Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Simulation and Control Aspects of FHT
M. V. Sivaselvan
CO-PI CU-NEES
Assistant Professor
Dept. of Civil, Environmental and Architectural Eng.
University of Colorado at Boulder
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Multi-story Building
Earthquake
Most damage happens here – need better understanding by experimentation
Rest of the structure is undamaged – does not have to be physically build in the laboratory
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Wall Specimen
Actuators
Physical Experiment
Computer Model`
Interact during the experiment to mimic testing the whole building
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Use of hybrid simulationLaboratory Testing
For QualificationFor Discovery
Develop or calibrate Material/system models
• Examine the performance of a
component in its host
environment
• Proof of concept tests
• Interaction with surroundings
may significantly modify input
• Hybrid simulation is useful
• Hybrid simulation not very
useful for this purpose
• Some kind of computation-in-
the-loop with geometric
reasoning about state-space
may be possible
Hybrid Simulation is useful for qualification/proof-of-concept testing
when the interaction of a component with its surroundings
needs to be accurately represented
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Feedback interaction in reality
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
In hybrid simulation however …
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
In hybrid simulation however …
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Sensor
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
In hybrid simulation however …
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Sensor Natural Physical
Feedback
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
In hybrid simulation however …
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Sensor Natural Physical
Feedback
Actuator
Feedback
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
In hybrid simulation however …
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Sensor Natural Physical
Feedback
Actuator
Feedback
NEW DYNAMICS
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Challenges• These additional dynamics create significant problems
• When the structure to be simulated is lightly damped, almost always
renders the system unstable
• Need to develop control algorithms to make hybrid simulation possible
• Causality → Design of such algorithms requires knowledge about physical
substructure (predictive model, implicit integration etc.) → This is a conflict
→ Robustness of algorithm with respect to modeling of the physical
substructure
• A numerical algorithm need not be causal, a hybrid simulation algorithm
does
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Hybrid Simulation
Pseudo-dynamic Dynamic
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Has no inertia effects of interest
• Born from the displacement-based finite element –
one of the elements is now physical !
• Algorithms also reflect this
• If in addition, there are no frequency-dependent
behavior is the physical substructure – can be done
as slowly as we want to
Substructure 1
Computational
Substructure 2
Physical
External Input
(Eg. Ground Motion) Boundary Condition
Work Conjugate
Boundary Condition
Has significant inertia effects
• More practical applications necessitate
this form of hybrid simulation
• My research is in this area
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Hybrid Simulation
Pseudo-dynamic Dynamic
Real-time Slow
CU NEES Site
Hybrid simulation with Shaking Tables
CU NEES Site
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Recall
Substructure 1
Computational
Substructure 2
Physical
External Input
Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
SensorNatural Physical
Feedback
Actuator
Feedback
Motivation: Want actuator to behave the same way as Substructure 1
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Introduce a controller
Substructure 1
Computational
Substructure 2
Physical
External Input
Boundary Condition
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Natural Physical
Feedback
Actuator
Feedback
Controller
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Introduce a controller
Substructure 1
Computational
Substructure 2
Physical
External Input
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Natural Physical
Feedback
Actuator
Feedback
ControllerTakes the same input as Substructure 1
Boundary Condition
+
-
Tries to do the same Thing as Substructure 1
Error
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Model Reference Control• Controller designed so that
does the same thing as
• Part implemented in the computer
Substructure 1
Computational
Substructure 2
Physical
External Input
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Natural Physical
Feedback
Actuator
Feedback
ControllerTakes the same input as Substructure 1Takes the same input as Substructure 1
Boundary Condition
+
-
Tries to do the same Thing as Substructure 1
+
-
Tries to do the same Thing as Substructure 1
ErrorError
Substructure 1
Computational
Substructure 2
Physical
External Input
Work Conjugate
Boundary Condition
Actuator / Transfer Device
Natural Physical
Feedback
Actuator
Feedback
ControllerTakes the same input as Substructure 1Takes the same input as Substructure 1
Boundary Condition
+
-
Tries to do the same Thing as Substructure 1
+
-
Tries to do the same Thing as Substructure 1
ErrorError
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Computational Substructure
Physical Substructure
Actuator
External ExcitationInterface
Force
Interface Displacement
Physical Substructure
Model
ActuatorModel -
+Modeling
Error
+
+
Another Approach
Internal Model Control - IMCPart implemented in the computer
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Equivalence of different approaches
• The two approaches can be shown to be shown to be different
parametrizations of a 2 DOF controller
• Each offers a different perspective
– MRC useful in design
– IMC useful in robustness analysis
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
CU FHT Algorithm
A R(u)+
-P
u
A +-
R u K
2
1
ms cs K F
Numerical Integration
Modeling Error Computation
• Computer implementation of IMC
Discretize at 10 ms Discretize at 1 ms
CU FHT Algorithm !!(Shing et. al., 2005)
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Hybrid Simulation with a Shaking Table
• Necessary when physical substructure has distributed mass
`Masses simulated
Physical substructure has no masses of significance, hence no inertia effects(Hence pseudo-dynamic)
• In many cases of practical interest for hybrid simulation, mass is distributed and there is no such natural way of lumping the mass for substructuring.
• Examples:
– Nonstructural components in civil structures
– Payloads in aerospace structures
– Machine components
– Dams, chimneys and other continuum civil structures
– Soil / fluid-structure interaction
• The interface device must be able to dynamically excited a system with distributed mass – shaking table
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Outline• What is hybrid simulation?
• Why do it?
• Challenges in implementing a hybrid simulation system
• Types of hybrid simulation
• Hybrid simulation algorithms – architecture and equivalence
• Hybrid testing with shaking tables
• Current and planned work, Conclusions
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Hybrid Testing with Shaking Tables• 1.5 m x 1.5 m working area• +/- 200 mm dynamic stroke• Frequency range – 0-50 Hz• Maximum payload – 2000 kg• Maximum Acceleration – 1.0-2.9 g• Maximum Velocity – 1 m/s• Will give CU structures lab capability to perform such hybrid
simulations as listed in the previous slide• Collaboration with MTS Systems
Physical Substructure
Computational Substructure
Shaking Table
External Actuator
Computational Substructure
Physical Substructure
Response Feedback `
Reaction Wall
Shaking Table
Physical Substructure
Response Feedback `
Hybrid Simulation ConfigurationsCombination of Shaking Table and
External Actuator Shaking Table Only
Feb. 19, 2008 CU-NEES 2008 FHT WorkshopNEES at CU Boulder
The George E Brown, Jr. Network for Earthquake Engineering Simulation
01000110 01001000 01010100
Conclusions• Hybrid simulation – online combination of computation and physical experimentation
• Useful for qualification/proof-of-concept testing when the interaction of a component
with its surroundings needs to be accurately represented
• Challenge – added dynamics and feedback paths created by the transfer
system/actuator applying that applied interface conditions between the two
substructures.
• More difficult in dynamic hybrid simulation where physical substructure has
significant inertia (as opposed to pseudo-dynamic)
• Algorithms based on a control-systems perspective offer more promise than those
motivated by the finite element method
Top Related