You are on page 1of 10

Capturing Nonlinear Vibratory Roller Compactor Behavior

through Lumped Parameter Modeling


Paul J. van Susante, S.M.ASCE1; and Michael A. Mooney, Ph.D., P.E., M.ASCE2

Abstract: Continuous monitoring of soil properties using an instrumented roller compactor requires models that can capture the essential
features observed during drum/soil vibration. This paper presents the results of lumped parameter modeling of the drum/soil system
Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

together with data from complex nonlinear behavior observed experimentally during operation on sandy soil. Model parameters and
response were developed using experimental data collected over a wide range of operating frequencies. Three and four-degree-of-freedom
共DOF兲 models with linear and nonlinear soil elements were investigated. The results showed that a 3DOF model incorporating the soil,
drum, and frame of the roller was successful in capturing behavior during coupled drum/soil vibration and during decoupling 共i.e., loss of
contact between drum and soil兲. Modeling the drum/soil decoupling accounted for most of the experimentally observed nonlinearity. The
addition of nonlinear soil stiffness due to the curved drum effect and due to strain hardening soil behavior accounted for additional
nonlinearity observed experimentally. Experimentally observed drum rocking during coupled drum/soil vibration was successfully mod-
eled with a 4DOF drum-frame model. The analysis also revealed that commonly observed heterogeneous soil conditions give rise to a
transient response that can have a significant influence on vibration behavior.
DOI: 10.1061/共ASCE兲0733-9399共2008兲134:8共684兲
CE Database subject headings: Data analysis; Geotechnical models; Dynamic models; Nonlinear systems; Soil compaction; Soil
dynamics; Parameters.

Introduction paction efficiency, prevent overcompaction, and improve


homogeneity.
Vibratory roller compactors are commonly used to transform The instrumentation of vibratory rollers has been addressed in
loosely placed granular and mildly cohesive soils into densely the literature 共Adam 1996; Adam and Kopf 2004; Brandl et al.
packed load bearing earth structures. Given the trend towards 2005; Rinehart and Mooney 2008兲 and is increasingly being
assessment of mechanistic soil properties 共e.g., modulus兲 and adopted in practice. Recent experimental data collected with in-
the need to improve earthwork quality control/quality assurance strumented roller compactors 共Anderegg and Kaufman 2004;
共QC/QA兲, there is increased interest in vibration-based monitor- Adam and Kopf 2004; Mooney et al. 2006; Mooney and Rinehart
ing of roller compactors to provide continuous assessment of 2007兲 reveal fairly complex nonlinear behavior, including loss of
the soil’s stiffness and/or modulus. Continuous vibration moni- contact between the drum and soil, drum and frame rocking, and
toring is essentially a system identification-type inverse prob- chaotic behavior. In addition, a traveling roller interacting with
lem wherein the soil parameters are estimated via measured underlying soil heterogeneity and employing a variety of excita-
inputs 共eccentric excitation forcing function兲 and measured tion frequencies and amplitudes gives rise to transient behavior.
outputs 共roller drum and frame vibration兲. In this context, a robust The development of mass-spring-dashpot lumped parameter
model structure is critical to effective system identification. models to characterize roller-soil behavior has evolved per the
An accurate model of the soil-roller system is also required findings from experimental studies. Early efforts 共Yoo and Selig
for so-called “intelligent soil compaction” wherein the roller’s 1979, 1980兲 proposed a 2-degree-of-freedom 共DOF兲 model to
excitation 共forcing兲 frequency and/or amplitude are automatically represent steady-state vertical drum and frame kinematics. These
adjusted via feedback control in an attempt to improve com- investigators restricted their modeling to contact behavior 共ne-
glecting when the drum loses contact with the soil, which occurs
1 in practice more than 50% of the time兲 and used a linear spring
Ph.D. Candidate, Division of Engineering, Colorado School of
Mines, 1610 Illinois St., Golden, CO 80401. E-mail: paulvans@ 共and viscous damping兲 to model soil stiffness. Limited experi-
mines.edu mental data at frequencies much greater than resonance were pre-
2
Associate Professor, Division of Engineering, Colorado School of sented; hence, the model could not be validated over the usable
Mines, 1610 Illinois St., Golden, CO 80401 共corresponding author兲. frequency range and at resonance. Yoo and Selig 共1979兲 model
E-mail: mooney@mines.edu damping ratio 共⬃15% 兲 and soil stiffness 共⬃5 MN/ m兲 are much
Note. Associate Editor: Anil Misra. Discussion open until January 1, lower than the 20– 30% damping and 20– 80 MN/ m stiffness
2009. Separate discussions must be submitted for individual papers. To that will be shown here and have been observed elsewhere 共e.g.,
extend the closing date by one month, a written request must be filed with
Quibel 1980; Adam 1996; Anderegg 1997兲. Drum/soil decoupling
the ASCE Managing Editor. The manuscript for this paper was submitted
for review and possible publication on October 17, 2006; approved on has been modeled by including a DOF to represent the involved
January 31, 2008. This paper is part of the Journal of Engineering Me- soil 共Quibel 1980; Machet and Sanejouand 1980; Kröber 1988;
chanics, Vol. 134, No. 8, August 1, 2008. ©ASCE, ISSN 0733-9399/ Pietsch and Poppy 1993; Adam 1996兲. These modeling efforts
2008/8-684–693/$25.00. revealed nonlinear drum response due to decoupling from the

684 / JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008

J. Eng. Mech., 2008, 134(8): 684-693


Fig. 1. 共a兲 Tandem drum; 共b兲 single drum; and 共c兲 vibratory roller compactors and coordinate axes
Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

ground; however, very little loss of contact experimental data was bratory excitation is created by an eccentric mass configuration
presented to develop and/or validate model behavior over the that rotates about the drum axle 共see Fig. 1兲 and creates a cen-
broad frequency range of interest 共15– 40 Hz兲. trifugal force m0e0⍀2, where m0e0⫽eccentric mass moment and
Adam 共1996兲 and Anderegg 共1997兲 used lumped parameter ⍀⫽circular frequency. Various eccentric mass configurations
modeling to explore the various operational modes of roller vi- exist, e.g., eccentric mass configurations may rotate unidirection-
bration, including nonlinear and chaotic vibration. In addition to ally creating both horizontal and vertical excitation components.
drum/soil coupled behavior 共i.e., full contact throughout兲, Adam Conversely, counter-rotating eccentric masses can serve to cancel
共1996兲 characterized both partial loss of contact where the drum the horizontal excitation component or provide directional ampli-
decouples from the soil for a portion of each loading cycle and tude if the entire assembly is rotated. The resultant eccentric mass
“jump” mode 共also referred to as double jump兲 wherein the drum moment 共m0e0兲 and excitation frequency ⍀ are often variable;
loses contact for more than one cycle of vibration at a time. The some roller compactors have multiple m0e0 settings that must be
resulting nonlinear signal in jump mode includes a subharmonic adjusted manually while others have electronically controlled
at one half the excitation frequency 共Adam 1996; Adam and Kopf m0e0 variability allowing for an unlimited number of settings
2004兲. Anderegg 共1997兲 and Anderegg and Kaufmann 共2004兲 between minimum and maximum. Excitation frequencies typi-
described jump mode and rocking mode vibration as chaotic cally range from 20 to 40 Hz during soil compaction and can
states. Employing chaos theory, Anderegg 共1997兲 showed that often be electronically controlled. With medium and high m0e0
rocking and jump mode vibration states occur above a certain settings, typical operating frequencies, and stiff soils, the vertical
centrifugal force and soil stiffness combination 共roller parameter excitation force can exceed the static weight of the roller at the
specific兲. In current practice, vibratory roller compactors employ drum location and cause drum/soil decoupling during a portion of
feedback control of the centrifugal force to prevent chaotic mo- each cycle. Only contact and partial loss of contact operation
tion 共Anderegg and Kaufmann 2004兲 because these motions are 共from here on called “loss of contact”兲 will be considered during
harmful to the machines and dangerous for the operator. To this modeling.
end, the modeling effort presented here does not explore these When considering roller-soil modeling, a few key issues are
chaotic processes. worth noting:
This paper explores the capability of 3 and 4DOF lumped 1. A vibratory roller compactor travels in forward or reverse
parameter models with linear and nonlinear elements to reproduce mode at speeds of 0.5– 1.5 m / s. Given these speeds and the
the vibratory roller behavior observed experimentally. Compre- typical 20– 40 Hz excitation frequency range, the spatial dis-
hensive drum/soil contact and loss of contact data collected over tance between excitation cycles 共often called impact spacing兲
a 10– 45 Hz operating range is presented and used to develop and varies from approximately 20 mm 共0.5 m / s , 40 Hz兲 to
analyze model performance. In addition to loss of contact induced 75 mm 共1.5 m / s , 20 Hz兲;
nonlinearity, the nonlinear interaction between a curved drum and 2. The depth of influence or measurement depth of an instru-
soil is incorporated into the model, as is strain hardening nonlin- mented vibratory is on the order of 1 – 2 m 共Anderegg and
ear soil stiffness. The various forms of model nonlinearity are Kaufman 2004兲. Given that the typical thickness of lifts
investigated and compared with the experimental data. This paper being compacted is 150– 300 mm, the roller response is also
also explores the modeling of drum and frame rocking, and the influenced by material underlying the lift being compacted.
comparison with experimental data that reveals rocking vibration. This subsurface can be quite variable and can have a signifi-
The paper also addresses the role of transient behavior. All the cant effect if it is much softer or stiffer than the compaction
model results presented are validated with experimental data. lift, e.g., near surface bedrock; and
3. Spatial heterogeneity in soil density, moisture content, and
lift thickness, all as a result of typical construction practice,
Model Development lead to heterogeneity in soil stiffness experienced by the
roller compactor. Furthermore, soil stiffness and modulus are
Typical earthwork vibratory roller compactors have total machine very sensitive to subtle changes in moisture and density
masses ranging from 7 to 20 Mg. Drum diameter and lengths 共Nazarian et al. 2002兲, and therefore can vary considerably
approach 1.6 and 2.1 m, respectively. Single drum rollers such on a project site.
as the one depicted in Fig. 1 are commonly used on soils. Dual With these issues in mind, the general framework for the roller-
or tandem drum rollers 共see Fig. 1兲, typically reserved for asphalt soil model is shown in Fig. 2 where both the roller and soil are
compaction, are sometimes also used for soil compaction. Vi- modeled with mass-spring-dashpot components. While x and y

JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008 / 685

J. Eng. Mech., 2008, 134(8): 684-693


Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. 共a兲 3DOF lumped mass model of vibratory roller compactor and soil; 共b兲 free-body diagrams used to generate Eqs. 共1兲–共6兲

motion is kinematically possible, only vertical 共z兲 motion is con- roller forces. The numerical modeling efforts and experimental
sidered here because it directly contributes to the determination of results presented here largely reflect roller vibration behavior on
soil stiffness. Vertical 共z兲 motion requires 3DOF, including the compacted soil, where applications such as “proof rolling” are
vertical motion of the drum, frame, and soil, where the soil is employed to assess soil properties. To this end, plastic soil defor-
modeled as a DOF to enable loss of contact. The drum mass mation is deemed insignificant and is not included in the model.
includes the eccentric assembly while the frame mass is deter- Additional model nonlinearity is introduced due to drum curva-
mined by subtracting the drum weight from the total static weight ture 共Lundberg 1939兲. For analysis of transient response, the
measured under the drum. Rubber mounts isolate the drum from spring element is also considered time varying to simulate spatial
the frame. These bearings were modeled with a complex stiffness heterogeneity. The viscous dashpot is used to represent both ra-
relationship 共static stiffness plus viscous dashpot兲 provided by the diation and material damping, with the former dominating the
manufacturer Lord Corporation. energy dissipation of the system 共Lysmer and Richart 1966;
The soil model employed here—mass, spring, dashpot—is a Mooney and Rinehart 2007兲.
simple analog that has been proven effective in modeling surface The equations of motion for contact and loss of contact behav-
vibration of foundations resting on homogeneous and layered ior were determined via force equilibrium using free body dia-
elastic half spaces 共Lysmer and Richart 1966; Gazetas 1983; grams of frame, drum, and soil 共see Fig. 2兲. Eqs. 共1兲 and 共2兲
Baidya et al. 2006; Wolf 1994兲. During coupled drum/soil vibra- represent drum and frame behavior during drum/soil contact
tion, a portion of the involved soil is in synchronous motion with mode vibration. All variables are defined in the “Notation”
the drum. The amount of added or apparent soil mass has been section at the end of this paper. Note that the soil moves as one
debated throughout the foundation vibration literature with with the drum 共zd = zs兲 while in contact and as separate DOFs
no clear consensus 共e.g., Richart et al. 1970; Barkan 1962; during loss of contact. Both contact and loss of contact mode
Balakrishna Rao and Nagaraj 1960; Crockett and Hammond commonly occur within each cycle of vibration. The contact force
1949兲. This lack of consensus also pervades the roller-soil mod- Fc between the drum and soil can be determined from both
eling literature. A number of roller modeling studies neglected the soil equilibrium and drum equilibrium 关Eq. 共3兲兴. Eqs. 共1兲–共3兲
soil mass 共Yoo and Selig 1979; Anderegg 1997兲. Quibel 共1980兲 are valid while the drum and soil are in contact 共Fc ⬎ 0兲. During
used an apparent mass equal to 62% of the drum mass while loss of contact behavior 共Fc ⬍ 0兲, frame, drum, and soil behavior
Kröber 共1988兲 used 10% of the drum mass. Pietzsch and Poppy are described by Eqs. 共4兲–共6兲. For all equations, displacement,
共1993兲 adopted an apparent soil mass equal to 36% of the drum
mass. From our data and numerical simulations, the inclusion of
an apparent soil mass equal to 20% of the drum mass provided a
good amplitude match between the models and the experimental
data. This can be seen in Fig. 3 in which 3DOF drum-frame-soil
model results with 0, 20, 40, and 60% added soil are shown
together with experimental data. Here, the experimental and
model results involve fully coupled drum/soil behavior 共hence,
the 3DOF model reduces to 2DOF兲. Fig. 3 presents the peak
vibration amplitudes 兩Z̈d 兩 / ⍀2 when the drum is at the top and
bottom of its vertical cycle. Fig. 3 modeling results employ a
linear soil stiffness and yield the same top and bottom amplitudes.
The soil spring element is a composite stiffness that represents
both the lift being compacted and the underlying layers within the
depth of influence 共Odemark 1949; Anderegg and Kaufmann
2004兲. The soil stiffness should be considered stress dependent Fig. 3. Influence of added or apparent soil mass on model drum
and thus nonlinear given the significant stresses created by the response during contact mode vibration

686 / JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008

J. Eng. Mech., 2008, 134(8): 684-693


velocity, and acceleration are downward positive as shown in msg − kszs − csżs
Figs. 1 and 2. Zero displacement is defined as the location of z̈s = 共5兲
ms
the masses before any force 共including gravity兲 is applied. Values
for the system characteristic constants 共provided by the manufac-
turers兲 are as follows: drum mass 共md兲 = 2,638 kg, frame mass m f g + kdf 共zd − z f 兲 + cdf 共żd − ż f 兲
z̈ f = 共6兲
共m f 兲 = 4,565 kg, the stiffness coefficient between the drum and mf
frame 共kdf 兲 = 3.4 MN/ m, and the damping coefficient between the
A Runga–Kutta finite difference numerical method was employed
drum and frame 共cdf 兲 is determined by multiplying kdf with the
to evaluate both transient and steady-state roller vibration, and to
loss factor 共␩ = 0.16兲 and dividing by the eccentric rotational fre-
accommodate nonlinear soil stiffness and loss of contact behavior.
quency 共⍀兲
This fourth-order numerical scheme is very stable but requires
z̈s = z̈d many evaluations per simulated time step 共0.5 ms in most cases兲.
m0e0⍀2 sin共⍀t兲 + mdg + msg − kszd − csżd − kdf 共zd − z f 兲 − cdf 共żd − ż f 兲 Within the numerical scheme, when Fc 艋 0 per Eq. 共3兲, the model
= response is governed by Eqs. 共4兲–共6兲. The displacement criterion
Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

md + ms
zd 艌 zs is used to determine when contact is regained. Upon
共1兲
regaining contact 共zd 艌 zs兲, md and ms combine and require the
m f g + kdf 共zd − z f 兲 + cdf 共żd − ż f 兲 calculation of the combined values of the velocity, displacement,
z̈ f = 共2兲 and acceleration using conservation of momentum as shown in
mf Eqs. 共7兲–共9兲, respectively

Fc = msz̈d − msg + kszd + csżd = m0e0⍀2 sin共⍀t兲 + mdg + m f g msżs + mdżd


żd−new = 共7兲
− mdz̈d − m f z̈ f 共3兲 ms + md

m0e0⍀2 sin共⍀t兲 − kdf 共zd − z f 兲 − cdf 共żd − ż f 兲 zs + zd


z̈d = +g 共4兲 zd−new = 共8兲
md 2

m0e0⍀2 sin共⍀t兲 + mdg + msg − kszd-new − csżd-new − kdf 共zd-new − z f 兲 − cdf 共żd-new − ż f 兲
z̈d-new = 共9兲
md + ms

Results used for both the contact and loss of contact behavior. The ks and
cs values were selected based on a best fit analysis with the ex-
Model results are presented together with experimental data to perimental data. Figs. 4共a and b兲 illustrate typical vertical drum
illustrate the efficacy of 3 and 4DOF lumped parameter modeling. inertia, frame inertia, and eccentric excitation behavior observed
The experimental results were determined from an instrumented experimentally for contact and loss of contact response, respec-
double smooth drum vibratory roller 共Ingersoll Rand DD138兲. tively. Model components of Fc are shown in Figs. 4共c and d兲
The roller traveled over lifts of poorly graded sandy soil 共SP per while Figs. 4共e and f兲 present the resultant model and experi-
Unified Soil Classification System; A-1-b per American Associa- mental contact forces. Note how Fc ⬇ 0 during loss of contact.
tion of State Highway and Transportation Officials兲 at forward The sign convention considers downward positive displacement,
velocities of 0.5– 1.5 m / s. A detailed presentation of the experi- velocity, and acceleration; positive contact force implies com-
mental program and roller instrumentation can be found else- pression between the drum and soil. The contact mode data
where 共Mooney and Rinehart 2007; Rinehart and Mooney 2008兲. 关Figs. 4共a and e兲兴 were recorded during 21 Hz vibration with
The presentation of model results begins with steady-state results m0e0 = 0.98 kg m, hence the eccentric force oscillates sinusoidally
from the 3DOF soil, drum, and frame system with linear soil with an amplitude of ⫾17 kN. The loss of contact mode data
stiffness 共and damping兲, and progresses to nonlinear soil behavior 关Figs. 4共b and f兲兴 was recorded during 27 Hz vibration with
and transient effects and to the 4DOF model for rocking mode m0e0 = 2.32 kg m; therefore, the eccentric force oscillates sinusoi-
vibration. dally with an amplitude of ⫾65 kN. The static weight is equal to
70 kN 关not shown in Fig. 4共a兲 due to scaling兴.
Drum inertia is fairly sinusoidal in both cases. Frame inertia is
Drum Response, Contact, and Loss of Contact
smaller in magnitude than the other components and is nonsinu-
The critical time history aspects of modeling roller-soil response soidal, particularly during loss of contact. The drum acceleration,
are captured in the four components of the contact force Fc reflected by the negative of the drum inertia divided by md, pro-
关Eq. 共3兲兴, namely eccentric force m0e0⍀2 sin共⍀t兲, drum inertia vides insight into the nonlinearity of the system. Figs. 4共a and b兲
共−mdz̈d兲, frame inertia 共−m f z̈ f 兲, and static weight 共mdg + m f g兲. show that the peak negative drum acceleration 共peak positive in-
Experimental data and model response of these four components ertia兲 exceeds the peak positive drum acceleration during both
are shown in Fig. 4 for both contact and loss of contact operation. contact and loss of contact response. This peak negative response
A 3DOF roller-soil model 共ms = 0.2md兲 with linear soil stiffness occurs when the drum is at the bottom in its trajectory, i.e., during
共ks = 72 MN/ m兲 and viscous soil damping 共cs = 230 kN s / m兲 was rebound from the ground. Figs. 4共c–f兲 present the model contact

JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008 / 687

J. Eng. Mech., 2008, 134(8): 684-693


Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 4. Comparison of experimental and 3DOF model of contact force and its components for typical contact mode 共left兲 and loss of contact mode
共right兲

force components and resultant Fc. For the contact case, Fc and frequency response for vertical drum vibration while traveling
components 共including phase differences兲 agree well with the ex- over the sand test bed. The experimental data in Figs. 5共a and c兲
perimental data 关Fig. 4共e兲兴. This 3DOF model response does not represent low amplitude eccentric force 共m0e0 = 0.98 kg m兲 behav-
include soil nonlinearity and thus overpredicts the peak positive ior and Figs. 5共b and d兲 represent high amplitude eccentric force
drum acceleration and thus the minimum Fc. For the loss of con- 共m0e0 = 2.32 kg m兲 behavior. As the roller traversed a 100-m long
tact mode, the model also effectively matches the resultant Fc and testbed, the excitation frequency was stepped up over 2 – 4 Hz
components with the exception of the nonsinusoidal frame re- increments; each frequency was held for 3 – 5 s to ensure steady-
sponse 关Fig. 4共f兲兴. The model also reveals greater distortion in the state vibration. Hence, both experimental and model results in
drum inertia than the experimental data as the drum loses and Fig. 5 represent steady-state response. Figs. 5共a and b兲 illustrate
regains contact with the ground 关see circles in Fig. 4共d兲兴. the peak acceleration amplitudes 共normalized by ⍀2兲 that occur
Broadening to behavior over a greater frequency range, Fig. 5 when the drum is at the top and bottom of its vibration cycle. The
presents both experimental and 3DOF 共drum-frame-soil兲 model peak upward drum acceleration 共occurring at the bottom of the

688 / JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008

J. Eng. Mech., 2008, 134(8): 684-693


Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Frequency response of drum experimental and 3DOF model amplitude response and phase lag for low eccentric excitation 共left兲 and high
eccentric excitation 共right兲

drum’s cycle兲 was always greater than the downward drum accel- 关Fig. 5共b兲兴. Consistent with the findings of Anderegg 共1997兲,
eration 共occurring at the top of the drum’s cycle兲. During low Figs. 4 and 5 illustrate that the majority of nonlinearity observed
eccentric force excitation the drum and soil remained in contact. is due to drum/soil decoupling.
However, the drum decoupled from the soil for most frequencies One area where the drum-frame-soil vertical DOF model
during high eccentric force excitation. proved insufficient was in capturing what is believed to be
The observed phase lag between the eccentric mass and drum rocking-induced vertical drum vibration observed 共rotation about
positions is shown in Figs. 5共c and d兲. This phase lag reflects the the x-axis in Fig. 1兲 around 30 Hz in Figs. 5共a and b兲. The ex-
difference in time between the eccentric mass and drum when at perimental data presented in Fig. 5 were captured from sensors
their bottom positions. Note that phase lag is not constant within mounted to a nonrotating element on one end of the drum. Rock-
each cycle of vibration due to system nonlinearity, e.g., compare ing can magnify the vertical acceleration on one end of the drum
␾top with ␾bot in Fig. 4. It can be seen from Fig. 5 that ␾bot 关as shown in Fig. 5共a兲兴 and diminish it on the other end 关as shown
measured experimentally exhibits a gradual increase from a low in Fig. 5共b兲兴. Unfortunately, experimental data from both ends of
of 30° on the left side of resonance to a high of 180° on the right the drum are unavailable 共only one end has a nonrotating mount
side of resonance. ␾bot was approximately 100° during low ec- to which an accelerometer can be attached兲; therefore, drum rock-
centric force resonant excitation and approximately 115° during ing cannot be confirmed. However, frame acceleration data reveal
high eccentric force resonant excitation. frame rocking near 30 Hz for both low and high amplitude eccen-
The 3DOF model responses are superimposed on Fig. 5. tric force, as evidenced by the left and right side frame accelera-
Model parameters used for both low and high excitation force tion being out of phase 共see Fig. 6兲.
include constant ks = 58 MN/ m and cs = 210 kN s / m, determined To model drum and frame rocking during drum/soil full
via best fit analysis over the entire frequency spectrum. With a contact mode, rotational DOFs were added to the vertical drum
constant ks, the model is linear during drum/soil contact and thus and frame DOFs. Constant ks = 58 MN/ m and cs = 180 kN s / m
does not convey different top and bottom drum acceleration peaks were used in combination with mass moments of inertia around
during contact 关Figs. 5共a and b兲兴. The model does capture the the x axis for the drum and frame of Ixxd = 400 kg m2 and
frequency response behavior during contact mode with the ex- Ixxf = 600 kg m2, respectively. Because only contact mode rocking
ception of the dip in amplitude that occurs near 30 Hz 共discussed was explored, the soil and drum DOFs merge into one. The
in next paragraph兲. The constant ks model does accurately reflect results from the 4DOF model 共soil and drum DOFs merge into
the frequencies where loss of contact occurs and does match one for contact mode兲 are presented in Fig. 7. The right and
the top and bottom drum accelerations observed experimentally left end peak amplitudes from the model analysis indicate rocking

JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008 / 689

J. Eng. Mech., 2008, 134(8): 684-693


Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Experimental frame response data from around 30 Hz for low eccentric force 共left兲 and high eccentric force 共right兲 showing left and right
side of frame vibrating out of phase

resonance occurs around 29 Hz, and that an increase or decrease results of Fig. 7, the occurrence of drum rocking cannot be
can occur depending on which end the sensor is located 共due fully confirmed without experimental data from both ends of the
to coupled vertical translation plus rotation兲. Though probable, drum.
based on the experimental results in Figs. 5 and 6 and the model

Nonlinearity
As was illustrated in Figs. 4 and 5, the roller-soil system exhibits
nonlinear response during contact and loss of contact operation.
Simply accounting for drum-soil separation in the model captures
much of the nonlinearity. However, there remains unaccounted
for nonlinearity, e.g., as evidenced in contact mode nonlinearity.
Two additional sources of nonlinearity were modeled: 共1兲 curva-
ture of the drum; and 共2兲 stress dependency of soil stiffness.
The curved surface of the drum leads to a nonlinearly varying
contact force-surface deflection relationship. Lundberg 共1939兲
developed a solution for a force applied through an infinitely
long rigid cylinder resting on an elastic half space using Hertzian
contact theory. Kröber et al. 共2001兲 presented Lundberg’s solution
as a relationship between the resulting soil stiffness ks and the
elastic properties of the half space, i.e., Young’s modulus E
Fig. 7. Experimental and 4DOF model drum rocking during contact and Poisson’s ratio ␯ 关see Eq. 共10兲兴. As exhibited in Eq. 共10兲
mode vibration and shown in Fig. 8共a兲, the curvature of the drum has a mild

Fig. 8. Nonlinearity in model soil stiffness due to: 共a兲 curvature of drum; 共b兲 strain hardening or softening behavior of soil

690 / JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008

J. Eng. Mech., 2008, 134(8): 684-693


Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

Fig. 9. Matching drum vibration by modeling constant ks, hardening ks, and curved drum during: 共a兲 contact behavior; 共b兲 loss of contact behavior

hardening effect on the relationship between contact force and 25%. The addition of curved drum nonlinearity 关Eq. 共10兲 with
soil displacement E = 100 MN/ m2 and ␯ = 0.25兴 improves the model’s ability to
match bottom drum acceleration amplitudes only slightly. The
EL␲ use of curved drum and mild stress hardening soil behavior
ks = 共10兲

冢 冣
L 共k2 = 1.03 for contact, k2 = 1.06 for loss of contact兲 in the model


2共1 − ␯ 兲 1.8864 + ln
2
matches the bottom drum acceleration response observed ex-
16 R共1 − ␯2兲 FC perimentally. Each model still overpredicts top drum accelera-
␲ E L tion. Other sources of nonlinearity that are not addressed here
An additional source of nonlinearity stems from the stress depen- might include the layering of media 共Adam 1996; Anderegg and
dency of soil stiffness. It is well documented that soil stiffness Kaufmann 2004兲 and any plastic deformation that may occur at
increases with increasing confining stress and decreases with in- the drum/soil contact area.
creasing shear stress. Within each cycle of vibration, the shear
stress is increasing markedly due to the applied eccentric force, Transient and Steady-State Response
and the confining stress is increasing due to lateral confinement.
To evaluate both stress hardening and stress softening, a simple Thus far, the experimental data and model responses presented
power law stiffness kszkd2 共Andrei et al. 2004兲 was adopted 关see here have assumed steady-state behavior. Until recently, vibratory
Fig. 8共b兲兴. roller compactors have typically operated at constant ⍀ and m0e0
Fig. 9 presents 3DOF model results for contact 共low eccentric while traveling at constant velocity. Consequently, all modeling
amplitude兲 and loss of contact 共high eccentric amplitude兲 behav- efforts have assumed spatially homogeneous soil conditions and
ior using constant ks and nonlinear ks due to the curved drum and thus considered the roller vibration to be steady state. There are
soil stress dependency. The nonlinearity in the experimental data two important deviations from this where transient behavior
in Fig. 9 is characterized by 15–25% greater bottom drum accel- may be important. First, roller compactors are increasingly being
eration amplitudes. Using constant ks = 62 MN/ m 共k2 = 1.0兲 and designed to vary ⍀ and m0e0 using real-time feedback control.
cs = 180 kN s / m, the model underestimates bottom drum accelera- Second, recent studies with instrumented rollers reveal that sub-
tion in both cases by approximately 10–20% and overestimates surface heterogeneity 共e.g., near surface bedrock兲 can cause sig-
top drum acceleration during loss of contact by approximately nificant variability in measured soil stiffness 共Mooney et al.

Fig. 10. Roller measured soil stiffness where local stiffness variation can be as much as 50%

JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008 / 691

J. Eng. Mech., 2008, 134(8): 684-693


Conclusions

This paper investigated the ability of nonlinear lumped parameter


models to capture the often complex nonlinear behavior observed
during vibratory roller/soil interaction. Comprehensive experi-
mental results over a wide frequency range 共10– 45 Hz兲 were
presented to compare with model results. A 3DOF model with
constant ks and cs captured the trends in drum vibration ampli-
tude and drum phase lag for low and high eccentric excitation
over a wide frequency range 共10– 45 Hz兲 with the exception of
rocking-induced excitation observed experimentally near 29 Hz.
This 3DOF model also captured the salient roller response
Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

during contact and partial loss of contact behaviors observed


experimentally.
Drum/soil decoupling accounts for a considerable portion of
the nonlinearity observed; however, there remains noticeable
Fig. 11. Generation of transient response from steady state by step nonlinearity in contact mode and loss of contact mode vibration.
change in soil stiffness Introducing the curved drum effect into the 3DOF model was
successful in accounting for some of this nonlinearity, and in-
cluding stress dependent soil hardening accounted for the
2006兲. For example, Fig. 10 illustrates considerable variability in rest. Therefore, a 3DOF lumped parameter model with non-
ks measured along a 40-m testbed with fairly abrupt changes in ks. linear ks to account for drum curvature and stress dependent
To this end, the effects of transient response were investigated
hardening proved effective in modeling the behavior observed
by introducing step increases and decreases in ks into the numeri-
experimentally.
cal simulation. Fig. 11 illustrates the process and the terms used
Introducing rotational DOFs into the model was effective in
here to describe the results. Under steady-state vibration, a step
change in soil stiffness ⫾⌬ks was introduced. The resulting tran- capturing perceived drum rocking and measured frame rocking;
however, more substantial experimental data are required to in-
sient amplitude Z̈d共tr兲 was determined as was the settling time 共␶兲
vestigate and verify drum rocking.
required to restore steady-state vibration and the new steady-state
Transient effects were investigated numerically by introducing
acceleration amplitude Z̈d共ss兲 共see Fig. 11兲. Because transient be- step changes in ks and observing drum vibration amplitude
havior depends on when ⌬ks was introduced within a cycle, simu- changes and settling time. For up to 50% changes in ks, 3DOF
lations were performed for eight different introduction times model simulations revealed amplitude changes from 20 to 150%
within a cycle. Fig. 12 presents the findings of numerical simula-
of steady-state amplitude and settling times from one to seven
tions performed with the 2DOF model 共roller/soil contact兲 using
cycles. Therefore, for drum/soil interaction in heterogeneous sub-
ks = 30 MN/ m and cs = 210 kN s / m. Values of ⌬ks / ks were varied
surface conditions 共e.g., variable near surface bedrock兲, it is im-
⫾50% in correspondence with measurements 共Fig. 10兲. The
analysis showed that the transient amplitude varied from 20 to portant to incorporate transient response to interpret the resulting
150% of the steady-state amplitude while the settling time nor- soil stiffness values. The results presented herein indicate that
malized by the period of excitation T ranged from one to seven drum-frame-soil lumped parameter models with nonlinear soil
cycles. Considering an average roller velocity of 1.5 m / s and stiffness are capable of capturing complex drum/soil interaction.
30 Hz excitation, the length of travel required to settle to within These models can therefore be used to extract important soil pa-
2% of steady state is 0 – 0.3 m. These results suggest that transient rameters 共e.g., ks兲 from roller vibration data. It should be noted
response changes can be significant in the case of soil heteroge- that the experimental results matched here were from roller vibra-
neity, and should be considered in modeling and model-based tion on sand. Further research is required to evaluate model effi-
extraction of soil properties. cacy on other soils.

Fig. 12. Transient drum vibration amplitude 共a兲 and settling time 共b兲 caused by step change in ks

692 / JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008

J. Eng. Mech., 2008, 134(8): 684-693


Acknowledgments Transportation Research Board, Washington, D.C., 124–134.
Andrei, D., Wiczak, M. W., Schwartz, C. W., and Uzan, J. 共2004兲.
The writers wish to thank the NSF 共Grant No. CMS-0327509兲 for “Harmonized resilient modulus test method for unbound pavement
materials.” Transportation Research Record. 1874, Transportation Re-
their financial support and Ingersoll-Rand for the use of their
search Board, Washington, D.C., 29–37.
DD138 roller. They also thank Robert V. Rinehart and Patrick K. Baidya, D. K., Muralikrishna, G., and Pradhan, P. K. 共2006兲. “Investiga-
Miller for their assistance with data gathering and for their in- tion of foundation vibrations resting on a layered soil system.”
sights in discussing results. J. Geotech. Geoenviron. Eng., 132共1兲, 116–123.
Balakrishna Rao, H. A., and Nagaraj, C. N. 共1960兲. “A new method for
predicting the natural frequency of foundation-soil systems.” Struct.
Eng., 38共10兲, 310–316.
Notation Barkan, D. D. 共1962兲. Dynamics of bases and foundations, McGraw-Hill,
New York.
The following symbols are used in this paper: Brandl, H., Kopf, F., and Adam, D. 共2005兲. “Continuous compaction
cs ⫽ damping of soil 共Ns/m兲;
Downloaded from ascelibrary.org by Central South University on 09/16/22. Copyright ASCE. For personal use only; all rights reserved.

control with differently excited dynamic rollers.” Strassenforschung-


E ⫽ Young’s modulus of soil 共N / m2兲; sauftrag Nr. 3.176 des Bundesministeriums für Verkehr, Innovation
Fc ⫽ contact force 共N兲; und Technologie, Heft 553, Vienna, Austria.
g ⫽ acceleration due to gravity 共9.81 m / s2兲; Crockett, J. H. A., and Hammond, R. E. R. 共1949兲. “The dynamic prin-
ks ⫽ soil stiffness 共N/m兲; ciples of machine foundations and ground.” Proc. Inst. Mech. Eng.,
160共4兲, 512–523.
k2 ⫽ nonlinear soil stiffness coefficient 共unitless兲;
Gazetas, G. 共1983兲. “Analysis of machine foundations: State of the art.”
md ⫽ mass of drum 共kg兲; Int. J. Soil Dyn. Earthquake Eng., 2共1兲, 2–42.
m f ⫽ effective mass of frame 共kg兲; Kröber, W. 共1988兲. “Untersuchung der dynamischen vorgäge bei der vi-
ms ⫽ apparent mass of soil 共kg兲; brationsverdichtung von böden.” Dissertation, Lehrstuhl und Prüfamt
R ⫽ radius of drum 共m兲; fur Grundbau, Bodenmechanik und Felsmechanik der Technischen
t ⫽ time 共s兲; Univ. München, München, Germany.
x ⫽ position along test bed 共m兲; Kröber, W., Floss, R., and Wallrath, W. 共2001兲. “Dynamic soil stiffness as
Z̈d ⫽ vertical drum acceleration amplitude 共m / s2兲; quality criterion for soil compaction.” Geotechnics for roads, rail
tracks, and earth structures, Balkema, Lisse, The Netherlands.
Z̈d共ss兲 ⫽ vertical drum amplitude in steady state; Lundberg, G. 共1939兲. “Elastische beruehrung zweier halbraeume.”
Z̈d共tr兲 ⫽
vertical drum amplitude during transient response; Forsch. Geb. Ingenieurwes., 10共5兲, 201–211.
zd ⫽
vertical drum displacement 共m兲; Lysmer, J., and Richart, F. E. 共1966兲. “Dynamic response of footings to
żd ⫽
vertical drum velocity 共m/s兲; vertical loading.” J. Soil Mech. and Found. Div., 92共1兲, 65–91.
z̈d ⫽
vertical drum acceleration 共m / s2兲; Machet, J. M., and Sanejouand, R. 共1980兲. “Modeles mathematiques
zd-new ⫽
vertical drum displacement after impact 共m兲; dans le domaine du compactage par vibration.” Proc., Int. Conf. on
żd-new ⫽
vertical drum velocity after impact 共m/s兲; Compaction, Session VII Compaction Equipment, ENPC, LCPC,
Paris, 651–657.
z̈d-new ⫽
vertical drum acceleration after impact 共m / s2兲;
Mooney, M. A., and Rinehart, R. V. 共2007兲. “Field monitoring of roller
zf ⫽
vertical frame displacement 共m兲; vibration during compaction of subgrade soil.” J. Geotech. Geoenvi-
ż f ⫽
vertical frame velocity 共m/s兲; ron. Eng., 133共3兲, 257–265.
z̈ f ⫽
vertical frame acceleration 共m / s2兲; Mooney, M. A., Rinehart, R. V., and van Susante, P. 共2006兲. “The influ-
zs ⫽
vertical soil displacement 共m兲; ence of heterogeneity on vibratory roller compactor response.” Proc.,
żs ⫽
vertical soil velocity 共m/s兲; GeoCongress 2006 共CD-ROM兲, ASCE, Atlanta.
z̈s ⫽
vertical soil acceleration 共m / s2兲; Nazarian, S., Yuan, D., and Arellano, M. 共2002兲. “Quality management of
␯ ⫽
Poissons’ ratio 共unitless兲; base and subgrade materials with seismic methods.” Transportation
␾bot ⫽
phase lag determined at bottom drum location Research Record. 1786, Transportation Research Board, Washington,
共degrees兲; D.C., 1–15.
␾top ⫽ phase lag determined at top drum location 共degrees兲; Odemark, N. 共1949兲. “Investigations as to the elastic properties of soils
and design of pavements according to the theory of elasticity.” Statens
and
Väginstitut, Mitteilung No. 77, Stockholm, Sweden.
⍀ ⫽ rotational frequency from data 共rad/s兲. Pietzsch, D., and Poppy, W. 共1993兲. “Simulation of soil compaction with
vibratory rollers.” Anal. Proc., 29共6兲, 585–597.
Quibel, A. 共1980兲. “Le comportement vibratoire: Trait d’union entre le
References choix des parameters et l’efficacite des rouleaux vibrants.” Proc., Int.
Conf. on Compaction, Session VII, Compaction Equipment, ENPC,
Adam, D. 共1996兲. “Flächendeckende dynamische verdichtungskontrolle LCPC, Paris, 671–676.
共FDVK兲 mit vibrationswalzen.” Dissertation, Technische Univ. Wien, Richart, F. E., Woods, R. D., and Hall, J. R. 共1970兲. Vibrations of soils
Fakultät für Bauingenieurwesen, Wien, Austria. and foundations, Prentice-Hall, Englewood Cliffs, N.J.
Adam, D., and Kopf, F. 共2004兲. “Operational devices for compaction Rinehart, R. V., and Mooney, M. A. 共2008兲. “Instrumentation of a roller
optimization and quality control.” Proc., Int. Seminar on Geotechnic compactor to monitor vibration behavior during earthwork compac-
Pavement and Railway Design and Construction, Athens, Greece, tion.” Autom. Constr., 17共2兲, 144–150.
Millpress, Rotterdam, The Netherlands, 97–106. Wolf, J. P. 共1994兲. Foundation vibration analysis using simple physical
Anderegg, R. 共1997兲. “Nichtlineare schwingungen bei dynamischen models, Prentice-Hall, Englewood Cliffs, N.J.
bodenverdichtern.” Dissertation 12419, Eidgenössische Technische Yoo, T.-S., and Selig, E. T. 共1979兲. “Dynamics of vibratory-roller com-
Hochschule Zürich, Schweiz, Switzerland. paction.” J. Geotech. Engrg. Div., 105共10兲, 1211–1231.
Anderegg, R., and Kaufmann, K. 共2004兲. “Intelligent compaction with Yoo, T.-S., and Selig, E. T. 共1980兲. “New concepts for vibratory compac-
vibratory rollers: Feedback control systems in automatic compaction tion of soil.” Proc., Int. Conf. on Compaction, Vol. II, ENPC, LCPC,
and compaction control.” Transportation Research Record. 1868, Paris, 703–707.

JOURNAL OF ENGINEERING MECHANICS © ASCE / AUGUST 2008 / 693

J. Eng. Mech., 2008, 134(8): 684-693

You might also like