A review of the UK masonry arch Proc. Instn Civ.

Engrs Structs &

assessment methods Bldgs, 1997, 122,

Aug., 305 – 315

T. G. Hughes, BSc, PhD, MICE, and M. J. Blackler, BE, PhD Paper 11302

Written discussion
closes 15 October
■ A critical review of current UK methods of (a) correctly predict the load carr ying 1997
masonr y arch bridge load assessment is capacity
presented. The paper considers the criteria (b) allow assessment at a reasonable cost
for assessment procedures and examines (c) correctly model the failure modes
the results of recent large-scale experimen- (d) properly consider those parameters which
tal programmes, and the information these influence load capacity
give on typical failure mechanisms. The (e) minimize the use of additional global
background to each of the methods of factors
analysis which are being used for assess- (f) provide additional appreciation of
ment of load carr ying capacity are pre- behaviour
sented. These being the modified Militar y (g) allow progressively more detailed investi-
Engineering Experimental Establishment gation of borderline bridges
(MEXE), pinned-elastic, mechanism and (h) model remedial work
cracking-elastic methods. The data (i) allow the modelling of ser vice loads
requirements for the dif ferent methods are (j) be suitable for improvement with addi-
considered. Each method is judged against tional knowledge.
the selected criteria and conclusions
drawn on their suitability for assessments.
This paper excludes any direct considera-
Arch behaviour
tion of the condition factor.
4. The principal components of an arch
bridge are detailed in Fig. 1. Arches are com-
Introduction plex three-dimensional (3-D) structures, fre-
Early predictions on the number of masonr y
quently skewed and composed of complex
arches likely to fail the national assessment
composite materials surrounded by fill. A sig-
programme prompted a review of the method
nificant proportion of bridges are of multi-span
of assessment of masonr y arch bridges. The
construction and a large number have been
structures are generally ver y old, have geo-
widened, at least once, as part of their histor y.
metries and materials which are poorly
The geometric requirements of the specific
defined and have been subjected to long-term
sites coupled with the variability of construc-
historical loadings and movements. Research
tion materials means that most arches are
efforts have focused on the monitoring and
unique structures.
then loading to failure of real arches supple-
mented by model studies, and in the develop- 5. Masonr y in the form of the arch ring
ment of methods of analysis.1 Considerable provides the main load bearing element to
arch bridges although the spandrel and wing T. G. Hughes,
progress has been made in understanding Senior Lecturer,
arch behaviour. walls contribute. The fill provides load dis-
Cardiff School of
2. The methods of masonr y arch assess- tribution, additional dead weight, restraint Engineering,
ment considered in this paper are the modified to the arch, and in some situations structural University
Militar y Engineering Experimental Establish- arching action. Because the spandrel walls of Wales, Cardiff
ment (MEXE) method,2,3 the pinned-elastic cannot be relied on, and because of the
method,2,3 the mechanism methods4,5 and the difficulty in 3-D modelling, current assess-
cracking-elastic methods.6,7 ment techniques consider masonr y arches as
two-dimensional (2-D) plane strain
structures.
Selection criteria for an assessment 6. There is no clear distinction in assess-
procedure ment methods between the behaviour of stone
3. It was considered necessar y to look at and brick bridges; the term masonr y being
the fundamentals of each assessment tech- applied to both. Stone masonr y and brick
nique, their methodologies and how they com- masonr y are similarly composed of generally
pared with actual structural behaviour. There stiffer, stronger blocks jointed by softer
are a number of criteria which can be weaker mortar. The significant difference in M. J. Blackler,
appraised to judge the suitability of the avail- their behaviour, in arches, is largely related to Associate, Special
able methods. In the authors’ view an assess- the frequent absence of inter-ring bonding in Services Division,
ment method should: the regular multi-ring brick barrel. Stone Mott MacDonald

305
 HUGHES ET AL.

Extrados of arch ring

Parapet

Fill

Wing wall
Spandrel wall
Springing

Abutment

Intrados of arch ring

Voussoirs
Fig. 1. General
arrangement of a
masonry arch bridge
showing the main
features

arches, by comparison, either have a single Table 1. Basic arch geometry and properties used in
ring or the inter-ring bonding allows no dis- study
cernible plane of weakness between adjacent Arch property Value
rings.
7. It is usual to assume that masonr y has Intrados span 10·000 m
zero tensile strength. This is a conser vative Rise at mid-span 2·500 m
but reasonable assumption particularly given Fill depth at crown (inc. road) 0·400 m
Arch thickness 0·400 m
the age of most masonr y bridges. Some of the
Arch shape Circular
more complex models can include tensile Arch elasticity 5.0 3 106 kN/m2
strength but it is considered unlikely that once Monitoring stress 10·0 3 103 kN/m2
a joint has started to open there will be any Arch unit weight 20·0 KN/m3
significant residual tensile strength. Fill unit weight 20·0 KN/m3
8. Details of the geometr y and material Position of load (from springing) 2·500 m
properties of an arch bridge, used throughout Single axle loaded length 0·300 m
Distribution angle 1H : 2V
this study to examine various parameters, are
given in Table 1. Although no single bridge
can be seen as representative of the entire UK Failure modes
stock the structure detailed could be consid-
ered typical. Mechanism
Full and small scale test bridges have pro- 10. This is the most readily identified mode
vided a good insight into the likely failure of failure of masonr y arches that results from
modes of masonr y arch bridges. Video record- the formation and rotation of hinges as
ings were made of the full scale failures and a depicted in Fig. 2 for both a near centre (a)
good summar y of the structural action of and quarter-point loading (b). In a well propor-
arches can be obtained from studying the TRL tioned arch the dead load line of thrust would
video.8 be near central in the arch ring. As the live
306
 UK MASONRY ARCH
ASSESSMENT METHODS

load increases, the moment in sections of the

arch increase and this is balanced by an
increasingly eccentric line of thrust, causing
the hinges to form progressively. With the for-
mation of the fourth hinge a ‘three bar’ mecha-
nism is invoked and the hinges rotate to
failure.

Snap-through
11. Full-scale experimental studies1 have
identified ‘snap through’ as a failure mode for
(a)
arches. In ‘snap through’ the hinges start to
form but instead of the final gradual formation
and rotation of the hinges a rapid change of
the local geometr y occurs with the section of
arch under the load snapping through. The
arch barrel can be viewed as a shell having a
large axial thrust and with the effective load-
carr ying area becoming increasingly small as
hinges develop. Snap-through is considered Fig. 2. Hinge positions
more likely for shallow thin arches but as any in masonry arches
arch moves towards a mechanism state there with (a) near centre
are substantial loads carried on relatively and (b) quarter point
small parts of the barrel section. (b) loading

Shear load. The ratio of the shear to axial force

12. There are few examples of the failure increases as the load is applied but, even as
of masonr y arches from shear. Known exam- failure occurs, the ratio is still quite low.
ples include the testing of an arch without fill
using a simulated wheel load which resulted in
punching shear failure of the arch ring.9 In a Compression
well-proportioned and well-maintained arch 13. Compression failure of the masonr y
shear failure is difficult to envisage although forming the arch ring was reported during the
punching failure can occur where loss of mor- full scale tests1,10 usually in combination with
tar leaves individual blocks unsupported by hinge development. General guidance on
axial thrust. Figure 3 shows the variation of allowable compressive strength is given
the ratio of the shear to axial force along an elsewhere.11
arch, the properties of which are given in 14. Arch forms which are more likely to
Table 1, at various percentages of ultimate suffer compression failures are short stocky

0·20 0%

28%
0·15
56%

0·10 83%

100%
Shear/axial ratio

0·05

0
2- 4 6 8 10
Distance along arch: m
–0·05

–0·10 Fig. 3. Variation of

shear/axial force ratio
–0·15 along arch with
increasing percentages
–0·20 of failure load
307
 HUGHES ET AL.

arches where the line of thrust tends to form Table 2. Modes of failure identified in full-scale tests
more direct paths to the abutments.
Name Feature Failure mode

Bargower Longitudinal cracks Material crushing

Barlae Skewed Snap-through
Full-scale tests Bolton New construction Mechanism
15. Tests have been undertaken on real Bridgemill Shallow arch Snap-through
arch bridges together with medium and large Dundee New construction Mechanism
scale models.1 Because of the dominance of Preston Elliptic Crushing
Prestwood No parapets Mechanism
gravity in determining arch behaviour, model
Shinafoot Detached spandrel Mechanism
tests cannot accurately replicate full-scale phe- Strathmashie Spandrel retained at failure Progressive
nomena but are an important data source for Torksey Spandrel retained at failure Snap-through
general behaviour. For this reason most of the
published comparisons of numerical results
have been made against the behaviour of the
full-scale tests. Full-scale arch test results typi- Table 3. Simulation of failure modes, arch geometry, materials and
cally report overall structural movements conditions
together with some masonr y strain measure-
MEXE Pinned- Mechanism Cracking-
ments. There are limited data on the beha-
elastic elastic
viour of the fill.
16. In developing any assessment method Failure modes
it would be unwise to rely too heavily on the Compression j r r r
Mechanism r r
single criterion of ultimate load without con-
Snap-through r
sidering the significant problems of full-scale
testing. The individual test reports provide Parameters
invaluable evidence that can support or dimin- Arch distortion j r r r
ish the value of the strict numerical data val- Abutment movement r
ues. Table 2 contains brief details of the failure Cracking: longitudinal j u u u
transverse j j j r
modes identified for each of the full-scale
diagonal j j j j
bridges. In summar y, the dominant collapse Bridge skew u u u
mode was the formation of hinges, either with Material degradation j j u r
or without snap through, sometimes including Mortar loss j j r r
the effects of material failure. Haunching r r r
Fill: type r
density r r r
passive r r
resistance r
Assessment analyses
stiffness
17. Not all of the methods of analysis of Arch: material type r
masonr y arch bridges have been exploited as strength r r r
assessment tools and some of the earlier elasticity r
assessment methods have been revisited using Arch condition r r r r
computers. It was considered that computer
r can be directly input into data
technology would not reduce the time taken to
u can be included in analysis
undertake assessment calculations, but it j covered by factors.
would allow more load cases and additional
influences to be considered and thus provide
more realistic predictions of the load carr ying
MEXE
capacities. Table 3 summarizes the failure
19. MEXE is based on an unknown amalga-
modes that each assessment method covers
mation of classic elastic arch theor y coupled
and also lists which site parameters of real
with the results of a series of tests. The basic
arches are simulated. Site values can either be assumptions are that the arch is parabolic,
put into the analysis input data, to affect the pinned at both ends, span/rise ratio of 4,
result directly, or can be used to quantify soundly built, specified density and symmet-
changes in the analysis results. Alternatively rically loaded at the crown with a transverse
they may be included by way of a suitable load distribution. These assumptions allow a
choice of global factors. closed form solution of the stresses. Consid-
18. For the MEXE, pinned-elastic and eration was given to rules regarding the posi-
cracking-elastic analysis the basis of the analy- tion of the line of thrust and limits on the
sis is a cur ved beam methods and where the allowable compressive (1·4 N/mm2) and ten-
distribution of live load is to the centreline of sile stress (0·7 N/mm2). The exact method by
the arch barrel. which the nomogram was developed remains
308
 UK MASONRY ARCH
ASSESSMENT METHODS

1·0

0·9

0·8

0·7

0·6
Profile factor

0·5
MEXE
0·4
Pinned-elastic
0·3 Mechanism (with soil)

0·2 Cracking-elastic (with soil)

0·1
Fig. 4. Variation of
0 effective profile factor
0·7 0·75 0·8 0·85 with ratio of quarter
Quarter point rise/crown rise point to central rise

Table 4. Factors used to determine the modified axle (b) Modifying factors are considered inde-
load pendent of each other and of the geo-
metr y yet they must be correlated. For
Factor Value example, the strength of an arch is influ-
Span/rise factor Fsr 0·6–1·0 enced by the fill and MEXE takes account
Profile factor Fp 0·0–1·0 of this by a factor related to the backfill
Material factor Fm material type. However, a deep arch will
(a) Barrel factor 0·7–1·5 be more influenced by the fill than a shal-
(b) Fill factor 0·5–1·0 low arch, yet the factor used is the same.
Joint factor Fj Factor values are also open to wide inter-
(a) Width factor 0·8–1·0 pretation with some having large variation
(b) Mortar factor 0·8–0·9* in their values.2
(c) Depth factor 0·9–1·0
(c) The basis of the modifying factors are
Condition factor Fc 0·3–1·0
unknown and have not been the subject of
*Mortar factor can be replaced by reducing barrel recent scrutiny. The profile factor, which
thickness used in material factor. is based on the ratio of the rise at quarter
span to that at arch crown, illustrates the
conflicts that can arise. The effects of
changes in profile were investigated for
unknown but the provisional axle load (PAL) is the arch detailed in Table 1. The shape
dependent only on the span and the depth of was changed by var ying the profile ratios
the arch ring plus fill. What is significant are between 0·71 and 0·87, keeping all other
the factors that are then adopted to revise the dimensions the same. Figure 4 shows the
PAL into an assessment load. The factors2 in effective profile factor obtained using all
Table 4 are used to determine the modified the assessment methods. As shown in Fig.
axle load with the range of the factors for 4, the MEXE profile factor takes the value
which identified values are known. 1·0 for rise ratios less than 0·75, and
20. The modified axle load is subsequently becomes smaller as the rise ratio
multiplied by axle factors to give the allowable approaches 1·0. By comparison, the other
single and multiple axle loads which are then methods predict significantly different val-
translated into gross vehicle weights. ues for the profile factor. The rise ratio of
an arch formed from a conical section has
values of 0·75 (parabola) and 0·87 (semi-
Points of concern circle). It is difficult to imagine an arch
21. with a rise ratio much above 0·87 or below
0·7, the latter representative of arch forms
(a) The PAL is equally dependent on the arch having a point at the crown. An arch com-
barrel and fill depth. posed of two walls leaning against each
309
 HUGHES ET AL.

other would have a profile factor of 1·0. Mechanism

The tendency to favour pointed arches 25. The basic assumptions are that the
may be a direct result of the central load- arch is soundly built, fails by the formation of
ing position used in the original derivation plastic hinges, has zero tensile strength and
of MEXE. allows passive soil forces to develop between
the two hinges most remote from the applied
load.
26. Mechanism methods are almost exclu-
Pinned-elastic analysis sively undertaken using computer software.4,5
22. The popularity of MEXE, coupled with The critical loading position is assumed and an
its ease of use and experience in application, estimate made of the four hinge positions.
led the Department of Transport to update the This allows the load to be determined, for a
method using a modern frame analysis given arch thickness, or alternatively deter-
approach.3 The arch geometr y and material mines the arch thickness appropriate for the
densities could be used directly, allowing the loading. Each hinge position is then relocated
geometric factors to be dropped whilst retain- until a minimum capacity is determined. The
ing the material and condition factors. The axle pattern is then moved and the process
basic assumptions of the analyses are that the repeated until a set of hinge positions asso-
arch is pinned, soundly built and fails in com- ciated with a critical mode is determined.
pression whilst retaining the full arch section. 27. The analysis has been extended to per-
The tensile stress, which appears in the analy- mit allowable plastic yield at hinges by modify-
sis, is simply ignored in the procedure. The ing the position of the point at which the hinge
compressive strength of the arch material rep- rotates to the centre of the plastic zone. This
position is determined from the axial thrust at
resents the most significant variable that the
the hinge and the allowable yield stress. The
engineer has to consider yet it is one of the
thrust is not known a priori and this process
most difficult parameters to properly assess.
therefore involves additional iterations.
23. Separate, unit width, factored dead and
28. Limited soil modelling has been intro-
unit live load analyses are undertaken and
duced by assuming that soil pressures develop
superposition is used. The assessment method between the hinges, remote from the load.
uses the MEXE Joint and Condition Factors to These pressures are determined from the
determine the allowable load which is then depth, density and passive coefficient of the
transformed to a permissible vehicle loading. soil. In deep arches these pressures can be
Alternatively the axle patterns of particular large enough to remove entirely the horizontal
vehicles can be considered directly. thrust at the abutment and the methods need
to deal with this anomaly.

Points of concern
24. Points of concern
29.
(a) The method is not a natural extension of
the working stress MEXE method. MEXE (a) The passive soil forces can be ver y large
included in its development a combination for deep arches and the deflections neces-
of a middle half rule and tensile strength sar y to develop these forces are signifi-
cant. The values recommended are
considerations, with 0·70 N/mm2 as a ten-
therefore not those associated with the fill
sile stress limit. The pinned-elastic
material with the results being sensitive to
method allows tensile stresses to increase
the passive coefficient values adopted.
to values which can be an order of magni-
(b) The method cannot deal with snap-
tude greater than this.
through (this also applies to MEXE and
(b) The method determines the compressive the pinned elastic method).
strength of the masonr y following consid-
eration of the compressive strength of the
masonr y unit and mortar, and geometr y of Cracking-elastic analysis
the masonr y unit. There seems no reason 30. The cracking-elastic analysis is based
to therefore include the MEXE joint factor on classic elastic theor y6 or by way of a finite
Fj. element approach.7 Loads are applied to the
(c) The dead load stress is generally insignif- arch barrel and the resulting member forces
icant in this method, less than 5% in the determined. These are used to determine the
worked example quoted,3 and in fact stress state and deformations. The stresses
detracts from the arch’s strength. The are used to identify the areas of tension and, if
arch’s strength is therefore almost linear appropriate masonr y yielding, and the defor-
with masonr y strength. mations subsequently used to modify the
310
 UK MASONRY ARCH
ASSESSMENT METHODS

geometr y, which enables the modelling of which the methods have been developed, have
snap-through behaviour and changes in soil been summarized. It remains to measure each
pressures. The arch ring area, inertia, co- method against the criteria selected. Table 5
ordinates and the applied loads are updated summarizes the authors’ judgement based on
and reapplied with the process cycled to the commentar y below.
convergence.
31. The soil models used var y but in gen-
eral they are assumed to provide dead weight Correctly predict the load carrying capacity
to the arch and lateral, deflection dependent, 34. The correct determination of the ulti-
restraint. These forces are linearly increased mate load can be a proxy for the assessment
(or decreased) with arch deflection into (or load for all methods except MEXE which is
away from) the soil being limited by the pas- not an ultimate limit analysis. A simple com-
sive (or active) coefficients. The coefficients parison has previously been undertaken3 and
used are based on the normally adopted for- ‘all the methods seem to give reasonably safe
mulae modified to account for the cur ved estimates of collapse loads’. Given the limited
geometr y. data it is not considered that undue weight
should be given to the outcome of such veri-
fication studies.
Points of concern
32.
Allow assessment at reasonable cost
(a) The methods use soil and arch stiffness 35. The cost of undertaking an assessment
which are not traditionally considered rel- is largely associated with the manpower
evant parameters, nor fully understood requirements. This typically includes a site
from known model and full-scale tests. visit to inspect and sur vey the arch and iden-
(b) The methods require quite complex tify the materials and their condition. The cost
purpose-written software for efficient is increased significantly if it is necessar y to
implementation of the technique. undertake boreholes or trial pits to establish
the ring thickness and backfill materials. The
time taken in the analysis stage can be a rela-
Assessment procedures – discussion of tively small element in the overall cost of the
selection criteria assessment.
33. At the start of the paper, the selection 36. The site information required for all
criteria for an assessment method were methods is detailed in Table 3. The MEXE
detailed. The behaviour of real arches, includ- method is insensitive to ring thickness and
ing the modes of failure, have been discussed. because this is expensive to determine it is
The background to the methods of assess- sometimes assumed at a conser vative value.
ment, including the important assumptions on The engineer assumes a significant factor of

Table 5. Overall assessment method’s performance

No. Criteria MEXE Pinned- Mechanism Cracking-

elastic elastic

1 Correctly predict the load carrying A A A A

capacity
2 Allow assessment at reasonable cost G A A P
3 Correctly model the failure modes P P A G
4 Properly consider those parameters P P A G
which influence load capacity
5 Minimize the use of additional global P A G G
factors
6 Provide additional appreciation of P P A G
behaviour
7 Allow progressively more detailed P P A G
investigation of borderline bridges
8 Model remedial works P A G G
9 Allow the modelling of service loads P A P G
10 Be suitable for improvement with P P A G
additional knowledge

G good
A average
P poor
311
 HUGHES ET AL.

h
3·0
d
MEXE
2·5
Pinned-elastic

2·0 Mechanism
Failure load ratio

Cracking-elastic
1·5

1·0

0·5

Fig. 5. Variation of
0
failure load ratio with
0 0·2 0·4 0·6 0·8 1·0 arch ring and soil
d /(d+h) depth

2·5

MEXE
2·0
Pinned-elastic
Failure load ratio

Mechanism
1·5
Cracking-elastic

1·0

0·5

0 Fig. 6. Variation of
5 10 15 20 25 failure load ratio with
Span: m arch span

safety has been introduced but the PAL will be There are a number of mechanism programs
the same and the effect on the material factor available at little cost but there is only limited
is small. This gives a sense of security without availability of the cracking-elastic programs.
reducing the assessed load. The arch ring Both methods use an array of axle patterns
thickness is important in all other methods, as and take longer to complete an assessment.
can be seen from Fig. 5 which shows, for the
arch detailed in Table 1, the variation of failure
load ratio resulting from different assumed
ring thicknesses. Correctly model the failure modes
37. MEXE has total availability and 38. The identified failure modes, detailed in
although limited to 18 m span and unable to Table 2, generally include the formation of
consider current UK bridge design loadings or mechanisms with or without snap-through.
even specified loading patterns, it is the quick- Table 3 details the failure modes allowed in
est to complete. The pinned-elastic method each method. The pinned-elastic method,
uses frame analysis but determining the crit- based solely on the compressive strength,
ical load position can be tedious. The time does not model the dominant failure mode.
taken depends on whether a variety of axle Mechanism methods model the majority of
patterns are checked or the single axle is used arch failure modes but only the cracking-
with conversions to double and triple axles. elastic models can model all modes.
312
 UK MASONRY ARCH
ASSESSMENT METHODS

1·6

1·4

1·2
Failure load ratio

1·0

0·8

0·6 Pinned-elastic

Mechanism
0·4
Cracking-elastic
0·2 Fig. 7. Cumulative
frequency distribution
0 for the failure load
0 0·2 0·4 0·6 0·8 1·0 ratio for assessment
Cumulative frequency methods

Properly consider those parameters known to two parameter values were varied for each.
influence load capacity The pinned-elastic analysis produced a signifi-
39. Previous criticisms of MEXE have cantly higher variability. MEXE is considered
included giving insufficient weight to the arch likely to have the highest overall variability
ring thickness and lack of conser vatism at since the increments of the modifying factors
large spans. The reason behind the 18 m limit are large and applied directly to the load.
is unknown, and a study on the arch detailed
in Table 1 was carried out to examine this
limit using all assessment methods. The Minimize use of additional global factors
results, shown in Fig. 6, indicate a similar 42. The use of multiple factors is consid-
trend for the other three methods but MEXE ered inappropriate since they cannot act inde-
does allow relatively larger loads at longer pendently. Table 3 details the way in which
spans. each method simulates arch geometr y, materi-
40. The significance of the fill on load als and condition. The modified MEXE method
capacity has been clearly demonstrated in a has eight factors covering both geometr y and
number of the experimental investigations. materials. The pinned-elastic analysis retains
The pinned-elastic analysis does not include both the joint and condition factors whereas a
consideration of different fill materials and in fill and condition factor may be more appro-
this respect it is inferior to MEXE. priate with the basic theor y of the method.
41. A sensitivity study of the various The mechanism methods generally only need
methods12 has quantified the effects of reason- to retain the condition factor. The cracking-
able variations in the material parameter val- elastic methods have been developed where
ues. Analyses were repeatedly undertaken and the condition factor only relates to cracking,
the effects of changes in the assumed material other considerations such as deformations
parameters quantified. MEXE was not included being directly included in the analysis.
because its variability is largely associated
with the modifying factors. For the pinned-
elastic analysis the compressive strength of Provide additional appreciation of the structural
the masonr y was varied, for the mechanism behaviour
analysis the yield stress of the masonr y and 43. Appreciation of structural action follows
the friction angle of the soil were varied and largely from consideration of the mechanisms
for the cracking-elastic analysis the arch ring of failure. The more complex cracking-elastic
stiffness and friction angle of the soil were programs provide additional, and some would
changed. A realistic probability density for say too much, information. The identified dan-
each material parameter was assumed and the ger is that the amount and apparent accuracy
analyses repeated at representative points of the data provides a false sense of security.
taken from the probability densities. The However, engineers generally like to under-
results were nondimensionalized against the take limited sensitivity studies to get a feel for
average material failure loads. The overall the significance of parameters and it is con-
cumulative frequency distribution is shown in sidered that the more complex models allow
Fig. 7. The cracking-elastic and mechanism this education to be achieved. The cross-
analysis give similar variability even though sections depicted in Fig. 8 which show the
313
 HUGHES ET AL.

predicted line of thrust for the pinned-elastic,

mechanism, and cracking-elastic methods are
not considered equal in aiding the appreciation
of structural action.

,



,
Allow progressively more detailed investigation
of borderline bridges
44. Bridges that are to be weight-restricted (a)
require considerably more assessment effort
than those that are not. In undertaking assess-
ments, conser vative assumptions can initially
be made regarding certain parameter values
and these can later be modified following site
investigations. The MEXE process tends to

,

cause significant incremental jumps since
changes in the factors are at a minimum of
10% and can be 30% of the assessment load.
(b)
The opportunity to undertake site testing of
the arch barrel material is considered unlikely
to bring significant changes in any prediction.
The method by which masonr y unit strengths
are transferred to masonr y strength is also
somewhat speculative in relation to compres-
sive failures in masonr y arches. The main ben- Fig. 8. Line of thrust
eficial effect of site work is in the correct for various assessment
determination of the arch geometr y to include methods: (a) pinned-
the presence of haunching which reduces the elastic; (b) mechanism;
effective arch span. Proper characterization of (c) (c) cracking-elastic
the fill will also be important in those models
that utilize fill behaviour. Care must be taken
to ensure that the engineer remains sceptical
of numbers during this process, since it is too
easy to tr y for that last few per cent needed
whereas 10% is realistic in modelling terms. current ser viceability considerations. Future
considerations are likely to include stress,
joint opening and deflections, all of which are
Model remedial works modelled by the cracking-elastic methods.13
45. The traditional ways of strengthening 47. MEXE is a working stress method but
arches are by repointing, saddling, lining and provides little information suitable for a proper
grouting. Repointing is unlikely to increase the consideration of ser viceability. The pinned-
short term strength of an arch and all meth- elastic is appropriate as a ser viceability model
ods can directly or indirectly include it. Sad- since structural behaviour is linear at lower
dling and lining of an arch ser ves either to loads. The method does however not simulate
strengthen or to replace the arch. Plain con- the opening and closing of the joints that may
crete can be considered as additional arch be significant for ser viceability.
thickness by all methods except the MEXE
method. Reinforced concrete cannot be readily
included in the arch analysis since the two Be suitable for improvement with additional
rings carr y load substantially differently. knowledge
Grouting the fill is included in the MEXE 48. MEXE and the pinned-elastic method
material factor and can be incorporated in the already greatly simplify existing known influ-
soil models in the cracking-elastic analysis. ences and do not purport to model real behav-
New methods of strengthening are likely to be iour; they therefore have little opportunity to
better considered by the more complex benefit from additional knowledge. The mecha-
cracking-elastic models since they better rep- nism methods, since they do simulate real
resent structural behaviour. behaviour, can benefit from better understand-
ing; this is likely to be by way of improved
characterization of parameter values. The
Allow the modelling of serviceability cracking-elastic models have the opportunity
46. The full scale test results indicated that to benefit from better soil models and better
damage generally started to occur at 50% of understanding of the behaviour of masonr y
the ultimate load and this forms the basis of under extreme fibre stress. Their room-for-
314
 UK MASONRY ARCH
ASSESSMENT METHODS

improvement is limited by the inability to forms of failure and allow site data to be more
include 2-D and 3-D effects. The models neces- fully incorporated in the principal analysis.
sar y to include these effects would likely be
unsuitable for regular assessments.
Recommendation
54. Overall the authors would suggest a
Conclusions multistage approach using MEXE for initial
49. The selection criteria are not all of appraisal, followed by either mechanism or
equal weight; most would consider the correct cracking-elastic where this indicates a clear
prediction of the load carr ying capacity as lack of capacity. Such an approach would allow
being the most important. Although numerical initial rapid assessment at reasonable cost yet
values were not given to the placement of each permitting the gradual introduction of the new,
method in Table 4 consideration of the results more fundamentally-based, techniques so that
suggests cracking-elastic methods have a expertise in their use can be developed.
slightly better assessment potential than
mechanism techniques with MEXE and
pinned-elastic analysis falling further behind. References
There are a number of conclusions that can be 1. PAGE, J. Masonry arch bridges - TRL state of the
drawn for each method. art review. HMSO, London, 1993.
50. MEXE method is economic and has 2. DEPAR TMENT OF TRANSPOR T. The assessment of
ser ved the bridge assessment engineer well highway bridges and structures. Department of
for many years. No specific software is Transport, London, 1993, Department Standard
required and little is to be gained from expen- BD21/93.
3. DEPAR TMENT OF TRANSPOR T. The assessment of
sive site investigations. The method does,
highway bridges and structures. Department of
however, have a large number of factors which Transport, London, 1993, Department Advice
tend to obscure the results of the original anal- Note BA16/93.
ysis. It has little opportunity for improvement. 4. HARVEY W. J. Application of the mechanism anal-
51. Pinned-elastic analysis was seen as a ysis to masonr y arches. Struct. Eng., 1988, 66,
modern replacement for MEXE. The software No. 5, 77–84.
is generally available but the process of load 5. CRISFIELD M. A. Finite element and mechanism
distribution and critical load point determina- methods for the analysis of masonry and brickwork
tion can be time consuming. In suggesting that arches. Transport Research Laborator y, Crow-
an arch’s strength is inversely related to dead thorne, Berks., 1985, TRRL Research Report 19.
6. HUGHES T. G. and BRIDLE R. J. Energy method
load the method contradicts all previous con-
for arch bridge analysis. Proc. Inst. Civ. Engrs,
sideration of masonr y arch behaviour. Its reli- Part 2, 1990, 89, 375–385.
ance on the compressive masonr y strength for 7. CHOO B. S., COUTIE M. G. and GONG N. G. Finite
determining capacity, with the implicit assump- element analysis of masonr y arch bridges using
tion of a large tensile strength, and its rejec- tapered elements. Proc. Inst. Civ. Engrs, Part 2,
tion of significant fill contribution contradicts 1991, 91, 755–770.
recent experimental evidence. The adoption of 8. TRL. TRL video, fallen arches – load tests on
the MEXE assessment factors does not con- masonry arch bridges. TRL, Crowthorne, V248,
form to the basis of the analysis and there is 1992.
little justification for their application. 9. DAVEY N. Test on road bridges. HMSO, London,
1953, National Building Studies, Research Paper
52. Simple mechanism methods are rela-
No. 16.
tively common and can be programmed quite 10. HENDRY A. W., DAVIES S. R. and ROYLES R. Load
quickly. They are fundamentally correct in test to collapse on a masonry arch bridge at Bar-
their assumed failure mode and with minor gower, Strathclyde. TRL, Crowthorne, TRL Con-
extensions can provide useful assessment tractor Report 26.
tools. There are problems with application to 11. HENDRY A. W. Masonry properties for assessing
arches with deep fill and they cannot readily arch bridges. TRL, Crowthorne, TRL Contractor
model snap-through. They have little potential Report 244.
for further development although parametric 12. HUGHES T. G. and BLACKLER M. Sensitivity analy-
studies may provide better guidance of soil sis of current assessment techniques. Proc. 1st
Int. Symp. on Computer Methods in Structural
parameters.
Masonry, Swansea, 1991, 207–215.
53. There are two main drawbacks of 13. HUGHES T. G. Modelling the effect of initial
cracking elastic analysis, the software cur- stress state on the ser viceability of masonr y
rently has limited availability and provides arches. Proc. 2nd Int. Symp. on Computer
large volumes of data to be considered. The Methods in Structural Masonry, 1993, Swansea,
method does provide useful information on all 50–60.

315