Professional Documents
Culture Documents
.-..
... '
P~per
~t al
;:,rn1u1
:1,II
j.
'~,
~f
Professor }\.E.
Vardy, BSc, PhD, CEng, MICE, MASCE
Wolfson Bridge Research Unit, University of Dundee
~lf
synopsis
r:
Introduction J
Since 1676 1, there have been attempts to determine how arches support
the loads they[~re called on to carry. In the case of masonry arch bridges,
the arch ring rr\ay be greatly stiffened by spandrel walls and by fill between
II
the pavement surface and the arch. Nevertheless, these features are by no
means decisivJin poorly maintained structures, and engineers usually base
judgments onjlthe capacity of the arch .alone (or on the arch and known
fill material)
In many areas of structural destgn, attention is focused primarily on the
behaviour of ~!structure at ultimate load. Serviceability criteria are evaluated
1
independentlYt and are assumed to have no direct bearing on the safety of
the structure,ieven when they lead to more onerous requirements. With
masonry consthiction, however, and with old masonry bridges in particular,
this may not ge a sound approach. Certainly, there is a need to know the
ultimate cap~~ity of a structure, but this may be strongly influenced by
behaviour ~t Jruch smaller loads. Alexander & Thompson2 , for example,
drew attention to the consequences of the continual opening and closing
'"
.
of crack.
s between
a d'Jacent voussoirs. In effect,
the most pressing need
may be for a fuethod of defining and measuring serviceability limits rather
11
than ultimatel:limits.
ll:
~noadsu
..
{'
::
ends. As loads were applied, tension developed in certain areas. THe crosssections at these locations were reduced to eliminate the cracked arhs, and
successive analyses were obtained, approaching a final soluti~b in an
rfac;e
I
'
et al
300
(a)
500
i!
.t
~{
/:1'
i:
!/
.t
(c)
(d)
iterative Lanner. The necessary calculations must have been tedious, but
the meth~d is potentially sound in principle, provided that proper account
is takenilof abutment movements (to which it is highly sensitive). Its
disadvantage today lies in the difficulty of incorporating interactions with
the fill. Hughes' has. developed a microcomputer-based analysis, treating
the fillfs ail elastic medium. -
.
in use today are
The Jlost
popular methods of analysis
#
.
-the MEXE method
-the m~chanism method
-finite1~Iemeni methods ,, "'
The K.fEXE method is based on Pippard's8 use of Castigliano's
theorenl~. It has a sound theoretical basis, but it is unduly sensitive to
parame~ers whose values cannot be determined accurately in many cases,
especiaiiy with existing structures.
The fuechanism method, .revived by Heyman\ is elegant and easily
underst'i>od. The method enables the locations of the four hinges shown
in Fig 2(d) to be determined and gives the load at which these will be formed.
Heyma~ did not claim that the method would provide an accurate prediction
of the failure load, although this would naturally be possible if all the
necessJ\-y parameters were known with sufficient accuracy. Instead, he
propos~d that the safe load should be deemed to be the ultimate load for
an arcij of the same shape, but half the thickness. Harvey 9 proposed an
alternalive interpretation of the same analysis, i.e. determining the minimum
thickn~ss of arch rlng capable of supporting the known (factored) applied
1
~~~
'
'
s8
,.
'
',
'
Thr~~e-hinge. analysis
''
'
'
intrados at the springings and at the extrados between the crbwn''and the
load, but anywhere in principle. The vertical and horizonta!Jomponents
of the thrust at any hinge can be determined by static equilibrium. For
example, by taking moments about each of the hinges A and ~in Fig 3(a),
:!
the components at C are found to satisfy
:r
~{v(x-x~)}+
:::{h(y-yA)}- V,(x,-xA)-H,({-;yA)=O
il .... (I)
' '
do~nwards)
Paper: Smith! et al
~t
11
-11-
r,1!;
II
I[
reveale~
after collale
some additional backing (haunching) near the
springings.ii
A 750mm-wide line load was applied at a single quarter point across
the roadwa~ (between the spandrels) by means of hydraulic jacks acting
on steel rods anchored in the. bed of the canal. Deflections and cracking
at successive load increments
until failure occurred
at
a load
were record~d
-lh '
.
'
.: . " '
of 140kN!J:h width across the roadway. Only the latter parameter can be
compared directly with the proposed analysis.
Fig 4 shors computer simulations at successively increased loads, based
on the abo~e analysis. With dead load only (Fig 4(a)), the hinges are at
the crown ~hd the springings. At a relatively small load of lOkN/m width
(Fig 4(b )), t'he central hinge has moved towards the load, thus highlighting
the concerl expressed by Alexander & Thompson 2 about the successive
opening an'U closing of joints between adjacent voussoirs. At a load of
32 kN/m width (noi shown), the hinge has reached the load point, and this
configurati~n obtains until the right-hand hinge begins t~ move along the
' the springing at loads in excess of about 40 kN/m width.
arch, away)from
As the loa~;increases, the hinge moves successively further to the left and
the line of thrust at the right-hand springing moves away from the intrados
'
\7"''
'1
J '
When the load is increased further, the line of thrust at the right"hand
springing eventually reaches the extrados. At this stage, a fo~rth hinge
forms, the arch ring becomes a mechanism, and collapse must follow. In
this particular example, the formation of the fourth hinge o'tcurs at a
predicted load of 140kN/m width which is in agreementJwith the
experimentally measured collapse load.
!
Since the underlying methods of analysis are identical, the tftree~hinge
and four-hinge (mechanism) analyses predict exactly the same ~ond,itions
at failure. The advantage of the three-hinge method is that il provides
information about the behaviour of the structure prior to fail~re, Both
methods depend on the assumed load distribution around the.textrados,
and the three-hinge predictions prior to failure also depend on the
assumption of minimum thrust at the springings. Nevertheless, t~e aii'alyses
are approximately valid and, at the very least, provide excellent ~ualitative
information. By inspection, the behaviour at low loads differsJmarkedly
from that at high loads.
~~
,
. Two featur~s of t_he presentation o~ Fig 4 merit com~e~t. wrstly,_ the
!me of thrust IS depicted as a double !me - although this 1s n0t obviOus
at low loads. This is a consequence of extending the analysis tJ allow for
local crushing of the masonry. The necessary details are explaihed :in the
1'
1 '.
I
'I
;f
I
(a)
140
li
I
:t
i'
,,
:II
'
'I'
PapenlhI' Smith et al
I!
!i
It
~~----]\
~I
'II
;[
------=
l1
I'
I,I
'I
f
'r,
-::-----r---=x_----~
1_'[-----
:1.
foliowin~
.-
~I
=- ;F_
1 ou
....
H .:
(3)
1 denotes the local thrust and a" is the assumed crushing str~s~:
in whicnlt
This is considerable simplification of the true state of stress, but it is
much bdtter than assuming infinite stresses.
The v~lue of a" used will depend on the needs of the analyst. In most
cases, it ~}viii be sufficient to accept the characteristic values derived from
graphs in the Code of Practice for the bridge assessment BD2l/84 12, and
apply sJltable partial factors of safety.
The dbuble
lines of thrust in Fig 4 and in subsequent figures depict the
,,
t depth ofilmaterial which would be required locally to form a hinge capable
_ -' of transrhitting the local thrust. They provide a visual means of interpreting
,--...
'II
'II
II
Ji
~~
.lr
11,
j[
.!
370
200
IIIi
I
I
II
il
the behaviour of the structure under increasing load (or varying loads).
By inspection, one effect is to decrease the effective ring thickpess; :stability
exists only when both lines are contained wholly''within the ring:iThis reduces
the load at which a fourth hinge forms, causing the structure to become
a mechanism.
t.
i.:~'""'"-~-~.,
Paper:
~'
lI
'!
::
Fig 7.
r
il
,.
'
I
II
,j;
'I
r
L
,I
[.
1,
(a) Before movement
'
,[1
.I:
.I
11
:.II
:11
..
:II
.... -r .
., .
-1~
J'
n..,
-lie
---~'~ 'L--~--
L! _ _ _ ---...l!.o.!-- ----L!-1....
- - - L - ------1-...1
'
'
,'
' '"'
'
-~
(
as two rigid bars, LA and AR. If the abutments L and R moyhe apart, the
new position of the intermediate hinge must lie on the arc AI\ centred on
L and on the arc BB centred on R (Fig 8(b). These arcs ha~e a unique
inters~ction point an? so thene~ position of the ring can b~tdete.rm.ined
'i!i :_
by a ~1mple geometnc.constructiOn:
This effect has been mcorporated mto the authors: computer programs,
which allow movement of the abutments in proportion to th~~thr~st. The
iterative procedure for each increment is as follows: .
1l ':
(i) determine the line of thrust assuming no additional movement;
(ii) using the predicted thrust at the springings, evaluate the nbw positions
of the abutments and hence the new position of the interm'~diate hinge;
(iii) repeat steps (i) and (ii) as often as necessary (usually dbly once)
In practice, spreading of the abutments usually causes at :feast one of
the outer hinges to move away from the springings and into'ihe ring. To
simulate this case, it is necessary to know the rotational compbnent of the
abutment movement as well as the linear displacements. Thl~ is because
the portion of the ring beyond the outer hinge effectively acts':~ompositely
with the abutment itself. The programs assume that this portiC:h of the ring
undergoes the same linear displacement as the abutment and also rotates
.
about the springing.
'