EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS

Earthquake Engng Struct. Dyn. 2011; 40:195213

Published online 4 June 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/eqe.1015

The development of shaking tablesA historical note

R. T. Severn,
CBE FREng., University of Bristol, U.K.

SUMMARY
The earliest known shaking table, driven by hand-power, was constructed in Japan at the end of the 19th
century. At the beginning of the 20th century developments had moved to the Stanford University in
the U.S. with the introduction of an electric motor to produce a more refined oscillatory motion in one
direction, the response of the testpiece being recorded mechanically by pens on a rotating drum. Major
earthquakes in the 1920s prompted renewed interest at Stanford resulting in a uni-directional table moving
on rails, activated either by a pendulum striking at one endthe other being resisted by springsor by
a wheel with an eccentric-mass attached to the table. A valuable feature here was that the size of the
eccentric mass could be varied as the harmonic motion continued, thereby providing a method of control.
In the 1950s, a similar pendulum input was used on a table constructed at the University of California,
but instead of rails, it was supported by a group of vertical bars flexible in one direction only, and the
19391945 war had resulted in the availability of electrical devices for measuring response. Also, in Italy
at this time the use of pendulums was augmented by contra-rotating mass input devices giving better
frequency control; arrays of several electrodynamic exciters were also used. In Japan, motion was induced
by the release of compressed springs.
The idea of producing input by an oil-filled piston was introduced at MIT after the 1933 Long Beach
earthquake to a table suspended from above by wires. Two other innovations here were of the greatest
significance. First was an analogue device for using an actual earthquake record as input, and the second
was control of the motion by an error-driven electrically controlled feedback loop. The development
of these ideas into the shaking tables, which we use today, had to wait upon the general development
of control engineering during the 19391945 war, followed by progressively greater speeds in digital
computation. This history ends (c.1985) after the continuation of these advances made possible full 6-DOF
control using many oil-filled actuators, but before they became able to give us real-time control with the
attendant abilities of multi-support input and the experimental study of inelastic behaviour.
Received 16 November 2009; Revised 1 April 2010; Accepted 1 April 2010
KEY WORDS:

shaking tables; early history of shaking tables; earthquake simulation

If I have seen further it is by standing on the shoulders of Giants.

(extract from a letter by Sir Isaac Newton to Robert Hooke, June, 1662)

1. INTRODUCTION
The above quotation comes from correspondence between two illustrious early researchers in engineering science. I have included it here so as to recall what we owe to those who have gone before us.
This paper, one of a series being presented on the history of early developments in earthquake
engineering, is aimed at readers who wish to become familiar with the contribution that the
development of shaking tables has made to the advancement of our subject. It traces their history,
basically in the U.S., Japan and Europe, from a first appearance at the end of the nineteenth century,
Correspondence

E-mail:

to: R. T. Severn, CBE FREng., University of Bristol, U.K.

[email protected]

Copyright 2010 John Wiley & Sons, Ltd.

196

R. T. SEVERN

through developments during the greater part of the twentieth. In those early years their contribution
to research and product qualification was primarily of a qualitative nature, but the subsequent
advances in control technology and in digital computation produced experimental results that were
of an accuracy such that they could be used in assessing the validity of theoretical approaches,
and in qualifying industrial products as fit-for-purpose if subjected to earthquakes. This record
of developments ends before real-time adaptive (as opposed to fixed-gain) control and very fast
computer speeds in the late 1980s made it possible to begin to study non-linear behaviour and the
sub-structuring of testpieces.
A significant difference between this paper and those in this historical series that have dealt with
theoretical aspects is that in the latter, major developments were usually the work of individuals or
a small group of co-workers, whereas, except in the very early days, shaking tables were the result
of a team effort involving researchers in a number of fields in engineering science. Moreover, since
the 1970s, large industrial companies, notably MTS in the U.S. and Mitsubishi in Japan, have
dominated the field, their substantial investments requiring them to be less willing than academic
researchers to publish details of their achievements.
2. THE REQUIREMENTS
Shaking tables are of value in earthquake engineering because they are the only experimental
device that attempts to replicate the true nature of the earthquake input. This they do by applying a
ground motion to the base of a structure, thereby inducing realistic inertia forces in every element
of its mass. It is these forces that generate the response displacements and stresses. As with other
experimental devices, the history of shaking tables is that of progression toward the satisfaction
of ideal requirements, making it necessary here to specify what these requirements are, but before
doing so it is noted that the progression started by using rotating wheels with eccentric driving
rods to produce some form of oscillatory motion, to be followed by a pendulum striking a table
resisted by springs, which gave a partial indication of real earthquake motion, because after the
initial impact a decaying vibration was produced. It was not until the mid-1930s that the first usable
record of an actual earthquake (Long Beach, U.S., 1933) became available, to be used by Ruge [1]
and [2] in an analogue form as an input to his table. The second world war period (19391945)
gave impetus to power transmission devices of many forms, one of which was available by the
mid-1960s with satisfactory servo-valve control to be used to input recorded earthquakes into
shaking tables in digital form.
2.1. Types of input
The ground motion produced by an earthquake can have components in all six degrees of freedom
(DOF), although only the time-histories of three of themtwo horizontal and one verticalmay
have been measured at the earthquake site, and for subsequent discussion of shaking table control,
it is important to note that it is acceleration that has been recorded, rather than displacement or
velocity. Quite often it is only one of these measured acceleration components that is specified for
testing purposes, and it might therefore be considered that a table capable of controlled motion
in one axis would be satisfactory, but this begs the question of how motion in the remaining five
is to be removed. Early tables were of this single-axis variety, with physical restraints on the
other DOF such as guide rails or roller-bearing supports. These restraints introduce a pattern of
forces into the table that is unknown and that modifies the intended single-axis motion. It can be
argued that for purely research purposes, this is not important; the investigator simply measures
the motion actually input into the testpiece and uses this for comparison with a corresponding
theoretical analysis. Indeed, the point can be inferred that real earthquakes do not repeat themselves
precisely, so that it is reasonable to subject the testpiece to any representative seismic motion.
However, in many instancessatisfying Code requirements and validation of industrial products
for exampleshaking tables are used to validate testpieces against precisely defined input motions;
these can be recorded time-histories of acceleration, or displacement derived therefrom, artificial
time-histories which have been obtained from specified response spectra, precise sinusoidal motions
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

197

of the sine-sweep or sine-dwell variety and random noise. All these types of input are used for
specific purposes, such as obtaining the natural frequencies and damping of the testpiece, and no
corruption of the input can be allowed.
2.2. Control of input
The laissez-faire attitude to input to a shaking table suggested above is not acceptable even for
basic research, for the practical reason that input of acceleration does not define displacement of
the table, and lack of control of this parameter could result in displacement values beyond the
physical limits of the table, resulting in serious damage to the installation. A further point is that
the testpiece itself is probably of considerable value for which an extensive study programme is
envisaged; early adventitious destruction is not therefore acceptable!
Ideally, what is required is control of acceleration, velocity and displacement of all six DOF of
the shaking table, and here earthquake engineering is greatly indebted to developments in control
engineering for eventually achieving this, and that the requirements just stated became available
for the first time in the mid-1970s for experiments, which remained in the linear range. Because
displacement transducers lose accuracy above about 10 Hz, and accelerometers are most accurate
above this value, a judicious choice of differentiated displacement and integrated acceleration is
used to produce velocity, the resulting control system being referred to as TVCthree variable
control. The search for satisfactory control when the testpiece becomes non-linear in its behaviour
began in the mid 1980s, and has met with some success [3] when the non-linearity occurs at a
rate that is within the frequency bandwidth of servo-valves and the actuators driving the table.
Difficulties also still arise when the mass of the testpiece is comparable to, or greater than, the
mass of the table itself.
2.3. Scale effects
The fact that energy in actual earthquakes lies mostly in the 08 Hz range means that shaking table
testing of a full-scale testpiece must occur in this range. But if the testpiece is at a scale of 1/n
of the real structure, dynamic similitude requires that the frequency content of the input must be
factored by n, leading to the fact that all but the largest tables operate in the 0100 Hz range.
Other effects of scaling are, first, that both the Froude and the Cauchy numbers for the model
and prototype must matchthe former being the ratio between the gravity and the dynamic inertia
forces, and the latter the ratio between these dynamic inertia forces and the elastic restoring forces.
Second, it is generally impossible to introduce realistic detail into a structural model, making it
particularly dangerous to accept any non-linear effects as being relevant to the prototype; this, the
fact that theoretical treatments of non-linearities are equally suspect, was one justification for the
recent construction of the worlds largest shaking table (1200T testpiece capacity and 015 Hz
frequency range) at Miki City in Japan [4].

3. SHAKING TABLES WITH SIMPLE OSCILLATORY MOTION

The first attempt at laboratory experiments of the effects of earthquakes on structural models
was made in Japan, c.1890, by an Englishman, John Milne , and a younger Japanese colleague,
Fusakichi Omori [5]. Their railway truck device (Figure 1) produced oscillatory motion in the
testpiece by means of a bar attached eccentrically to a hand-driven wheel, the motion being recorded
by pens on a rotating drum.
Somewhat more sophisticated was the table built in 1906 by F. J. Rogers, Assistant Professor of
Physics at Stanford University, which is shown in Figure 2. His very stiff table [6] had a wooden
box firmly attached to it, having dimensions 1008630 cm with a sheet metal lining. The table
John

Milne (18501913) was Professor of Geology and Mining at the Imperial College of Engineering in Tokyo.
He introduced the first international system of seismological recording stations, using his own seismographs.

Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

198

R. T. SEVERN

Figure 1. The MilneOmori railway truck shaking table ( c.1890s).

Figure 2. The shaking table designed and used by F. J. Rogers for testing the
behaviour of wet and dry sand.

plus box were separated from a fixed platform by 4 cm diameter steel rollers that constrained the
box to move in the single horizontal direction. A DC motor drove a 75 kg balancing wheel, which
in turn drove a connecting rod between the box and a crank connected eccentrically to the hub of
the wheel. The motion of the box was therefore basically sinusoidal (Figure 3), with frequency
variable within 0.54.6 Hz, and with a 10 cm maximum horizontal motion, which could be varied
by modification to the crank connection.
The response measurements made by Rogers were recorded on a hand-operated paper-covered
drum (indicated by G in Figure 2) mounted independently of the table, using pencils (H) to record
the motion of the box, and the motion of a small block (F) firmly embedded in the sand by
means of side pieces running down into it. In addition, the beats of an electromagnet connected
to a seconds pendulum were recordeda device that obviated any errors induced by hand-turning
the drum. Figure 3 is an example of the recordings made. They are given here to illustrate the
point that Rogers had no instruments for measuring either velocity or acceleration, and hence
estimated them from measurements on the displacement traces of Figure 3. He argued that the
square of the ratio of the times represented by cd and cddrawn at the same proportion (0.9)
of the peak-to-peak amplitudedivided into the ratio of the two amplitudes, gives the ratio of
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

199

Figure 3. Amplitude and time recordings made by F. J. Rogers; the lowest record is that of the
box itself; the middle one that of the block embedded in the sand; and the top trace is that of the
timing mechanism giving 1 s blips.

the two accelerations during the motion cPd. Thus, from Figure 3, the ratio of accelerations at
the reversal of motion direction is around three. From this resultforce being proportional to
accelerationRogers explains the destructive force of earthquakes.

4. EXCITATION USING SPRINGS AND PENDULUMS

As is customary in earthquake engineering research, it required a major earthquake to stimulate
both interest and funding, and this occurred after the 1925 Santa Barbara event and the 1923 Kanto
earthquake, which destroyed the greater part of Tokyo city. In the U.S. the Stanford University
was again the location of interest, where a committee of engineers and academics was formed to
consider how best to carry out vibration experiments with models of engineering structures. At
the first meeting, a decision was made that a laboratory should be built in which the principal
facility would be a shaking table, a task taken up by Dr Jacobsen and coworkers [710]. The
early discussions concerning its design centered on two aspectswhat types of input should be
attempted and how should these be achieved? For the first of these, the interesting use of the
words savage and civilised were used to describe what was known to be the characteristics of
real earthquakes on the one hand, and on the other the continuous and oscillatory disturbances,
which had already been achieved in the laboratory. Although it was supposed that the savage
motions might be achievable with complicated machinery, the inability to repeat any particular
experiment ruled it out of further consideration. But in accepting the use of civilised inputs in
one horizontal direction only, the production of continuous shaking by means of an adjustable
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

200

R. T. SEVERN

Figure 4. Jacobsens shaking table being used with pendulum input mechanism to study the added-mass
effect of water vibrating in a tank.

Figure 5. Theoretical traces of the two types of input used in the Jacobsen shaking table; the upper and
the lower ones are produced by the pendulum and the unbalanced flywheel, respectively.

Figure 6. The unbalanced flywheel mechanism of the Jacobsen shaking table.

crank and connecting rodas had been used, for example, by Rogers [6]was thought to be
inappropriate. Instead, two civilised inputs were implemented, a pendulum (Figure 4) making a
single impact through a bumper spring attached to the table, which was resisted by a group of
springs at the other end, giving an initial shock followed by a decaying motion as shown in the
upper part of Figure 5. Alternatively, an unbalanced flywheel mounted on the table itself produced
the harmonic motion shown in the lower part of Figure 5. Details of the flywheel can be seen
in Figure 6, which shows four symmetrically attached weights and a single eccentric box that
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

201

contains lead shot. An ingenious arrangement allowed the frequency and the amplitude of the
motion to be changed by increasing, or decreasing, the amount of shot, even with the wheel in
motion. It was appreciated that this civilised motion would become less civilized if resonance
occurred between the natural frequencies produced by the flywheel and those of the table itself,
and that precautions must be taken to avoid this. This eccentric box was therefore an early form
of shaking table frequency control.
The plan dimensions of the table were 1012 ft, fabricated from 8 in steel H-beams bolted and
welded together and weighing 600 lbs. Figure 4 indicates that it was carried on wheels running
along two rails, at one end concrete abutments were constructed to which the table was fastened
by helical springs, the stiffness of which could be varied so as to modify its behaviour.
Over a period of 25 years Jacobsen made many improvements to his table and its recording
devices, which still had no means of measuring either velocity or acceleration. In particular, he
introduced force and moment dynamometers (Figure 7) that required the measurement of deflections
less than 0.07 in during a time interval of 0.05 s, and this led to two devices, one being a stack of
carbon discs under pressure, and the other magnetic reluctance of a varying air gap, both operating
in conjunction with an oscillograph; only the second of these two devices was successful in practice.
An important feature of all of Jacobsens papers on shaking table developments is that they were
accompanied by mathematical analysis of the problems for which they were created, and he held
strongly to the view that experiments should be guided by analysis. Though development of such a
thesis is outside the scope of this paper, it is considered valuable to record Jacobsens own comment:
In order to make inferences from model experiments reliable, a considerable acquaintance with
the theory underlying the similitude problem must be acquired; a mere experimentation without
a conscious theoretical foundation cannot be expected to yield results on which an engineer is
willing to risk the safety of human beings and invested capital.
The Jacobsen design of shaking tables was copied by several researchers around the world. The
production of harmonic excitation of a shaking table by an eccentricmass vibrator was used in
Japan, and in 1936 Mononobe et al. [11] refer to 1:100 scale models of earth dams being tested in
which its displacements were measured by optical methods, which included motion pictures. In
Italy, as late as 1956 Oberti [12] refers specifically to the construction of a Jacobsen-type table at
ISMES, of plan dimensions 1015 ft, which was suspended from a very stiff gantry by wires, and
could be excited in three waysby a combination of a pendulum and springs, by an eccentric- mass
device producing either horizontal or vertical motion, or by four synchronized electromagnetic
vibrators producing harmonic motion. Figure 8 illustrates the first two of these options.
In the early 1950s a 710 ft table of the Jacobsen type was built at the University of California,
Berkeley at the instigation of Clough and Pirtz [13] in order to study the behaviour of rockfill
dams. It had an 8 in thick reinforced concrete slab mounted on four steel-plate legs (Figure 9),
so oriented as to be flexible in the single allowed direction of horizontal motion. One end of the
slab was connected to a heavy spring, which provided the frequency characteristics of the system,
with motion being initially introduced by a 150 lb pendulum striking the other end. By now, as
a result of developments during the 19391945 war, accelerometers and LVDT were available
for measuring accelerations and displacements, and these were used to record the movements of
rockfill dams with the aid of a two-channel direct recording accelerograph.

Figure 7. Illustrating the force and the moment dynamometers used by Jacobsen.
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

202

R. T. SEVERN

Figure 8. Obertis Jacobsen-type shaking table for testing concrete dam models: (a) shows the Jacobsen
arrangement and (b) the eccentric-mass vibrator in the vertical position.

Figure 9. The Jacobsen-type shaking table used by Clough and Pirtz to study rockfill dams.

Before leaving this description of tables actuated by a pendulum and springs, it is noted that
in Japan, c.1958, the pendulum was replaced by a mechanism that pre-loaded a set of springs,
subsequently releasing them to produce a damped vibration (Figure 10). It was used to study the
dynamic behaviour of the graphite core of the TokaiMura nuclear power station, the first such
facility to be built in Japan, using the U.K. Calder Hall station as its prototype [14].

5. EXCITATION USING LIQUID-FILLED ACTUATORSTHE BIRTH OF THE MODERN

SHAKING TABLE
5.1. The innovations of A. C. Ruge
The collapse of elevated water tanks in the 1933 Long Beach earthquake gave A. C. Ruge, a
Research Associate in seismology at MIT, the opportunity to construct a table suspended from
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

203

Figure 10. The spring-loaded Japanese shaking table (c.1958) testing a model of the graphite core of the
Tokai-Mura nuclear power station.

Figure 11. Schematic diagram of Ruges innovatory shaking table.

above by piano wires and prevented from rotation by an array of wires attached to the floor [1, 2].
The table was initially controlled by springs, with a driving system of the eccentric-mass variety
to produce sinusoidal motion, but of greater importance in the development of shaking tables was
Ruges realization that the mechanical systems then used gave an inadequate representation of
real earthquake motions, and were never likely to be capable of doing so. As a consequence he
began research into the kind of input devices that, after more than 70 years of enhancements, have
given us the shaking tables that we use today. Instead of springs, pendulums and eccentric-mass
flywheels, Ruge introduced an oil-driven piston attached to the table (Figure 11), the position of
which was controlled by another of Ruges original concepts, a device that allowed displacements
obtained from an actual earthquake record to be the input displacements of the table.
Referring to Figure 11, the flow of oil to the piston is controlled by a floating cylindrical valve,
to which a driving coil is attached, the position of whichand hence the valveis effected by
a powerful electromagnet. The current in this magnet is determined by the second of Ruges
inventionswhich he refers to as an electric eye; it is indicated diagrammatically at the bottom
right-hand side of Figure 11. Here, in a circular disc rotated by a small motor, the perimetral
serrations have been cut to replicate the displacements obtained from an actual earthquake record
by double integration of its recorded acceleration. As this disc rotates at a required speed, light
from the cell is directed at the serrated perimeter allowing a current to be generated which
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

204

R. T. SEVERN

copies in magnitude the displacement record, and hence ultimately drives the piston. The practical
requirement is to keep the spot of light exactly at the serrated edge of the disc, a demand which
control engineers call an error-drive process; in other words, it is an error (e in Figure 11) in
failing to achieve the desired condition, which is fed backin the form of an electrical signalto
the driving coil. However, if e was used by itself, the control would not be stable and the system
would hunt violently about the null position. To avoid this, the control system must anticipate what
is likely to happen by measuring the rate-of-change (de/dt) of the error, and Ruge incorporated
this into his system in addition to a damping force V proportional to the valve velocity. He also
anticipates that for very fast inputs, the second and third time-derivatives of e would also be
required, and as Figure 11 indicates, he also contemplated an additional input in the orthogonal
directionreferred to in Figure 11 as Future Drive. It is not known whether these intentions
were actually realized, and todays control engineers are skeptical that Ruge could have succeeded
with them, but he did invent a novel arrangement for measuring response displacement of the table
using a swiveled mirror and lens.
The account of this ground-breaking research [1] ran to 60 pages and an extensive discussion.
It is essential reading for those interested in the early days of laboratory studies in earthquake
engineering. Before leaving these major contributions by Ruge, it is noted that he preceded his
experimental work by a thorough investigation of the similitude requirements of the models he
intended to use.
5.2. The water-supported table at Jassy (Romania)
The 19391945 war and its aftermath interrupted the development of shaking tables, but a compensation was the rapid innovation in control of hydraulic power systems for military purposes,
particularly servo-valve technology, and these were soon applied to shaking table control. Possibly
the first to do so was the Building Research Institute at Jassy in Romania [15], where in the early
1960s two electro-hydraulic actuators were used to drive a 1010 m table with acceleration up
to 0.4g, although it also had an eccentric-mass driving device for producing sinusoidal motion in
the frequency range 015 cps. But the need for testing full-scale structures prompted a return to
the earlier option of supporting the table by fluid pressurein this case waterthe novel feature
at Jassy being a subdivision of the fluid bearing surface into 16 compartments each of which was
provided with a nozzle indicated by the numeral 9 in Figure 12. The separating grid, indicated
by numeral 11, was made from compressible polystyrene foam that had an appreciable hydraulic
resistance, allowing the weight of both table and testpiece to be taken initially by the fluid pressure
from a 10 m high concrete water tower. The innovatory feature of this compartmentalized system
was the ability of each compartment to experience a different pressure for a short time before the
self-stabilising effect of the fluid system took place, therebywithin the compressibility limit of
the polystyrenereducing rotations of the table about the two horizontal axes (pitch and roll), but
rotations about the vertical axis (yaw) could not be controlled in this way.
5.3. The single-axis shaking table at the University of Illinois, Urbana
In its mechanical characteristics this single-axis shaking table built in 1967 at the University of
Illinois by the Los Angeles Company, Ormond [16, 17], owed something to that devised by Clough
and Pirtz [13] 10 years earlier, in that its connection to the rigid base was through an arrangement
of 16, 38 in thick steel plates that had flexure-joints at the top to the table and at the bottom to the
base, allowing a swaying motion of 5 in in one horizontal direction only. But a significant difference
was its use of a controlled electro-hydraulic actuatordescribed as a ram in Figure 13(a)to
drive it, instead of the pendulum used by Clough and Pirtz. The coupling between the 1212 ft
tablemade from a 38 in steel plate and 5 in I-beamsand the actuator was described as a flexure
link, having a 4 in diameter steel bar reduced to 1.75 in at each end, which, in principle, allowed
for the small vertical movement of the table as it moved horizontally.
A second, and very important, advance on the Clough and Pirtz table was that Illinois had
the benefit of advice on its control system (Figure 14) from Herb Johnson at the newly created
(1966) MTS Systems Corporation, which allowed input of three kindssteady-state motion (C2 in
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

205

Figure 12. (a) The water-supported shaking table at Jassy, Romania (1968); plan view of the table showing
pipework and 8 of the 16 jets(12) and (b) the water-supported shaking table at Jassy, Romania (1968);
table cross-section showing water nozzles (9) and polystyrene grid (11).

Figure 14), real or simulated earthquakes (C4), and an arbitrary wave-form (C3). Displacement,
velocity, or acceleration time-histories could all be input, but herefor the first timethe embarrassment of choice was resolved in favour of displacement because it was realized that displacement
must be controlled to avoid physical damage to the table, and an LVDT was attached to it to
allow comparison between input and out put displacements. Figure 13(b) shows the performance
envelope of the table, indicating maximum displacement, velocity, and acceleration of 2 in, 15 in/s,
and 7.0g, respectively.
As with all new shaking table developments, the crucial question to be answered was how closely
it could represent a real earthquake. It will be recalled that Clough and Pirtz had, faute de mieux,
relied on a comparison of spectral curves, but available to Sozen and Otani were records from
both the El Centro (1940) and Taft (1952) earthquakes, and both were used with timecompression
ratiosas is required for model testsof 2.5, 5.0, and 10.0, as well as corresponding response
spectra for 0, 5, and 10% critical damping.
For the actual performance assessment of the Illinois table, a concrete block weighing 4100 lbs
and approximately 1.52.5 ft in cross-section was securely attached to the table with its length
in-line with the actuator and its shortest side in contact with the table. The table itself weighed
5500 lbs. Many combinations of the three possible types of input, timecompression ratios, and
viscous damping were studied, as was acceleration as opposed to displacement input, although the
former was actually integrated twice in the control software in order to provide a displacement
command to the actuator. The important result of these last tests was that acceleration input gave
better results for low timecompression ratios, but when this ratio was greater than 5, displacement
input was preferable. Overall, it was concluded that the Illinois table . . . can reproduce in the range
of structural engineering interest, scaled acceleration histories with characteristics quite similar
(my italics) to those of measured earthquake records. This reserved claim was realistic, because
not only were displacement measured on the table itself used for this assessmentrather than the
actual input to the actuatorbut the test specimen used was so arranged, and of such a shape,
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

206

R. T. SEVERN

Figure 13. (a) Side elevation of the University of Illinois shaking table and (b) performance characteristics of the University of the Illinois shaking table. (Acceleration is plotted diagonally to the right
and displacement to the left.)

that it did not excite, to any major extent, the other five DOF, which would have attracted some of the
input energy and distorted its shape. Even so, we must accept the Illinois table as a major milestone
in shaking table developments, and in the broad field of earthquake engineering the research carried
out on it was a significant contribution to the development of displacement-based design methods.
5.4. The shaking table of Penzien and Rea
In 1965 a group at the University of California, Berkeley, led by Prof. J. Penzien, embarked on
an ambitious feasibility study of a 100100 ft table [18], which would have had 52 hydraulic
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

207

Figure 14. Input and control components of the University of Illinois shaking table.

actuators in total, 20 in each horizontal direction and 12 vertically. Its estimated cost was $19.55M,
with an annual running cost of $3M. At this time laboratories had been content for their shaking
tables to produce some semblance of an earthquake motion in one direction only, without concern
about inputs from the other five degrees of freedom, and only the most advanced had attempted
to control them by passive hydraulic or mechanical devices. For the proposed UCB table, it was
intended to take advantage of developments in control and computing technologies by introducing
for the first time an active control system, which measured what we now refer to as spurious
table motions, feeding them back to an electro-hydraulic control system that made appropriate
corrections to the input signals at all actuators. But the technology to achieve this was not available
in 1965, and even after a 2-year research programme, which involved the MTS Corporation, the
uncertainty of success caused the 100100 ft project to be abandoned. Instead, a 2020 ft table
was built at UCB (Figure 15), principally by Rea and Penzien [19] with MTS having the contract to
design the hydraulic actuators, their servo-valves, and the electronic control system, an experience
which led them, with Penziens involvement, to initiate their role as the main international supplier
of shaking tables, replicating the UCB table at the Kajima Corporation in Tokyo, at UNAM in
Mexico City, and many other laboratories throughout the world.
The UCB table, formally opened in 1972, was constructed from a combination of reinforced
and prestressed concrete to form a square plate 1 ft thick, stiffened by transverse ribs of width
1 ft extending 1.75 ft below the bottom surface of the plate (Figure 15). The three horizontal
actuators, each producing 50 kips and all acting in the same horizontal direction, were attached to
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

208

R. T. SEVERN

Figure 15. A diagram indicating the essential features of the 2020 ft UCB shaking table.

the central rib, and were positioned such that any yawing motion of the table could be resisted
by the control system. The four vertical actuators, each of 25 kips, were attached to the table at
the locations shown in Figure 15 by means of prestressing rods in 2 in diameter pipes running
vertically through the table on a 3 ft square grid. The pipes also served as attachment points for
the structure being tested. The actuators had a swivel joint at each end, of such a length that they
made a significant contribution to de-coupling the vertical and the horizontal motions of the table,
with further de-coupling provided by the electronic control system.
Although these vertical actuators were controlled to produce a common vertical motion, some
small pitch and roll motions did occur. These motions also occurred with a horizontal input, and to
keep them to a minimum two pairs of passive hydraulic devices were provided acting in parallel.
Each pair consisted of two actuator-type cylinders with the top chamber of each connected to the
lower chamber of the other. By this mechanism, a high resistance to pitching motion was achieved.
Because the input was an acceleration time-history a mini-computer was used, as a prelude to the
actual test, to obtain by integration the time-histories of both velocity and displacement in order to
check that neither of them would have exceeded the performance limits of the table. If they did, the
original time-history was factored in its intensity and the process repeated iteratively. Referring to
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

209

Figure 15, the tripartite performance plot shows an important feature of the UCB table, a relatively
high maximum velocity of 25 in/s, with maximum displacement and acceleration of 5 in and 1.5g,
respectively, achieved by a high oil-flow rate and large capacity servo-valves.
Another novel feature of the UCB design was the insertion between the table itself and the
surrounding 1580 kips reaction block, of a 2 ft wide strip of vinyl-covered nylon fabric. This
enabled a 1.5 psi air pressure to be maintained to support both the weight of the table and of any
testpiece (the maximum being 100 kips), which it carried. By this device, the four vertical actuators
could give the tablewithout testpiecea 1g vertical acceleration. In the horizontal direction, the
three actuators produced a 1.5g acceleration in the unloaded condition.
Although the table was able to produce motion in the vertical and the one horizontal direction
simultaneously, the capacity of the servo-valves was not sufficient for simultaneous achievement
of the two maximum values, it being assumed during the design stage that maximum vertical and
horizontal velocities would not occur together in an actual earthquake. It was used in a number of
unique tests on steel and concrete frames, and on the Ruck-a-Chucky cable-stayed bridge model
[20] where for the best comparison with the calculated values, the measured table motions were
used rather than the actual input values.
In the overall development of shaking tables, the control system of the UCB table was greatly
significant. It was supplied by MTS and consisted of controlling five of the six DOF, with the
sixthtranslation orthogonal to the horizontal actuator directionbeing controlled by a sliding
mechanism. Transducers were installed in each actuator to measure displacements and forces. From
these displacements, feedback signals representing the average horizontal and vertical displacements, as well as roll, pitch, and yaw, were derived on the reasonable assumption that the table
was a rigid body. The aim of the control system was to reduce the three rotations to zero, and
the two displacements to their specified values by an adjustment process that required considerable skill from the operator, and an approach towards the requirement that the limits on velocity
and displacement were not exceeded. Because this meant that the time taken for the test was
considerably longer than the duration of the recorded earthquake, the control process can be said
to be out of real time.
It is noted that UCB intended to provide three DOF motion by mounting a smaller one DOF
table on the larger one, specifically for research in soil dynamics, and also . . . to study non-linear
behaviour specimens. . ... but other developments overtook these intentions.
5.5. The development of a large shaking table in Japan, 19701986
By 1970, in Japan as elsewhere, the developments in electronics and computing had produced servohydraulic-electronic control of actuators. At universities, private companies and research institutes
many small- and medium-sized tables had become available, driven by such actuators. The largest,
at 1515 m, was constructed by the National Centre for Disaster Prevention in Tsukuba Science
City [21], capable of either horizontal motion of a testpiece of 500 tons, or vertical motion of one
of 200 ton. Its frequency range was 0.150 Hz., maximum displacement, velocity and acceleration
being 60 mm, peak-to-peak 30 cm/s, and 0.55g for horizontal motion and 1.0g for vertical motion,
respectively. The primary motive for building this table, at a cost of $3M, was soil and foundation
research after the 1964 Niigata earthquake, which caused many instances of soil liquefaction.
Figure 16 shows the essential components of this table, which was composed of a grid of
I-beams set at 1 m pitch with a covering plate 25 mm. thick. Its maximum depth was 2.7 m., with
four actuators in the vertical direction and four in one horizontal direction. Because this was the
largest shaking table to be built up to this time anywhere in the world, an extensive preliminary
study was made, using a 1:12 scale plastic model, and a 1:2 steel model specifically to study local
behaviour in the region of the vertical actuators, with model studies being accompanied by finite
element analysis, then in its early stages of development. Referring to Figure 16, four balance
cylinders support the weight of the table including the weight of any testpiece, and 12 hydrostatic
bearings support it by oil pressure, while the function of the guide rollers was to restrict swaying
and/or turning motion of the table, which is taken to mean translation orthogonal to the horizontal
actuator direction, and the yaw rotation. It would seem to be clear, however, that control as
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

210

R. T. SEVERN

Figure 16. A diagram showing the components of the 1515 m shaking table built by Mitsubishi
Heavy Industries Ltd. in 1970.

we now understand the termand as was being developed in the UCB table at about the same
timewas not an element in the design of this Japanese table, reliance being placed on mechanical
systems to remove unwanted displacements.
At the end of that decade, and also in Tsukuba, the Public Works Research Institute constructed
a 68 m. table costing $20M, having a maximum load capacity of 100 tons, with maximum
displacement, velocity, and acceleration of 150 mm, peak-to-peak 60 cm/s, and 1.7g, respectively.
The same Institute constructed four tables, each 22 m, for what is now referred to as multiple
support excitation, in this case for applying different inputs to the piers of a long-span bridge.
As previously noted, many Japanese shaking tables are to be found in laboratories owned by
large industries, one such being at the Nuclear Power Engineering Corporation at Todotsu, Shikoku,
completed in 1982 at a cost of $200M and a monthly operating cost of $1M. The table is 1515 m
with a maximum load capacity of 1000 tons. It can be excited in one horizontal and the vertical
direction simultaneously, with maximum peak-to-peak displacement of 400 mm horizontally and
200 mm vertically. The maximum velocity was 75 cm/s horizontally and half that vertically, with
corresponding acceleration of 1.84 and 0.92g in a frequency range 030 Hz.

6. THE DEVELOPMENT OF SIX DOF CONTROL

The construction of nuclear power stations in Europe, which began in the mid-1950s, caused
an interest in shaking table construction, manifesting itself first at Julich, near Cologne, in what
was then the Federal Republic of Germany, where the SAMSON 55 m steel-box table was
built in the mid-1970s (Figure 17). Leaving aside the actuator whose function was to take the
deadweight of the table and testpiece, the Julich table was an innovation, in that one end of each of
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

211

Figure 17. The 6-axis SAMSON table at Julich, Germany.

its four horizontal actuators applied its force at a mid-side of the table, which meant that together
they could control yaw, as well as providing forces in the two horizontal axes. Also, there were
four vertical actuators, each one attached at a corner, thereby providing pitch and roll control, as
well as the vertical forces required in the tests. Thus, it is seen that 8 is the minimum number
of actuators required for shaking table testing that is able to exercise control over the six DOF,
and at the same time utilize the available actuator forces to maximum effect in the directions in
which they are most needed in earthquake engineering studies, that is, in the three translational
DOF. But this essential practical fact of eight (and sometimes more) actuators controlling six DOF
introduces two significant extra steps in the control system itself, the first being described as the
inverse kinematic problem in which the required values of three translations and three rotations
are converted into eight instructionsone for each actuator. The computationally more difficult
converse requirement, described as the direct kinematic problem, converts eight recorded actuator
values into the six displacement components. The Julich table incorporated both these innovations,
but in the mid-1970s the time taken for the calculations was such that the control process was
out-of real-time and also required that the properties of the testpiece did not change during the
test, for example by any non-linear behaviour.
A political decision by the German Government to abandon nuclear power generation meant that
the Julich facility saw little use and was dismantled after a few years. It was, however, copied in
principle in the early 1980s when the Bristol University table was constructed, the main differences
being a smaller size (33 m), the dead load of the specimen being taken by the vertical actuators
themselves, and a table constructed of four equal sections of cast aluminium bolted and glued
together in the shape of an inverted pyramid, giving it a high fundamental natural frequency and a
higher-performance ratio of testpiece to table weight. However, as previously noted, a high value
of this ratio is not always an advantage, and has sometimes to be lowered by adding deadweight
to the table, particularly when inelastic behaviour is being studied.
Despite the German withdrawal, other European countries pressed ahead with nuclear power
programmes requiring shaking table test facilities. In Italy, the ISMES (now CESI) laboratory at
Bergamo began using a 44 m, six DOF table in 1984, supplementing it with a 42.5 m, one
DOF steel table in 1986, and at the Technical University of Athens a 44 m steel table began
operation in 1986. In Italy also, ENEA commissioned MTS to provide it with a 44 m, six DOF
aluminium table, which was ready to begin tests in 1986.
Without doubt, the most important European shaking table installation is at Commissariat a
lEnergie Atomique (CEA), Saclay, near Paris. It consists of four tables with sizes 3.13.8 m (one
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

212

R. T. SEVERN

DOF), 22 m (one DOF), 22 m (two DOF), and a 66 m six DOF aluminium table. Like all
other European tables, it had the linear, fixed-gain PID (proportional integral derivative) type of
actuator control until the early 1990s.
For the above description of six DOF control, the use was made of European activities, but it is
important to note that parallel advances were being made in other parts of the world, notably in the
U.S. and Japan, the continuation of which has eventually led to real-time active control systems,
but such developments are outside the scope of this paper.
ACKNOWLEDGEMENTS

The author wishes to acknowledge first the invitation from the Editor, Prof. Anil Chopra, to write this
paper, and then to Professors Joseph Penzien, Mete Sozen, Chuck James, and Masayoshi Nakashima for,
respectively, providing details of early U.S. and Japanese shaking tables. We have made appreciable use
of the information, which they provided. We must also thank the libraries of the University of Bristol
and of the Institution of Civil Engineers for searching out many references and for obtaining copies of
papers in the Bulletin of the Seismological Society of America from the British Library. Our colleague
Dr Adam Crewe gave us unstinting help with the illustrations, as did Prof. Gabriela Atanasiu with details
of Romanian shaking tables.
As with the history of development of any subject, the information presented here can be neither
complete nor totally accurate, although we have tried to make it so. If readers would kindly alert us to
errors and omissions we shall receive them with pleasure.
While the historical character of the illustrations has been preserved as far as possible, where components
have been referred to in the text, for the sake of clarity enlargement of the lettering used has been made.
REFERENCES
1. Ruge AC. A machine for reproducing earthquake motion direct from a shadowgraph of the earthquake. Bulletin
of the Seismological Society of America 1936; 26(3):201205.
2. Ruge AC. The determination of earthquake stresses in elastic structures by means of models. Bulletin of the
Seismological Society of America 1934; 24(3):169230.
3. Stoten DP, Gomez EG. Adaptive control of shaking tables using the minimal control synthesis algorithm.
Philosophical Transactions of the Royal Society of London, Series A 2001; 359(1786):16971723.
4. Ogawa N, Ohtani K, Katayama T, Shibata H. Construction of a three-dimensional, large-scale shaking table
and development of core technology. Philosophical Transactions of the Royal Society of London 2001;
359(1786):17251752.
5. Muir-Wood R. Robert Mallet and John Milneearthquakes incorporated in Victorian Britain. Journal of
Earthquake Engineering and Structural Dynamics 1988; 17(1):107142.
6. Rogers FJ. Experiments with a shaking machine. The Californian Earthquake of April 1906, Report of the State
Investigation Commission, vol. 1, part II, 1906; 326335. Reprinted in Bulletin of the Seismological Society of
America 1930; 20:147159.
7. Hoskins LM, Jacobsen SL. Water pressure in a tank caused by a simulated earthquake. Bulletin of the Seismological
Society of America 1934; 24(1):122.
8. Jacobsen SL. Experimental study of the dynamic behaviour of models of timber walls. Bulletin of the Seismological
Society of America 1930; 20:115146.
9. Jacobsen SL. Motion of soil subjected to a simple harmonic ground vibration. Bulletin of the Seismological
Society of America 1930; 20:160195.
10. Jacobsen SL, Lydik S, Roberts SA. Hydrodynamic experiments with rigid cylindrical tank subjected to transient
motions. Bulletin of the Seismological Society of America 1951; 41:313345.
11. Monobe N, Takata A, Matumura H. Seismic stability of earth dams. Second World Congree on Large Dams,
Washington, DC, Question VII, 1936.
12. Oberti G. Development of aseismic design and construction in Italy by means of research on large model tests.
Proceedings of First World Conference on Earthquake Engineering, Berkeley, CA, 1936.
13. Clough RW, Pirtz D. Earthquake resistance of rockfill dams. Transactions of the American Society of Civil
Engineers 1958; 123:792816.
14. Muto K, Bailey RN, Mitchell KJ. Special requirements for the design of nuclear power stations to withstand
earthquakes. Nominated Lecture to the Nuclear Power Group, Institution of Mechanical Engineers, 10 October
1962; 146.
15. Sesan A, DAlbon G, Congradi I, Strat L. A new shaking table used to test structure at seismic actions. Bulletunal
Institulu Politehnic Din Iasi, Tomul XV, 1969.
16. Sozen MA, Otani S. Performance of the University of Illinois earthquake simulator for reproducing scaled
earthquake motion. Proceedings of U.S.Japan Seminar in Earthquake Engineering, Sendai, Japan, September
1970.
Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe

THE DEVELOPMENT OF SHAKING TABLES

213

17. Sozen MA, Otani S, Gulkan P, Nielsen NN. The University of Illinois earthquake simulator. Proceedings of
Fourth WCEE, Santiago de Chile, vol. 111(Session B-5), 1969; 140149.
18. Penzien J, Bouwkamp JG, Clough RW, Rea D. Feasibility study large earthquake simulator. Report No. EERC-67-1,
University of California, Berkeley, 1967.
19. Rea D, Penzien J. Dynamic response of a 2020ft. shaking table. Proceedings of Fifth WCEE, Rome, vol. 2,
June 1973; 14471456.
20. Godden WG, Aslam M. Dynamic model studies of Ruck-A-Chucky bridge. Proceedings of the American Society
of Civil Engineers (Structural Division) 1978; 104(ST12):18271844.
21. Sawada K et al. Strength and vibration characteristics of large type shaking table. Technical Review No. 3,
Mitsubishi Heavy Industries Ltd., September 1970.

Copyright 2010 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2011; 40:195213

DOI: 10.1002/eqe