CPT-based evaluation of liquefaction potential for fine- grained soils

Recent ground failure case histories after 1994 Northridge, 1999 Kocaeli and 1999 Chi-Chi earthquakes revealed that low-plasticity silt-clay mixtures generate significant cyclic pore pressures and can exhibit a strain-softening response, which may cause significant damage to overlying structural systems. In this study, results of cyclic tests performed on undisturbed specimens of ML, CL, MH and CH types were used to study cyclic shear strain and excess pore water pressure generation response of fine-grained soils. Based on comparisons with the cyclic response of saturated clean sands, a shift in pore water pressure ratio (ru) vs. shear strain (max) response is observed, which is identified to be a function of PI, LL and (wc/LL). Within the confines of this study, i) probabilistic based boundary curves identifying liquefaction triggering potential in the ru vs. max domain were proposed as a function of PI, LL and wc/LL, ii) these boundaries were then mapped on to the normalized net tip resistance (qt,1,net) vs. friction ratio (FR) domain, consistent with the work of Cetin & Ozan (2009). The proposed framework enabled CPT-based assessment of liquefaction triggering potential of fine-grained low plasticity soils, differentiating clearly both cyclic mobility and liquefaction type soil responses. 1999 Chi-Chi once again showed that silty and clayey soils can undergo seismicallyinduced soil liquefaction. Consistent with the advances in seismic soil liquefaction engineering, susceptibility assessments of fine grained soils revolved from pioneering Chinese Criteria 1979, to the methodologies of Seed & Idriss 1982, Seed et al. 2003, Bray & Sancio 2006). Considering the limitations of these studies, which will be presented later in this paper, an alternative framework with a theoretical background is proposed to assess the liquefaction susceptibility of fine-grained soils. The inspiration behind the proposed framework is due to the observation that cohesionless soils have a unique pore water pressure ratio (ru) vs. shear strain (max) response, and compared to saturated clean sands, a shift in pore water pressure ratio (ru) vs. shear strain (max) response is observed in cohesive soil samples. This shift is identified to be a function of PI, LL and (wc/LL). Thus, (ru) vs. shear strain (max) domain is decided to be used to differentiate ‘sand-like’ and ‘clay-like’ responses. The proposed framework provides liquefaction susceptibility boundary curves as a function of soil index parameters (PI, LL, wc/LL). The boundary curves developed in ru vs. γmax domain are then mapped on to CPT domain (qt,1,net vs. FR), consistent with the recent study of Cetin & Ozan (2009). After a brief review of existing methodologies, data compilation and process efforts, and development of proposed framework will be discussed in the following sections of this manuscript. 2 EXISTING LIQUEFACTION SUSCEPTIBILITY CRITERIA Based on liquefaction-induced ground failure case histories compiled from predominantly fine grained soils sites after 1975 Haicheng and 1976 Tangshan earthquakes, Wang (1979) proposed liquefaction susceptibility assessment rules widely referred to as Chinese Criteria. Chinese Criteria and its improved versions have been widely used (e.g. Seed & Idriss 1982, Andrews & Martin 2000) in practice. However, the ground failure case histories after 1989 Loma-Prieta, 1994 Northridge, and especially 1999 Kocaeli and 1999 Chi-Chi earthquakes revealed that neither Chinese Criteria nor these improved versions can successfully discriminate potentially liquefiable and non-liquefiable fine grained soils. Inspired from this gap, Seed et al. (2003), Bray & Sancio (2006), and Boulanger & Idriss (2006) proposed new liquefaction susceptibility criteria based on field observations and laboratory test results. A summary of these criteria is presented in Table 1. For the assessment of liquefaction triggering potential, first step is to determine whether the soil is potentially liquefiable or not. For this purpose, “Chinese Criteria” had been widely used for many years. However, contrary to Chinese Criteria, recent advances revealed that i) non-plastic fine grained soils can also liquefy, ii) PI is a major controlling factor in the cyclic response of fine grained soils. These criteria are then modified by Andrews & Martin (2000) for USCS-based silt and clay definitions. Bray et al. (2001) has concluded that the use of Chinese Criteria percent “clay-size” definition may be misleading and rather than the % of clay size material, their activities are judged to be more important. Seed et al. (2003) recommended a new criterion inspired from case histories and cyclic testing of “undisturbed” fine-grained soils compiled after 1999 Kocaeli and Chi-Chi earthquakes. These criteria classify saturated soils with a PI < 12 and LL < 37 as potentially liquefiable, provided that the wc is greater than 80% of the LL (0.8·LL). Recently, Bray & Sancio (2006), based on mostly cyclic triaxial and some simple shear test results performed on Adapazari undisturbed fine grained soils developed their liquefaction susceptibility criteria, summarized in Table 1. Valid for both Seed et al. and Bray & Sancio methodologies, laboratory test-based liquefaction triggering definition was not clearly presented. Bray& Sancio (2006) adopted 4 % axial strain as liquefaction triggering criterion. However, tests were performed under CSR levels of 0.3, 0.4 and 0.5 and loading cycles were continued if and until this strain level was reached. Thus, their conclusions are judged to be constrained by CSR and durational levels adopted for their testing program. The most recent attempt for determining potentially liquefiable soils was by Boulanger & Idriss (2006). Based on cyclic laboratory test results and an extensive engineering judgment, they have recommended new criteria summarized in Table1. As part of this new methodology, deformation behavior of fine-grained soils are grouped as “Sand-Like” and “Clay-Like”, where soils within the sand-like behavior region are judged to be susceptible to liquefaction and have substantially lower values of cyclic resistance ratio, CRR, than those within the clay-like behavior region. The main drawback of the methodology is the fact that the y-axis of Figure 1 is not to scale, thus a direct comparison between CRR of “clay-like” and “sand-like” responses is not possible. Also, very little, to an extent of none, is known about if and how identical or comparable “sand-like” and “clay-like” samples were prepared. Table 1. A summary of available liquefaction susceptibility criteria for fine grained soils Assessment Method Potentially Liquefiable Test for a Decision Non-liquefiable Chinese Criteria Wang (1979) FC ≤15% LL ≤35% wc ≥ (0.9*LL)% Otherwise Andrews and Martin (2000) Clay content, CC<10% LL < 32% CC<10% & LL≥32% -CC ≥ 10% & LL < 32% CC≥10% & LL≥32% Seed et al. (2003) PI < 12% LL < 37% wc/LL > 0.8 12 < PI < 20 37 < LL < 47 wc/LL > 0.85 Otherwise Bray and Sancio (2006) PI<12% wc/LL > 0.85 12 < PI < 18 wc/LL > 0.80 Otherwise Boulanger and Idriss (2006) PI < 3% 3≤ PI ≤ 6 PI ≥ 7 Although these studies are judged to be improvement over earlier studies, they suffer from one or more of the following issues: (i) there is no unique definition of liquefaction and hence, each criterion is developed based on different understandings regarding what liquefaction response is, (ii) the amplitude of cyclic loading is not specified in maxor ru-based exceedence of threshold definition; as a consequence there exist ambiguity under which cyclic stress conditions these criteria are applicable, and iii) most of these studies fail to differentiate cyclic liquefaction and mobility type soil responses. 3 DATABASE COMPILATION EFFORTS For the purpose of discriminate between the responses of cohesionless and cohesive soils, cyclic test results of both types of soils were studied. The databases studied and compiled consist of tests performed on: i) laboratory reconstituted clean sands (Wu et al. 2003 and Bilge 2005) and ii) “undisturbed” fine-grained soils (Pekcan 2001, Sancio 2003, and Bilge, in prep.). The compiled database is composed of 158 cyclic test results including ru vs. max histories, Atterberg limits along with moisture content of specimens, consolidation and applied cyclic shear stress conditions. Table 2 briefly summarizes the data sources used in this study, and compiled data is presented on ru vs. max domain in Figure 2. More detailed information regarding these data sources and details of data processing can be found in the original references. Figure 1. Criteria for discriminate between sandand clay-like soil behavior (Boulanger & Idriss 2006) max (%) 0.001 0.01 0.1 1 10 100 ru 0.0 0.2 0.4 0.6 0.8 1.0 Pekcan (2001) Bilge (in prep.) Wu et al. (2003) Bilge (2005) Sancio (2003) Figure 2. Compiled database in ru vs. max domain Table 2. Summary of the data sources used in this study Type Data Source # of data Test Type Tested Material Coarsegrained Wu et al. (2003) 50 Simple Shear Monterey No.0/30 Sand Bilge (2005) 36 Cyclic Triaxial Kizilirmak Sand Finegrained Pekcan (2001) 7 Cyclic Triaxial “undisturbed” Adapazari Sancio (2003) 15 Cyclic Triaxial “undisturbed” Adapazari Bilge (in prep.) 50 Cyclic Triaxial “undisturbed” Adapazari, Ordu 4 DEVELOPMENT OF PORE WATER PRESSURE GENERATİON MODEL As discussed earlier, the shift in (ru) vs. shear strain (max) response of cohesive soils relative to cohesionless ones is identified to be a function of PI, LL and (wc/LL). Thus, (ru) vs. shear strain (max) domain is decided to be used to differentiate ‘sandlike’ and ‘clay-like’ responses. The proposed framework provides liquefaction susceptibility boundary curves as a function of soil index parameters (PI, LL, wc/LL). Selection of a limit state model capturing the important features of the observed behavior is the first step for development of a probabilistic model. The limit state function has the general form of g = g (x, Θ) where x is a set of descriptive parameters and Θ is set of unknown model coefficients. Inspired by previous studies and observed trends from tests, it is concluded that for cohesive soils, key parameters affecting ru response are max, PI, LL and wc/LL. Inspired mainly by the recent study of Cetin & Bilge (in prep.), given for cohesionless soils, various functional forms have been tested (Pehlivan 2009) and consistent with maximum likelihood methodology the following functional form is selected as the limit state model as it results in greater likelihood value and smaller model error, which are the indications of a superior model.   ) ln( max ) exp( 1 ln ) ln( ) , ( ˆ u r u u r r g        (1)       10 7 4 1 1 1 9 8 6 5 3 2 1                          ) LL / w ln( ) LL ln( ) PI ln( c max (2) where  is a random model correction term to account for possibilities of i) missing descriptive variables, and ii) imperfection of the adopted mathematical expression. It is reasonable and also convenient to assume that  follows a normal distribution with a mean of zero for the aim of producing an unbiased model. The standard deviation of  () is unknown and must be estimated. Both the unknown coefficients and  were determined via maximum likelihood analysis and their corresponding values are presented in Table 3. Figure 3 presents the boundary curves developed for the mean values of compiled database, PI=22, LL=45, wc/LL=0.82 along with + one standard deviation () curves and compiled data. This figure revealed that proposed model and the suggested error bands captures the observed soil response successfully. Rather than considering only soil index parameters, this correlation also accounts for the accumulated shear strain which is related to amplitude and duration of cyclic loading. Moreover, by this way the mechanisms governing cyclic response of soils can be taken into account. Cyclic stress-strain relationships of soils are usually defined through degradation of shear modulus as a function of cyclic shear strain. As pointed out previously (e.g. Seed & Idriss 1970) remolding (i.e. strain accumulation) and loss in effective stress (i.e. pore water pressure generation) play an integral role in stiffness degradation. The other factors affecting this degradation response have been studied by various other researchers (e.g. Vucetic & Dobry 1991). Founding on this theoretical background, a robust relationship between ru and max is developed and this relationship will be the basis of our framework which will be consequently valid for any liquefaction definition, take into account the significance of stress amplitudes and also be able to differentiate cyclic liquefaction and mobility type soil responses. 5 NEW LIQUEFACTION CRITERIA Development of new liquefaction susceptibility criteria requires a definition for triggering of liquefaction. Considering both previous efforts and trends observed from available experimental data, for fine-grained soils liquefaction is defined as follows: For max = 7.5%, if induced ru is between 0.85 and 1.0 then soil is classified as potentially liquefiable (sand-like). If ru is less than 0.7 at max = 7.5%, then it is classified as potentially nonliquefiable (clay-like) and in between these limits further testing may be required. Validity of these criteria is assessed by using available test data and it was observed that the error in identification of cohesive soils susceptible to liquefaction was not greater than 10%. Figure 4 presents the proposed liquefaction susceptibility criteria for wc/LL=1.0 condition on plasticity chart. Table 3. Model coefficients CoarseGrained FineGrained