MGK – Consulting Report December 20, 2011

Information on the Duncan and Duncan/Selig Soil Models in CANDE

Purpose. This document shows the comparative performance of Duncan and Duncan/Selig soil models
that are available (canned) in the CANDE program. The intent is to provide the CANDE user with
reference information on the comparative stiffness and strength of each canned model in CANDE.
(For those not interested in background information, key results are shown in Tables 3 and 4 and
Appendix A.)

Background. The Duncan and Duncan/Selig soil models are variable-modulus elasticity formulations
using stress-dependent equations for Young’s modulus and bulk modulus. The Duncan form and
Duncan/Selig form are very similar, differing slightly in the expression of the bulk modulus function.
Both forms of the soil models are considered to be very representative of actual soil behavior, particularly
for representing the stress-dependent behavior of backfill soil in culvert installations. Basically, the
models exhibit stiffening when confining stress increases and softening when shear stress increases.

Duncan’s original work for the Young’s modulus formulation, which is based on a hyperbolic stress-
strain relationship, dates back to 1970 and remains today as originally developed with 6 model parameters
to be specified. In 1978 Duncan and his colleagues introduced a variable bulk modulus to serve as the
second constitutive parameter, replacing the previous assumption of a constant Poisson ratio. The bulk
modulus function is based on a power law function of confining pressure and is characterized with two
parameters. Today, Duncan’s original Young’s modulus formulation together with the power-law bulk
modulus function is referred to as the Duncan soil model with a total of 8 model parameters.

Based on extensive tri-axial soil tests, Duncan and his colleagues investigated four different types of soil,
classified as Course Aggregates (CA), Silty Sand (SM), Silty Clayey Sand (SC) and Silty Clay (CL).
Each soil type was tested at three different percentages of relative compaction (e.g., 85%, 95% and 100%)
based on AASHTO T-99. Table 1 shows the values of the model parameters for each soil type and
percentage of relative compaction. For example, CA95 stands for course aggregate compacted to 95%
relative density.

Selig proposed an alternative form of the bulk modulus function in 1988 again using two parameters.
Selig’s bulk modulus function is based on an observed hyperbolic relationship between volumetric strain
and hydrostatic pressure from soil specimens in hydrostatic compression. Thus, the so-called
Duncan/Selig soil model is based on Duncan’s original Young’s modulus formulation (6 parameters) and
Selig’s hyperbolic bulk modulus formulation (2 parameters). Selig and his students performed
independent triaxial and hydrostatic tests to characterize the parameters of the Duncan/Selig model for
Gravelly Sand (SW), Sandy Silt (ML) and Silty Clay (CL) at several percentages of relative compaction
again based on AASHTO T-99. Table 2 shows the model parameters thus determined for the
Duncan/Selig model.

To summarize the background information, Tables 1 and 2 identify the canned Duncan and Duncan/Selig
soil model parameters that are stored in CANDE. For more detailed information on the soil model
formulations and parameter identification methods, the reader is referred to CANDE’s Formulations and
Solution Methods manual. Here, the focus on soil model behavior and performance of the canned models.

1
MGK – Consulting Report December 20, 2011

Table 1. Duncan soil model parameters (as stored in CANDE)

Model name: Young’s Tangent Modulus Parameters Bulk Parameters
Soil type and K n C φ0 Δφ Rf Kb m
compaction (‐‐) (‐‐) (psi) (deg) (deg) (‐‐) (‐‐) (‐‐)
CA105 600 0.40 0.0 42 9 0.7 175 0.2
CA95 300 0.40 0.0 36 5 0.7 75 0.2
CA90 200 0.40 0.0 33 3 0.7 50 0.2
SM100 600 0.25 0.0 36 8 0.7 450 0.0
SM90 300 0.25 0.0 32 4 0.7 250 0.0
SM85 150 0.25 0.0 30 2 0.7 150 0.0
SC100 400 0.60 3.5 33 0 0.7 200 0.5
SC90 150 0.60 2.1 33 0 0.7 75 0.5
SC85 100 0.60 1.4 33 0 0.7 50 0.5
CL100 150 0.45 2.8 30 0 0.7 140 0.2
CL90 90 0.45 1.4 30 0 0.7 80 0.2
CL85 60 0.45 0.7 30 0 0.7 50 0.2
*Soil type is defined as follows: CA = Course Aggregates, SM = Silty Sand, SC = Silty-Clayey Sand and CL =
Silty Clay. The compaction number is percent relative compaction, per AASHTO T-99. As an example, SM100
means silty sand compacted to 100% relative density.

Table 2. Duncan/Selig soil model parameters (as stored in CANDE)

Soil type and Young’s Tangent Modulus Parameters Bulk Parameters
compaction K n C φ0 Δφ Rf B i/ P a εu
(‐‐) (‐‐) (psi) (deg) (deg) (‐‐) (‐‐) (‐‐)
SW100 1300 0.90 0.0 54 15 0.65 108.8 0.01
SW95 950 0.60 0.0 48 8.0 0.70 74.8 0.02
SW90 640 0.43 0.0 42 4.0 0.75 40.8 0.05
SW85 450 0.35 0.0 38 2.0 0.80 12.7 0.08
SW80 320 0.35 0.0 36 1.0 0.90 6.1 0.11
ML95 440 0.40 4.0 34 0.0 0.95 48.3 0.06
ML90 200 0.26 3.5 32 0.0 0.89 18.4 0.10
ML85 110 0.25 3.0 30 0.0 0.85 9.5 0.14
ML80 75 0.25 2.5 28 0.0 0.80 5.1 0.19
ML50 16 0.95 0.0 23 0.0 0.55 1.3 0.43
CL95 120 0.45 9.0 15 4.0 1.00 21.2 0.13
CL90 75 0.54 7.0 17 7.0 0.94 10.2 0.17
CL85 50 0.60 6.0 18 8.0 0.90 5.2 0.21
CL80 35 0.66 5.0 19 8.5 0.87 3.5 0.25
* Soil type is defined as follows: SW = Gravelly sand, ML = Sandy silt, and CL = Silty Clay. The compaction
number is percent relative compaction, per AASHTO T-99. As an example, SW95 means gravelly sand compacted
to 95% relative density.

2
MGK – Consulting Report December 20, 2011

Fundamental Soil Behavior. Laboratory soil tests can produce either stiffening or softening stress-strain
behavior depending on the stress paths and loading conditions. Stiffening behavior occurs when the
maximum and minimum principal stresses increase in roughly the same proportion during the loading
schedule such as in the case of hydrostatic pressure loading or as in the case of uniaxial strain loading.
Figure 1 depicts a uniaxial strain test with an applied axial stress σ1 and a measured axial strain ε1. Since
this test is laterally confined such the lateral strains equal zero, we find that lateral stress σ3 increases in
near proportion to σ1 by means of the so-called Poisson effect, or equivalently, the lateral coefficient
effect. As illustrated in Figure 1, the secant value of the confined modulus, which is defined as Ms = σ1/
ε1, increases its stiffness with applied stress.

Figure 1. Depiction of uniaxial-strain test with confined modulus, Ms

In contrast to the above, the same soil specimen will exhibit a softening behavior if the deviator stress (σ1
– σ3) increases at a faster rate than the minimum confining stress σ3 such as in the case of standard triaxial
tests. Figure 2 depicts a standard triaxial soil test conducted for two levels of confining pressures, σ3a and
σ3b. In the first experiment, σ3a is the applied hydrostatic pressure that is held constant while the axial
stress σ1 is steadily increased and axial strain ε1 is measured. Next, the entire process is repeated on a
duplicate sample in which the confining pressure is reset to a larger value, σ3b. As shown in Figure 3, the
plots of deviator stress (shear stress) versus axial strain exhibit softening behavior because the slope of the
curves (tangent Young’s modulus) decrease as the deviator stress decreases. We also observe that there
exists a maximum deviator stress called the failure stress which increases with confining pressure.

3
MGK – Consulting Report December 20, 2011

Figure 2. Depiction of standard triaxial test for two confining pressures.

The Duncan and Duncan/Selig models are formulated to replicate the behaviors observed in Figures 1 and
2 as well as combinations of these behaviors representing more general stress paths. When we use these
soil models to represent the backfill soil, we don’t know how the soil model will behave in different
backfill zones because we don’t know a priori the stress paths in each locale of the backfill. However, for
backfill zones at some distance from the structure, i.e., outside the zone of soil-structure interaction, the
soil is essentially in free-field conditions, which is equivalent to the uniaxial strain loading condition in
Figure 1. Thus, the confined modulus Ms in uniaxial strain is a good measure for the backfill stiffness in
free-field zones. On the other hand, within the zone of soil-structure interaction we have very little a
priori knowledge on the state of stress, and in some locales the deviator stress may dominate the
minimum principal stress so that shear failure as depicted in Figure 2 is a possibility. Thus, the shear
strength of soil is an important measure of backfill quality in the zone of soil-structure interaction.

Comparative Performance of Soil Models. Based on the forgoing discussion, we may assess the merits
of one soil model versus another based on the relative magnitudes of the confined modulus and the
respective shear strengths. These comparative values are determined by obtaining CANDE solutions for
each Duncan and Duncan/Selig soil model under two loading conditions; uniaxial strain and standard
triaxial loading. The results are summarized in Tables 3 and 4.

4
MGK – Consulting Report December 20, 2011

Table 3. Duncan Model stiffness and strength measures

Ultimate deviator
Duncan Average Elastic Parameters in uniaxial strain strength
Model for σ3 = Pa
name Ms Ko E ν (σ1 – σ3)u
(psi) (‐‐) (psi) (‐‐) (psi)
CA105 3838 0.34 3218 0.24 85
CA95 1613 0.35 1308 0.26 60
CA90 1051 0.39 822 0.28 50
SM100 9301 0.57 5494 0.36 60
SM90 4985 0.61 2709 0.38 47
SM85 2882 0.65 1414 0.39 42
SC100 3133 0.43 2270 0.30 69
SC90 1177 0.45 834 0.31 61
SC85 786 0.46 550 0.37 58
CL100 2527 0.61 1368 0.38 56
CL90 1444 0.61 784 0.38 49
CL85 907 0.61 499 0.38 45

Table 4. Duncan/Selig Model stiffness and strength measures

Ultimate deviator
Duncan/Selig Average Elastic Parameters in uniaxial strain strength
Model name for σ3 = Pa
Ms k E ν (σ1 – σ3)u
(psi) (‐‐) (psi) (‐‐) (psi)
SW100 7881 0.37 6236 0.27 192
SW95 4786 0.28 4039 0.22 122
SW90 2402 0.27 2078 0.21 79
SW85 1067 0.28 911 0.21 59
SW80 783 0.32 629 0.23 47
ML95 2246 0.34 1798 0.24 55
ML90 1156 0.36 886 0.25 51
ML85 707 0.40 498 0.27 47
ML80 482 0.43 319 0.28 43
ML50 146 0.65 66 0.39 34
CL95 883 0.52 525 0.33 34
CL90 539 0.53 303 0.34 33
CL85 370 0.54 194 0.34 33
CL80 288 0.57 139 0.36 32

5
MGK – Consulting Report December 20, 2011

The elastic parameters shown in Tables 3 and 4 are average values over the axial stress ranging from 0 to
100 psi. Appendix A shows the actual stress-strain plots for each model in uniaxial strain loading
conditions. For reference, the definition of elastic parameters is given below.

 Ms = secant confined modulus = σ1/ε1 in uniaxial strain - (averaged over all load levels)
 Ko = lateral coefficient = σ3/ σ1 in uniaxial strain - (averaged over all load levels)
 E = Young’s modulus = Ms(1+2 Ko)(1- Ko)/(1+ Ko) - (elastic parameter identity)
 ν = Poisson ratio = Ko /(1+ Ko) - (elastic parameter identity)

In the last column of Table 3 and 4, the ultimate deviator strength is presented for the case σ3 = 14.7 psi
(atmospheric pressure). The ultimate deviator strength may be obtained from CANDE solutions of the
triaxial test. Alternatively, strength values may be calculated directly from Duncan’s strength formula
shown below using the parameters in Tables 1 and 2.

2 C cos    3 sin 
 (σ1 – σ3)u = ( )
Rf 1  sin 

Summary Soil stiffness and strength are very important properties in sustaining the structural integrity of
culvert installations. Hence, soil models with large values for the confined modulus (Ms) and large values
for shear strength (σ1 – σ3)u are better than soil models with lesser values. Tables 3 and 4 allow us to
compare the structural robustness of one soil type versus another as well as one compaction level versus
another such as 90% versus 95% relative density. Finally, Appendix A provides a graphical view of
model behavior and offers further insights into soil stiffness.

6
MGK – Consulting Report December 20, 2011

Appendix A

Stress-strain plots of Duncan and Duncan/Selig soil models for uniaxial-strain loading.

Duncan Models contained herein:

 CA-series
 SM-series
 SC-series
 CL-series

Duncan/Selig Models contained herein:

 SW-series
 ML-series
 CL-series

7
MGK – Consulting Report December 20, 2011

Duncan model confined compression behavior for CA soil type compacted to 3 different densities.
100
CA Series ‐ Uniaxial Strain
90

80

70
Axial pressure (psi)

60

50

40

30 CA90 ‐Uniax
CA95 ‐Uniax
20 CA105‐Uniax

10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
Axial strain (percent)

8
MGK – Consulting Report December 20, 2011

Duncan model confined compression behavior for SM soil type compacted to 3 different densities.

100

SM Series ‐ Uniaxial Strain

90

80

70
SM85 ‐ Uniaxial
Axial pressure (psi)

60
SM90 ‐ Uniaxial
50
SM100 ‐ Uniaxial
40

30

20

10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%
Axial strain (percent)

9
MGK – Consulting Report December 20, 2011

Duncan model confined compression behavior for SC soil type compacted to 3 different densities.

100

SC Series ‐ Uniaxial Strain

90

80

70
Axial pressure (psi)

60

50

40
SC85 ‐ Uniaxial
30
SC90 ‐ Uniaxial
20
SC100 ‐ Uniaxial
10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Axial strain (percent)

10
MGK – Consulting Report December 20, 2011

Duncan model confined compression behavior for CL soil type compacted to 3 different densities.

100

CL Series ‐ Uniaxial Strain

90

80

70
Axial pressure (psi)

60

50

40
CL85 ‐ Uniaxial
30
CL90 ‐ Uniaxial
20
CL100 ‐ Uniaxial
10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Axial strain (percent)

11
MGK – Consulting Report December 20, 2011

Duncan/Selig model in confined compression for SW soil type compacted to 5 different densities.

100
SW Series ‐ Uniaxial Strain
90

80

70
Axial pressure (psi)

60

50

40

30
SW80 ‐Uniax
SW85 ‐Uniax
20
SW90‐Uniax
SW95 ‐Uniax
10
SW100 ‐Uniax

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Axial strain (percent)

12
MGK – Consulting Report December 20, 2011

Duncan/Selig model in confined compression for ML soil type compacted to 5 different densities.

100
ML Series ‐ Uniaxial Strain
90

80

70
Axial pressure (psi)

60 ML50 ‐Uniax
ML80 ‐Uniax
50 ML85‐Uniax
ML90 ‐Uniax
40 ML95 ‐Uniax

30

20

10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Axial strain (percent)

13
MGK – Consulting Report December 20, 2011

Duncan/Selig model in confined compression for CL soil type compacted to 4 different densities.

100
CL Series ‐ Uniaxial Strain
90

80
CL80 ‐Uniax
70 CL85‐Uniax

CL90 ‐Uniax
Axial pressure (psi)

60
CL95 ‐Uniax

50

40

30

20

10

0
0% 1% 2% 3% 4% 5% 6% 7% 8% 9% 10%

Axial strain (percent)

14