Professional Documents
Culture Documents
Lusas Implementation PDF
Lusas Implementation PDF
SYSTEMS
Terry M J Cakebread
Biography: Since 2004 Terry Cakebread has been the Vice President (North America) of
LUSAS and is in charge of American Operations involving training, sales and support to
clients. He has a BSc Honours Degree in Civil Engineering from Southampton University
and is a Chartered Civil Engineer and a Member of the Institution of Civil Engineers (MICE)
in the UK. He has delivered many lectures on Finite Element Analysis and the benefits that it
ABSTRACT
This paper describes and illustrates how advanced Finite Element Analysis (FEA) software
has been used to model joints, bearings and seismic systems on a variety of projects
worldwide. Reference will be made to the software program, LUSAS1, which has been
developed progressively over the last 30 years and become one of the leading FE software
The paper will illustrate the different ways that bearings can be modeled and cover:
The relevance of using different finite element joint models in differing situations
Different ways to model lift off behavior (smooth contact or elastic-plastic joints)
The use of more advanced joints for modeling lead rubber bearings or
friction/pendulum bearings
It will discuss why dampers and other seismic systems are employed and cover the
1
It will also show some more detailed bearing models including carrying out bearing
repairs in situ and ways of modeling detailed bearing models with full contact
behavior.
Keywords: Finite, Element, Analysis, Global, Modeling, Local, Modeling, Joints, Bearings,
Seismic, Systems
INTRODUCTION
Finite element analysis first really came into use in the 1960s where analysis was carried out
on mainframe computers. It has come a long way since then. On todays range of PCs and
laptops engineers want to be able to model and predict with a fair degree of accuracy the
response of structures that incorporate increasingly sophisticated joints and bearings to ensure
that they articulate properly under a range of static and dynamic loading. Traditionally
bearings and joints were not modeled well, meaning that often basic assumptions were made
about their behavior. As a result the finite element boundary conditions that were used were
often poorly defined leading to incorrect results. To give one example, consider the case of
lift-off from a bearing occurring for part of a structure that was ignored in the modeling. A
better solution would be to remove the bearing support if the reaction was negative. This
would then allow for lift off but this solution would mean that in a subsequent loadcase, if
the support took load, it would no longer be there. A better solution in this case would be to
use a full nonlinear joint which allows for lift off and re-contact as necessary.
A range of basic and advanced joint models are described in this paper. Modeling of the
response of a structure can be done by using global modeling with assigned joint properties.
Localized modeling of the joint or bearing itself can also be carried out either to derive joint
2
RESEARCH SIGNIFICANCE
Finite element analysis provides researchers and practicing engineers with the tools to
accurately model the behavior of test specimens and real structures in the field. Finite
element models can be calibrated or fine-tuned against either measured or experimental data
to enable more accurate predictions of response to be made for a wide range of anticipated or
unexpected situations.
In their simplest form joint elements can be used to connect two or more nodes in a finite
element model with springs having translational and rotational stiffness. They may have
initial gaps, contact properties, an associated mass and damping, and other nonlinear behavior.
Joint material models are used in conjunction with joint elements to define the material
properties for linear and nonlinear joint models. Linear joint models can be defined by a
spring stiffness that corresponds to each local freedom or by specifying a set of general
properties for spring stiffness, mass, coefficient of linear expansion and damping factor.
Nonlinear joint models2 typically provided in finite element software, and as shown in Fig. 1,
allow for elasto-plastic uniform tension and compression with isotropic hardening where
equal tension and compression yield conditions are assumed; elasto-plastic general joints
with isotropic hardening for unequal tension and compression yield conditions; smooth
contact with an initial gap and frictional contact with an initial gap. Both smooth contact and
frictional contact joints can be used for lift-off or hook contact by using appropriate
3
Seismic Isolators
These more complex joint models exist to control the damage impact of seismic activity on
structures. These joint types may be summarised as being used for seismic isolation, energy
dissipation, or to model an active control system. Various types of isolator are available, as
shown in Fig. 2, including High Damping Rubber Bearings (HDRB) the most commonly
used elastomeric bearings; Lead Rubber Bearings (LRB) with plastic yield and biaxial
hysteretic behavior as modeled using the Bouc-Wen3 model; and Sliding/Frictional Pendulum
Systems (FPS) with pressure and velocity dependent friction coefficient and biaxial hysteretic
behavior. The idealised behavior of an FPS bearing is shown but in reality this follows the
hysteric behavior of lead rubber bearings. Hysteresis is that highly nonlinear phenomenon
that occurs in systems that possess memory and, as a result, all isolator types shown are
Viscous Dampers
Visco-elastic dampers can be modeled using the four parameter solid model shown in Fig. 3
which comprises 3 springs and a dashpot. If only K1 exists then this becomes the Kelvin-
Voigt or Kelvin Model. If all springs are absent it then reverts to a simple dash-pot damping
model. If K1 does not exist and K2 and/or K3 exist it becomes a Maxwell model.
Halving/Hanger Joints
Halving/hanger joints, as used with drop-in concrete or steel spans, are a fairly simple joint to
model in the global sense. The global steel truss bridge model, as shown in Fig. 4,
incorporates a joint-supported drop-in span. In these cases, normally all that is required is that
the joint acts as a support when the load on the joint is downward, then when the load is lifted
4
it allows the bridge to articulate upwards. It can also be used to restrict horizontal movement
when it locks up but this is often not modeled. Normally only a check is carried out to ensure
that the movement does not exceed the travel allowed for the joint.
To model lift-off and frictional sliding nonlinear contact joint elements are used. The
used in a dock closure system in the UK (see Fig. 5) were assessed in order to guarantee its
safety under seismic loading. Additional joint elements were used to provide hydrodynamic
mass and damping actions on the walls and base-interface respectively. Thin shell elements
modeled the caisson cell walls and thick-shell elements modeled the base. Ground
acceleration history for a UK hard site provided the seismic input with increments of 0.005
second being used for each time step. Hydrostatic pressure and self-weight were applied as
initial static loads. Hydrodynamic forces from the water enclosed in the cells were simulated
by locating joint elements at each node on each wall and assigning directional masses
calculated using the Westergaard4 model. Acceleration histories were applied to the
foundation to drive the ensuing dynamic analysis. Values of frictional damping at the contact
interface of 3%; structural damping of 5%; and interface damping of 2% to simulate the
effect of the fluid between the base and the dock floor were used in the analyses. The analysis
clearly showed the caisson had adequate structural capacity to withstand a seismic event and
Viscous Damping
Nonlinear joint elements modelled the elastomeric bearings and seismic dampers of a 1108m
long, multi-span bridge structure in the Mediterranean region, as shown in Fig. 6, and
5
enabled design forces to be expected in the case of an earthquake to be assessed to Eurocode
EC8. This prestressed reinforced concrete road bridge comprised both straight and curved
sections with an expansion joint midway along its length. In LUSAS, engineering thick beam
elements defined at the respective centroid of each structural component modelled the
reinforced concrete deck. Connection between deck and elastomeric bearings and between
the top of the piers and elastomeric bearings was made using nominally stiff members of
negligible mass. These represented rigid links between the centroids of components and were
defined with negligible mass so as not to contribute to the dynamic behavior of the bridge.
Two longitudinal dampers were located at the 1st abutment and transverse dampers, located
at every 3rd pier along the bridge, required an additional stiff member arrangement.
Eigenvalue analyses on both bridge structures found that 225 structural modes were required
to meet the 95% mass participation factor value prior to carrying out a subsequent spectral
response analyses using EC8 design spectra. Three nonlinear transient dynamic analyses were
performed on each bridge using combinations of acceleration time-history dataset pairs in the
longitudinal and transverse directions, as used by the bridge designers. Fig. 7 shows a typical
transverse force time history plot produced. Good correlation of results was achieved for both
the spectral response and transient dynamics analyses, verifying the modeling techniques
used by the original designers and the viscous damping capabilities of LUSAS.
Half-joints, initially introduced into concrete bridge decks as a means of simplifying design
deterioration from chloride attack in the event of deck expansion joint failure, and also cause
concern because they are not easily accessible for inspection or maintenance. In addition, on
6
older structures, the half-joints as designed may not be code-compliant with todays standards
and may require assessment for increased modern vehicle loadings. The Kingston Bridge in
Glasgow, UK, is one such bridge with half-joints that attracted investigation.5 The bridge
carries an average of around 180,000 vehicles per day, and is one of the busiest in Europe.
The post-tensioned, table-top spans and reinforced concrete box girder suspended spans of
the approach ramps include numerous half-joints designed in accordance with late 1960s
standards. These are shown schematically in Fig. 8. Dimensions of half-joint nibs vary but
are generally in the order of 24 (600mm) deep x 18 (450mm) wide. An assessment showed
that some of the half-joints were not compliant with modern codes and so, in light of a
potential inadequacy, a destructive load test was undertaken on a typical half-joint on a ramp
that was being demolished and replaced as part of other work taking place on the structure.
The data obtained demonstrated significant capacity for the half-joint above that predicted by
the assessment codes. The load test results where then used to calibrate a LUSAS nonlinear
finite element model of the tested half-joint using a multi-crack concrete model. Once proved,
various derivative models were used to reassess all half-joints in the Kingston Bridge
Complex, showing actual capacities were significantly higher than those calculated from the
When joint properties cannot be easily defined localized joints models may be used to derive
suitable modeling values for use in a global seismic analysis, as carried out for the concrete
encased riveted steel frame structure of the Cathedral Building in San Francisco6. The
particular beam /column connections of the structure result in it being classified by FEMA7 as
a Partially Restrained moment frame. FEMA-356 does not provide explicit guidance on
appropriate moment-rotation properties for the minor axis joints in the transverse direction,
7
where the beams frame into the webs of the columns. With this arrangement, a moment at the
end of the beam will impose a twist on the column web before it is transferred to the column
flanges. This adds flexibility to the joint and also provides an additional inelastic mechanism
(web yield) which can affect the overall response of the joint. In order to determine the nature
of this effect, a series of finite element models of typical minor axis joints were developed in
LUSAS. Triangular thick shell elements with quadratic formulation were used to construct
these models and nonlinearity was incorporated by way of Von-Mises yield criteria. The
effect of the concrete encasement in preventing inward movement of the column flanges was
incorporated in the model by way of compression-only strut elements. These models were
used to determine nonlinear moment-rotation curves representing the isolated effect of web
flexibility. The moment-rotation curves were then incorporated in the definition of the overall
moment-rotation of the minor axis joints determined using the FEMA procedure.
Research work carried out by Jon Lindsey of HNTB at the University of Kansas into the
potential use of alternative moment-resisting connections8 shows very clearly the benefit in
using FE analysis alongside experimental testing. The projects aim was to make design
connection were analyzed that typically included wide-flange steel shape sizes intended for
use in multi-storey structures in moderate to high seismic zones. Both exterior (one-sided)
and interior (two-sided, cruciform) connections, underwent assessment. Loading was applied
applying a predefined increasing displacement to the end of the beam member. These
displacement parameters were based on prescribed inter-storey drift increments. The single-
8
cycle loadcurve was proportioned such that the model would experience significant yielding
at the peak displacements of the cycle. By using nonlinear solid modeling with slideline
surface contact (to be explained later in this paper) excellent correlation was achieved
between the results predicted by LUSAS and measurements obtained from detailed
experimental testing. Stress time history plots, of the type as seen in Fig. 9, showed the
formation of stress concentration zones and areas of yield. Load versus total displacement
(Fig. 10) and time history data was graphed to record the correlation of the experimental
results.
Even back in 1995 bearings were being analysed using finite element analysis. Then, UK
Consultant Hyder had to carry out collapse analysis of fabricated steel trestle bridge
bearings, as used on the M5 road bridge at Avonmouth, and predict their ultimate strength
both with and without strengthening modifications. Initial FE models assessed the
performance of both shell and solid element idealisations. Final all-solid models similar in
nature to Fig. 11 included geometric, material and contact nonlinear effects. With
experimental data (load-strain measurements) being supplied very close agreement between
measured and calculated values of ultimate load could be seen. The analysis also clearly
showed that the failure mode was plastic collapse with elastic buckling occurring at a much
higher load. Results were used by Hyder to help determine which bearings would require
Two other ways to model the interaction of structural parts or components concern the use of
9
freedom in a particular way, and slidelines - also known as slidesurfaces - which model the
Constraint Equations
Constraint equations allow linear relationships between nodal freedoms to be set up.
Constraint equations can be used to allow plane surfaces to remain plane while they may
translate and/or rotate in space. Similarly straight lines can be constrained to remain straight,
and different parts of a model can be connected so as to behave as if connected by rigid links.
These geometric constraints are only valid for small displacements. Principal constraint types
Slidelines/Slidesurfaces
Slidelines/slidesurfaces can be used to tie dissimilar finite element meshes together and to
model contact and impact problems in both 2D and 3D. They can be used as an alternative to
joint elements or constraint equations and have advantages when there is no prior knowledge
of the contact point. The properties of a slideline such as the contact stiffness, friction
coefficient, temperature dependency etc are used to model the contact interaction between
master and slave features. Fig. 12 shows an example that includes both tied slidelines (to join
the dissimilar meshes) and frictional slidelines to model the contact between the components.
The former avoids the need for stepped mesh refinements between different mesh densities.
Fig. 13 shows a simplified contact application for a floating pontoon restrained by cables to
two anchor blocks sitting on the sea bed. For this, only a frictional slideline is required.
10
COMPARISON OF PREDICTED AND EXPERIMENTAL RESULTS
When measured or experimental data is available, as was the case in the beam/column
connection and bridge bearing examples described previously, results predicted by finite
element analysis can be readily correlated. Once verified, fine-tuning of a model can be done
or more advanced what-if modeling can take place safe in the knowledge that the base
model is accurate. However, comparing predictions made using the range of analytical joint
models developed for use by finite element analysis software against measured data is often
more difficult, and sometimes this is because the manufacturers of various bearings and
dampers are reluctant to put their detailed behavior into the public domain. If manufacturers
were to make public relevant data relating to the behavior of their bearings it would
particularly help engineers in carrying out structural assessments of existing structures where
CONCLUSIONS
With the advanced finite element analysis tools available today it is possible to model all
different types of joint and bearing conditions. These can be used in conjunction with line
beam models when global modeling is carried out or they can be modeled in detail using
plane stress or solid localized models. Often, because of the very nature of the problems to be
solved it will require nonlinear analysis. Verification of finite element modeling results for a
structure against measured or test data is useful when additional modeling is to be carried out.
Bearing and damper manufacturers need to give more information to designers to enable
them to model the structural systems that use their devices easier.
REFERENCES
11
2. Advanced Nonlinear Joints, Software Development Report SDR/LUSAS/381, LUSAS,
3. Wen, Y.K. Equivalent linearization for hysteretic systems under random excitations.
5. Boothman, D., Leckie, S., MacGregor, I., Brodie, A., Assessment of Concrete Half-Joints
using Non-linear Analysis, Proceedings of the ICE - Bridge Engineering, UK, Volume 161,
6. Chambers J.D., Kelly T.E., Dreyer R.C., Fully Integrated Performance Based Design and
Retrofit of Existing Buildings: A Case Study, 2007 NZSEE Conference, New Zealand.
7. FEMA, 2000, Prestandard and Commentary for the Seismic Rehabilitation of Buildings,
FIGURES
List of Figures:
Fig. 7 Typical Longitudinal Force / Time History Plot for a Selected Pier
12
Fig. 10 Load / Displacement Curve for First Yield Cycle of Beam / Column Connection
13
+ : tension F F
- : compression
+ strain hardening
stif f ness
Lif t-of f
= 2 - 1 f orce
-
strain hardening K - contact spring
stif f ness stif f ness
Fy or Fz
Fx
Fo Fo
xx = x2 - x1 xz = z2 - z1 Fx
K - contact
spring stif f ness
-Fo
Frictional Contact
14
Force
b
Rubber cd
Steel
Plate Displacement
Xg Bouc-Wen
Elastomeric Bearing
Force - post-yield
stiffness
Fyld
b
Lead Core
Rubber
Steel
Displacement
Plate
cd
K - elastic Xg
Lead Rubber Bearing spring stiffness
Force
Bearing Material
Spherical Sliding N b
Surface
Xb
Displacement
15
K3 Cd
dd
F K1 F
K2
d1 d-d1
16
Fig. 6 Global Model of Viscous Damped Road Bridge
Fig. 7 Typical Transverse Force / Time History Plot for Selected Pier
17
Table-top span Table-top span Table-top span
A
Bearing strip Upper nib of half-joint
Lower nib
A-A
Lower nib of half-joint Diaphragm
A (omitting upper nib)
18
Fig. 10 Load / Displacement Curve for First Yield Cycle of Beam / Column Connection
19
Tied slidelines Friction slidelines
Friction slidelines
20