gravel and gravelly soil has liquefied. As a result, much research has been focused on assessing the liquefaction potential of gravelly soils both in situ and in the laboratory. Laboratory liquefaction assess- ment typically includes performing undrained, cyclic triaxial tests to determine liquefaction resistance and dynamic material properties. However, the potentially adverse effects of membrane penetration and compliance must be considered. Membrane compliance may result in pore fluid redistribution, soil densification, and increased liquefaction resistance in the undrained triaxial test, making it difficult to properly evaluate the performance of the prototype material in situ. This paper presents results of undrained, cyclic triaxial tests performed on gravel specimens and sluiced gravel specimens. This paper will show the corrected cyclic strength of gravel specimens to be as low as 60% to 70% of the uncorrected strength. Increases in gravel specimen density of up to 20 percentage points or more due to membrane compliance are also documented. The bases for these determinations are presented in this paper. INTRODUCTION The phenomenon of liquefaction of sandy soil is fairly well understood by the geotechnical profession. Many researchers have investigated the phenomenon over the last 30 years (Seed and Lee*, Castro^, Seed^, and many others), and it is generally well accepted that loose, clean sands may develop excess pore pressures during earthquake loading, and may undergo extreme strength loss and large deformations. Liquefaction of gravelly soils, however, may not be so well estab- lished in the geotechnical community. For years, gravels were considered to be completely free draining and no consideration was given to assessing liquefaction Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 318 Soil Dynamics and Earthquake Engineering potential. Gravelly soils may be free-draining under ideal conditions, but drainage is often impeded. A gravelly soil may be bounded by a low-permeability cap, for example, temporarily restricting drainage. Such formations may occur in an alluvial fan containing gravelly soil interlayered with finer material, or in a gravel embankment where drainage is impeded by silt deposition on the upstream face (Evans et al.4). Evans and Harder^ summarized several case histories where gravelly soil has liquefied in situ, including gravelly soil in two embankment dams. Grain size distributions for some of these soils are shown in Figure 1. Since gravelly soil liquefaction has been documented, more attention has been devoted to assessing the liquefaction potential of such soils. Indeed, many researchers have investigated various aspects of gravel liquefaction in the triaxiai test (Wong&, Banerjee et al7, Evans and Seed*, Hynes^, Seed et al.^, and Evans^). However, gravel-sized particles present unique complications to conventional sampling and laboratory testing techniques. Gravel particles create membrane compliance problems in triaxiai tests, artificially reducing the laboratory liquefaction potential, .resulting in an unconservative assessment of in situ liquefaction potential. Thus, assessing the liquefaction potential of a gravelly soil presents unique challenges to the design engineer. This paper will address how some of the challenges associated with laboratory assessment of liquefaction potential may be overcome. Sand Gravel Rockfdl (1) Shimen Dam (Wang, 1984) " (2) Pence Ranch (Harder, 1988) (3) Whiskey Springs (Harder, 1988; Andrus et al., 1986) (4) Baihe Dam (Tamura and Lin, 1983) I _1_ 10 100 1000 Grain Size (mm) Figure 1: Grain size curves for gravelly soils that have liquefied. Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 Soil Dynamics and Earthquake Engineering 319 CYCLIC TRIAXIAL TESTING IN LIQUEFACTION ANALYSIS Introduction It is always preferable to sample and test high quality, undisturbed samples from the soil layer of interest. Sampling gravelly soil can be extremely difficult, however, due to lack of cohesion and large particle sizes. Therefore, specimens are typically reconstituted in the laboratory to accurately model in situ conditions, especially density, structure, and stress history (Mulilis et al.^). The specimens are installed in the triaxial cell, subjected to lateral and axial stresses representative of the in situ effective stress, allowed to consolidate, and then subjected to a cyclic deviator stress, a^c, under undrained conditions until the sample liquefies. Several test specimens are subjected to various cyclic stress levels to define a relationship between cyclic stress ratio, Gfo/2G-$c, and number of cycles required to cause liquefaction, N/. Test Program In this paper, a comparative study was made between the results of sluiced and unsluiced gravel specimens. Approximately half the specimens were tested in a conventional, compliant system; and the other half were tested in a specially pre- pared, low-compliance system prepared by sluicing, or washing sand into the voids of the gravel specimens. This procedure filled the peripheral specimen voids with sand, significantly reducing the amount of membrane penetration that occurred during consolidation, also minimizing membrane compliance effects during undrained loading. Grain size distributions of the gravels and sluicing sands used in this study may be found in Figure 2. Specimens were constructed by dry pluviation following generally accepted procedures. A detailed description of the sluicing procedure and control is presented by Evans and Seed&. The membranes used to confine the 71-mm diameter specimens were manufactured of latex rubber by 3-D Polymers of Gardena, California. They were 69 mm in diameter, 230 mm tall, 0.30 mm thick, with an elastic modulus of about 1330 kPa. Drained hydro- static compression and rebound tests and undrained cyclic triaxial loading tests were performed on sluiced and unsluiced 71-mm and 3 05-mm diameter specimens composed of uniformly-graded gravel at various relative densities. The results of some of these tests are presented below. Membrane Penetration and Compliance Before any effective confining pressure is applied to the triaxial specimen, the confining membrane is stretched flat over the surface of the specimen, bridging the peripheral sample voids. The membrane will penetrate into the peripheral voids, continuing to penetrate further with each effective pressure increase until no more penetration is possible. Figure 3 shows a photograph of a 71-mm diameter gravel specimen (9.5-mm by 4.75-mm particles) confined with a single membrane. The severe degree of membrane penetration is apparent in this figure. Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 320 Soil Dynamics and Earthquake Engineering Sand , Gravel Rockfill 100 , / San Francisco 80 TT- Dune Sand /Aswan High Dam • 9.5-mm by 4.75-mm / 50-mm Maximum Gravel I / Parallel Gradation 60 .= 40 , 'Aswan High Dam - / 50-mm Maximum 20 /Modified Gradation ^wan High Dam - Rockfill Field Gradation 0.1 1 10 100 1000 Grain Size (mm) Figure 2: Field and laboratory material gradations. Figure 3: 71-mm diameter gravel triaxial specimen Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 Soil Dynamics and Earthquake Engineering 321 Unit membrane penetration curves for 71-mm diameter triaxial sand speci- mens and unsluiced 9.5-mm by 4.75-mm gravel specimens are plotted in Figure 4. Unit membrane penetration in the gravel specimens is significantly greater than values for sand. It may also be seen that about 20% of the unit membrane pene- tration in two membrane systems is not recovered during unloading. Thus, unit membrane penetration could be overestimated by up to 20% by using the load portion of the curve rather than the unload portion. Lin and Selig^ found no significant difference between load and unload unit membrane penetration for specimens of medium to coarse sand. The sluiced gravel specimens tested in this study exhibited unit membrane penetration values in the range shown for sand in Figure 4 and showed no appreciable difference in behavior during loading versus unloading. The use of load versus unload curves should, therefore, be determined on a case by case basis. When the investigator is in doubt, unit membrane pene- tration should preferably be determined from unload curves. Effective Confining Pressure (kPa) 20 50 100 150 200 1.4 9.5-mm x 4.75-mm Gravel 71-mm Diameter Specimens 1.2 I— Sanaanda (btoy othersotnersj) 0.0 Figure 4: Unit membrane penetration curves for gravel (after Evans**). Transactions on the Built Environment vol 3, © 1993 WIT Press, www.witpress.com, ISSN 1743-3509 322 Soil Dynamics and Earthquake Engineering DENSITY CHANGES DUE TO MEMBRANE COMPLIANCE Introduction During undrained cyclic loading, the effective confining pressure is reduced as pore pressure develops and the membrane rebounds from the penetration sites. Water drains from interior voids and migrates to the peripheral voids previously occupied by the membrane as the membrane