SECED 2015 Conference: Earthquake Risk and Engineering towards a Resilient World
9-10 July 2015, Cambridge UK
OVERVIEW OF SEISMIC HYBRID TESTING
Daniel McCRUM1
Abstract: Even though computational power used for structural analysis is ever increasing,
there is still a fundamental need for testing in structural engineering, either for validation of
complex numerical models or material behaviour. Many structural engineers/researchers are
aware of cyclic and shake table test methods, but less so hybrid testing. Over the past 40
years, hybrid testing of engineering structures has developed from concept through to
maturity to become a reliable and accurate dynamic testing technique. In particular, the
application of hybrid testing as a seismic testing technique in recent years has increased
notably. The hybrid test method provides users with some additional benefits that standard
dynamic testing methods do not, and the method is much more cost effective in comparison
to shake table testing. This paper aims to provide the reader with a basic understanding of
the hybrid test method and its potential as a dynamic testing technique.
Introduction
Physical testing in a structural engineering context can be; monotonic, cyclic or dynamic. A
monotonic test is essentially a tension or compression test under increasing load. For
example, the steel coupon test consists of a steel bar being pulled under tension and results
provide the Young’s Modulus, yield stress and ultimate stress of the steel. A cyclic test
applies repeated tension and compression cycles to a specimen. Cyclic tests are quasi-static
in that they are performed at a slow loading rate, and provide vital information about the
fatigue performance and energy dissipation capacity of the material/structural component
under investigation. Figure 1 provides typical load cycles for example in a moment resisting
connection according an American Institute for Steel Construction code (ANSI/AISC (2005)).
Importantly, cyclic tests are easily repeatable.
Figure 1. Typical load cycles used in cyclic testing of moment resisting connections
Dynamic tests replicate dynamic loading and can vary from free-vibration, impact,
earthquake to wind loading. In earthquake engineering, shake table test facilities have been
developed to provide fully dynamic testing capabilities and give the most realistic laboratory
replication of dynamic effects. A shake table is a rigid platform that is moved by dynamic
actuators to mimic ground motion during an earthquake event. The first ever shake table was
wheel operated and used to characterise construction types, taking place at the University of
Tokyo, Japan in 1893 (Reitherman (2012)). The first ever hydraulically powered shake table
was conceived by Professors Clough and Penzien at UC Berkely, California, United States in
1969 (PEER (2014)). The rigid table was first used in 1972 and measuring 6m sq. is still the
largest multi-directional shake table in the United States (PEER (2014)). Shake table test
1
Lecturer in Structural Engineering, Queen’s University Belfast, UK, d.mccrum@qub.ac.uk
D.P. McCrum
facilities are expensive to construct and maintain whilst the limitations of payload capacity of
the table often result in experimental structures having to be scaled. Subsequently, the
length, force and time need to be scaled appropriately to ensure similitude between scaled
and unscaled results.
In dynamic time history analysis, building structures are often idealised as frame elements
with lumped masses at the nodes. Numerical integration is performed on the equation of
motion to solve for the acceleration, velocity and then displacement of the structure at each
incremental time step. However, what if the stiffness of the structure is provided by a physical
test rather than numerically calculated? This would mean that the displacement being solved
for would be more accurate as the stiffness is the real stiffness of the structure. This is the
fundamental principle behind the hybrid test method.
The hybrid test method involves the physical testing of a critical part of a structure whilst
simultaneously numerically modelling the remainder of the structure. The critical part of the
structure’s response may be difficult to numerically model e.g. highly nonlinear material
response or else rate dependent behaviour, and the numerically modelled part of the
structure has a more predictable response e.g. linear or only slightly nonlinear. Therefore,
rather than testing the entire structure, it may only be necessary to test the critical part of the
structure. The results from the physical test feed into the numerical model of the remainder of
the structure to provide a more accurate overall prediction of the displacement structural
response during dynamic excitation. Essentially, a hybrid test replaces numerically calculated
structural response with physically tested structural response. In most cases, only part of the
structure is physically tested, however in some experiments, the entire structure has been
tested. This will be discussed in more detail later on in this article. Importantly, tests can be
run at full-scale thus removing all potential issues associated with scaling dynamically tested
structures. Shake table tests often need to be performed on scaled specimens as fully scaled
specimens would exceed the capacity of most shake table facilities.
The development of different forms of hybrid testing such as; pseudo dynamic testing,
substructured testing, real-time hybrid testing and geographically distributed hybrid testing
will be discussed in this paper. The content of this paper is based on McCrum (2014).
Pseudo Dynamic Testing
Dynamic test methods have developed over the past 40 years to make use of computational
power and advances in hardware to provide more cost effective method of full-scale dynamic
testing as compared to shake table testing. Rather than using multiple hydraulic actuators to
push a shake table the actuators have been more efficiently employed in hybrid testing, see
Figure 2. The method combines physical testing with simultaneous computational modelling.
The method has been widely implemented in earthquake structural engineering testing and is
discussed in general terms in this paper. Physical testing remains an important part of the
progression of engineering science and the hybrid test method has advanced with
developing computational power to achieve a cost effective full-scale dynamic testing
method.
2
D.P. McCrum
Figure 2. Schematic of the hybrid test method (Carrion & Spencer (2007))
The pseudo dynamic test method is a computational dynamic analysis procedure in which
the stiffness term is physically measured. The assumption of lumped masses at the nodes is
necessary as dynamic actuation can only be applied at lumped mass nodal locations (refer to
Figure 1). If we consider the equation of motion below;
(1)
in which M is the mass matrix,
is the nodal acceleration vector, C is the damping matrix,
is the nodal velocity vector,
is the restoring force vector (which is equal to
,
where K is the stiffness matrix and
is the nodal displacement vector) and Fi+1 is the
external excitation force applied to the system (i is the current time step and i+1 is the next
time step). Time steps are usually 0.01s as this is the time increments at which typical
earthquake time histories are recorded at. As per Figure 1, if the excitation is an earthquake
ground acceleration of , then
. Figure 1 indicates how the displacement at the
next time step,
is calculated using the equation of motion by the numerical component.
The displacement of the structure at the next time step (i+1) is then sent to the actuator. The
actuators then apply the displacement, xi+1 to the structure. The restoring forces are
measured by the actuators, R1 and R2. These are then sent back to the numerical
component. The restoring forces are then applied to the equation of motion and used to
calculate the acceleration, velocity and subsequently the displacement for the next time step.
This process is referred to as closed looped. In Figure 2, the structural masses (M) and
damping (C) are numerically modelled, whilst the entire two-storey structural frame is
experimentally tested.
Concept of Substructuring
The concept of substructuring is critical to the overall cost effectiveness of the hybrid test
method. It was first introduced in the United States in 1985 (Dermitzakis & Mahin (1985)) and
subsequently implemented by many others. An example of substructuring can be seen in
Figure 3. Figure 3 shows a steel concentrically braced frame structure in which the response
of the ground floor braced frame on one side of the building is critical to the overall dynamic
response of the structure. Rather than testing the entire structure (Figure 3(a)), the critical
ground floor braced frame that has been identified through numerical simulation is then
physically tested (Figure 3(b)). The physically tested part of the structure is referred to as the
physical substructure. Figure 3(c) shows the numerical part of the structure, referred to as
the numerical substructure. Results from the physical test are sent back to the numerical
substructure to solve for the overall structural response. Substructuring prevents the need for
scaling test structures or the testing of entire structures.
3
D.P. McCrum
Figure 3. Concept of substructuring applied to concentrically braced frame structure; (a) entire
structure; (b) critical elements tested; and (c) numerical model
Real-time Hybrid Testing
The challenge to overcome when a pseudo dynamic test that is run close to or at real-time is
that the computation of the displacement for the next time step needs to be performed very
quickly. Considerable research has been undertaken to improve the hydraulic actuator
systems and numerical methods associated with accurate and timely actuation. The first
system capable of performing a real-time hybrid test was performed in Japan in 1992
(Nakashima & Kato (1992)). In a real-time hybrid test the structure is tested at or close to
real-time therefore velocity dependent behaviour is physically measured. The damping and
inertia response of the tested structure are measured in the force feedback and not defined
computationally as in the pseudo dynamic test method.
Figure 4. Schematic of real-time hybrid test system
Figure 4 presents a schematic of a real-time hybrid test system. The analysis is initiated by
the Computational Structural Model in which the command displacement (referred to as
command displacement as the command is being sent to the actuator) for the first time step
is calculated through numerical integration of the equations of motion. The structural model is
separated into numerical and physical substructures by the user. The digital command
displacement signal is converted to an analogue signal for use by the actuator. The analogue
command displacement is sent to the Actuator Controller. This controller ensures accurate
application of the command displacement to the test specimen. An actuator command
displacement is sent by the Actuator Controller to the actuator and applied to the Test
System. Data acquisition devices provide the measured displacement that is sent back to the
4
D.P. McCrum
Actuator Controller to check that the displacement has been accurately applied to the
structure (referred to as an inner control loop). With the displacement error tolerance of the
inner control loop satisfied, the measured restoring force and measured displacement from
the actuator are converted back from an analogue to digital signals for use by the
Computational Structural Model. The measured displacement and restoring force of the
experimental substructure is combined with the numerical substructure and the next time
step command displacement is calculated. This procedure is repeated until the end of the
test which is typically the end of the input time history (e.g. recorded earthquake).
Distributed Hybrid Testing
Hybrid testing is ideally suited to performing tests in geographically separate locations as
communication between the physical and numerical substructures can take place over the
internet rather than through a cable in a laboratory. Researchers can benefit greatly from the
combination of one or more experimental sites. The first implementation of a geographically
distributed hybrid test was performed between Kyoto University and Osaka City University,
Japan (Watanabe et al. (2001)). Since then, geographically distributed hybrid tests have
been conducted internally within the United States (Spencer et al. (2004)), United Kingdom
(Saleem et al. (2008)) and Taiwan (Yang et al. (2003)) and between a number of countries,
notably Korea/Japan (Watanabe et al. (2001)), the United States/Japan (Park et al. (2005)),
the United States/Taiwan Wang et al. (2007)) and New Zealand/United States (Ma et al.
(2007)).
Figure 5. Schematic of client/server configuration and data communication for a distributed hybrid test
(adapted from Park et al. (2005))
A number of distributed hybrid test frameworks have been developed over the past 15 years
to improve the communication between the physical and numerical substructure locations.
Figure 5 presents a standard client/server framework that forms the foundation of many of
the frameworks developed to date. The framework shown consists of two experimental sites
(Institution I & II) and one main computational site. The command displacements calculated
by the Analysis Engine (typically by a finite element package) on the Main Computer are
transferred across the internet and controlled locally by each Local Server as shown in
Figure 5. The command displacement is then applied to the local test structure. The restoring
force and measured displacements are then sent back to the Client System. The data
communication between the sites is co-ordinated over the internet using the standard
transmission control protocol/internet protocol. The protocols control how two computers talk
to each other over the internet. The protocols are responsible for sending, receiving and
checking the contents of the data packets. The way in which data is sent and received varies
between the different distributed hybrid test frameworks that have been developed, but
typically a protocol is set-up to ensure application of the command displacement has been
achieved at each experimental site before sending back to the Main Computer site. The data
5
D.P. McCrum
communication can be paused to allow all sites to complete the test time step and is
particularly important if there are network interruptions i.e. communication delays over the
internet. Improvements on the transmission and storage of data over the internet have been
a large focus of research in distributed hybrid testing, in particular different approaches to
dealing with random delays in internet communication.
More recently, Ojaghi et al. (2014) demonstrated the feasibility of performing distributed
hybrid tests in real time over the internet, through a series of experiments conducted
between Oxford and Bristol universities in the UK. The tests used existing hardware and
control systems at both sites, with modifications designed to minimise local delays and to
prioritise real-time communications over other processes. Real-time hybrid testing was
achieved across a variety of relatively simple test set-ups. However, the method was tailored
to the particular laboratories involved and was not readily transferable to other sites.
Recent Applications of the Hybrid Test Method
In earthquake engineering research, the pseudo dynamic test method has been applied to
structures that do not have any rate dependent behaviour, whilst real-time hybrid tests have
been applied to rate dependent devices such as viscous dampers. In order to provide
readers of this article with an understanding of the capability of hybrid testing, Figures 6-9
present a small selection of some of the hybrid experiments undertaken to date.
Figure 6. Pseudo dynamic test of a full-scale three-storey reinforced concrete plan irregular structure
(Negro et al. (2004))
Figure 7. Full-scale substructure pseudo dynamic test of a full-scale three-storey plan irregular steel
concentrically braced frame structure (McCrum & Broderick (2013))
6
D.P. McCrum
The experiment shown in Figure 6 was a pseudo dynamic test performed on a full-scale
three-storey moment resisting reinforced concrete framed structure at the European
Laboratory for Structural Assessment in Ispra, Italy. The structure was constructed as per
1970’s design typical of southern Europe and had plan irregularity. The entire structural
frame was tested including barrels containing water to provide live loading. The white frame
is the test structure and the orange frame is the reference frame for measuring
displacements. Figure 7 presents a substructured pseudo dynamic test performed on a
three-storey two-by-one bay steel braced plan irregular structure at the Structures Laboratory
in Trinity College Dublin, Ireland. The ground floor braced bay frame of the steel
concentrically braced frame structure was tested at full-scale as shown in Figure 7. The
remainder of the structure was numerically modelled.
Figure 8. Schematic of full-scale substructured real-time hybrid test of a full-scale semi-active control
device (Carrion & Spencer (2007))
The experiment shown in Figure 8 presents a schematic of a real-time hybrid test performed
on a full-scale magnetorheological damper within a three-storey steel moment resisting
frame undertaken at University of Illinois at Urbana-Champaign, United States. Figure 9
shows the physical substructure of a real-time hybrid test of a three-storey, one-bay steel
braced frame structure with viscous dampers (referred to as DBF in Figure 9). Tests were
undertaken at the Advanced Technology for Large Structural Systems Research Center at
Lehigh University, United States. Performance based design of the steel framed structure
with viscous dampers was assessed.
Figure 9. Substructured real-time hybrid test of large-scale steel frame with nonlinear viscous dampers
(a) photo of test set-up; and (b) schematic of damper (Dong et al. (2014))
7
D.P. McCrum
Within the discipline of structural engineering in the United Kingdom & Ireland, hybrid test
facilities exist at University of Oxford, University of Bristol, University of Cambridge and
Trinity College Dublin. Within the United States the George E. Brown, Jr. Network for
Earthquake Engineering Simulation (NEES (2014)) has brought together 14 experimental
sites across the United States. Other major facilities are located in Italy at the Joint Research
Centre (JRC (2015)), South Korea at the Korea Advanced Institute of Science & Technology
(KAIST (2015)), Taiwan at the National Centre for Research on Earthquake Engineering
(NCREE (2015)) and the Disaster Prevention Research Institute in Japan (DPRI (2015)).
Conclusions
The need for physical dynamic testing still exists in the progression of structural engineering
science. In structural engineering, cyclic tests were developed to investigate the effects of
low cycle fatigue with loading applied at a very slow rate. To overcome issues related to slow
loading rates, shake table and hybrid test methods were developed towards the later end of
the 20th century to investigate true dynamic behaviour.
The hybrid test method is a dynamic testing technique that replaces numerically calculated
structural response with physically tested structural response. Hybrid tests can be performed
at an extended timescale (pseudo dynamic test) for structures with a non-velocity dependent
response e.g. steel framed structures, or else at real-time (real-time hybrid test) for structures
with velocity dependent response e.g. damping device. The concept of substructuring within
hybrid testing greatly reduces the cost of full-scale testing as only the critical part of the
structure and not the entire structure needs to be physically tested.
An interesting development is geographically distributed hybrid testing which involves testing
the physical substructure(s) in one or more locations and solving the numerical substructure
in another separate geographical location. Data from the physical test is sent over the
internet to solve the equation of motion of the combined physical and numerical
substructures. This offers great opportunities in the future to share testing facilities and
expertise. This test method still requires some further research and development before it
becomes reliable and robust.
The hybrid test method has reached a level of maturity over the past 40 years to become a
reliable, robust and accurate dynamic test method that can be used to perform full-scale
seismic tests of engineering structures in a cost effective manner.
REFERENCES
ANSI/AISC 341-05 (2005) Seismic Provisions for Structural Steel Buildings. American Institute of Steel
Construction (AISC), Illinois, United States
Carrion JE and Spencer FB Jr. (2007) Model-based Strategies for Real-time Hybrid Testing, NSEL
Report Series. Report No. NSEL-006 ed., University of Illinois at Urbana-Champaign
Dermitzakis SN and Mahin SA (1985) Development of substructuring techniques for on-line computer
controlled seismic performance testing. Report UCB/EERC - 85/04, Earthquake Engineering Research
Center, University of California, Berkeley
Dong A., Sause R, Ricles JM (2014) Real-time Hybrid Simulation: Validation of PBD Procedure for
Steel Frames with Nonlinear Viscous Dampers, Tenth U.S. National Conference on Earthquake
Engineering, Anchorage, Alaska
DPRI (2015) [online] Available at: http://www.dpri.kyoto-u.ac.jp/web_e/index_topics.html (Accessed
January 2015)
JRC (2015) [online] Available at: https://ec.europa.eu/jrc/en/about/jrc-site/ispra (Accessed January
2015)
KAIST (2015) http://www.kaist.edu/html/en/index.html (Accessed January 2015)
8
D.P. McCrum
Ma QT, Omenzetter P, Ingham JM, Butterworth JW, Pender MJ (2007) Overview of
nznees@auckland, Proceedings of 2007 New Zealand Society for Earthquake Engineers Conference,
Palmerston North, New Zealand
McCrum DP (2014) Development of dynamic structural testing to current state-of-the-art hybrid
testing, The Structural Engineer, November 2014
McCrum DP and Broderick BM (2013) An experimental and numerical investigation into the seismic
performance of a multi-storey concentrically braced plan irregular structure, Bulletin of Earthquake
Engineering, Vol. 11, Iss. 6, pp. 2363-2385
Nakashima M, Kato H, Takaoka E (1992) Development of real-time pseudo dynamic testing.
Earthquake Engineering & Structural Dynamics, 21, pp. 79-92
NCREE (2015) [online] Available at: http://www.ncree.org.tw/ENG/ (Accessed January 2015)
NEES (2015) [online] Available at: https://nees.org/ (Accessed January 2015)
Negro P, Mola E, Molina FJ, Magonette G (2004) Full-scale psd testing of a torsionally unbalanced
three-storey non-seismic rc frame, Proceeding of 13th World Conference on Earthquake Engineering,
Vancouver, B. C., Canada
Ojaghi M, Williams MS, Dietz MS, Blakeborough A, Lamata Martinez I (2014) Real-time distributed
hybrid testing: coupling geographically distributed scientific equipment across the internet to extend
seismic testing capabilities. Earthquake Engng Struct. Dyn., 43(7), 1023-1043
Park D-U, Yun C-B, Lee J-W, Nagata K, Watanabe E, Sugiura K (2005) On-line pseudo-dynamic
network testing on base-isolated bridges using internet and wireless internet, Experimental
Mechanics, 45, pp. 331-343
PEER (2014) [online] Available at: http://peer.berkeley.edu/laboratories/shaking_table.html. (Accessed
October 2014)
Saleem K, Parra-Fuente J, Ojaghi M, Williams M, Blakeborough A (2008) Uk-nees: Grid services
architecture for earthquake, Proceedings of the WSEAS International Conference on Applied
Computing Conference, Istanbul, Turkey, World Scientific and Engineering Academy and Society
(WSEAS)
Spencer BF, Finholt TA, Foster I, Kesselman C, Beldica C, Futrelle J, Gullapalli S, Hubbard P, Liming
L, Marcusiu D, Pearlman L, Severance C, Yang G (2004) Neesgrid: A distributed collaboratory for
advanced earthquake engineering experiment and simulation, 13th World Conference on Earthquake
Engineering, Vancouver, BC, Canada
Reitherman R (2012). Earthquakes and Engineers: An International History. Reston, VA: ASCE Press.
pp. 126–127
Wang KJ, Tsai KC, Wang SJ, Cheng WC, Yang YS (2007) Isee: Internet-based simulations for
earthquake engineering part ii: The application protocol appraoch, Earthquake Engineering &
Structural Dynamics, 36, pp. 2307-2323
Watanabe E, Kitada T, Sugiura K, Nagata K (2001) Parallel pseudo-dynamic seismic loading test on
elevated bridge system through the internet, The Eighth East Asia-Pacific Conference on Structural
Engineering and Construction. Singapore
Watanabe E, Yun C, Sugiura K., Park D, Nagata K (2001) On-line interactive testing between kaist
and kyoto university, Proceedings of the 14th KKNN Symposium on Civil Engineering, Kyoto, Japan
Yang YS, Wang SJ, Wang KJ, Tsai KC, Hsieh SH (2003) Isee: Internet-based simulations for
earthquake engineering part i: The database approach, International Workshop on Steel and Concrete
Composite Construction. Taipei, Taiwan
9