Post on 12-Mar-2018
Dynamic Analysis of Civil
Engineering Structures
Jan Walczak
ADINA R&D, Inc., USA
www.adina.com Jan.Walczak@adina.com
Krynica 2012, Copyright ADINA R&D Inc. 2012
Content of Presentation
•Philosophy in the development of the ADINA System
•Bathe method for implicit time integration
•Analysis of simple problems – stability and accuracy
of the Bathe method
•Analysis of civil engineering structures
•Concluding remarks
Structures
CFD
Electromagnetics
Multi-Physics (TMC, FSI, TFSI, …)
The ADINA System
The ADINA System offers analysis capabilities in
in ONE system.
This makes the ADINA System unique in the market.
Note: ADINA is also the nonlinear solver for
NX Nastran (SOL 601)
The ADINA System
RELIABILTY is most important !
EFFICIENCY is also important !
We need to achieve both !
The ADINA System
for Structures, CFD and Multi-Physics
The philosophy of the ADINA development
• What we believe to be important
The need for use of reliable finite element
methods
Using the state-of-the-art techniques
2d and 3d solids, shells,...
Plasticity, Contact,….
CFD, …
Multi-Physics,…
• Strong theoretical foundation is essential in the
development of reliable finite element program
• ADINA – from mechanical to biomedical applications
ADINA Development
Strong robot picking up weight
ADINA in multibody dynamics
Weak robot picking up weight
Magnum Shock Absorber Analyzed using
ADINA
Courtesy of
Gabriel, India
FSI Analysis of Shock Absorber
Model created from Nastran input
Thermo-mechanical coupling, multi-layer shell
Thermo-fluid-structure interaction (TFSI)
TFSI – fluid velocity in a manifold
ADINA in biomedical applications -
Analysis of Carpal Tunnel Syndrome
Model used
Fluid response
Solid response
Fluid pressure and solid stress
ADINA in Civil Engineering
Applications
• Our involvement with civil engineering started
after the 1989 San Francisco earthquake
• All major codes were tested for reliability
• The ADINA program has been chosen by
CALTRANS as the main nonlinear analysis tool for
all California tall bridges
Courtesy of Caltrans, Division of Structures
Analysis of the Bay Bridge
San Francisco
Oakland Bay
Bridge
West
Span
East
Span
San
Francisco
Oakland
West Span
East Span
Courtesy of Caltrans, Division of Structures
Analysis of San Francisco Oakland Bay Bridge
using ADINA
Total length – 23,556 ft (7,180m)
Height – 190 ft (58 m)
Construction started 1933, finished 1936
Analysis of San Francisco Oakland Bay Bridge
East Span
San Francisco Oakland Bay Bridge
Damage caused by 1989 earthquake
• Safety issues – tremendous responsibility
• Size of civil structures – modeling issues
• Lack of experimental data, reliability at most
important:
- Reliable finite elements
- Reliable and robust solution techniques
- Efficient and stable implicit time integration
method for long durations (Bathe method)
Civil engineering specific
requirements
Bathe time integration method
Evaluating velocities in terms of
displacements and accelerations we have:
+t t t t t t t t M U C U R F
1 2 3
t t t t t t tc c c U U U U
1 2 3
t t t t t t tc c c U U U U
Where are displacement and velocity
solutions at time and:
,t t t t U U
t t
1
(1 )c
t
2
1
(1 )c
t
3
(2 )
(1 )c
t
Using the above expressions, the equilibrium equation (1)
can be written at time t+dt in the following form:
( 1) ( )
3 3 3
( 1)
1 2
( 1)
3 1 3 2 3 3
( 1)
1 2 3
( _ )
(
)
( )
t t i i
t t t t i t t t
t t t t t i
t t t t t i
c c c
c c
c c c c c c
c c c
K M C ΔU
R F M U U
U U U
C U U U
Properties:
-- no parameter to adjust, simply the time
step has to be sufficiently small for accuracy
-- solves in nonlinear analysis when the TR fails
-- shows excellent accuracy/ dissipation
Effective in structural dynamics and
in wave propagations
A simple pendulum – implicit solutions
• Newmark method is unstable for large
deformation analysis over a long duration
Newmark Method
Pendulum under gravity load
Bathe Method
Pendulum under gravity load
Model problem: three degrees of
freedom spring system
7
1 2
1 2
10 ; 1
1; 1
k k
m m
Acceleration at node 2
Close-up of acceleration at node 2;
trapezoidal rule results of order 800
1D bar impact problem
1u
1E u t0u
0u
10L
0
Trapezoidal
rule
100 elements
Bathe
method
1u
1E u t0u
0u
10L
0 1D bar impact problem
100 elements
1D bar impact problem
Acceleration
Bathe method vs. trapezoidal method (TR)
• We have to distinguish between stability and
accuracy – loss of stability means huge error, totally
bad results
• The TR method is stable in linear analysis but not in
nonlinear
• The Bathe method is stable in linear and nonlinear
analysis
• The TR method has no amplitude decay
•The Bathe method has a small amplitude decay
(which can be reduced by selecting a reasonable
small time step)
•For FE solutions, the Bathe method is much better
than TR and Newmark methods:
- always stable
- larger steps can be used
- better convergence in Newton iterations
- good accelerations, that means good reactions
Reservoir cross section, 471 ft long and 21.8 ft high Courtesy of Alexander Kozak, SC Solutions, USA
Bathe method vs Newmark method –
a sensitivity study
Wall pressure envelopes – horizontal and vertical ground motions,
potential-based fluid elements with Newmark method compared
With Navier-Stokes solutions
blue line – potential based fluid elements with Newmark method
red line – potential-based fluid elements with Bathe method
green line – CFD (Navier-Stokes fluids) solution
Reservoir – refined mesh for the Bathe method
Bathe method solution compared with Navier-Stokes solution
Solution times:
Bathe method -5.26 sec, Navier-Stokes – 158.36 sec
Schematic of the problem
Bathe method, solution accuracy of a pipe
breaking system - a simple problem
Courtesy of Onsala Ingenyorsbyra, Sweden
Large scale real application
Checking Solution Accuracy
• test problem with the same features as
a nuclear container
• elastic shell fully clamped at its base, and
a fluid surrounding it
• MITC4 shell elements and potential-based fluid
elements were used
• the model is subjected to a sudden fluid flux
representing a pipe break
• solutions for Newmark and Bathe methods
are compared
Solution using standard Newmark method, spurious high
frequency oscillations, non-smooth contact (on and off),
parasitic pressure distribution
To overcome these problems, different
techniques can be used:
• Adding physical damping, e.g. Rayleigh damping,
to the structure only (difficult to predict how much
damping need to be added)
• Adding numerical damping to the Newmark method
(that reduces oscillations but also reduces response)
• Using Bathe time integration method
(no parameters needed to be adjust, the method is
second order and effectively damps out higher
frequencies )
Newmark method with Rayleigh damping, C=0.001K
Newmark method with numerical damping,
a0.3025 d0.6
Bathe method, no physical damping
Bathe method Newmark method,
without damping
Retrofit Seismic Analysis of
the AURORA bridge
Courtesy of Tim Ingham, T.Y. Lin International, USA
Workshop on Seismic Assessment and Retrofit Techniques for Freeway Bridges 57
Aurora Avenue Bridge
• official name – George Washington Memorial
Bridge, Seattle, WA
• arch-truss bridge, constructed in 1929-1932
• 2,945 ft (898 m) long, 70 ft (21 m) wide,
167 ft (51 m) high (above water level)
• following the collapse of Minneapolis I-35 arch-truss
bridge in 2007 inspections have been ordered
• after inspections, the bridge has
been determined functionally obsolete with marginal
sufficiency rating of 55.2 %
• The bridge consists of 2 V-shape cantilevers,
each 325 ft (99 m), balanced on large concrete
pilings on opposite site of the ship canal
• A 150 ft (46 m) long Warren truss suspended span
connects the two cantilevers in the middle
•Starting June 2011, the bridge has been undergoing
seismic retrofitting
• The bridge’s height and pedestrian access make it a
popular location for suicide jumpers
Bridge layout
June 25, 2008 Analysis and Vulnerability Study
Layout
North Approach
June 25, 2008 Analysis and Vulnerability Study
Layout
South Approach
Geotechnical Investigations
•Field Investigations
–New borings at N-15, N-1, S-1, and S-5
–Sampling for geotech index properties
–Crosshole seismic testing
•Develop new ground motions
–Based on a 475-year Return Period
–Acceleration time histories at each pier
•Foundation Springs
Evaluation Criteria - Standards
•AASHTO (American Association of State Highway and
Transportation Office) Seismic Guide Spec
•AASHTO LRFD (Load and Resistance Factor Design) 4th Ed
–Strength capacity calculations (except shear)
•FHWA (Federal Highway Administration) Seismic Retrofitting
Manual
–Shear provisions
•Priestley ( M.J.N. Priestley and F. Seible, Seismic Design and
Retrofit for Bridges, www.amazon.com)
–Detail checks and methods, consistent with Seismic Guide Spec and
Seismic Retrofitting Manual
Modeling
•New north and south approach models
•Inelastic M-C for columns, crossbeams/floorbeams
•Foundation stiffness and input based on Golder’s work
(Golder Associates, www.golder.com)
•Modal analysis
•Pushover (nonlinear) analysis
•Nonlinear dynamic, direct time integration analysis
Modal analysis using ADINA
• static dead load calculations of a full model with material
nonlinear models (moment-curvature relations for beams,
elasto-plastic material for shells,…)
• contact conditions are included in static and
frequency/modal solutions
• restart to frequency and mode superposition solutions
• frequencies calculated using Bathe’s Subspace iteration
method
Moment-curvature relations for
beam elements
• modeling of nonlinear elastic and elasto-plastic
beams with arbitrary cross sections
• small or large displacements
• bending moments vs. curvatures and torsional
moments vs. twists are functions of the axial forces
Moment-curvature relations - input into ADINA
0 1 103
2 103
3 103
4 103
0
5 103
1 104
1.5 104
Column
Splice
Curvature, /ft
Mom
ent, f
t-kip
0.000364
Moment-curvature relation used in the Aurora
N8 – splice at column base (longitudinal demands)
0 1 103
2 103
3 103
4 103
0
5 103
1 104
1.5 104
Column
Splice
Curvature, /ft
Mom
ent, f
t-kip
0.000364
Moment curvature relation used in the Aurora
S8 – splice at column base (longitudinal demands)
Fundamental Transverse Mode, South Frame
Modal analysis using ADINA
Longitudinal Mode – North Approach, Tallest Frame
Fundamental Transverse Mode – North Approach
Fundamental Longitudinal Mode, South Approach
Pushover analysis using ADINA
•South Frame S7-S8
–Longitudinal
–Transverse
•North Frame N6-N9
–Longitudinal
–Transverse
Workshop on Seismic Assessment and Retrofit Techniques for Freeway Bridges 76
Overturning of Structure
ADINA time history model
Prescribed displacements at the North
main pillar
prescribed displacements at the South
main pillar
Modeling of friction pendulum bearings
Friction pendulum bearing, FE model
Workshop on Seismic Assessment and Retrofit Techniques for Freeway Bridges 82
Friction Pendulum Bearings
Workshop on Seismic Assessment and Retrofit Techniques for Freeway Bridges 83
Friction Pendulum Bearings
• Mechanically simple
• Mathematically complex
(sgn )N
F D N DR
F – lateral force,
N – vertical force acting on the bearing
(in practice it is the dead load supported by the
bearing),
R – radius of bearing surface curvature,
D – lateral displacement,
D(dot) - is a relative velocity between top an bottom
parts of the bearing
• implemented to ADINA as a user-supplied friction
Friction pendulum mathematical model
Workshop on Seismic Assessment and Retrofit Techniques for Freeway Bridges 85
Contact Surface Model
Dish - Contact Surface
Slider - Contact Surface
Slider - Contact Point
Solid
Element
Rigid Link,
typ.
Normal reactions
Lateral reactions
Longitudinal reactions
Top of the bridge, longitudinal displacements
Top of the bridge, vertical displacements
Top of the bridge, lateral displacements
June 25, 2008 Analysis and Vulnerability Study
Proposed Retrofit Scheme –
North Approach
• Retrofit columns N9-N15 using FRP
(fiber reinforced plastic)
• Split column modification N6, N9, N11, N14; and
wrap with FRP
• Longitudinal girder strengthening
• Confinement of N6-N9 crossbeams
FRP Retrofit Testing
June 25, 2008 Analysis and Vulnerability Study
June 25, 2008 Analysis and Vulnerability Study
Proposed Retrofit Scheme –
South Approach
• Strengthen deficient elements in 75’ truss spans
• Strengthen bracing in steel bent S4 / S5
• Retrofit S6 bearing / Retrofit S6 backwall for shear
with concrete or FRP
• Steel casings at split columns S7, S8 (eliminate
split) and S9
Amarube Railway Bridge (Japan)
• old steel bridge replaced in 2010 by a pre-stressed
concrete bridge
• 93-meter concrete girder moved with slow sliding to be
adjusted to the tunel entrance
• ADINA has been used for the analysis
courtesy of Kozo Keikaku, Japan
FSI analysis in nuclear power plant
assessments
• Forsmarks Kraftsgrupp operates three nuclear power plants in Sweden, all boiling water reactors.
Courtesy of A. Thorsson, B. Olsson, J, Sundqvist – Forsmark Kraftsgrup
BWR nuclear power plant schematic
Main steam line: 70 atm, 286 oC
1100 MW(e)
Safety issues • Forsmark is responsible for the safe operation of
the plants.
• The plants must operate safely under normal operating conditions.
• The plants must be shut down safely when an emergency occurs:
– Turbine trip
– Earthquake
– LOCA (loss of coolant accident)
– Blowdown
• The plants are being upgraded during the next several years. The plants must operate safely,
and be shut down safely, after the upgrades.
Numerical modeling at Forsmark
• Forsmark creates numerical models of the
reactor pressure vessels, piping systems,
containment buildings, etc.
• These models are routinely used to analyze the
plants.
– Analysis corresponding to normal operation,
and to anticipated emergencies, to show
authorities that the plants are safe.
– Analysis after emergency, to determine if the
plant can be safely restarted.
• These models are continually maintained and
upgraded, as the plants are upgraded.
Need for FSI analysis • Many of the plant components contain water and steam,
which must be considered in the analyses.
• Decoupled fluid analysis: approximately include fluid
effects in structural analysis, e.g., include added mass of
water as extra density in structural analysis.
• Decoupled fluid analysis gives in most cases an
overconservative design, but also sometimes a
nonconservative assessment.
• FSI-based models are more accurate than decoupled
fluid models, because FSI effects such as reduced speed
of sound, wave propagation in fluid, etc. are directly
included.
FSI vs without FSI
Measured
FSI
w/o FSI
from
Andersson
et. al.,
“Numerical
simulation
of the HDR
blowdown
experiment
V31.1
at
Karlsruhe”,
2002
Forsmark 3 reactor pressure
vessel model
Forsmark 3 reactor pressure
vessel model
• Model built in the AUI
(ADINA User
Interface)
• About 65000 nodes
• Shells, beams, fluid
elements
Forsmark 3 reactor pressure
vessel model
Cutaway of reactor pressure
vessel model Steam dryer
Core cover
Core grid
Steam separators
Core shroud, fuel
Core stand
Control rods
Core stand
Core shroud
Core grid
Core cover
Steam dryer
Steam separators
Modeling of water
FSI solution procedure • Potential-based fluid
elements
– Unknown is the velocity potential (one DOF per fluid node)
– Inviscid irrotational fluid with constant density and bulk modulus
– Fluid velocities can be subsonic (nonlinear element) or small (linear element)
– Structural boundary motions are coupled to fluid.
– Small structural boundary motions.
FSI solution procedure • System matrices are symmetric and sparse.
• Frequency analysis, harmonic/random/response spectrum analysis are possible.
• Results from potential-based elements are comparable to results from Navier-Stokes based CFD elements (when the same modeling assumptions are used).
• Primary reason to use potential-based elements: speed with reasonable accuracy.
TUUU UUUU FU
FF FFFU
RM 0 K 0u u uC C
R0 M 0 KC 0
Division of water into element
groups • The water is divided into
7 element groups.
• Each element group has different physical properties (bulk modulus and density).
• Physical properties based on analysis of steam/water mixture under normal operating temperatures and pressures using RELAP5.
Modeling of water - detail
• The water is separated by structural elements.
• The adjacent water groups are connected with fluid-fluid interfaces (need continuity of pressure between adjacent groups).
• Pressure is applied on the free surfaces, corresponding to the pressure in the steam.
Loading - Pressure from steam
Initial conditions • Solved for in one static
load step.
• Loads include gravity, pressures from steam, pressure from recirculation pump
• Effect of fluid flow during normal operation is neglected; water is modeled “at rest”.
Pipe break analysis
Low pressure coolant
injection inlet
Prescribed mass flux, calculated
from RELAP5 analysis:
0
200
400
600
800
1000
1200
1400
1600
9.999 10.049 10.099 10.149 10.199
Time (sec)
Ma
ss
flu
x (
kg
/s)
Assume pipe break opening
time approx. 16 ms
Forces applied to ends of pipe
Solution procedure
• Implicit dynamic analysis, Bathe method, time step size 10-5 sec.
– Step size chosen to accurately integrate the highest frequency of interest in the structure.
• 15000 time steps.
• Linear analysis. Linear structural elements, linear
potential-based fluid elements (neglect
effect).
21v
2
Justification for using linear potential-based elements
• In the linear potential-based elements, we compute the pressure using
and neglect the nonlinear term .
• Analysis show that is significant only close to the
pipe break, when the outflow velocity is developed.
• Linear potential-based elements can be used only because
the mass flux is prescribed at the pipe break. If the pressure
had been prescribed instead, then the term cannot
be neglected.
21v
2
p
21v
2
21v
2
Results
Results – detail near pipe break
Use of model results • Results are used
– to verify that the stresses/forces are less than allowable values
– to generate loads for more detailed analyses of reactor internals, e.g. new core shroud lid to be installed during uprate.
– to generate floor response spectra (acceleration response spectra), used as loadings for more detailed analyses.
Example - dynamic membrane stresses
in core shroud
Pipe break analysis - summary
• The pipe break analysis is one of many similar
analyses routinely performed at Forsmark.
• This model is also used for
– earthquake analysis
– blowdown analysis, for different configurations of safety
relief valves
• This model can be locally refined as necessary to
perform a more detailed analysis.
Validation of FSI solution
procedure • HDR blowdown experiment V31.1
– Experiments performed in late 1970s to provide data for
verification of numerical codes used in reactor pressure
vessel analysis.
Validation of FSI solution
procedure • Typical results for strains on outside of core barrel:
Dynamic analysis of a nuclear steel
container
Courtesy of GRS mbH, Germany
A complex FE model, including hatches, pipe penetrations,
and variation of wall thicknesses. 3000 steps of 0.00005s
Effective stress distribution at time 24ms
(view from outside)
Nuclear concrete dome subjected to an
impact loading – direct implicit solution
cracks on the external and internal surfaces
of the dome
Damaged areas of the concrete dome
vertical and horizontal reactions of the dome
Conclusions
and a look into the future
• Very powerful capabilities are now available to perform finite element analyses of civil engineering structures
• Any “new” development should be measured against the already existing techniques
• Further significant challenges are still before us and major advances must still be expected