Site characterisation for the Ballina field testing facility Free

The comprehensive laboratory and in situ measurements reported by Kelly et al. (2017) and also by Pineda et al. (2016) on geotechnical characteristics of Ballina soft clay in New South Wales, Australia, deserve further interpretation. The stratigraphy is uniform between reduced levels of −2 m to −10 m; therefore, this interpretation is limited to Ballina soft clay at mid-depth of −6 m. At this depth, plasticity index, I_p, is near 70%; clay size fraction, CF, is near 70%; and in situ effective vertical stress is 42 kPa. Ballina soft clay was deposited about 8000 years before present (BP) in saline water in an estuarine environment.

Before introduction of the constant rate of strain consolidation test (CRS) most data on preconsolidation pressure of soft clay deposits had originated from incremental loading (IL) consolidation tests, and corresponded to an end-of-primary (EOP) e–log σ′_v relation of 20 mm thick specimens. Therefore, an expression was developed for strain rate imposed on the CRS test that would lead to the EOP e–log σ′_v relationship (Mesri & Feng, 1992; Mesri et al., 1994a)

where k_v0 is the initial permeability in vertical direction; C_c/C_k has typical values of ½ to 2 (Mesri & Rokhsar, 1974); H is maximum drainage distance; σ′_p is preconsolidation pressure; γ_w is unit weight of water; and, for soft clays, C_α/C_c = 0·04 ± 0·01 (Terzaghi et al., 1996). An imposed vertical strain rate equal to $ε˙p$ leads to excess pore water pressures near zero at the bottom (impermeable boundary) of a CRS specimen. For soft clay deposits, the most typical value of $ε˙p$ is equal to 2·5 × 10⁻⁷ s⁻¹ (Mesri et al., 1994a).

An imposed vertical strain rate equal to $ε˙p$ results in a rather long duration of the CRS test, and very small, near-zero excess pore water pressures do not allow the coefficient of permeability to be computed using the CRS porewater pressure measurements. Therefore, Mesri & Feng (1992) proposed for the CRS test an imposed vertical strain rate

which produces u′_b/σ_v values in the range of 3– 5% (where u′_b is the excess pore water pressure at the bottom of the specimen and σ_v is the imposed total vertical stress), and allows the Darcy flow equation to be used for reliable calculation of the coefficient of permeability (Mesri et al., 1994a, 1994b). However, the resulting preconsolidation pressure must be corrected to obtain EOP σ′_p (Mesri & Choi, 1979, 1984; Mesri, 1987; Terzaghi et al., 1996).

For Ballina clay, the value of σ′_p/σ′_v0 is near 1·77 from the CRS oedometer data in Fig. 4(b). However, the imposed strain rate was 3·3 × 10⁻⁶ s⁻¹ for the CRS tests on Ballina soft clay, compared to $ε˙p$ = 3·4 × 10⁻⁷ s⁻¹ from equation (2) for k_vo = 10⁻⁹ m/s, C_k/C_c = 1, C_α/C_c = 0·04 and σ′_p = 68 kPa. Therefore, using equation (4) together with C_α/C_c = 0·04, leads to σ′_p/σ′_vo = 1·61, and for σ′_v0 = 42 kPa, to σ′_p = 68 kPa.

For cone penetration tests, according to Mesri (2001) 

From the cone penetration tests on Ballina clay, at −6 m, q_t = 327 kPa and σ_v = 100 kPa lead to σ′_p = 64 kPa.

For a geologically normally consolidated clay, the value of σ′_p/σ′_v0 resulting from secondary compression and thixotropic ageing alone is (Mesri & Choi, 1979, 1985; Mesri, 1993)

Using t = 8000 years, t_p = 10 years, C_α/C_c = 0·04, C_r/C_c = 0·1, and thixotropic hardening parameter β = 0·02, it is possible to calculate σ′_p/σ′_v0 = 1·53. The difference between 1·61 and 1·53 has probably resulted from groundwater fluctuations.

For Ballina clay, the undrained shear strength was estimated from field vane tests (FV), triaxial compression tests (TC) and triaxial extension tests (TE). At −6 m depth, the value of s_uo (FV) is 20·0 kPa from Fig. 15. Thus s_uo (FV)/σ′_p = 0·294. However, from the Bjerrum–Mesri s_uo (FV)/σ′_p against I_p relationship, for I_p = 70%, s_u0 (FV)/σ′_p = 0·32 (Terzaghi et al., 1996, fig. 20·20). For Ballina clay at −6 m depth s_uo (TC)/σ′_p is 0·300, and s_uo (TE)/σ′_p is 0·263.

The undrained shear strength mobilised in failure of embankments, footings and excavations is

• (a)

  Bjerrum (1972, 1973) with Bjerrum correction μ_B = 0·70 for I_p = 70% (Terzaghi et al., 1996, fig. 20·21).

  $su0(mob)σ′p=μBsu0(FV)σ′p$

  7

  = 0·21

  s_u0 (mob) = 0·21 × 68 = 14·3 kPa

• (b)

  Mesri (1975), for inorganic soft clays

  $su0(mob)σ′p=0⋅22$

  8

  s_u0 (mob) = 0·22 × 68 = 15·0 kPa
• (c)

  Mesri (2001), mobilised undrained shear strength from cone penetration test

  $su0(mob)=qt−σvNk(mob)$

  9

  where q_t is the corrected cone tip resistance, σ_v is the total vertical stress and N_k (mob) = 16.
  $su0(mob)=327−10016=14⋅2kPa$
• (d)

  Mesri (1989) and Mesri & Huvaj (2007) mobilised undrained shear strength from laboratory tests. In the absence of s_u0 (DSS)/σ′_p

  $su0(mob)σ′p=12su0(TC)σ′p+su0(TE)σ′pμt$

  10

  where μ_t is a time to failure correction factor as a function of plasticity index, I_p, equal to 0·83 for I_p = 70% (Terzaghi et al., 1996, fig. 20·24).

  Note that shear strength from laboratory tests from Jamiolkowski et al. (1985) and Ladd (1991) have been incorrectly reproduced in Fig. 21.

  $su0(mob)σ′p=120⋅300+0⋅2630⋅83$

   = 0·23

  s_u0 (mob) = 0·23 × 68 = 15·6 kPa

Therefore, for Ballina soft clay at −6 m depth, the most typical value of s_uo (mob)/σ′_p = 0·22, and s_uo (mob) = 15 kPa.

According to Fig. 11(d), K₀ = 0·560, and according to Fig. 13, ϕ′_cv = 34°, where ϕ′_cv is the constant volume friction angle (Terzaghi et al., 1996).

The Schmidt (1967) equation for K₀ = σ′_h0/σ′_v0 is (Mesri & Hayat, 1993)

where K_0p is the coefficient of lateral pressure in the compression range beyond σ′_p, and according to Jaky (1948) 

thus K_0p = 0·4408 and K₀ = 0·575.

The Mesri & Hayat (1993) equation for K₀ of soft clay and silt deposits is

using K_0p = 0·4408 and σ′_p/σ′_v0 = 1·61, K₀ = 0·575.

Thus the predictions of K₀ = σ′_h0/σ′_v0 by Schmidt (1967) and Mesri & Hayat (1993) empirical equations are in agreement with values reported in Fig. 11(d).

The compressibility of soft clay deposits is completely defined by the EOP e–log σ′_v relationship of each sublayer, and C_α/C_c. For strongly non-linear EOP e–log σ′_v relations in the compression range beyond σ′_p, such as those for Ballina clay in figs 6 and 7 of Pineda et al. (2016), the values of tangent compression index, C_c, such as those in Fig. 9 of the paper under discussion cannot be readily incorporated into a simple settlement equation. The strongly non-linear compression curve is best described in terms of values of secant compression index, C′_c, as a function of log σ′_v/σ′_p (Mesri & Choi, 1985, Terzaghi et al., 1996: figs 16·16 and 16·7, equation (16·9); Mesri & Funk, 2015: figs 7 and 8).

The values of C_α/C_c for Ballina soft clay are within the range of 0·04 ± 0·01 for soft clay and silt deposits (Terzaghi et al., 1996, table 16·1). Because, in the recompression range C_α increases with time, frequently, an incorrect value of C_c is paired with C_α, thus resulting in high values of C_α/C_c as in fig. 12 of Pineda et al. (2016).

According to Fig. 18(b) a typical value of k_v is 10⁻⁹ m/s for Ballina soft clay, which is also the most common k_v0 for soft clay deposits. An empirical equation for k_v0 (Mesri et al., 1994b), in terms of e₀/CF and activity, A_c = I_p/CF, for e₀ = 3·0, CF = 70% and A_c = 1·0, predicts k_v0 = 1·4 × 10⁻⁹ m/s. However, for Ballina soft clay, the value of C_k is near 1·0, based on data in fig. 11(b) of Pineda et al. (2016), whereas for most soft clays C_k is close to e₀/2, which for Ballina soft clay is 1·5 (Tavenas et al., 1983; Mesri et al., 1994b). According to Fig. 4(d) a typical value of c_v for Ballina soft clay is 3 m²/year, which is a common value for many soft clays; however, the lower range of 0·3–0·5 m²/year is more consistent with the Ballina soft clay liquid limit of 110%.

In summary, the observed behaviour of Ballina soft clay, in general, is remarkably consistent with those of most soft clay deposits. Therefore, an excellent selection has been made for a field testing facility in Australia.

The authors would like to thank Professor Mesri and Mr Kane (the discussion contributors) for their valuable contribution to the analysis of the behaviour of Ballina clay. In this reply the authors wish to provide some additional information and clarifications on the points raised by the discussion contributors.

To facilitate the discussion, the authors also refer to the interpretation of the information presented in Kelly et al. (2017) and Pineda et al. (2016) by academics and practitioners, to predict the behaviour of a trial embankment built on Ballina clay improved by prefabricated vertical drains. These predictions were presented in a symposium organised by the ARC Centre of Excellence for Geotechnical Science and Engineering in Newcastle (Australia) in September 2016 (CGSE, 2016). One lesson learned from this symposium was that the interpretation of the same set of laboratory and in situ data by different predictors may result in large variations in adopted soil properties, even among experienced predictors who used the same analysis approach. Hence the authors believe that it is worth elaborating more on the derivation of some key Ballina clay parameters, in the light of the discussion contributors’ analysis and limited additional data. Detailed outcomes of the above-mentioned symposium will be reported in a special issue to be published in Computers and Geotechnics later in 2017. Note that, in the following, the same notation is used as employed by the discussion contributors, for the purpose of consistency.

The discussion contributors have estimated values of preconsolidation pressure using three different methods proposed by Professor Mesri and his co-workers. Values reported by Pineda et al. (2016) (Table 2) were obtained at strain rates larger than the value imposed during IL tests which, for soft clays, typically ranges between 1·0 × 10⁻⁷ s⁻¹ (Watabe et al., 2012) and 2·5 × 10⁻⁷ s⁻¹ (Mesri et al., 1994a; Terzaghi et al., 1996). Although a correction factor was suggested based on the work by Watabe et al. (2012), uncorrected values were intentionally reported in Pineda et al. (2016), allowing the predictors to make their own estimates.

Most of the predictors did not account for rate effects and that was, in the authors’ opinion, a major contributor to under-prediction of embankment settlement. The authors attribute this to some of the predictors not being familiar with the interpretation of the CRS test. For the depth analysed by the discussion contributors (5·5 < z < 6 m), the corrected preconsolidation pressure leads to σ′_p/σ′_v0 = 1·49 (similar to the lower bound estimated by the discussion contributors). Fig. 23(a) presents recent measurements from a creep-IL test carried out on a Sherbrooke (block) specimen obtained from borehole BH1 (depth 5·24–5·61 m), drilled 100 m north from borehole inclo 2. The results shown in Fig. 23 were obtained from long-term loading steps (14 days), and the maximum vertical effective stress reached was 500 kPa. Small loading increments were applied in order to obtain reliable estimates of the preconsolidation pressure. Fig. 23(b) shows the ε_v–log σ′_v curves for constant rates of strain between 10⁻⁸ and 10⁻⁶ s⁻¹, back-analysed from the creep test following the procedure described by Leroueil et al. (1985). The preconsolidation pressure ranges from 57 kPa to 68 kPa for strain rates of 10⁻⁸ s⁻¹ and 10⁻⁶ s⁻¹, respectively, and a value of σ′_p ≈ 62 kPa is obtained for a strain rate equal to 10⁻⁷ s⁻¹. This value is in agreement with the lower bound estimate proposed by the discussion contributors, as well as with the values reported in Pineda et al. (2016), once corrected by rate effects.

The discussion contributors are right in the fact that laboratory tests from Jamiolkowski et al. (1985) and Ladd (1991) have been incorrectly replicated in Fig. 21. Owing to an error when digitising the corresponding figures, values of s_u/σ′_yield as well as the trend lines were unintentionally shifted down compared with the original data. The corrected Fig. 24 presents the original test results together with triaxial data obtained from Ballina clay. Note that the general trends remain unchanged, but the triaxial extension tests from Ballina clay lie below the trend line drawn by Ladd (1991).

Various approaches have been used to estimate the coefficient of earth pressure at rest from seismic dilatometer tests (Marchetti, 1980; Powell & Uglow, 1988; Lacasse & Lunne, 1988; Kouretzis et al., 2015). A range of K₀ profiles were obtained from each interpretation method, including independent estimations from a self-boring pressuremeter test (Gaone et al., 2016). Interpretation of push-in pressure cell (PiPCs) measurements showed good agreement with the lower bound profiles obtained using the expressions proposed by Lacasse & Lunne (1988) and Kouretzis et al. (2015). Independent checks were made using similar expressions to those presented by the discussion contributors. From the analysis of the triaxial compression tests, a constant volume friction angle equal to 36° was adopted. This value is similar to the constant volume friction angle reported by Mayne (2016) for a wide range of natural soft soils. A comprehensive laboratory study using high-quality Sherbrooke (block) specimens is currently underway at the University of Newcastle in order to assess K₀ for Ballina clay specimens corresponding to various depths.

The authors agree with the discussion contributors that the strongly non-linear compressibility curves should be incorporated into settlement assessments – for example, using secant values of C_c over the appropriate stress range as proposed by Mesri & Choi (1985). For the prediction of the trial embankment behaviour, most symposium contributors adopted conventional bi-linear void ratio–log stress curves in their analyses and used ‘average’ C_c values based on their own judgement. Although a wide range of values was finally employed by the various predictors, there was a tendency to use values measured at stress levels much higher than the one imposed by the embankment (excluding analyses where advanced constitutive models incorporating destructuration effects were employed). In the authors’ opinion, this also contributed to under-prediction of embankment settlement. The authors agree with the discussion contributors that explicitly accounting for destructuration is preferable to the use of average values and judgement, particularly when high-quality laboratory tests are available.