Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.

com, ISSN 1743-3509

Models of Ancient Columns and Colonnades

Subjected to Horizontal Base Motions - Study of
their Dynamic and Earthquake Behaviour
G C. Manos, M. Demosthenous

Czw/. [/mvcrwfy,

1. Introduction
Ancient Greek and Roman structures
composed of large heavy members that
simply lie on top of each other in a perfect-
fit construction without the use of
connecting mortar, are distinctly different
from relatively flexible contemporary
structures. The colonnade (including free
standing monolithic columns or columns
with drums) is the typical structural form
of ancient Greek or Roman temples. The
columns are connected at the top with the
epistyle (entablature), also composed of
monolithic orthogonal blocks, spanning the
distance between two columns (figure 1). Figure 1 Typical ancient colonnades
The seismic response mechanisms that develop on this solid block structural system
during strong ground motions can include sliding and rocking, thus dissipating the
seismic energy in a different way from that of conventional contemporary buildings
This paper presents results and conclusions from an extensive experimental study that
examines the dynamic response of rigid bodies, representing simple models of ancient
columns or colonnades. These models are subjected to various types of horizontal base
motions (including sinusoidal as well as earthquake base motions), reproduced by the
Earthquake Simulator Facility of Aristotle University. The employed in this study rigid
bodies were made of steel and are assumed to be models of prototype structures 20
times larger. Two basic configurations are examined here; the first is that of a single
steel truncate cone assumed to be a model of a monolithic free-standing column whereas
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
290 Structural Studies, Repairs and Maintenance of Historical Buildings

the second is formed by two steel truncate cones, of the same geometry as the one
examined individually, supporting at their top a rectangular solid steel, representing in
this way the simplest unit of a colonnade. In thefirstpart, dynamic response results are
presented, from an extensive investigation employing one monolithic free-standing rigid
column These are next compared with predictions from numerical simulations that
employed software developed at Aristotle University. In the second part, the dynamic
response of two steel columns, supporting at their top a rectangular steel block are
investigate experimentally under simulated base motions.

2. Investigation of the dynamic response of a solid rigid body

The dimensions of the examined rigid body, a monolithic steel truncate cone, are shown
in figure 2. This specimen was placed on the earthquake simulator simply free-standing
on the horizontal moving platform of the shaking table. A very stiff, light metal frame
was built around the studied steel truncate cone in order to carry the displacement
transducers that measured the rocking angle; this metal frame also provided temporary
support to specimen during excessive rocking displacements indicating overturning.
The dynamic behavior of this specimen is
governed by its rocking response. The
experimental investigation of its rocking
response included periodic sine-wave
tests with a chosen range of frequencies
and amplitudes as well as earthquake
simulated tests using the El Centro 1940
record. A simple way to estimate the
dissipation of energy during rocking is to
obtain the value of the coefficient of
restitution by comparing free vibration
rocking response measurements with
those predicted numerically. From such
comparisons the coefficient of restitution
is obtained with the value that yields the
Figure 2. Dimensions of the best agreement between numerical and
steel truncate cone. measured response parameters.
The numerical solutions employed were verified for their stability and convergence
through an extensive parametric study that examined various methods of numerical
integration and time step. This correlation was next extended with measurements from
sinusoidal as well as earthquake simulated tests. The numerical studies in this case
employed the same base excitations recorded at the Earthquake Simulator during the
tests and used as values for the coefficient of restitution the ones determined from the
same experimental sequence and the best numerical solution that was found from the
above mentioned study (Demosthenous 1994)
2.1 Free Vibration Tests The sequence of tests for the solid truncate cone included
free vibrations tests, with the objective of assessing the coefficient of restitution of the
test structure. The following figure (figure 3) depicts the rocking angle response (solid
line) from such a free vibration test. Supenmposed on this graph with a dash line is the
numerically predicted rocking response, with a coefficient of restitution value that gave
the best fit to the experimental measurements. The predicted rocking response was
derived from the linearised equations of the rocking motion (Aslam et al 1980). Certain
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Structural Studies, Repairs and Maintenance of Historical Buildings 291

discrepancies are apparent at the initial large amplitude rocking stages. This must be
attributed to three-dimensional (3-D) rocking (Koh 1990). since rocking in a perfect
plane for a 3-D model requires exacting initial conditions and the slightest out-of-plane
disturbance will cause rocking to be 3-D
Truncate Cone (Solid) When the specimen tested is
Initial non-dim angle = 0.475 prismatic instead of truncate
cone the 3-D response during
rocking is minimized and good
correlation can be obtained
coef. of restitution ( e ) = 0.982
between observed and predicted
: O 5 TIME (s«c) 10 behavior thus assessing more
Figure 3 Measured and Predicted accurately the coefficient of
Free Vibration Rocking restitution (Manos and
Demosthenous 1990).
a) max. non dim angle : 0.012 CONE (solid) 2.2 Sinusoidal Tests During
these tests the frequency of
motion was varied from IHz to
7Hz in steps of IHz from test to
max. base disp. : 4mm FREQ. EXCITATION 3Hz test. This resulted in groups of
tests with constant frequency for
b) max. non dim. angle : 0.209 CONE (solid)
the horizontal sinusoidal motion
for each test. In the various tests
belonging to the same group of
constant frequency, the amplitude
max. base disp.: 4mm FREQ. EXCITATION 4Hz of the excitation was varied
c) max. non dim. angle : 0.104 CONE (solid) progressively from test to test
Figures 4a, 4b and 4c depict the
rocking response of the solid
specimen during this sequence of
tests for three different excitation
max, base disp : 4mm FREQ. EXCITATION 6Hz frequency cases (3Hz, 4Hz and
O 6 TIME ( sec ) 12 6Hz). The following points can
Figure 4 Rocking during sinusoidal base be made from the observed
excitations. behavior during these tests:
- For small amplitude tests the rocking behavior is not present; the motion of the
specimen in this case follows that of the base.
- As the horizontal base motion is increased in amplitude, rocking is initiated This
rocking appears to be sub-harmonic in the initial stages and becomes harmonic at the
later stages.
- Further increase in the amplitude of the base motion results in excessive rocking
response, which after certain buildup leads to the overturning of the specimen. At this
stage the rocking response is also accompanied by some significant sliding at the base
as well as by rotation and rocking response out-of-plane of the excitation axis
The above stages are graphically portrayed in the plot of figure 5 (together with the
numerical results as explained below). The ordinates in this plot represent the non
dimensional amplitude of the base motion whereas the absiscae represent the non-
dimensional frequency given by formulae (1). The following observations summarize
the mam points:
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
292 Structural Studies, Repairs and Maintenance of Historical Buildings

- The non-dimensional stable-unstable limit rocking amplitude increases rapidly with th -

non-dimensional frequency.
- For small values of the non-dimensional frequency the transition stage from no-
rocking to overturning, in terms of non-dimensional amplitude, is very small and it
occurs with minor amplitude increase.
A further numerical study has been performed, aimed to simulate the dynamic response
of the solid specimen, using the non-dimensional linear equations for sinusoidal
excitation as given by Spanos and Koh (1984) and with a value for the coefficient of
restitution as obtained from the free vibration tests. The numerical analyses have been
performed for various combinations of amplitude and frequency. Figure 5 presents these
results in a summary form. Fairly good agreement can be seen between numerical
predictions and the corresponding experimental measurements.
A = x / ( 6cr * g ),
H = co/p, (1)
p2 = W Re / lo
SINUSOIDAL BASE EXCITATION
A =Non-dimensional base TRUNCATE CONE LINEAR EQUATIONS
acceleration amplitude e = 0.982
experimental
x = Actual honz. peak acceleration
of the sinusoidal motion
6cr ^Critical angle(rad) indicating
overturning of the specimen
g =The gravitational acceleration
Q =Non dimensional frequency
to =Actual sinusoidal freq. (Hz)
p =The natural rocking frequency
of the specimen (Hz)
W =The weight of the block 2 3 4 5
Rc=The distance of the center of EXCITATION FREQUENCY
gravity from the rocking pole
lo =The mass moment of inertia Figure 5. Summary from numerical analysis

: EXPCI~i mental Max value : 8.589 Min value : -8 575

: Hunei*ical Max value : 8.685 Min value : -8 .597
Specimen : COME /' M.F.D.
e : 8.973
Method : RUNGE KUTTA
Non Dimentional Rocking Angle

Non 1 inear equat ions

A Constant time stc:p
1 ;'; .': (DT=8.8881 )
J
I! fJS !/| i\f\ .ftftft'j>_fl'JV f'1'Yi VjVA'riiijV^AlLf^ViWjO/A'SJM^iviiiiaW

f: EL CEMTRO 1948
Span : 2.5
max base accel = 8.568 g
r
J8 16.81
T I M E ( sec )
(jq co

6 Correlation of the predicted and measured rocking angle re sponse

 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Structural Studies, Repairs and Maintenance of Historical Buildings 293

A number of tests were performed during this

sequence with progressively increasing intensity, based on the 1940 El Centre
Earthquake record. The base acceleration and displacement response was measured
together with the rocking response of the tested specimen. The principal objective of
these tests was to observe again the stable-unstable behavior of the model. The base
excitations recorded at the Earthquake Simulator were used in the numerical simulation
of the earthquake response; the value for the coefficient of restitution was taken from
the free vibration tests. The full time history of the predicted rocking angle response is
compared with the corresponding measured rocking angle in figure 6. As can be seen
from this correlation, good agreement was achieved in this case by the employed
numerical simulation.

3. Dynamic Response of two model column with a top-block

17 Genera/ D2jrr%?fzbM The research effort presented before for individual free
standing rigid bodies (models of free-standing individual columns) was complemented
by this study aiming to investigate the dynamic response of two individual columns
supporting at their top a rectangular prismatic rigid block (representing part of the
entablature). This formation, shown in figures 7 and 8. is the simplest possible
structural unit that combines more than one free standing column. In this investigation
the examined structural unit was
formed of two model columns with the
axis of form of solid truncate cones with the
symmetry
same lower part, having dimensions
identical to the single column presented
Instrumentation in section 2. This lower part was
complemented at the top by an upper
» Horiontal displ.
part formed by a small truncate cone
and a prism (see figures 1. 7 and X).
522mm Vertical displ.
These lower and upper parts for each
CONE CONE column were not constructed
1 Accelerometer separately; instead they were shaped in
2
one piece from a single steel block
I SHAKING
BASETABLE resulting in two identical solid rigid
-215mni bodies. Each free-standing column was
522.5mm high and had a circular base
Figure 7. Studied structural formation and of 97mm diameter and a 103mm by
employed instrumentation 100mm almost square top.

These model columns were placed on the horizontal moving platform of the shaking
table at a distance of 215mm between their axes of symmetry. A rectangular prismatic
rigid block was placed in a symmetric way on top of these columns, as shown in figures
7 and 8; it was also made of the same steel that the two columns were made of and it
had a length of 215mm. a height of 68mm and a width of 87mm. This block was not
connected in any other way with the two columns but was simply free standing on their
top square sections. The longitudinal horizontal axis of this rectangular block coincided
with the vertical plane formed by the two axes of symmetry of the columns; its left
and right end sections (figures 7 and 8) were located at the center planes of the
left and right column, respectively.
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
294 Structural Studies, Repairs and Maintenance of Historical Buildings
3.2 Expected Modes of Response - Instrumentation
Instrumentation was provided in order to measure
the expected complex response of this structural
formation during a series of tests that included free
rocking vibrations as well as sinusoidal and
simulated earthquake excitations of the base. This
instrumentation is depicted in figure 7; it included
accelerometers at the top and bottom of each column
placed in a way as to coincide with the longitudinal
vertical plane of symmetry of this system. Additional
accelerometers were placed at the longitudinal
Figure 8. Rigid Block Formation horizontal axis of the rigid top-block at both ends
on the Shaking Table again at the longitudinal vertical plane of
symmetry. All these accelerometers had the purpose
of recording the acceleration response of the various parts of this system. Horizontal
base excitations used during this sequence with direction coinciding with this
longitudinal vertical plane of symmetry. However, an accelerometer was placed at the
mid-side of the rigid top-block perpendicular to this longitudinal plane of symmetry in
order to identify out-of-plane response modes of the block. Apart from the
accelerometers, displacement sensors were also used, as indicated with the arrows in the
same figure, aimed at recording the rocking displacement response of either the columns
or the top-block. Finally, the acceleration of the horizontal motion of the shaking
table moving platform was ui ui
also recorded. Figure 9a
depicts the ideal rocking
response of this complex a) 1 2
system. This response is 0-
derived with the assumption i If i!
that it occurs only in the 1 —
vertical plane of symmetry IDEAL ROCKING RESPONSE OF THE SYSTEM
and that there is no COLONNADE WITH 2 COLUMNS (TRUNCATE CONES)
differential sliding of the two b ) UPLIFT OF THE BLOCK VS ROCKING ANGLE OF THE COLUMN
Uplift over the 1st Column Uplift over the 2nd column
free-standing columns at the
base. Moreover, the top- "J
U1 ( O < 0 ) U2 ( 0 > 0 ) ..."•"
block is assumed to rock on
top of the columns, as shown P U
in figure 9a. without either L
sliding or losing contact with IF U2 (0 < 0 )
the supporting columns As a T U1 ( 0 > 0 )
result from these assumptions (mm)
both columns develop the
same rocking angle during all -Gcr ROCKING ANGLE +Gcr
stages of the response. Figure 9. Ideal Rocking Response of the System.
The rocking and uplifting response of the top-block can be obtained as a function of the
rocking angles of the supporting cones based on these assumptions. Figure 9b depicts
the ideal uplift displacements of the left and right ends of the top-block for rocking
angles of the columns from - Ocr to Ocr (where Ocr the critical rocking angle beyond
which each one of these two cones studied individually overturns).
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Structural Studies, Repairs and Maintenance of Historical Buildings 295
non dim. rocking angle of the 1st col. ( 0 = 8 / Gcr ) / max. 0.281, mln. - 0.236
Uplift of the block over the 1st col. (U1) / max. 4.45mm, min. -0.493mm

3.95 T I M E ( sec ) 12.88

Figure lOa Left-cone and top-block response (3Hz)
• non dim. rocking angle of the 1st col. ( 0 = 8 / Gcr ) / max. 0.037, mln. • 0.043
-Uplift of the block over the 1st col. (U1) / max. 0.223mm, min. -0.113mm

2.99 TIME
Figure 1 Ob Left-cone and top-block response (7Hz)
-Top acceleration of the 1st column / max accel. 0.32 g
Hor. accel. of the block over the 1st column / max. accel. 0.35g
c ) j SIN useMDAl. BASE EXCITATIC)N - FRE Q. EXC TATIC N 3.0 Hz
j /\ * A fl
1 \ ft \ i
I/ \ 1 f 1 I Plf
1 / V

'I

II 1 y
V V y
1 !\ r- V II ft 1)111 e \? ^111
'* T I M E ( sec ) 11.99
Figure lOc Left-cone and top-block response (3Hz)

j. j 5"z/z^ozWa/ 7^^ A limited number of results from this experimental sequence is

presented here and discussed Figures lOa to lOc depict the measured response during
this sinusoidal base excitation sequence. The peak base acceleration during this
sequence was kept equal to 160 gals The excitation frequency was varied from 3Hz to
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
296 Structural Studies, Repairs and Maintenance of Historical Buildings

7Hz in steps of IHz from test to test. From the measured response the rocking angle of
the left column (1) was derived and is depicted with a solid line in the plots of figures
lOa (3Hz). 10b(7Hz). In these figures the uplift displacement of the left end of the top-
block is also plotted with a dashed line. As can be seen, considerable rocking angle and
block uplift response develops when the excitation frequency is equal to 3Hz (fig. lOa).
However, when the excitation frequency becomes equal to 7Hz (fig. lOb) the observed
cone and top-block response becomes almost 10 times smaller. This observation agrees
very well with a similar conclusion reached from the extensive study of the rocking
response of the individual solid rigid columns (see section 2.2 and figure 4). Figure lOc
depicts the acceleration response for the left cone (1) and the left side of the top-block.
The base motion is in this case 3Hz sinusoidal excitation with peak base acceleration
160gals. As can be seen in this figure, the acceleration response of the top-block
exhibits strong correlation with that of the supporting cone. The peak block uplift.
measured during the sinusoidal test rocking stages, is plotted with asterisks in figure 11
against the ideal response curves described in section 4.2 (see also figure 9a and 9b). As
can be seen there is reasonable agreement between measured and expected values.
COLONNADE WITH 2 COLUMNS (TRUNCATE CONES)
UPLIFT OF THE BLOCK AS ESTIMATED FROM THE ROCKING ANGLE
Corellation with the Uplift from measurements
5.55
mm SINUSOIDAL BASE EXCITATION
U2 (0 > 0 )
U U1 (O < 0 )
p
L
I
F
T
(mm) U2 (G < 0 )
U1 ( 0 > 0 )

• 0.3 0 cr ROCKING ANGLE + 0.3 G cr

Figure 11. Observed and expected top-block response
non dim. rocking angle of the P' column (0) / max.. 0.065 min.-0.10
Uplift of the block over the 2"** column (U2) / max. 1.23mm
EL CENTRO BASE EXCITATION-SPAN 1 - max. base accel. 0.304g
a)

00 T I M E ( sec ) 8.05
Figure 12 Left-cone and top-block right displacement response (Span 1)
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509
Structural Studies, Repairs and Maintenance of Historical Buildings 297
non dim. rocking angle of the 1" column (0) / max. 0.54 min. - 0.30
Uplift of the block over the 2"^ column (U2) / max. 7.3 mm
EL CENTRO BASE EXCITATION-SPAN 3 - max. base accel. 0.935g

00 T I M E (sec) 8.05
Figure 13 Left-cone and top-block right displacement response (Span 3)
COLONNADE WITH 2 COLUMNS
UPLIFT OF THE BLOCK AS ESTIMATED FROM THE ROCKING ANGLE
Correlation with the Uplift from measurements
EL CENTRO BASE EXCITATION

(mm)

-0.6 0cr ROCKING ANGLE - 0.6 0cr

Figure 14 Observed and expected top-block response

3.4 Simulated Earthquake tests A number of tests were performed during this
sequence with progressively increasing intensity, indicated by the parameter Span.
based on the 1940 El Centro Earthquake record. The larger the value of the parameter
Span the more intense the simulated earthquake motion. Again the principal objective
here was to observe the top-block response relative to the response of the supporting
free-standing cones (see figures 12 and 13). Measured peak top-block uplift response
values are listed in Table 1 together with the peak base acceleration and the peak non-
dimensional rocking angular response for the columns. As can be seen from this table
the more intense simulated earthquake motions result in larger rocking angular response
for the columns and consequently to larger uplift response for the top-block. Moreover.
the columns together with the top-block exhibit rocking response during all the duration
of the intense earthquake motion, although this does not lead to unstable response for
this particular earthquake motion. Finally, figure 14 depicts the measured peak top-
block uplift (circles) against the ideal response curves described in section 4.2 (see also
figure 9a and 9b). As can be seen the correlation between measured and predicted top-
block response exhibits similar trends with the similar correlation that was observed
during the sinusoidal tests (figure 11)
 Transactions on the Built Environment vol 26, © 1997 WIT Press, www.witpress.com, ISSN 1743-3509

298 Structural Studies, Repairs and Maintenance of Historical Buildings

4. Discussion of the results - conclusions

a. The coefficient of restitution can be defined with sufficient accuracy from the free
vibration test results in combination with the performed numerical analysis, despite
complications from the three dimensional rocking, which was present for both solid as
well as sliced specimens.
b. The numerically predicted upper-bounds for the stable-unstable rocking of the solid
prismatic specimen, subjected to sinusoidal base motions, agrees well with the observed
behavior.
c. The behavior of the examined two columns with the rigid top-block subjected to
sinusoidal base excitations exhibits similar trends to those observed for the individual
column, with regard to the influence that the excitation frequency exerts on the rocking
amplitude. That is, for the same non-dimensional excitation amplitude, the higher the
frequency of the excitation the smaller the rocking amplitude.
d. For not very large rocking angles the observed response of the two columns with the
top-block system tends to remain in-plane. It also exhibits good correlation with the one
predicted under the assumptions of no differential sliding of the columns at their base
and no sliding between the top-block and the supporting columns. This observation is
valid for both the sinusoidal as well as the simulated earthquake excitations.
Table 1. Measured peak top-block uplift
during the simulated earthquake sequence.
Intensit Peak e/ Ul U2
yof Base 6cr (mm) (mm)
ACKNOWLEDGEMENTS : The
Base Ace. max max max
Motion partial support of the Greek
(8) mm nun min
Organization for Seismic Planning
Span 1 0.304 0.065 0.40 1.23
and Protection (OASP) as well as the
-0.100 1.18 1.00
European Commission (DIR XII ,
Span 2 0.614 0.183 1.39 3.28
-0 190 3.65 2.43 ENV.4-CT.96-0106, ISTECH),
is gratefully acknowledged.
Span 3 0.935 0.540 2 JO 7.30
-0.300 5.90 397
0 Non Dimensional Cone Rocking Angle
Ul Top-Block Left Uplift Displacement
U2 Top-Block Right Uplift Displacement
References
As lam . MM. et.all. 1978. Rocking and Overturning response of rigid bodies to
earthquake motions. Report No. LBL-7539, L B Lab. Univ. Of California. Berkeley.
Demosthenous M. 1994. Experimental and numerical study of the dynamic response of
solid or sliced rigid bodies. Ph. D. Thesis, Dep. Civil Eng . Aristotle University
Koh. A.S. and Mustafa G. 1990. Free rocking of cylindrical structures. J of Eng
Mechanics. Vol. 116. NO 1. pp.35-54.
Manos. G. C. and Demosthenous. M . 1990. The behavior of solid or sliced rigid bodies
when subjected to horizontal base motion. Proc. 4^ U.S. Nat. Confer. Earth. Eng, Vol.
3, pp. 41-50
Manos, G.C.. and Demosthenous, M. 1991. Comparative study of the dynamic response
of solid and sliced rigid bodies, Proc , Florence Modal Analysis Confer., pp.443-449.
Spanos, P.D. and Koh A.S., 1984 Rocking of rigid blocks due to harmonic shaking. J of
Engm. Mech.. ASCE 110 (11), pp. 1627-1642