Discussion of “Evaluation of Flow Liquefaction and Liquefied Strength Using the Cone Penetration Test” by P. K. Robertson

Abstract

Loose sands and silts are a situation of practical interest not only for mining companies with their large tailings dams but also for hydropower companies with some surprisingly large water retaining dams on poor foundations. In this context, the flow liquefaction strength sr is a basic input to any engineering assessment of stability. No other considerations can be entertained unless the dam is stable on a residual strength basis, which is an easy limit equilibrium calculation but with enormous (and partly hidden) complexity in the assessed strength. Robertson’s paper is a substantive contribution to the literature on sr . However, this discussion will suggest that the cone penetration test (CPT) is more valuable than just being a screening-level test. There are two aspects to this suggestion: the repeatability and detail of the CPT data and the range of data processing options. Engineers who use the CPT rapidly come to appreciate the repeatability and accuracy of the test—in essence, different testing companies can test the same ground and measure the same tip resistance and piezometric pressure profile to within the tolerances set by the electronic transducer nonlinearity, hysteresis, and accuracy—overall, about 1% full scale with available commercial equipment (e.g., ASTM 2007; Jefferies and Been 2006; Sandven 2010; Lunne 2010). As a comparison, even university research laboratories cannot achieve this repeatability in simple triaxial compression testing of reconstituted samples. The CPT’s repeatability in routine engineering practice, on its own, is a strong argument that the CPT should be viewed as the basic reference test for any site— and particularly in preference to standard (onshore) North American practice, which adopts the standard penetration test (SPT). A consequence of the CPT’s repeatability, and the continuous data recorded, is that it allows evaluation of soil variability at a site. The Lower San Francisco Dam (LSFD) case history shows the importance of such an evaluation. There are a number of studies of sr from the LSFD based on the SPT data. What is less appreciated is that the reconstructed dam section used in these SPT-based back analyses to estimate sr is inconsistent with the expected configuration on the basis of how the dam was constructed—they show a near-zoned dam, not the gradational layering and systematic segregation inherent in hydraulic fill construction that was used. Nor has the associated characteristic SPT penetration resistance allowed for the layering and gradation distribution. The CPT data, in contrast, with its far greater detail and resolution, reveals a dam section in accord with its known construction method, and where the effect of layering and gradation change can be included in estimating the characteristic penetration resistance. The inference is clear: a SPT-based case history approach for sr can mislead in general, and has at LSFD (Jefferies and Been 2006). The CPT should be the basic reference test for all liquefaction assessments, in contradiction to the recommendations of Youd et al. (2001). The traditional objection to the CPT has been the lack of soil samples obtained. Such an objection misunderstands penetration test data. Neither the CPT nor the SPT measure soil strength or state (density). Strength and/or state have to be recovered from the test data, which require a mechanistic framework. The usual mechanistic framework is cavity expansion theory, either using the simple dilating-frictional model (“nonassociated” Mohr Coulomb) or some advanced elastoplastic representation of the detailed stress strain behavior. But, regardless of the approach, mechanical parameters such as critical friction angle, in situ elastic shear modulus, and plastic compressibility control the penetration resistance (e.g., Shuttle and Jefferies 1998; Russell and Khalili 2004; Shuttle and Cunning 2007). Geological descriptors (e.g., “silty sand”) are completely absent. This is a further reason why the CPT is preferable to the SPT: geological descriptors, from the disturbed SPT sample, are a crude basis for estimating the needed mechanical properties. In contrast to the SPT, the CPT directly measures further mechanical response data in the Fs and Bq data streams. The addition of the seismic downhole measurements, routine with modern CPT equipment, cements the advantage of the CPT for liquefaction assessment. The question then becomes: how best to use the CPT in assessing sr . Clearly, the case history approach remains the gold standard, as such an approach includes attributes like partial drainage, scale, and variability. Accordingly, the author’s Fig. 5 is an important contribution to the profession. But, perhaps Figs. 6 and 7 would be improved if presented at 1σ rather than mean Qtn values (with an associated requirement to process data the same way) so as to capture the influence of soil variability, and it would be useful if the author could suggest the minimum number of CPT soundings at a site that are appropriatewhen invoking the proposeddesigncharts. As a screening-level assessment, the paper’s Figs. 5, 6, and 7 tacitly assume average soil properties with some indexation with soil type. But the accuracy of the CPT data warrants more formal treatment, and the discusser does not accept that high-risk projects should move beyond the CPT. Indeed, Figs. 5, 6, and 7 would become even more of a contribution if associated Gmax, compressibility, and critical friction data could be provided for the various case histories, as that additional information would allow the first steps in a rigorous assessment. However, for high-risk projects, there is no question that supporting laboratory data become needed—but this does not mean undisturbed samples; in the discusser’s view, what is needed is a few drained triaxial tests on reconstituted samples to establish the plastic hardening (normal compression index), critical friction ratio, and dilatancy behavior. Once these mechanical properties are known, formal processing of the CPT data offers a substantial improvement in understanding without a great deal of effort (see Jefferies and Been 2006). Equally important, however, is the measurement of in situ shear modulus Gmax by seismic methods; elastic modulus is as important as any other soil property in understanding the relationship between soil strength/state and penetration resistance. In summary, the remarkable repeatability, accuracy, and detail offered by modern electronic seismic-piezocones warrants far greater use of the CPT in practice—it should be the reference test for most soil testing. Although a screening-level evaluation of the data involves little effort, if greater precision in site characterization