You are on page 1of 2

viewpoint: harvey

Arch bridge assessment:


an issue of confidence
Bill Harvey argues that it is time for a thorough Bill Harvey
(pictured right)
reappraisal of the assessment process for arch has studied arch
bridges for nearly
bridges and its codification 25 years. He left
his academic

I
n setting out my views on arch which are intended to be structural, career in 2000 to
bridge assessment I hope to stimu- some not. All the elements are rela- become a
late discussion and possibly argu- tively stiff in some senses and therefore consultant and to
ment. Most importantly, I hope to become structural as the bridge develop and sell a
break through the complacency that responds to load. For example, the arch new version of the
surrounds some aspects of the process, itself resists compression round the Archie programs
without increasing the fears of engi- ring, and shear round the ring and for arch bridge
neers working in the field. I believe across the width. It has relatively little analysis (Email
that many engineers, and indeed code resistance to bending. The natural him at
drafters, are unreasonably confident in response of an arch to concentrated bill@obvis.com).
some assessment tools and in the test load is to sway but not to shorten. A much more
results which are used to support them. The load is typically applied from a detailed mented by general knowledge of struc-
I acknowledge that, as the proprietor vehicle which is itself a structure and discussion of arch tural behaviour and analysis, of arch
of a tool widely used in the analytical which modifies the way its weight is bridge test results bridges and of construction practice.
element of arch assessment, I have a applied to the bridge. That weight is can be found at Particular knowledge, with varying
vested interest in the issues I am about applied to surfacing which has a www.obvis.com levels of confidence and expense, can be
to raise. I raise them in the firm belief distributing as well as a weatherproof- achieved through inspection, measure-
that the concerns are justified. ing function. Under the surfacing is ment, reference to history and invasion
usually some form of earth fill. The fill of the structure.
What is assessment? is typically contained between spandrel
Let me begin by saying that I believe walls which stand on the arch. The arch Understanding – One may hope that an
that assessment is an art. It is a good is supported on piers and abutments assessor will understand the likely
example of the definition of structural which have foundations. Because the behaviour of the bridge and use that
engineering that has featured so much arch generates thrust, the abutments understanding as a basis for calcula-
in Verulam over recent weeks - ‘The art push both down and out into the tion. In fact, as with most structures, it
of modeling materials we do not wholly supporting ground. It is unlikely that is more realistic to hope that the asses-
understand, into shapes we cannot the ground is stiff enough to ensure sor has a reasonable grasp of the limits
precisely analyse, so as to withstand that a realistic analysis can treat the of knowledge of the bridge and the rela-
forces we cannot properly assess, in abutments as rigid. The presence of tionship between analysis and behav-
such a way that the public at large has wing walls, with foundations of their iour or performance.
no reason to suspect the extent of our own, will modify the effective stiffness
ignorance.’ of the foundation across the width of an Analysis – Since the information avail-
The most important aspect of the abutment. The contributions of fill and able is very limited, the value of calcu-
practice of the art is that we must spandrel walls to resisting sway move- lation is necessarily bounded. The aim
satisfy our clients without compromis- ment are almost impossible to quantify. of calculation is to increase confidence.
ing ourselves. My concern is that we The internal geometry of a bridge is If engineers have a model for some
are all too eager to accept glib often different from that expressed in potential load paths and are confident
pronouncements in a set of codified the external elevation. A flat spandrel that they can trust that model, they
rules. wall and a constant thickness ring edge may be prepared to venture a little in
Assessment is the application of may conceal a tapered ring, solid judging the benefits which can accrue
knowledge, understanding and judg- masonry backing topped with fill, inter- from load paths which are clearly
ment to confidently predict the future nal spandrel walls capped with slabs or present but less calculable.
performance of an existing structure. with arches, or even a mini viaduct
The performance required is that it standing on the arch. Judgment – If the engineer knows that
does not deteriorate excessively under a bridge has been carrying for many
the application of approved loads. The basis of assessment years past similar traffic to that which
Assessors must achieve a measure of A good assessment will be based on is expected in the future, the assess-
economy in not underrating bridges but knowledge, understanding, calculation ment will begin from a standpoint of
they must be confident of safety if they and judgment. Because there is so confidence. That confidence may be
are to sleep at night. much that we cannot economically manifested in the way an inspection is
discover, the most important aspect will carried out, in the way the results are
What is an arch bridge always be judgement. Good judgment conveyed into analysis and the value
An arch is a simple structure. A requires rational thought about the placed on the analytical results. When
masonry arch is more complex than limits of knowledge. the loads are increasing, as is nearly
some, but its lack of any substantial always the case in modern assess-
tensile strength modifies its behaviour Knowledge – An assessor will require ments, the comfort of history is no
in a way that simplifies analysis. as much specific knowledge about a longer available.
An arch bridge is somewhat more particular bridge as he/she can reason- Judgment must be exercised in
complex. It has many elements, some of ably acquire. That will be supple- considering the geometry and materials

5 March 2002 – The Structural Engineer|13


viewpoint: harvey

of the structure and the way in which moment the centres are struck. It is loads at which cracks were first
the various parts interact. What can be vain, however, to expect realistic predic- detected on all the bridges tested. Of
confidently predicted from inspection tions of deformation, and particularly course, the tests were carried out on
and experience and where must geome- deformation-related interactions, from real bridges, with spandrel walls and
try and material properties be meas- a model based on rigid abutments. many other features not included in the
ured? If measurements are made Because longitudinal cracks often analysis.
locally, what confidence do we have that form in an arch at the inner face of the Here, then, we have a basic model
they are valid generally? If we ignore spandrel walls, analysts are required to which is unrealistic. It does not repre-
certain aspects of behaviour in analysis ‘Authors ignore the stabilising effect of the span- sent, except in the most superficial way,
can we be certain that we will not over- drels. Evidence from tests suggests (but the structure it purports to model. To
estimate capacity? of the is far from proving) that the spandrel that model has been applied an unreal-
walls are so much stiffer than typical istic failure criterion. The result is
Analysis – The term analysis is a diffi- standards soil fill that the main load path in a widely reputed to be conservative and
cult one. It involves developing or sound arch involves the spandrels, economic. Traffic loads are increasing.
selecting a model for the structure then have failed which effectively prevent sway defor- Nearly 70 years of happy experience
applying appropriate numerical values mation of the arch under large concen- with MEXE may give an entirely false
to it. If the model itself, and some of the to trated loads. sense of security. Belief in MEXE is
numerical values, are effectively A 2-Dimensional model has very deep, though. In Ireland the allowable
concealed from the engineer, any confi- distinguish limited value in predicting failure of an axle loads derived from MEXE have
dence developed may be false. arch, though it should be possible to recently been doubled for sound
The range of models available now is between ensure that it provides a lower bound bridges. This act, in itself, underlines
very broad. Hand calculation may be to the bridge capacity. In doing so, the the arbitrary nature of the results
sufficient to deliver a thrust line analy- analysis model may imply, or predict, move- obtained.
sis. Some complexity can be added ments which could only take place if
using modest computing power. Elastic, and the spandrel walls were effectively Value of load tests on real bridges
elastic-no-tension and elasto-plastic detached. Ignoring stiffer load-paths is It should, by now, be clear that I have
models are available for the arch ring. assessment’ only dangerous if failure of that path grave doubts about the use of load tests
The fill may be modelled in a variety of would lead to collapse of the structure. to calibrate limited models. An analyti-
ways, simply as dead weight or as a Careful application of parameters in a cal model can only yield a capacity for
responsive medium contributing to the 2D analysis will ensure safety. the model not for the structure.
support of the arch as well as loading Engineers must aim to deliver a model
it. At the limit, discrete element models My problems with MEXE which gives a safe result.
can follow a loading cycle through The analysis on which MEXE is based Test results from physical models
failure of both the fill and the arch treats the arch as a two-hinged elastic which closely mimic the analytical ones
within the limits of the chosen parame- parabolic arch with rigid abutments. can be valuable in pinpointing faults in
ters. The fill is treated as vertical dead load. the development of the analytical
The assessment scheme commonly Live load is applied as two axles 1.8m model. Tests on real structures cannot
referred to as MEXE is not, strictly, an apart placed symmetrically on the sensibly be compared with limited
analysis but rather an amalgam of span. The failure criterion ignores analytical models, except to increase
computation and judgment set down tensile stress anywhere in the arch, no confidence in the conservatism of those
many years ago. It has accrued a large matter how high it might be. models. If a model that is thought to be
and faithful following. Compressive stress is considered only conservative produces results close to or
at the crown and the limit set is above those in a test where alternative
Interface between analysis and 200lb/sq in. I have quoted the old impe- load paths were known to be present,
assessment rial measure originally used, because it the model cannot be considered safe
In British practice, the constraints of gives a realistic impression of the accu- until the anomaly is explained. Even if
the Highways Agency Standards mean racy attributed to the calculation. The all the available test results are
that some aspects of behaviour are authors of MEXE did not mean 200 substantially higher than the equiva-
perforce ignored when using some rather than 200.1, but rather, at best, lent analytical ones, there is no guaran-
analytical tools, while they are effec- somewhere between 220 and 180. The tee that the next test would not reveal
tively taken into account in others. This SI form of 1.4MPa gives an impression flaws. The best an engineer can ever
is, at least partly, because the authors of much greater precision. hope for is a measure of confidence.
of the standards have failed to distin- The actual value used is less impor-
guish between analysis and assess- tant than the underlying criterion and Confidence – how do you sleep at
ment. the reason behind the value. There is night?
Abutment movement is not specifi- no good reason why an arch should fail Confidence is an incremental thing
cally excluded from analysis, but its when the compressive stress reaches which develops over time.
exclusion is implied by the recommen- 1.4MPa. Cracks usually will have been Understanding the model in use and its
dation that abutments are assessed by present in the arch before loading limitations will generate a real sense of
inspection only, unless the inspection began and certainly will have opened confidence. Experience of inspection
raises cause for concern about perform- by the time the live load approaches a and assessment and a study of history
ance. In a typical arch, very small safe limit. add to the feeling of confidence with
movements of the abutment will have a A series of large and expensive tests which assessments are delivered.
dramatic effect on peak stresses (espe- were used to calibrate the MEXE Regular use of particular models, espe-
cially tensile stresses) in the arch ring. method. A first, rational, limit was set cially if they offer some form of pictorial
They will, however, have negligible using a no tension rule in the analysis. output can enhance understanding of
effect on the short term load capacity, Real bridges, tested with much heavier why one bridge might be stronger than
and normally do not materially affect loads than predicted, showed no another. A model which is a black box,
the ability of the arch to withstand conspicuous cracks. A middle half rule even if the box carries a comforting
large range cyclic traffic loading. This is was then tried, but was also found to be label, offers only a false sense of secu-
because arches can crack and redistrib- too ‘conservative’. The load producing rity. An engineer must aim to learn how
ute load effects without suffering long 1.4MPa compressive stress at the a structure might behave and not be
term damage. Indeed, most small crown, ignoring all other stresses, was satisfied with a quick but poorly under-
arches have permanent cracks from the found to be lower than the actual test stood solution. se

14|The Structural Engineer – 5 March 2002

You might also like