Numerical simulation of a vibratory

roller on cohesionless soil

Åke Sandström, 9411

SIMVIB2.DOC 94-11-07

Numerical simulation of a vibratory roller on

cohesionless soil

Åke Sandström, Geodynamik AB

Abstract
Compaction of soil by a vibratory roller is a complex process. Various soil properties and
roller parameters interact making the dynamic drum/soil system a highly non-linear one. The
reasons for non-linearity are the plastic deformations in the soil and the fact that the contact
between the drum and the soil may be lost for a portion of the cycle.
Prediction of the dynamic behaviour of a vibrating drum acting on a soil layer is important for
several reasons. It can be used for optimising the parameters of the roller and also to study the
limits for stable vibration and to evaluate different compaction meter systems.
A numerical model of the drum/soil system is described in the paper. The model takes into
account the most important factors influencing the dynamics of the system. The soil is
characterised by an elastic parameter comparable to the shear modulus of a homogenous
isotropic half space and the plastic effect in the compaction zone is modelled by a second
parameter. The design of the model is described and results are presented for a few examples.
Predicted accelerations of the drum, force/displacement relations, maximum force generated
by the drum and vibration stability are presented. CMV-values are also calculated and
displayed as function of the compaction state.

Introduction
Vibratory rollers have been used for many years and the design has evolved step by step from towed
types to self propelled rollers of different weight classes (Forssblad [7]). The drums are smooth for
compaction of gravel or rock-fill. Sheepsfoot or padfoot drums are used for clay and other cohesive
material. For the compaction of thin gravel layers and asphalt paving the tandem type of roller having
two smooth drums is very common. The oscillatory roller is a new type of roller, gaining popularity
for use both in soil and pavement compaction.
The optimum design of a vibratory roller poses some difficult problems. The roller should compact
efficiently on different types of soils and in different stages of compaction from loose to fully
compacted. It is well known that a drum bouncing on the soil surface can enter undesirable vibration
states like double jump and rocking motion. Existing vibratory rollers normally have one or two
1
SIMVIB2.DOC 94-11-07

different frequency/amplitude settings and therefore represent a compromise regarding the

compaction efficiency under various soil conditions.
Rollers evolved step by step in a trial and error process. It was however soon realised that there is a
need for theoretical modelling and a number of investigations have been performed with the aim of
increasing the theoretical basis for the design of vibratory rollers. The first models were linear and
aimed at studying basic phenomena as conditions for drum lift off and relations between the eccentric
force and the force applied to the soil (Machet[15], Yoo and Selig [28]).
Work was started at Geodynamik around 1980, with the aim of developing a computer simulation
model that could take into account the non-linearities involved in the drum/soil system and the
important parameters of the roller. The aim was to create a tool for roller design, especially for the
development of automatically controlled rollers and for the study of compaction meters. It was
considered important that the model had a simple basic structure that could be extended to include
different aspects of the roller and the soil. The parameters of the model should be understandable,
measurable and comparable to standard parameters in soil mechanics. A simulation model was
developed along these lines and it was used in practice during the eighties. The model was used
primarily for parameter studies on existing rollers in consultancy work for different roller
manufacturers.
The basic idea behind the simulation model is the following. When loaded dynamically by a cylindrical
roller drum, the soil is modelled by two separate parts, acting in series. One part is a plastic one,
having a load/displacement relation that can be obtained from conventional bearing capacity theory
and the second is an elastic part. The modelling of the elastic part is solved by using results from
research on dynamics of foundations.
Work regarding the dynamic loading of an elastic half space has been going on for many decades.
These investigations aimed at predicting the dynamics of different structures resting on soil e.g.
machine foundation design, earthquake action on buildings and wave action on structures resting on
the sea bed. The knowledge has evolved from pure analytical solutions for circular plates resting on
an ideally elastic half space [3,20,23] to lumped parameter models [11,14] to numerical schemes for
taking into account hysteresis, rate dependence and irreversible deformations in real soil [2]. The
methods have also been extended to arbitrarily shaped loading areas and layered soil [5,6,16].
Results from this research is also applicable in the case of dynamic action of a roller drum and is used
in the modelling of the elastic part of the simulation model.
Pietzsch [19] developed a model, composed entirely of mass, spring and damper elements. He
included non-linear elements and also allowed the drum to lift off from the soil surface. Energy losses
by radiation were simulated by an added mass/spring/dashpot system. There are a number of
coefficients for the different elements included in the model and they have to be calculated from
known soil parameters. This may be difficult since there is no direct correspondence between the
element parameters and conventional soil parameters.
A method similar to the one presented in this paper was used by Grabe [9] to simulate the behaviour
of a vibrating roller on soil. He however used a different method to model the soil. In addition to an
elasticity module for deeper layers, Grabe used the loading stiffness and also the ratio between the
loading stiffness and the unloading stiffness as soil parameters.

2
SIMVIB2.DOC 94-11-07

Static vertical load from a cylinder on a soil surface

1. Plastic deformation

The loading of the flat surface of a thick layer of friction material from a smooth circular cylindrical
steel drum of a vibratory roller is considered.
There will be a contact zone, that will undergo plastic deformation down to a certain depth, the
extent of which is dependant on the strength of the material and of the line load and radius of the
drum. The width and depth of the contact zone will initially be zero and there will therefore always be
a zone of soil failure. The extent of the failure zone will increase as the load on the cylinder is
increased. An equilibrium will be reached when the total static line load of the roller is applied. The
total vertical displacement of the cylinder under influence of the static line load and with respect to
the originally flat surface is a characteristic measure of the strength of the soil.
The failure zone is extending downwards to a certain level, that is related to the width of the contact
area, the internal friction angle and the density of the soil material.
The problem of calculation of load versus settlement, with regard to plastic effects only, resembles
the classical problem of estimating the bearing capacity of a strip or plate foundation. Figure 1. The
main difference is that the loading body has a curved surface and that the geometry is changing with
increasing load.
The stress distribution under the drum is affected by the radius of the cylindrical loading body and
differs from the one present under a flat plate. The effect of this can be expected to be moderate
because of the active wedge forming under the contact area. It is assumed that the surface roughness
of the cylinder is high enough to prevent sliding
at the interface with the soil.
B B
The influence of a curved lower surface of a
σu strip foundation was investigated by
Széchy[24]. He made laboratory tests to
investigate the bearing capacity of rectangular
loading bodies having different length/width
ratio. Flat, convex and concave concrete
bodies were tested on compacted sand and
Figure 1: Conventional bearing capacity gravel. The results were that the bearing
theory. capacity is somewhat reduced for a convex
surface compared to a flat one.
The bearing capacity for a strip foundation
according to Figure 1 is often expressed as:
σu = cN c + qNq + B γ N γ (1)

with the bearing capacity expressed as a sum of the contributions by cohesion, surcharge and unit
weight of the soil material respectively.

3
 SIMVIB2.DOC 94-11-07

The so called bearing capacity factors Nc, Nq and Nγ are functions of the internal friction angle
ϕ alone.

Displacement Neglecting cohesion and assuming that there is no

surcharge, the total load capacity of a rectangular
area of width 2B and length W can be written:

Qu = γ B N γ ⋅ 2 BW = const ⋅ B2 (2)
p
Approximating the cylindrical cylinder with a
parabolic cylinder in the contact area, the relationship
between the settlement of the cylinder and the
mobilised bearing capacity of the plastic layer will be
Load linear.
Fstat
A plasticity parameter p is defined as the plastic
Figure 2: Plastic load/displacement deformation corresponding to the static load . See
relationship for a cylinder. Figure 2.

2. Elastic displacement

The plastified zone developing under the drum will determine the stress distribution acting at a level
below the zone, and it will be different from the one predicted by the theory of elasticity for the
contact between a cylinder and a flat surface of an elastic medium. The load will cause displacement
of the soil below the plastified zone that is mainly elastic.
In a theoretical sense the word "elastic" applies to a material which recovers completely after the
removal of an applied stress. This does not mean that the stress-strain relationship is linear. For the
superposition principle to be applicable, however, the stress-strain relationship must be linear and in
connection with analytical solutions the material is considered to be "linear elastic".
Soil cannot be considered as an ideal homogeneous isotropic elastic medium having constant values
of the elastic constants G and ν. Because of the nature of the formation processes involved, soils in
situ are far from homogenous and also far from isotropic. In the context of compaction of soils, the
structure is normally built in several layers using locally available material. The quality of the material
is generally highest near the final surface or foundation level while deeper layers are built from lower
cost material.
The dynamic behaviour of soil material is strongly dependant on
• void ratio or degree of compaction
• strain amplitude
• stress situation
• number of load cycles

The G-modulus determined by conventional geotechnical methods underestimate the actual modulus
at small strains. The maximum modulus at small strains can be determined by resonant column tests
or in situ seismic methods.

4
SIMVIB2.DOC 94-11-07

Hardin and Richart [10] determined G-moduli indirectly by measuring shear wave velocities in situ.
They present the initial modulus as

( C 2 − e) 2
Gmax = C1 σ0 (3)
1+ e

where e = void ratio, σ0 = mean principal effective stress and C1, C2 are parameters that are
functions of gradation and grain shape.
For a horizontal homogenous layer, the mean principal effective stress is related to the effective
vertical stress as

 1 + 2 K0 
σ0 =  σv (4)
 3 
where K0 is the ratio of lateral to vertical stress.
The effective vertical stress varies with depth and density of the soil material. The value of K0
generally increases with the degree of compaction.
To be able to model the roller behaviour and it's interaction with the soil, it is necessary to model the
load/deformation conditions of the soil as seen downwards from the level just below the plastified
contact zone. The deformation condition at this level is the integrated effect of the properties of all the
underlying layers. It is not the aim of this investigation to incorporate the wide field of aspects on
varying elastic properties with depth and stress/strain level, nor the layering of the soil.
The layers below the contact zone will be modelled as an elastic half space. The values for the elastic
soil constants G and ν that should be used are related to the real deformation parameters in the soil
mass. The relation is similar to the connection between measured plate load moduli and soil
parameters.
The elastic deformation of deeper layers is
assumed to be governed by a parabolic load as
shown in Figure 3.
The first step is to formulate the displacement
r for a total vertical static load Q distributed over
a rectangular area as shown in Figure 4.
B B
p
I The displacement is calculated by integration of
the expression for a point load according to
II elasticity theory and the result is expressed as a
kB non-dimensional parameter D.

Figure 3: Assumed vertical stress distribution.

δG B 1 3 L
D= = ⋅ ⋅ I (α) where α= (5)
Q (1 − ν) α 4π B

5
SIMVIB2.DOC 94-11-07

I(α) is the expression arising at the integration of the displacement- according to Boussinesq - from
the assumed parabolic stress distribution over a rectangular area with length/width-ratio α . The
value of the integral can be very closely approximated by:
I ( α) ≈ 0. 66 ln ( 7. 9α) (6 )

Figure 4: Parabolic load on a rectangular area.

The resulting dimensionless expression D

versus the length/width-ratio α is shown in
0.4

Figure 5.
0.3

D The displacement parameter D can be used

0.2 to calculate the displacement at the centre
of a rectangular area when the vertical
0.1 stress distribution is parabolic along the
short axis of the area and constant along
0 the long axis.
0 2 4 6 8 10 12 14 16
A more clear picture of the influence of the
α shape of the loaded area is obtained if the
displacement is compared to the one for a
Figure 5: Dimensionless displacement function square, keeping the magnitude of the area
and the total load constant. This result is
shown in Figure 6.
It is assumed that the stress distribution is constant along the whole width of the roller drum. This is
an approximation because the drum itself does not deform and therefore produce a constant strain
rather than a constant stress. The presence of the plastified contact zone, however evens out the
stress variation and the approximation is therefore considered to be justified.

6
SIMVIB2.DOC 94-11-07

1.0

Figure 6:
0.9

Displacement at the centre of a

0.8 rectangular area in relation to a
δ / δ1 square area. The total load and the
0.7
size of the area are constant. The
0.6 load distribution is parabolic along
the short axis.
0.5

0 2 4 6 8 10 12 14 16

Dynamic vertical load on an elastic half space

Considering a weightless plate resting on an elastic half space, the displacement at the centre of the
plate can be expressed as function of the load. When the load is constant, there is a linear relation
between the magnitude of the load and the resulting deformation determined by the elastic constants
of the continuum, the shape of the loaded area and the stress distribution on the loaded area.
z se = κ⋅ Q (7)

where zse is the elastic soil displacement and Q is the total load. The factor κ can be calculated from
equation (5).
When the load varies with time the amplitude relation between load and deformation changes, and
there is a time lag between the load and the deformation. Energy is flowing away from the loaded
area as elastic waves into the medium and along the surface of the half space. The important
parameter governing the problem is the speed of shear waves in the medium in relation to the driving
frequency and the size of the loaded area. This is expressed as the ratio between a linear dimension
of the area and the wavelength of shear waves.
d
a0 = (8)
vs / ω

where d is a characteristic linear dimension of the loaded area, ω is the circular frequency

G
and vs = is the shear wave velocity.
ρ

The frequency response of the this linear system can be characterised by a complex transfer function
of order n expressed in Laplace form as

an s n + an −1 s n −1 Κ + 1
H ( s) = H0 (9)
bn s n + bn −1s n −1 Κ + 1

7
SIMVIB2.DOC 94-11-07

The coefficients of Equation (9) are real valued and therefore complex poles and zeroes appear as
conjugate pairs. The poles have to be in the negative half of the s-plane for equation (9) to represent
a stable system.
The Laplace transform of the displacement can be written formally as the product of H(s) and the
Laplace transform of the driving force.
In our case we want to calculate numerically step by step the resulting deformation from a given load
input. In order to achieve this, equation (9) is mapped to a z-transform representing a sampled
version of the system having a constant time step of ∆t. The mapping is made by using a bi-linear
transform, equation (10). This means that an ideal integrator (1/s) is represented by a trapezoidal
approximation.
1 ∆t  z + 1
→ (10)
s 2  z − 1

This mapping converts equation (9) to a rational function of z, which can easily be reformulated as a
difference equation relating consecutive samples of the displacement to consecutive samples of the
load time series. Starting with a known state of the displacement, it is possible to calculate the
displacement step by step for any given time history of loads.
The problem remains to determine the poles and zeroes of the transfer function for the given half
space and loading area. This can be achieved by methods common in filter synthesis in electrical
engineering. The poles and zeroes of equation (9) are determined by linear programming in order to
approximate the frequency response with a sufficient accuracy. The number of poles and zeroes
necessary for a sufficient accuracy
1 depend on the complexity of the
frequency response function.
0.8
Analytical solutions of dynamically
f1 Uniform load
loaded areas on top of an elastic half
0.6
space can de found in the literature for
- f2 Rigid plate
vertical, horizontal and rocking mode
0.4
deformation of circular areas. Different
load distributions have also been
0.2
treated, corresponding to a rigid plate,
Parabolic load
a uniform load distribution and
0
parabolic load distribution [3,20,23].
0 0.5 1 1.5 2

a0 Because of the complicated expressions

arising, available closed form solutions
are limited to circular shapes.
Figure 7: Compliance functions for a circular plate,
ν = 0.25. The solutions are expressed as
compliance functions often denoted as
f1 and f2 after Reissner [21]. f1 is the
real part and f2 the imaginary part of H(s) evaluated along the imaginary s-axis.
Figure (7) shows results compiled by Sung [23] for circular load areas and different load distribu-
tions. As pointed out by Richart it seems possible to collapse the curves into one by assigning an
"effective radius" for each type of distribution. This would be equivalent to applying different scaling
8
SIMVIB2.DOC 94-11-07

factors to the horizontal scale for the different load distributions in Figure (7). The effective radius for
the rigid plate would be 1.273 r0 and for the parabolic load it would be 0.75 r0 to convert to
equivalent uniform load cases. Hsieh [11] transformed the Reissner functions into functions containing
a spring term and a damping term, both being frequency dependent. His idea was developed further
by Lysmer and Richart[14] and others and has been widely used in prediction of the dynamics of
different structures resting on soil.

1
Non-circular shapes are difficult to treat
4:th order analytically. Numerical solutions have
0.8 however been presented e.g. by
f
1
Dasgupta and Rao [5], Elorduy et al.
0.6 [6], Meek [16], Veletsos and Verbic
-f 2:nd order
2 [25] and Öner[18]. The most
0.4
interesting case from a practical point of
view is the one with a rectangular load
0.2

area.
0

0 0.5 1
Because of the variation of the elastic
1.5
a properties of the soil with depth as
0

described above, there is a need for

Figure 8: Compliance functions for a rectangular
methods to estimate the influence of
load area. Approximation of solution by
these effects. Dasgupta and Rao [5]
Elorduy et al. (points in the diagram) with
made 3D FEM-calculations for cases
2-pole and 4-pole functions.
having linearly increasing G-modulus
with depth. Öner [18] calculated
equivalent spring constants and damping ratios for cases with spatially varying elastic properties.
Borja [2] included hysteresis, rate dependence and irreversible deformation of the soil in 3D FEM-
calculations of circular and square foundations.

Roller simulation program

A system consisting of a vibratory roller
compacting a soil layer is fairly complex. A
simulation program therefore has to be simplified in
many respects and focused on the parameters that
are most important for the problem at hand.
The modelling of different details in the roller may
be important for the roller designer and may
require a modelling of for instance the exciting shaft
including the dynamic effects of couplings and the
hydraulic motor driving the shaft, while the
modelling of the soil may be simplified.
Figure 9: Self propelled vibratory roller.
On the other hand, when dealing with the compaction efficiency of a machine, the interaction between
the drum and the soil is very important and therefore has to be modelled in greater detail.

9
 SIMVIB2.DOC 94-11-07

A series of different simulation models has developed at Geodynamik over the last 15 years. They
were tailored for different needs and the programs have been used for basic research in connection
with development of compaction meters and automatic rollers and also for evaluation of the roller
designs of different companies.
In this article, the emphasise will be placed on the basic part of the problem, namely the interaction
between the cylindrical drum and the soil. The purpose is to demonstrate the basic ideas and
possibilities using a simplified model.
A conventional vibratory roller excited by a rotating eccentric mass will be studied. A system with
two counter-rotating eccentric masses adjusted for vertical forces can also be investigated with the
same model and will give similar results.
Calculations with models taking into account both vertical and horizontal motions as well as rotations
of the drum, have shown that there are some horizontal reaction forces at the drum/soil interface. The
vertical forces are, however, the most important for the compaction effect and for the other
phenomena we are going to study here.
We will therefore deal only with drum/soil forces in the vertical direction and neglect the horizontal
drum/soil forces, resulting from the rotation of the drum and the horizontal motion of the drum.
This is at least qualitatively justified by the curve in Figure 10, showing a typical result from
calculation of the drum acceleration in vertical direction vs. horizontal direction with time as
parameter. The result compares well with results obtained when measuring with biaxial
accelerometers on a roller drum.
The frame of the roller consist of several stiff parts that are
az
connected to each other with more or less soft connecting
elements. As seen from the drum, the frame has the main
effect of supplying a static load on it, in order to increase
the line load on the ground and to enable full exploitation of
dynamic compaction force.
ax
The connection between the drum and the frame is normally
very soft, the resonance frequency of the frame being of the
order of one tenth of the excitation frequency. The damping
in the suspension is also generally very low. The
transmission of dynamic forces from the drum to the frame
is therefore small compared to the other forces acting on the
Figure 10: Simulated drum drum during compaction.
acceleration in
horizontal and vertical
directions.

10
SIMVIB2.DOC 94-11-07

1. Roller model

The simplified model of the drum being used is shown in Figure 11.
Ff is the force acting on the drum from the frame. Only the
Ff static part of the force is taken into account.
Fe(ωt) is the force generated by the eccentric system. It is
Fe z assumed to be varying harmonically at one single frequency.
ωt md is the mass of the drum.
x
md g Fs is the reaction force from the soil, which is a complex
function of the motion of drum and the deformation of the soil.
Fs All the forces on the drum, except the soil force Fs, can readily
be calculated from the masses and from details of the
eccentric system.
Figure 11: Forces acting on the
drum. The motions of the drum in the vertical z-direction and
horizontal x-direction are governed by the following equations:

d 2z
md = Fe sin ( ωt ) + Fs − F f − md g (11)
dt 2

d2x
md = Fe cos ( ωt ) (12)
dt 2

2. Soil model

As was outlined above, the deformation properties of soil are very complex especially under dynamic
conditions. In addition, there are always plastic deformations taking place under the drum of the
roller. It is not realistic to model all the details of a set of layers at all stages of compaction.
The aim was instead to design a numerical model that will enable one to study
• a thick homogenous layer of granular material
• the nearly stationary conditions at the final stage of compaction
• forces and deformations at the drum/soil interface
• the influence of different soil parameters
• the influence of different roller parameters
• the possibilities and limitations of extracting information on soil properties from measured roller
parameters.
The real soil layer has to be modelled in such a way, that it describes with sufficient accuracy
• energy consumption through plastification
• flow of energy by elastic waves moving away from the area
• return of energy to the roller

11
SIMVIB2.DOC 94-11-07

This aim has to be achieved by using as few parameters as possible to describe the soil, and
furthermore by using parameters that have a geotechnical meaning and can be estimated for a
practical case. This rules out the use of a system of springs,
dampers and masses.
A simple system that meets the requirements consists of a 'plastic
layer' on top of an elastic half space. The plastic and elastic effects
are in this way separated and attributed to two separate layers.
r
This does not mean however that energy losses are confined to the
'plastic layer' - energy will also be lost in the elastic half space
through radiation and internal damping.
Plastic layer p The 'plastic layer' will deform according to the actual force applied
and is considered to have no mass and no elasticity. It's thickness
Elastic half space G ρν will actually vary with the magnitude of the contact force. The layer
is characterised by a single parameter p, defined as the depth of the
remaining indentation created by the roller drum when lowered
Figure 12: Definitions. upon the flat surface with it's static load and than lifted again. See
Figure 12. The value of p is mainly a function of the internal friction
angle and density of the soil and of the line load and radius of the drum.
As detailed above, the relationship between the vertical displacement of the roller drum and the force
will be close to linear when the cylinder is sinking into an originally flat surface, causing progressive
failure of the soil in the contact zone. For moderate deformations, the cylindrical drum surface can be
approximated by a parabolic cylinder and the width of the contact area will be increasing as the
square root of the vertical displacement. According to conventional bearing capacity theory, the
failure load is proportional to the square of the width. The resulting equation describing the
deformation of the plastic layer alone is therefore assumed to be:
∆Fs
∆zsp = − p ⋅ F ∆Fs > 0
stat (13)
∆zsp = 0 ∆Fs ≤ 0

where Fstat is the total static load acting on the soil surface when the roller is at rest.
The elastic half space assumed for the layers below the 'plastic layer' is loaded by a strip load
assumed to have a width equal to a factor k times the width of the contact area created by the plastic
deformation and a length equal to the width of the roller drum. The stress distribution is assumed to
be parabolic along the rolling direction and constant in the cross direction. See Figure 3.
The displacement of the elastic half space was expressed above as a transfer function H(s)
approximating the Reissner functions for the particular loaded area and stress distribution. The
relationship between the displacement and total force is thus formally written as
z se = H ( s ) ⋅ F~s
~ (14)

Equation (14) represent a differential equation which has to be solved together with equation (13).
Because of the constraints from equation (13) the system will be non-linear and must be solved
numerically. The solution has to be performed in phases because the roller drum may lift off from the
ground and it may even make a whole cycle in the air without contact with the ground. The instants

12
SIMVIB2.DOC 94-11-07

when the solution enters another phase or state is not known in advance and has to be determined as
part of the solution.
The elastic half space is characterised by the primary variable G, the shear modulus of the material.
The elastic material is further defined by the secondary parameters density ρ and Poisson's ratio ν
which have a smaller range of variation than G. These parameters enter into the term H(s) in equation
(14), deciding the static stiffness and the dynamic behaviour of the system.
The roller moves along the surface and it is assumed that the roller speed is high enough that the drum
will hit fresh soil every cycle.

3. Numerical solution

The interaction between the drum and the soil passes different phases during the excitation cycles.
The forces and the motions of the soil and the drum are described by different sets of equations in the
different phases. The numerical solution is done in short time steps and the state of the system is
checked in every step to decide if a change of state has occurred.
Three different states of the system are considered:
Loading. The plastic layer is successively deformed according to equation (13) and the load is
transmitted through the plastic layer and applied as
strip load with parabolic distribution to the surface of
the elastic half space. The surface of the elastic half
space is displaced according to equation (14). The
Loading
drum is moving according to equations (11) and (12)
and consistent with the soil contact force.
Unloading. The plastic layer is not deformed further.
Air phase Unloading The soil force applied by the drum is decreasing but
there is still contact between the drum and the soil. The
forces and displacements are determined by equations
(11) and (14).
Figure 13: State diagram for numerical
Air phase. The contact is lost between the drum and
calculation.
the soil. The drum and the soil move as two separate
systems. This state may never occur if the excitation is low or the soil is very soft.
The different phases and the possible transitions between phases are illustrated by the state diagram
of Figure 13. Each new time interval in the calculation represents a new state, that may or may not
belong to the same phase.
The differential equations are transformed into difference equations using a constant time step ∆t. An
IIR-method is applied and the output value for a new time step is calculated as a weighted sum of the
new input value and old output and input values. The weights are given by the coefficients of the
difference equations. The number of old values involved is equal to the order of the differential
equation being integrated.
The time step size necessary is dependant on the highest frequencies present in the results. A suitable
value for ∆t is one fifth of the shortest period of interest. The mapping of the transfer function to the z-
plane causes a certain warp of the frequency scale that is corrected for.
13
SIMVIB2.DOC 94-11-07

Example of roller simulation

Input data
An example will be presented of typical results that can be obtained with the kind of simulation model
that has been described. A typical 10 ton type self-propelled vibratory roller (Figure 9) will be
simulated.
Roller data:
Frame Mass acting vertically on the drum 2400 kg
Drum Mass 3200 kg
Diameter 1.50 m
Width 2.10 m
Eccentric Mass 50 kg
Frequency 15 - 40 Hz
Amplitude 0 - 2.5 mm

Soil data:
Density ρ = 2000 kg/m3
Poisson's ratio ν = 0.25
There is a certain relationship between p and G . A weak soil having a low G will generally have a
high value of p and vice versa. A typical co-
1.5 variation of the two parameters are shown in Figure
14.
p
1 There will be considerable variation in this
relationship for different soil conditions. A stiff layer
on top of weak sublayers will for instance have a p
0.5
that is below the curve in Figure 14.
The data for Figure 14 was obtained from test runs
0
0 20 40 60 80 100 with the simulation program and comparison of
G simulated drum accelerations with recorded ones.
When comparing different roller sizes the value of p
Figure 14: Assumed co-variation of soil has to be scaled to the radius of the drum.
parameters p and G with
varying degree of compaction. Elastic transfer function:
A 2-pole approximation of Reissner functions for a rectangular area computed by Elorduy et al was
used. In order to get better agreement with measured results, some internal damping was introduced.
The calculated Q-value of 0.6 was decreased to 0.4.

14
SIMVIB2.DOC 94-11-07

The length/width-ratio of the loaded area differs from the one valid for Elorduy's data. When
calculating the frequency scale for the actual loading area it was reshaped to the same length/width-
ratio, while keeping the magnitude of the area unchanged.
Time step size:
The time step was selected to give 128 samples for one revolution of the eccentric shaft, which
means that the size of the time step varies with the excitation frequency.

Simulation results

1. Drum accelerations

The simulated acceleration in the vertical direction is an important indicator of the validity of the
simulation result. The acceleration is easy to
150
measure on a roller and it is also used as the
35/0.80 input signal for the calculation of various
100
m/s2 compaction meter values.
50

0
Figure 15 shows simulated results for two
different soil conditions. The excitation
-50
frequency and amplitude have the nominal
-100
values for the roller, which are 28 Hz and
-150 1.6 mm respectively.
0 20 40 60 80 100 120
ms
In the first diagram (G=35 MPa,
p=0.8 mm), the vibration is stable with the
150 cycles repeating without change. There is a
80/0.50
100 typical asymmetry caused by the drum lifting
m/s2 off of the ground during a certain part of
50
each cycle. The drum hits the ground at a
0
certain phase angle of the vibration, which
-50 causes the upwardly directed acceleration
-100 peak.
-150
0 20 40 60 80 100 120
The second part of Figure 15 illustrates a
ms situation with a stiffer soil (G=80 MPa,
p=0.5 mm), when the roller is double-
Figure 15: Simulated vertical drum accelerations. jumping. The drum hits the ground very hard
f=28 Hz, A=1.6 mm. every other cycle and gets such a heavy
rebound, that it makes a full cycle in the air
before hitting the ground again.
The vibration is stable, but is still not desirable for various reasons. Firstly the very heavy blows will
hit the ground with double the normal separation and will cause reloosening of already compacted
material. Secondly, the fundamental frequency of the drum vibration will be halved and at the same
time the drum will become a very high amplitude. This will cause heavy excitation of the roller frame.

15
SIMVIB2.DOC 94-11-07

2. Vibration stability

Because of the non-linearities involved in the system, instability phenomena are likely to occur. When
the excitation amplitude is increased, period doubling will show up, first as double jump as illustrated
above and than as another doubling leading to quad jump i.e. a situation where the vibration pattern
has a period of 4 full revolutions of the eccentric shaft. When the excitation exceeds a certain limit the
vibration enters a chaotic state and the vibration will never repeat itself but will vibrate in a random
manner.
It is important to know the conditions under which these instabilities appear. Instabilities can be
investigated by running simulations for a large number of combinations of frequency and amplitudes
and plotting the resulting vibration modes as in Figure 16.
These simulations are rather time consuming because the program has to be run for a fairly long time
for each combination. The reason for this is that the vibration need many periods to settle to a stable
state.
The instability visible to the left in the diagram at frequencies of 15-18 Hz is related to the drum/soil
resonance. The frequency of this resonance decreases with the excitation amplitude, which is a well
known fact confirmed by measurements (Kröber [13] ).

2.5

Chaos

2.0 Quad

A (mm) Double

1.5 Single

1.0

0.5
20 30 40
f (Hz)

Figure 16: Vibration stability result for a soil with parameters G=80 MPa and
p=0.5 mm.

3. Force/displacement results

One important application for the simulation program is the assessment of the compaction efficiency
of the roller.
The compaction near the surface is accomplished by a displacement of the soil and is unfortunately
also accompanied by a re-loosening of previously compacted soil volumes. This is a well known

16
SIMVIB2.DOC 94-11-07

disadvantage for vibratory rollers that is difficult to avoid. The loose upper layer often has to be
compacted by a final static pass over the area.
The compaction effect on deeper layers is mainly caused by the creation of a vibration in the soil
material. The vibration enables the grains to relocate into a more dense arrangement under influence
by the combined effect of gravity and static load generated by the roller.
The simulation program generates output data for soil forces and displacements that can be used for
an estimation of the compaction potential. By varying the parameters of the roller it is then possible to
study the influence on compaction efficiency.
It is natural to present the results in the same way as the results from a plate load test, i.e. as drum
displacement versus applied force. A few examples are shown in Figure 17.

1 1
05/2.00 20/1.40
0 0

-1 -1
mm mm
-2 -2

-3 -3

-4 -4

-5 -5
0 100 200 300 400 0 100 200 300 400
kN kN

1 1
35/0.80 80/0.50
0 0

-1 -1
mm mm
-2 -2

-3 -3

-4 -4

-5 -5
0 100 200 300 400 0 100 200 300 400
kN kN

Figure 17: Drum displacement vs. soil force. The vertical dotted lines indicate the static load.
f=28 Hz, A=1.6 mm.

17
SIMVIB2.DOC 94-11-07

2.5 The maximum force Fs max transmitted to

the soil is an interesting parameter. As
seen in Figure 17, the maximum soil force
2
in those examples vary from twice the
A (mm)
5 static load to 6.9 times the static load. In
the last case the force occurs only every
1.5 other cycle with the drum doing a cycle in
the air in between.
4
1 The ratio Fs max / Fstat has been investi-
gated further for the whole range of
3 frequencies and amplitudes. In cases of
0.5 double jump the value of the quotient is
20 30 40 evaluated as an average value for a large
B f (Hz) number of cycles.
Figure 18: Maximum soil force in relation to the As seen from the contour plot in Figure
static force. G=80 MPa and p=0.5 mm. 18 the value Fs max / Fstat varies gradually
from 3 to 6 in the area of stable vibration.

4. Compaction meter results

The simulation program delivers the necessary output signals for the calculation of CompactoMeter
Values (CMV). An example of simulated
120 CMV is shown in Figure 19. The low
amplitude will never put the roller into
100
A= 0.8 mm double jump mode and the CMV are
CMV

80 increasing monotonously with the

60 stiffness of the soil.
A= 1.6 mm
40 The high amplitude on the other hand will
20 force the roller into double jump mode
and the CMV-curve will show a
0
0 20 40 60 80 100 120
discontinuity. The value of CMV will
G
then increase again monotonously as the
stiffness increases, however, the value
will never reach the level it has for the
Figure 19: CMV vs. G for two different amplitudes. low amplitude.
The value of the plastic parameter p was
It is evident from these results that it is
coupled to the value of G according to
very important, when using CMV for
Figure 14.
result verification, that the roller vibration
is stable and takes place in the same
mode and that the same amplitude is
used as at the calibration .

18
SIMVIB2.DOC 94-11-07

CMV varies also with frequency, with all other parameters kept constant. The variation over the
whole range of amplitudes and frequencies for a fixed combination of G and p is illustrated by Figure
20.
2.5

1.5

0.5
20 30 40
A

Figure 20: Variation of CMV with frequency and amplitude of the roller for constant soil
conditions (G=80 MPa, p=0.5 mm)

19
SIMVIB2.DOC 94-11-07

Notation
α length/width ratio of loaded area (=L/B)
A nominal amplitude of the roller drum
a0 dimensionless frequency (=dω/vs)
ax horizontal drum acceleration
az vertical drum acceleration
B half width of loaded area
c cohesive strength
CMV Compaction Meter Value
d characteristic linear dimension
D dimensionless displacement
δ vertical displacement
e void ratio
f excitation frequency
Ff Fs Fe forces (frame, soil, exciter)
Fstat static force
γ unit weight
G Gmax shear modulus, initial shear modulus
Η transfer function
ϕ internal friction angle
k factor for calculating width of the loaded area
κ static load factor
L half length of loaded area
md drum mass
ν Poisson´s ratio
Nc Nq Nγ bearing capacity factors
p parameter for plastic deformation
Q Qu load, ultimate load
r drum radius
ρ density
σu ultimate stress
vs shear wave velocity
ω angular frequency
W roller drum width
zse zsp vertical soil displacement (elastic, plastic)

20
SIMVIB2.DOC 94-11-07

References
1. Ahmad S., Bharadwaj A. : Horizontal Impedance of Embedded Strip Foundations in
Layered Soil. Journal of Geotechnical Engineering, ASCE, Vol. 117, No. 7, July 1991.
2. Borja R.I. et al. : Nonlinear response of vertically Oscillating Rigid Foundations.
Journal of Geotechnical Engineering, ASCE, Vol. 119, No. 5, May 1993.
3. Bycroft G. N. : Forced Vibrations of a Rigid Circular Plate on a Semi-infinite Elastic
Space and on an Elastic Stratum. Phil. Trans. of the Royal Society, London, Series A, Vol.
248, No. 948, 1956, pp. 327 - 368.
4. Chen E.F., Davidson H.L. : Bearing Capacity Determination by Limit Analysis. Journal
of the Soil Mechanics and Foundations Division, ASCE, Vol. 99, No. SM6, June 1973.
5. Dasgupta S. P., Rao K. : Dynamics of Rectangular Footings by Finite Elements.
Journal of the Geotechnical Engineering Division, ASCE, Vol. 104, No. GT5, Proc. Paper
13763, May, 1978, pp. 621-637
6. Elorduy, J. . Nieto, J.A. , Szekely. E.M. : Dynamic Response of Bases of Arbitrary
Shape Subjected to Periodic Vertical Loading. Proc. of the International Symposium on
Wave Propagation and Dynamic Properties of Earth materials, Univ. of New Mexico,
Albuquerque, N.M. , August 1967, pp. 105-121.
7. Forssblad L. : Vibratory Soil and Rock Fill Compaction, Dynapac Maskin AB, 1981.
8. Gazetas G. et al. : Vertical response of Arbitrarily Shaped Embedded Foundations.
Journal of Geotechnical Engineering, ASCE, Vol. 111, No. 6, June 1985.
9. Grabe J. : Experimentelle und theoretische Untersuchungen zur flächendeckenden
dynamischen Verdichtungskontrolle. Dissertation. Veröffentlichung des Institutes für
Bodenmechanik und Felsmechanik, Universität Karlsruhe, Heft 123, 1992.
10. Hardin B.O., Richart F.E. : Elastic Wave Velocities in Granular Soils. Journal of the
Soil Mechanics and Foundations Division, ASCE, Vol. 89, No. SM1, February, 1963.
11. Hsieh T.K. : Foundation Vibrations. Proceedings Institution of Civil Engineering, London,
Vol. 22, 1962.
12. Karafiath L.L., Nowatzki E.A. : Soil Mechanics for Off-road Vehicle Engineering.
Trans Tech Publications, Series on Rock and Soil Mechanics, Vol. 2, No. 5, 1978.
13. Kröber W. : Untersuchung der dynamischen Vorgänge bei der Vibrationsverdichtung
von Böden. Dissertation. Schriftenreihe des Lehrstuhls und Prüfamtes für Grundbau,
Bodenmechanik und Felsmechanik der TU München, Heft 11, 1988.
14. Lysmer J., Richart F.E. Jr : Dynamic Response of Footings to vertical Loading.
Journal of the Soil Mechanics and Foundations Division, ASCE, Vol. 92, No. SM1, January,
1966.

21
SIMVIB2.DOC 94-11-07

15. Machet J.M. : Interprétation de l´efficacité des compacteurs vibrants. Rapport de

recherche No 59, Laboratoires Central des Ponts et Chaussées, Paris, 1976.
16. Meek J.W, Wolf J.P. : Approximate Green´s Function for Surface Foundations.
Journal of the Geotechnical Engineering Division, ASCE, Vol. 119, No. 10, Oct. 1993.
17. Nii, Y. : Experimental Half Space Dynamic Stiffness. Journal of Geotechnical Engineering,
ASCE, Vol. 113, No. 11, Nov. 1987.
18. Öner M. : A Finite Element Approach to the Equivalent Spring Constant Problem in
Soil dynamics. Soil Mechanics and Foundation Engineering Division, The Norwegian
Institute of Technology, Trondheim, Norway, August 1975.
19. Pietzsch, D. : Untersuchungen zum Schwingungsverhalten und zur
Verdichtungswirkung von Vibrationswalzen. Fortschr.-Ber. VDI Reihe 4, Nr. 108,
Düsseldorf, VDI-Verlag 1991.
20. Quinlan P. : The Elastic theory of Soil Dynamics. Symposium on Dynamic Testing of
Soils, Atlantic City, N.J., 1953. ASTM Special Tech. Publ. No. 156.
21. Reissner V.E. : Stationäre, Axialsymmetrische, durch eine Schüttelnde Masse Erregte
Schwingungen eines Homogenen Elastischen Halbraumes. Ingenieur-Archiv, VII Band,
pp. 381-396, 1936.
22. Richart F.E., Woods R.D., Hall J.R. : Vibrations of Soil and Foundations. Prentice-
Hall, Inc., Englewood Cliffs, N.J., 1970.
23. Sung T.Y. : Vibrations in Semi-infinite Solids due to Periodic Surface Loading.
Symposium on Dynamic Testing of Soils, Atlantic City, N.J., 1953. ASTM Special Tech.
Publ. No. 156.
24. Széchy K. : Der Einfluss der Sohlflächenform von Streifenfundamenten auf die
Tragfähigkeit und Spannungsausbreitung. Verein Deutscher Ingenieure - Zeitschrift, Vol.
109, Nr.8, 1967, pp. 339-344.
25. Veletsos A., Verbic B. : Basic Response Functions for Elastic Foundations. Journal of
the Engineering Mechanics Division, ASCE, Vol. 100, No. EM2, April, 1974.
26. Vesic A. S. : Analysis of Ultimate Loads of Shallow Foundations. Journal of the Soil
Mechanics and Foundations Division, ASCE, Vol. 99, No. SM1, January, 1973.
27. Vrettos C., Prange, B. : Evaluation of In Situ Effective Shear Modulus from
Dispersion Measurements. Journal of Geotechnical Engineering, ASCE, Vol.116, No.10,
October 1990.
28. Yoo, T. Selig, E.T. : Dynamics of Vibratory Roller Compaction. Journal of the
Geotechnical Engineering Division, ASCE, No.105, 1979, pp. 1211-1231.

22