Strength I Limit State using minimum and maximum load factors, respectively, from Table 4 11.)
GEOTECHNICAL SITE CHARACTERIZATION
5.2 Planning Exploration and Testing Programs
5.2.2 Soil and Rock Variability
In developing exploration and testing programs, the geotechnical engineer should qualitatively assess the effects of variables such as the expected type and importance of the structure, magnitude and rate of loading, and viability of foundation alternatives relative to technical, economic and constructibility considerations. From the planning stage, information regarding land use and topographic, surficial soil and geologic conditions can be used to define guidelines for developing subsurface exploration and testing programs. Sources of information available for many sites include:
y Topographic Maps - Landforms, ground slopes and shapes, and stream locations
y Aerial Photographs - Information on landforms, soil types, rock structure and stream types
y Agricultural Soil Maps - Landform, soil associations, soil descriptions and approximate engineering characteristics for surficial soils
y Well Drilling Logs - General description of soil and rock, and ground water levels at the time of drilling
y Existing Borings - Information from subsurface explorations in the vicinity of a structure.
For example, given the project information that a heavily-loaded bridge pier is to be located adjacent to an old river with low velocity might suggest the need to use some type of deep foundation systems due to the likely presence of relatively soft, fine-grained alluvial soils. With this information, the geotechnical engineer can determine the types and number of in-situ and laboratory tests needed for the expected geology, foundation type(s) to be used, load conditions to be evaluated (i.e., undrained versus drained loading) and stress history of the foundation soils. A drained analysis should be based on effective stress parameters and an undrained analysis should be based on total stress parameters. The stress history of the foundation soils is also an important factor in the design process. For example, whereas the stability of an embankment constructed on a lightly overconsolidated clay will be controlled by undrained behavior representative of the short-term case, a similar embankment built over a deposit of heavily over consolidated clay will probably be controlled by drained behavior representative of the long term case. Accordingly, selection of an appropriate in-situ test method must reflect these considerations so that the test program will yield the necessary information about the subsurface conditions. For example, whereas a SPT can be expected to provide reasonable information about the effective stress friction angle for a clean sand, it has almost no value in evaluating the undrained shear strength of a clay. Likewise, a CPT will not provide meaningful information in a deposit of coarse gravel.
At present, the variability of subsurface conditions and the level of subsurface exploration (i.e., number of borings) or testing (i.e., number of in-situ or laboratory tests) are not explicitly related to resistance factors used in the AASHTO LRFD Specification. Instead, planning subsurface exploration and testing programs are based on guidelines such as those recommended by FHWA (1988) in Table 5-1, the availability of information from previous explorations in the vicinity of the site, and/or engineering experience and judgment.
A simple quantitative measure of the variability of data or an engineering property is the coefficient of variation, COV, which is defined as:
x / =
COV σ (Eq. 5-1)
where:
COV = Coefficient of variation (dim); σ = Standard deviation of the data; and x¯ = Mean value of data.
Table 5-1 - Guideline Minimum Boring and Sampling Criteria
(Modified After FHWA, 1988)
Geotechnical Feature Minimum Number of Borings Minimum Depth of Borings
Structure Foundation 1 per substructure unit for width # 30 m 2 per substructure unit for width > 30 m
Advance borings: (1) through unsuitable foundation soils into competent
v
material of suitable bearing capacity and; (2) to a depth where )F < 10% of existing effective soil overburden stress or; (3) a minimum of 3 m into bedrock if bedrock is encountered at shallower depth.
Retaining Walls Borings alternatively spaced every 30 to 60 m in front of and behind wall.
Extend borings to depth of 2 times wall height or a minimum of 3 m into bedrock.
Culverts Two borings depending on length. See structure foundations.
Bridge Approach
Embankments Over Soft Ground
For approach embankments placed over soft ground, one boring at each embankment to evaluate embankment stability and foundation settlement. (Note: Borings for approach embankments usually located at proposed abutment locations to serve a dual function.)
See structure foundations.
Shallow explorations at approach embankment locations are an economical means to determine depth of unsuitable surface soils.
Cuts and Embankments
Borings typically spaced every 60 m (erratic conditions) to 150 m (uniform conditions) with at least one boring taken in each separate landform.
For high cuts and fills, 2 borings along a line perpendicular to centerline or planned slope face to establish geologic cross-section for analysis.
Cut: 1) In stable materials, extend borings a minimum of 3 to 5 m below cut grade.
2) In weak soils, extend borings below cut grade to firm materials, or to depth of cut below grade whichever occurs first.
Embankment: Extend borings to firm material or to depth of twice the embankment height.
The greater the value of COV, the less reliable the data. An example of how the COV can be used as a guide in planning a subsurface exploration program is illustrated in Figure 5-1.
Figure 5-1
Reliability Variation with Sample Size for Indirect Testing for Mobilized Undrained Shear Strength
(after Teng, et al., 1992)
Figure 5-1 relates the COV of the mobilized undrained shear strength, Su, estimated using cone
penetrometer (CPT) and field vane shear (VST) in-situ tests with the number of samples tested for a slope stability problem (Teng, et al., 1992). These in-situ tests are compared with a correlation of Su
with the preconsolidation stress, σ'p, as determined by laboratory consolidation tests. As expected,
the COV decreases with increasing numbers of samples until a limiting value is reached. The figure also shows that the field VST is a more reliable method for estimating Su than the CPT, and that the
field VST and laboratory test methods provide comparable results. In the future, it is conceivable that relationships such as those shown in Figure 5-1 could be used in designing site exploration programs with the type and number of tests selected to achieve a specified βT.
In addition to considerations regarding the type and extent of exploration, another important factor in planning any subsurface exploration is the cost-benefit relationship of the exploration program relative to construction cost. In general, as the amount and quality of subsurface exploration increases, the uncertainties and resulting conservatisms in the design process decrease. Conversely, an inadequate geotechnical exploration program can result in significant project cost overruns during construction as shown in the Table 5-2 which summarizes a recent study of 58 highway projects in the United Kingdom.
Table 5-2
Comparison of Ground Investigation Cost to Project Cost Overruns
(Modified after Whyte, 1995) Ground Investigation Cost/
Total Project Cost (%)
Mean Cost Increase Due to Geotechnical Origins (% of Total Project Cost)
# 1.5 14
1.5 - 2.0 8
> 2.0 4
The process of optimization of subsurface exploration with respect to the project foundation costs is illustrated from a conceptual perspective in Figure 5-2.
Figure 5-2
Optimization of Foundation and Exploration Cost
Figure 5-2a shows the efficiencies of scale which may be realized by an expanded exploration program, and Figure 5-2b shows the expected trend of decreasing foundation cost with the increasing reliability associated with greater exploration. These two concepts are combined in Figure 5-2c which shows the optimization of the combined costs of exploration and foundations. While it may never be possible to quantify the combined costs to establish an optimal level of exploration, the concepts of reliability-based design provide a more rational framework for developing economical subsurface exploration programs than is currently practiced.
Optimization of the exploration program should also consider the reliability of the different methods available for engineering property assessment of soil and rock. The three primary sources of error which contribute to the uncertainty of material resistance estimates were identified in Chapter 3 as:
y Inherent Spatial Variability represented by the uncertainty in using point
measurements compared to measurements reflecting a larger volumetric extent
y Measurement Error due to equipment and testing procedures
y Model Error reflected by the uncertainty of the predictive method
To develop a resistance factor, φ for a particular design approach (e.g., bearing resistance of a spread footing on sand using SPT) or subsurface conditions, these sources of uncertainty must be combined with uncertainties in load and the level of safety required. As described in Section 3.4.6 of Chapter 3, these factors can be combined to develop a φ-factor for design using:
] ) COV + COV + )(1 COV + ln(1 [ exp ) + Q Q ( COV + 1 COV + COV + 1 ) + Q Q ( = 2 QL 2 QD 2 R T QL L D QD 2 R 2 QL 2 QD L L D D R β λ λ γ γ λ φ (Eq. 5-2)
where loads and uncertainty in loads are represented by:
γD, γL = Dead load factor and live load factor, respectively (dim) QD/QL = Ratio of dead load to live load (dim)
λQD, λQL = Bias on dead load and live load, respectively (dim)
COVQD = Coefficient of variation of the overall bias on dead load (dim) COVQL = Coefficient of variation of the overall bias on live load (dim) The desired level of safety is represented by:
βT = Reliability index (dim)
λR = Overall bias on the resistance (dim); and
COVR = Coefficient of variation of the overall bias on resistance (dim) The value of COVR, is determined using:
COV + COV + COV =
COVR 2MODEL 2MEASUREMENT 2INHERENT (Eq. 5-3)
where:
COVMODEL = Coefficient of variation of predictive model (dim) COVMEASUREMENT = Coefficient of variation of property measurement (dim) COVINHERENT = Coefficient of variation of inherent soil variability (dim)
Figure 5-3 shows the relationship between φ and COVR for QD/QL = 2.0 (i.e., characteristic of a medium-span structure) using Eq. 5-2. Representative values of the other variables are presented on the figure and were developed from information presented in Chapter 3.
Figure 5-3
The following section provides information about the inherent variability of important soil and rock properties and measurement errors resulting from various in-situ and laboratory test methods. Model error is discussed in Chapter 6.