Transportation Research Record 1757 ■ 75
Paper No. 01-0152
Permanent Deformation Behavior of
Granular Materials and the
Shakedown Concept
Sabine Werkmeister, Andrew R. Dawson, and Frohmut Wellner
The shakedown concept has been used to describe the behavior of con- concluded that the resulting permanent strains at low levels of addi-
ventional engineering structures under repeated cyclic loading. The pos- tional stress ratio (∆σ1/σ3, where σ1 is the vertical stress and σ3 is
sibility has been raised that a critical stress level exists between stable and the radial stress) eventually reach an equilibrium state after post-
unstable conditions in pavement. According to the “shakedown” con- compaction stabilization (i.e., no further increase in permanent
cept, this level is termed the “shakedown limit.” Several repeated load strain occurs with increasing number of loads). At higher levels of
triaxial tests were performed on crushed rock aggregates at different additional stress ratio, however, permanent deformation increases
stress levels. The resulting permanent deformation, which accumulated rapidly, does not stabilize, and eventually leads to failure. For design
with the repeated loading, was described and compared with the types of purposes, this implies that the maximum load level associated with
responses usually described by the shakedown approach. The existing a resilient response must be known and must not be exceeded if the
shakedown approach can describe some, but not all, of the observed onset of permanent deformation is to be prevented. The possibility
responses. Thus, a modified set of possible responses was defined in has been raised that a critical stress level exists between stable and
shakedown terms. The method of description could provide a powerful unstable conditions in a pavement. According to the “shakedown”
material assessment and pavement design tool for engineering unbound concept, this level is termed the “shakedown limit.”
pavement bases. A design chart derived from the data illustrates a The shakedown concept has been used to describe the behavior of
possible design approach. conventional engineering structures under repeated cyclic loading. It
was originally developed to analyze the behavior of pressure vessels
to cyclic thermal loading. Later, the theory was applied to analyze
Pavement design is a process intended to find the most economical the behavior of metal surfaces under repeated rolling or sliding load
combination of layer thicknesses and material types, taking into
(3). For more details of this concept as applied to pavements, see
account the properties of the soil foundation and the traffic to be car-
Collins et al. (4) and Sharp and Booker (2). The concept maintains
ried during the service life of the road. Traditional design methods are
that four categories of material response are under repeated loading
more or less empirical, but worldwide there is an increasing desire to
(Figure 1):
develop analytical approaches. The experimental measurement and
appropriate mathematical characterization of the permanent defor-
1. Purely elastic: the applied repeated stress is sufficiently small
mation behavior of unbound granular materials (UGM) is a prerequi-
such that no element of the material enters the yield condition. From
site of a successful analytical method. A model for the description and
the first stress-strain excursion, all deformations are fully recovered,
calculation of permanent behavior is currently under development by
and the response is termed purely elastic.
the authors. Triaxial tests to investigate the permanent deformation
2. Elastic shakedown: the applied repeated stress is slightly less
behavior are the basis of these studies.
than that required to produce plastic shakedown. The material
response is plastic for a finite number of stress-strain excursions.
THE SHAKEDOWN CONCEPT AND However the ultimate response is elastic. The material is said to
have “shaken down,” and the maximum stress level at which this
PAVEMENT DESIGN
condition occurs is termed the “elastic shakedown limit.”
In pavement design, the pavement must be able to resist permanent 3. Plastic shakedown: the applied repeated stress is slightly less
deformation. Essentially, only elastic deformations are permitted in than that required to produce a rapid incremental collapse. The ma-
the pavement. The permanent deformation of UGM (and other layers) terial achieves a long-term, steady-state response (i.e., no accumula-
leads to irreversible deformations at the pavement surface. Thus, in tion of plastic strain, and each response is hysteretic). This implies that
practice, a pavement construction should be designed in such a way a finite amount of energy is absorbed by the material on each stress-
that no or only small permanent deformations appear in each layer. strain excursion. Once a purely resilient response has been obtained,
Several researchers (1, 2) have related the magnitude of the accu- the material is said, once again, to have “shaken down” and the max-
mulated permanent (plastic) strain to shear stress level. They have imum stress level at which this condition is achieved is termed the
“plastic shakedown limit.”
4. Incremental collapse or ratcheting (3): the applied repeated
S. Werkmeister and F. Wellner, Technische Universität Dresden, Mommsen-
straße 13, 01069 Dresden, Germany. A. R. Dawson, School of Civil Engineering,
stress is relatively large. A significant zone of material is in a yielding
University of Nottingham, University Park, Nottingham, NG7 2RD, United King- condition, and the plastic strains accumulate rapidly, with failure
dom. occurring in the relatively short term.
76 Paper No. 01-0152 Transportation Research Record 1757
FIGURE 1 Elastic and plastic behavior under repeated cyclic pressure and tensile load (3).
With this understanding of the component material behavior, the defines the boundary between stable (nonrutting) and unstable (rut-
shakedown concept typically then adopts classical upper- or lower- ting) conditions in a pavement. To provide response data, the per-
bound limit theorems. These theorems incorporate the appropriate manent strain behavior of granular materials has been investigated
shakedown limit stress states (rather than the higher stress state using the repeated load triaxial test. Using the experimental results,
associated with monotonic rupture) to compute the load-carrying we described the development of permanent deformation of UGM
capacity of the structure if it is not to undergo excessive plastic strain. subjected to cyclic loads.
The possible use of the shakedown concept in pavement design
was first introduced by Sharp (5) and Sharp and Booker (2). The
application of the shakedown concept is justified by Sharp and Tested Materials and Experiment Execution
Booker with results of the AASHO Road Tests (6 ) when in some
Triaxial tests were carried out with UGM as part of a collaboration
cases the distress was reported to stabilize after a finite number of
with the University of Nottingham, England. A sandy gravel and
load applications.
a granodiorite were tested (Figure 2). The materials examined
A pavement is liable to show progressive accumulation of per-
were taken from the test pit at Dresden University of Technology,
manent strains (seen as rutting) under repeated loading if the mag-
Pavement Engineering Division. The samples had a water content
nitude of the applied loads exceeds the limiting value (i.e., similar
of 4.0 percent.
to Range 3 in Figure 1). If the applied traffic loads are lower than
For these tests, the constant confining pressure was set at levels
this limit, after the postcompaction stabilization, the permanent
of 70, 140, 210, and 280 kPa. After the confining pressure was
strains will level off. The pavement will come to a state of “shake-
reached, additional dynamic (frequency = 5 Hz) vertical stress
down” (i.e., similar to Ranges 1 and 2 in Figure 1) from which time
it undergoes only resilient deformation under additional traffic
loading (7). This implies an adaptation to the loads. However, the
ideal behavior illustrated in Figure 1 does not relate in a straight-
forward manner to behavior observed in laboratory testing, as will
be described in the following sections.
Note that the “additional stress ratio” (∆σ1/σ3) referred to earlier
also can be written as σD/σ3 (where σD is the deviatoric stress) and
in this paper is referred to as “the stress ratio.”
RESEARCH PROJECT AT DRESDEN
UNIVERSITY OF TECHNOLOGY
Research Targets
This paper reports on one aspect of an ongoing research project at
the Dresden University of Technology, Germany, which is aimed at
developing a model to calculate permanent deformation behavior.
Part of the project’s goal is to find the critical stress condition that FIGURE 2 Gradings of tested materials.
Werkmeister et al. Paper No. 01-0152 77
FIGURE 3 Permanent vertical strain of granodiorite (8).
(deviator stress) was applied. The triaxial tests were carried out with Range A: Plastic Shakedown Range
axial stress pulses reaching stress ratios of σD/σ3 = 0,5 − 11.
Figure 4 shows the permanent strain versus number of load cycles
for a Range A material. Here, the response is plastic for a finite num-
TEST RESULTS
ber of load applications, but after completion of the postcompaction
Permanent Strain Rate and the period, the response becomes entirely resilient, and no further per-
Shakedown Concept manent strain occurs. On Figure 3, this plots as a convex-downward
line (labeled “A”) because the plastic strain rate progressively
Dawson and Wellner (1) reported a new method of presenting per- decreases, effectively halting further accumulation of strain and
manent deformation results. Figure 3 shows the data for a granodio- leading to an asymptotic final permanent strain value. Figure 3 shows
rite plotted as permanent vertical strain rate versus permanent verti- that the level of accumulated strain depends on the load level. Also,
cal cumulative strain. Three types of permanent strain accumulation inspection of the individual test results shows that the decrease of
were observed (labeled A, B, and C in Figure 3), and the figure styles permanent strain rate to zero is related to an increasing number of
help to identify them. None of the test results showed Range 0— cycles as load level increases. This is probably associated with the
purely elastic—behavior as suggested by Figure 1. This type of end of the postcompaction behavior.
response probably does not occur in the UGMs in pavements, as evi- A pavement with material in this condition would come to a sta-
denced by the occurrence of postcompaction strain. ble equilibrium behavior in response to the loading (4, 2). It “shakes
FIGURE 4 Permanent vertical strain versus number of load cycles,
granodiorite (8).
78 Paper No. 01-0152 Transportation Research Record 1757
• If the load level approaches the (assumed) monotonic failure
load, there is only a small decrease in incremental strain rate during
the first few load cycles (Figure 3).
• In such cases, the beginning of the failure process can be
recognized by a resurgence of the permanent strain rate (Figure 6).
• There is no cessation of strain accumulation.
Thus, Range C behavior appears to be equivalent to Range 3
behavior in Figure 1. Range C behavior in a pavement would result
in the failure of the pavement by the formation of ruts. This range
should not appear in a well-designed pavement.
In Range C, both grain abrasion and grain crushing probably occur.
The permanent deformation behavior in Range C likely depends on
the friction of particles and the grain strength, whereas the resilient
deformation depends on the number of grain contacts.
FIGURE 5 Permanent vertical strain versus number of load
cycles, granodiorite (8).
Range B: Intermediate Response—Plastic Creep
down.” Range A behavior is, therefore, permitted in the pavement, Lines like Range B in Figure 3 show an intermediate response. Dur-
provided the total accumulated strain is sufficiently small. Obser- ing the first load cycles, the high level of plastic strain rate decreases
vations of the resilient response at the end of testing reveal a small for the time being to a low, nearly constant level. The number of
hysteresis loop. load cycles for reaching this constant level of strain rate depends on
The resistance to deformation of the grain assembly in Range A the material and the load level. Perhaps this number of load cycles
presumably depends on the number of grain contacts (well-graded marks the end of postcompaction. The strain rates at the end of the
material⇒low resilient deformation). Particle crushing probably test, as shown in Figure 3 were as follows:
does not occur, or is of minor importance, as evidenced by the small
Almost Constant Permanent
amount of and no total ongoing plastic strain. Material Strain Rate at End of Test
Granodiorite 10–8/load cycle
Sandy gravel 4 × 10–9/load cycle
Range C: Incremental Collapse
Because of the almost constant level of strain rate, a near-linear
Lines like those marked C on Figure 3 plot concavely outward, indi- rise of permanent strain is observed (Figure 7). Perhaps a further
cating continuing incremental plastic deformation with each load decrease of permanent deformation rate with increasing number of
cycle. Thus, at the highest load levels, the response is always plas- load cycles occurs beyond the rate shown in Figure 3.
tic, and each load application results in a progressive increment of In Range B, insignificant total distortion occurs; therefore, only
the permanent strain (2) (Figure 5). It is observed that low-grain abrasion is anticipated. The level of accumulated strains
depends on the load level and number of load cycles. The per-
• The permanent strain rate depends on load level (Figure 3). manent deformation behavior in Range B, therefore, presumably
• The strain rate decreases very slowly compared with Range A depends on the frictional characteristics of the particles (e.g., shape,
and B or not at all. texture) and the resilient deformation on the number of grain con-
FIGURE 6 Permanent vertical strain rate versus permanent strain, granodiorite (8).
Werkmeister et al. Paper No. 01-0152 79
Although Figure 3 shows a fairly clear distinction between Range
B and Range C behavior, this is not always the case. The differing
resilient stain response with number of cycles, as shown in Figure
8, allows the responses to be separated.
Classification of Behavior Ranges
By considering the observed behaviors just described, it became
clear that Figure 1 is not a good representation. Instead, Figure 9 is
introduced as a summary of the observed types:
A. Plastic shakedown,
FIGURE 7 Permanent vertical strain versus number of load
cycles (8). B. Plastic creep, and
C. Incremental collapse.
tacts (i.e., a well-graded material is preferred). The grain charac- Range B is an intermediate behavior observed over a range of
teristics and the grading are likely to be essential for determining stress states. Initially behavior is like Range A behavior, but a small
the stress-strain behavior of granular materials in Range B. residual incremental plastic strain is observed, yet without stiffen-
ing (strain hardening). This is a consequence of the ability of the
aggregate to absorb low amounts of energy by largely frictional
means without undergoing significant adaptation or damage as would
Resilient Strain and the Shakedown Concept be observed in a continuum (like steel) or as seen in Range C behav-
ior. It is likely that any damage is limited to slight particle wear and
Figure 8 shows that both Range A and Range B specimens exhibit that the small amount of fines generated can be stored within the
a constant level of resilient strain during a test and that the level of void space of the granular material without significantly affecting
resilient strain depends on the load level. However, a significant overall response.
decrease of resilient deformation with increasing number of load Note that the original shakedown explanation (Figure 1) assumes
cycles is observed in Range C (i.e., strain hardening appears to be that the material is loaded equally in tension and in compression
taking place in Range C specimens). in each cycle. Clearly, the no-tension ability of a UGM precludes
The permanent strain in Range C is thus expected to be associ- this. Thus, the plastic recovery part of the hysteresis loop does not
ated with considerable grain breakage and particle rearrangement. take place at the tension stress level, which “mirrors” the com-
The number of grain contacts and the dimension of the grain con- pressive stress level that caused plastic strain but at a very low
tacts is expected to increase. The state of compaction is also compression. Because of the material’s nonlinearity, it is unlikely
increasing. If a constant level of resilient strain is reached, then that the hysteresis loop will be symmetrical. Thus the recovery of
constant-volume permanent deformation is probable. plastic strain is unlikely to mirror the plastic strain generation.
FIGURE 8 Resilient vertical strains versus number of load cycles, granodiorite (8).
80 Paper No. 01-0152 Transportation Research Record 1757
FIGURE 9 Behavior of granular materials under repeated cyclic pressure load.
This introduces a further possibility of incremental plastic strain σ SD = Aσ 3 + B (1)
accumulation.
where
Shakedown-Limit Calculation σSD = bounds of range (kPa),
σ3 = cell pressure (kPa), and
Figure 10 shows the test results for a sandy gravel and a granodio- A, B = regression parameter (–).
rite with varying cell pressures. The relationship between cell pres-
sure level and the deviatoric stress at which permanent deformation A and B are likely to depend on the grading, particle shape, particle
begins to develop rapidly is obvious. Considering the data from Fig- surface, water content, and so forth of the materials. Further research
ure 10, labeled using Figure 3 (and other plots like it), it would be will be necessary on this point.
possible to define, by regression, a simple stress formula that defines
the boundaries between the different Ranges (A, B, and C) as a func-
tion of cell pressure, σ3. Figure 11 shows the results for sandy DISCUSSION AND CONCLUSIONS
gravel.
In fact, it is difficult to calculate the exact shakedown load. For We have been shown that the application of the shakedown concept
this reason an approach to defining the upper and lower bounds for to granular materials as used in road construction is possible, although
the range boundaries was developed: adaptations must be made to allow for the particular response of these
FIGURE 10 Permanent vertical strains after 100,000 load cycles for sandy gravel (K) and
granodiorite (G).
Werkmeister et al. Paper No. 01-0152 81
FIGURE 11 Elastic and plastic shakedown limits for a sandy gravel at 4 percent
water content.
materials to repeated loading. The shakedown Ranges A, B, and C are versity of Nottingham and of the Queen’s University, Belfast,
suspected to occur in all granular materials. Northern Ireland.
The permanent deformation response is affected by several factors.
Further research should concentrate on the influence of parameters
such as aggregate type and grading, as these factors appear to influ-
ence the modes of plastic strain observed. Furthermore, the determi- REFERENCES
nation of the range boundary parameters as a function of the values φ′
1. Dawson A. R., and F. Wellner. Plastic Behavior of Granular Materials.
and c′ or of grading, aggregate type, water content, and so forth should Report ARC Project 933 Reference PRG99014. University of Nottingham,
be investigated. United Kingdom, 1999.
If the unbound layers behave in a manner corresponding to Range 2. Sharp, R., and J. Booker. Shakedown of Pavements Under Moving
A, the pavement will “shake down.” After postcompaction defor- Surface Loads, Journal of Transportation Engineering, Vol. 110, No. 1,
Jan./Feb. 1984, pp. 1–14.
mations, no further permanent strains develop, and the material sub- 3. Johnson, K. L. Plastic Flow, Residual Stresses, and Shakedown in Rolling
sequently responds elastically. Thus Range A is permitted in the Contact. Proc., 2nd International Conference on Contact Mechanics and
pavement if the accumulated strain before the development of fully Wear of Rail/Wheel Systems, University of Rhode Island, Kingston,
resilient behavior is sufficiently small. University of Waterloo Press, Ontario, Canada, 1986.
4. Collins, I. F., A. P. Wang, and L. R. Saunders. Shakedown Theory and
The next step is to examine the application of material in the
the Design of Unbound Pavements. Road and Transport Research, Vol. 2,
pavement that responds according to Range B. It will be important No. 4, Dec. 1993, pp. 28–39.
to know the permitted maximum number of load cycles that will 5. Sharp, R. Shakedown Analyses and the Design of Pavement Under Moving
prevent distress in the pavement from occurring. Further tests with Surface Load. Ph.D. Thesis. University of Sydney, Australia, 1983.
load applications up to 1 million load cycles may be necessary. 6. Kent, M. F. AASHO Road Test Vehicle Operating Costs Related
to Gross Weight. In Special Report 73: Road Test Vehicle Operating
Range 3 behavior should not be allowed to occur in the pavement. Costs Related to Gross Weight, HRB, National Research Council, 1962,
pp. 149–165.
7. Sharp, R. W. Pavement Design Based on Shakedown Analyses. In Trans-
portation Research Record 1022, TRB, National Research Council,
ACKNOWLEDGMENT Washington, D.C., 1985, pp. 99–107.
8. Wellner, F., and S. Werkmeister. Beitrag zur Untersuchung des Ver-
formungsverhaltens ungebundener Gesteinskörnungen mit Hilfe der
The authors thank the Deutscher Akademischer Austauschdienst and
Shakedown-Theorie (in German). Strasse + Autobahn, No. 6, 2000,
the British Council for funding under the ARC program, which has pp. 369–375.
partly supported the work described here. They also gratefully
acknowledge the suggestions provided by Greg Arnold of the Uni- Publication of this paper sponsored by Committee on Mineral Aggregates.
