You are on page 1of 39

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/293193824

Structural analysis of flying buttresses

Article  in  European Journal of Environmental and Civil Engineering · February 2016


DOI: 10.1080/19648189.2015.1131201

CITATION READS

1 2,237

1 author:

Valentín Quintas
Universidad Politécnica de Madrid
6 PUBLICATIONS   1 CITATION   

SEE PROFILE

All content following this page was uploaded by Valentín Quintas on 18 September 2018.

The user has requested enhancement of the downloaded file.


European Journal of Environmental and Civil Engineering

ISSN: 1964-8189 (Print) 2116-7214 (Online) Journal homepage: http://www.tandfonline.com/loi/tece20

Structural analysis of flying buttresses

V. Quintas

To cite this article: V. Quintas (2016): Structural analysis of flying buttresses, European Journal
of Environmental and Civil Engineering, DOI: 10.1080/19648189.2015.1131201

To link to this article: http://dx.doi.org/10.1080/19648189.2015.1131201

Published online: 05 Feb 2016.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tece20

Download by: [Valentín Quintas Ripoll] Date: 05 February 2016, At: 10:05
European Journal of Environmental and Civil Engineering, 2016
http://dx.doi.org/10.1080/19648189.2015.1131201

Structural analysis of flying buttresses


V. Quintas

Departamento de Estructuras de Edificación, Universidad Politécnica de Madrid, Madrid, Spain


(Received 14 January 2015; accepted 4 December 2015)

In this work, the analysis of flying buttresses as rampant arches, and by means of the
classical Strength of Materials concepts, is performed. Since the elastic theory cannot
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

be applied to masonry arches, some considerations on limit analysis of masonry


arches and its boundary conditions must be exposed as well as on the equilibrium
mechanism by which masonry supports bending moments. The appliance of this
analysis to existing flying buttresses, allows obtaining some new conclusions on the
behavior and the real role of those structures.
Keywords: Flying buttresses; gothic structures; cross section efforts; limit analysis of
arches; rampant arches; masonry structures

Introduction
Masonry structures analysis has been the object of many studies in the last 20 years.
FEM method applied to no tension materials has been developed by Giuffrè, Pinto,
Bagi, Milano, Saetta, Fraternalli, Angelillo, Lourenço and Baratta. Huerta and Block
have generalised the use of line of thrust to arches and vaults. Analysis of cracks
appearing at buttresses and arches has been performed following Heyman’s “Leaning
Towers” (Heyman, 1992) by de Lorenzis, Ochsendorf, Dimitri, Hernando and Huerta.
These authors have also studied the collapse of straight masonry buttresses (Ochsendorf,
Hernando, & Huerta, 2004), buttresses with trapezoidal form (de Lorenzis, Dimitri, &
Ochsendorf, 2012a) and buttresses with a stepped shape (de Lorenzis, Dimitri, &
Ochsendorf, 2012b), submitted to an upper inclined punctual load and to their dead
load. Finally, Livesley’s method of analysis (Livesley, 1978) supposing masonry as
composed by rigid blocks applied to arches and buttresses has been used by Hernando
and Carocci. An extensive description of all these studies can be found in (Tralli,
Alessandri, & Milani, 2014).
Anyway, from the issue in 1966 of Jacques Heyman’s “The Stone Skeleton”
(Heyman, 1966), flying buttresses analysis continues to be considered as something
exclusively related to lines of thrust. However, flying buttresses can be analysed as what
they really are: as rampant arches, that is, as arches whose abutments lie at different
levels, and this way of analysis is only slightly more complex than the use of lines of
thrust. The fact that masonry arches cannot be analysed by elastic theory does not imply
that they cannot be analysed using the cross section efforts of bending moments, and
axial and shear forces. As happens with almost all types of structures, this use provides
a much better understanding on its behaviour.

Email: valentin.quintas@telefonica.net

© 2016 Taylor & Francis


2 V. Quintas

The use of limit analysis is, however, unavoidable. Some considerations on the
influence of the arch boundaries on the collapse mechanisms of the structure must be
exposed. As seen in the following, that influence is determinant for the yield patterns
developed by arches.
The statement of Heyman that masonry has not bending resistance, must be inter-
preted in the sense that masonry can only resist a bending moment if at the same time a
normal force acts. Some remarks on the bending resistance of masonry must be exposed
previously to the analysis of the different types of flying buttresses, their behaviour and
their role in gothic structures. Before entering on rampant arches analysis, it should be
useful to remind some basic concepts on inclined structures.

Reflections on a ladder
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(1) A straight ladder

Consider a beam inclined an angle “a,” simply supported at different heights (Figure 1).
As it is well known, global bending moments, Mv, and shear forces, Tv, are identical to
those of the horizontally projected beam, submitted to the horizontal plane projected
loads. By simple equilibrium, forces acting at the beam inclined cross section should be
(Figure 1)
M ¼ Mv

T ¼ Tv  cos a

N ¼ Tv  sin a

The only difference between cross section efforts of the inclined beam and those of the
projected one is that at the inclined beam Tv produces a shear T and a normal force N.
If the beam is a ladder, the equilibrium will be only possible if friction forces acting
at the wall and at the floor in which the ladder is supported exceed the vertical reac-
tions. If this does not happen, the beam will slide along the wall and the floor as a rigid
body. To prevent this movement, an external horizontal force, H – in words of Heyman,
an active thrust – must be applied at the lower support. This force must produce at the
upper support the same force with opposite sign, and therefore a pair of forces acting at
the whole structure. This pair of forces produces a global bending moment H f, if f is
the vertical distance between the two supports. That must be balanced by a pair of verti-
cal reactions RM, in such a way that, if l is the horizontal distance between the two sup-
ports, RMl = Hf. The existence of an active thrust H implies that the lower vertical
reaction grows H f/l while the upper vertical reaction diminishes in the same value
(Figure 2).
It must be noted that an inclined beam with low values of f/l needs a greater thrust
than one with high values of f/l.
Since fl ¼ tan a, the resultant of H and RM acts exactly along the beam axis, and
therefore, it only produces an additional compressive normal force tanH a, without any
change on the values of M and T. The efforts of an inclined beam with an active thrust
are therefore:
European Journal of Environmental and Civil Engineering 3

P/2
P

P/2 P
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

P/2
P/2

Mv

Tv

Tv sin a
a Tv
N
T
Tv cos a

Figure 1. Equilibrium of an inclined beam.

M ¼ Mv

T ¼ Tv  cos a

H
N ¼ Tv  sin a  (1)
tan a
4 V. Quintas

Hf/l
P
P/2

H a

Hf/l
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

P/2

H / tan a

Tv sin a
a Tv
N
Tv cos a
T

H
a

H / tan a
Hf/l

P/2

Figure 2. Equilibrium of an inclined beam with an active thrust.

If the inclined beam is built with a material without any tensile resistance, H can be
increased until the tensile stresses produced by bending moments are eliminated by the
compression produced by the normal force N.
European Journal of Environmental and Civil Engineering 5

Also, H can be increased until the upper vertical reaction due to vertical loads RU is
cancelled. The value of H for which this equilibrium state is performed will be
designated in what follows as “natural thrust,” HN. Really this is the case in which the
beam supports have no sliding resistance, and its equilibrium state is exactly that of half
a three-pinned arch (Figure 3). An inclined beam in this equilibrium state shares
therefore the properties of the full arch. The value of HN can be obtained simply by
equilibrium, if:
RU  H N tan a ¼ 0; H N ¼ RU = tan a

If H is increased beyond the value of HN, the upper reaction begins to push upwards
and the ladder turns itself into a strut. That is the reason because struts have an upper
support that prevents them sliding upwards (Figure 3).Finally, if the value of H makes
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

the stresses reach the yield limit, it would collapse with the mechanism of a simply sup-
ported beam, as bending moments continue unchanged (Figure 4).

(2) A clamped ladder

Let us suppose that the inclined beam is clamped at its supports. Two negative
bending moments should appear at the two ends of the beam. If the two clamping
moments are included in the function of Mv, and the reactions due to clamping moments
in that of Tv, the set of cross section efforts continues being that of (1).As shown by
Courbon (1964), this can be done making Mv and Tv the efforts of the projected hori-
zontal beam submitted to the projected load and the clamping bending moments.
If the beam is submitted to an uniform load, and the bending moment diagram is
drawn – as it shall be done in all what follows – with positive bending moments over
the beam axis and negative bending moments under that axis, it can be seen that bend-
ing moment diagram has the same form than the lines of thrust obtained by Heyman
(1966) for straight flying buttresses (Figure 4). It is well known, as it shall be remained
later, that there exists a linear relationship between lines of thrust and bending moment
diagrams. As happens with simply supported beams, this bending moment diagram
should not vary if H varies, and therefore the collapse mechanism is that of an inclined
clamped beam (Figure 4).

(3) A curved ladder

Suppose now that the inclined beam has a variable slope β along its axis. Since the
global equilibrium is the same that a straight beam, if an active thrust H is applied, the
resultant force, V, of H and RM will now act along the springing line of the arch. V now
induces a negative bending moment to the arch (Figure 5):
V  d ¼ cos ay cos a ¼ H  y. Unlike a straight beam, H modifies the bending
H

moments of the arch. The equilibrium of forces at a cross section between the efforts N
and T and TV, RM and H, gives the value of T and N (Figure 5), and the set of cross
section efforts of an inclined arch:
M ¼ Mv  Hy

T ¼ ðTv þ H tan aÞ cos b  H sin bð2Þ


6 V. Quintas

HN
(a)

HN a

l
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(b)
P P

HN a a HN

P P

l l

(c) H>HN

H>HN a

Figure 3. (a) Equilibrium at the natural state, (b) The statically equivalent arch, (c) Equilibrium
of a strut.
European Journal of Environmental and Civil Engineering 7

HN
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(a)

HN a

RL

(b)

(c)

Figure 4. (a) Line of thrust, (b) Bending moment diagram, (c) Collapse mechanism.

N ¼ ðTv þ H tan aÞ sin b  H cos b


In which clamping moments are included into the expression of Mv and Tv as in
Equation (1), For β = a, the set of equations became those of an inclined beam (1). For
a = 0, those of an arch with a horizontal springing line.
It must be noted that, as happens with straight beams, arches with low values of tan
need greater values of H to equilibrate RM, and this is an important feature in the design
of rampant arches as we shall see later.
It can be seen applying the first equation of (2) that, if H is gradually increased, M
begins to decrease, attains a minimum, and for values of H beyond that minimum,
8 V. Quintas

V Ru

d H tan a
a
β

V
H tan a

H
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Rpl

N
Tv + H tan a

H tan a

Rp

N
Tv + H tan a

β
β

Figure 5. Equilibrium of forces of an inclined arch submitted to a uniform load p.


European Journal of Environmental and Civil Engineering 9

changes its sign, begins to increase its value, and become indefinitely great. It can be
deduced from this behaviour that an inclined arch is not an appropriate structure if it is
used as a strut. The appropriate structure is a straight inclined beam, in which M does
not vary when N increases it value. However, flying buttresses, which obviously were
used as struts, are always curved arches. To explain this, a more deep study on rampant
arches must be done.

Rampant arches

(4) Passive thrust

Consider an inclined arch submitted at one of its points to a positive bending moment.
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

The cross section must rotate a certain angle θ, the material follows Hooke’s law or not,
simply because the upper part of the cross section is more compressed than its lower
part. This rotation, unlike an inclined straight beam, produces necessarily a horizontal
displacement outwards, ΔX at the lower support. If it is supposed fixed the upper
support, from Figure 6 it can be deduced that:
h  t  sin c h  h
DX ¼ ¼ ¼ h  y (3)
cos a cos a

For a = 0, we obtain Bresse’s Equation (2) of a horizontally supported arch, for y′ = 0,


the arch becomes an inclined straight beam and, ΔX = 0.
A straight inclined beam can perform a rotation without any movement at the abut-
ments. On the contrary, for a curved inclined arch, any positive moment and its associ-
ated rotation produces an outward displacement at the supports, and any negative
bending moment, and its associated rotation, produces an inward displacement at the
lower support. It must be noted that this phenomenon is not only an elastic one; it hap-
pens whatever are the arch material properties. A bending moment may produce a rota-
tion which is not linearly proportional or that depends also on the normal force, but
anyway has the same sign of the bending moment and produces a displacement at the
abutments.
If that displacement is prevented by a buttress or by the friction resistance of the
springing, a thrust H must appear in a natural way, not as an external force. That thrust
produced by the action of vertical loads will be named in what follows “passive thrust,”
as Heyman did. Flying buttresses were always curved simply because, due to its dead
load, they produce a passive thrust and, by Equation (2), H produces a normal force N
acting at the whole length of the arch that reduces, or even cancels, the tensile stresses
associated to bending moments.

(5) Equilibrium states and yield patterns

Three equilibrium states can be found depending on the value of H and, therefore,
on the capability of arch abutments to perform a determinate displacement. Each equi-
librium state has its corresponding yield pattern.
If the abutments move outwards, H has a relative low value (Figure 7) and the arch
works as a clamped beam with compressive force acting at its cross sections. This equi-
librium state will be named “beam state,” since the bending moment diagram is that of
10 V. Quintas

γ 90º
tθ γ
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

a
a
(a)
ΔX


tsi
h=
a

(b)

Figure 6. Horizontal displacement at a rampant arch lower springing.

a clamped beam, that is, with negative moments acting at its supports and positive
moments acting at the middle span (Figure 7). The yield pattern of the arch is also that
of a clamped beam (Figure 7). It is usual to find this state in flying buttresses that move
the façade wall up to the interior of the nave, as some on the Mallorca Cathedral. Of
course this equilibrium state would appear also if the lower support moves outwards,
due to a weak buttress, or even if the upper support moves upwards, or any combina-
tion of those movements. In any case, these yield patterns can be found simply moving
and rotating that of Figure 7 as a rigid body.
If all the movements at the abutments are entirely prevented by a very stiff buttress
and a very stiff wall with a prop at the upper support, the only possible yield pattern is
that of Figure 8 or that corresponding to rotations with opposite signs. That yield pattern
corresponds to the bending moment diagram of Figure 8. As the displacements at the
abutments are the sum of the products of the rotations and y′, see (3), and this sum
must be zero, the displacements produced by positive rotations must be cancelled by
those produced by negative rotations. The corresponding bending moments must more
or less follow the same law: positive moments must be approximately of the same mag-
nitude of negative bending moments. This can only be fulfilled if the bending moment
European Journal of Environmental and Civil Engineering 11

Ru
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Rl

Figure 7. Beam state of a rampant arch.


Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

12

Figure 8.
V. Quintas

Rl

Motionless state of a rampant arch.


H

Ru
European Journal of Environmental and Civil Engineering 13

diagram has four maximums, as in Figure 8. Of course, that diagram depends on the
values of y′, and on the relationship between rotations and bending moments, but in any
case two maximums of bending moments with opposite sign must appear along the arch
length, and two clamping moments of opposite sign must appear at the arch supports.
Using Equation (2), it can be seen that the thrust necessary to obtain that diagram is
greater than that of the “beam state.” In fact it corresponds to the value of H producing
the bending moment’s minimum described in the last section. This equilibrium state will
be designated in what follows as “motionless state,” and curiously flying buttresses
builders tried frequently to obtain it.
For values of H greater than that of the motionless state, the bending moments dia-
gram begins to change its sign, and finally turns itself into the beam state diagram with
opposite signs, that is, with positive clamping moments and negative moments at the
middle span. This state will be designated in what follows as “arch state,” since bending
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

moment diagram is that usually appearing in half a full arch (Figure 9).
If the value of H cancels the upper vertical reaction, the arch works with its natural
thrust HN; there is no vertical reaction at the upper support, and it works exactly as half
a complete arch. As it has been described in the first section, this can be obtained allow-
ing the upper support to slide vertically, and was also intended by flying buttresses
designers by making a vertical joint at the upper support. The bending moment diagram
and the yield pattern are usually that of “arch state,” as in Figure 9, but, if values of y′
are low, that equilibrium can arrive at beam state or near the motionless state. If the val-
ues of y′ are very low, by the first equation of (2), the term of H grows very slowly.
For values of H greater than HN, the thrust must be necessarily an active thrust, that
is it must be produced by an external force, for instance, by the thrust of a vault or
another arch. The rampant arch begins to work as a curved strut, and the upper vertical
reaction tries to move the arch upwards (Figure 10(a)), since the value of H exceeds that
of the upper reaction RU. However, it can be very rarely found at the upper springing of
flying buttresses devices designed to prevent this movement. This should be explained
by the following: If ΔH is the increment of H exceeding the natural thrust, the vertical
reaction acting upwards at the upper springing must be: DH  tan a, and if aR is the
angle of friction of the wall, the condition for the arch to slide it must be:
HN þDH
tan aR  DHtan a and therefore: DH  tan atan aR 1. For a reasonable value of aR = 30°, it
HN

can be obtained that when a = 60°, ΔH = ∞, that is, only for very sloping flying
buttresses, a ≫ 60°, the arch slides vertically.
In any case, bending moments grow as H grows, and became those of arch state
(Figure 10(b)). The yield pattern is that of arch state, but since necessarily one of the
abutments must move in the arch direction, it takes the form of Figure 10(c), in the
usual case that the active thrust is produced by the nave arches or nave vault.

(6) Fulfilment of the uniqueness theorem

From the study of these equilibrium states, it would be supposed that the uniqueness
theorem of limit analysis has been violated, since it has been found three collapse mech-
anisms for the same structure. Really it is not the case, because the capability of the
supports to perform vertical or horizontal displacements is in fact boundary conditions.
As it has been seen, to transform itself into a mechanism, it is necessary for the arch to
perform a determinate set of rotations, and these rotations are only possible if some
determinate displacements at the supports are possible. Each boundary condition is
therefore related to a yield mechanism as establishes uniqueness theorem.
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

14

Figure 9.
V. Quintas

HN

Rl

Arch state of a rampant arch.


HN

Ru =0
European Journal of Environmental and Civil Engineering 15

Ru

Rl
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 10. A rampant arch with H>>HN.


16 V. Quintas

(7) Lines of thrust and bending moment diagrams

As has been said, if we examine the lines of thrust obtained by Heyman in (1966)
for flying buttresses, it can be seen that nearly all correspond to the “beam state” bend-
ing moment diagrams. This happens because Heyman, when obtaining the passive
thrust, tried to find the minimum value of H for which masonry is capable to resist
bending moments without any tensile strength. In the only case in which Heyman found
the active thrust-that of Amiens flying buttresses – the line of thrust corresponds to
bending moments of “arch state,” with the maximum of all possible active thrusts. To
establish the limits of the strength capability of arches, the way in which masonry
resists bending moments must be revised in the following section.
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Masonry arches
Masonry has a very weak tensile strength in the perpendicular direction at the joints; it
is, therefore, reasonable to neglect tensile strength at least for “vousoir arches.” Anyway
they can resist bending it accomplishing the following criterions.

(8) Young’s criterion

Thomas Young explained for the first time how a material without any tensile
strength can resist a bending moment. If a compressive normal force acts at a cross
section at the same time than a bending moment, tensile stresses are eliminated. From
Figure 11, for a rectangular section of height h and depth b, the condition for the cross
2  bh  0, and therefore: h  N , which is the
6M N 6M
section of having no tensile stresses is: bh
expression of “mean third law,” in terms of bending moments and normal forces. This
condition will be designated in the following as “Young’s criterion.” This criterion is in
fact very conservative, and it leads to very thick structures. In any case, it was estab-
lished for the first time that the real factor for designing masonry are not the values of
M and N alone, but the “eccentricity” M/N.

(9) Heyman’s criterion

If Young’s criterion is not accomplished, a crack must appear at the arch joints.
However, as pointed by Heyman, many medieval arches show cracks nearly since their
construction and continue resisting loads. This can be explained, using yield analysis.
Consider (Figure 11), a cracked rectangular section in which the compressive strength fd
acts as a constant stress – in the same way as supposed in concrete – along the non-
cracked length h0, in order to counteract the ultimate bending moment Md, and the ulti-
mate normal force Nd. By the equilibrium of forces, it can be obtained readily h0:
h0 ¼ fNd bd
The ultimate bending moment Md can be obtained by equilibrium of moments:
Nd  e ¼ Md
Since e ¼ hh
2 ; the bending capacity of the cross section can be obtained as:
0

Md ¼ Nd  2 , and the minimum possible height: h  h0 þ 2M


hh0
N
This condition will be designated in what follows as “Heyman’s criterion.” Once
more, although this criterion is much permissive than Young’s one, the crucial value is
the eccentricity M/N. The minimum value of h, hmin, necessary to resist M and N can
European Journal of Environmental and Civil Engineering 17
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

M/N M/N M/N M/N M/N M/N


Young´s Criterion

M/N M/N
h0/2 h0/2
hmin

Heyman´s Criterion

Figure 11. Strength assessment criterions of masonry.

be found geometrically by placing (Figure 11) at each side of the arch axis:
hmin ¼ M
N þ 2 , as prescribed in many building codes.
h0

(10) Curve of pressure

In order to assess the safety of a masonry arch, a diagram must be found to


describe the accomplishment of Heyman’s criterion along the whole length of the
arch. If the value of the eccentricity M/N is drawn in the perpendicular direction and
at any point of the arch axis, we obtain a curve that usually is named “curve of
pressure” (Figure 12(c)). If at the end of each eccentricity is added the value of h0/2,
it can be obtained at any point of the arch half the minimum value of its height,
18 V. Quintas

(a) N (b) M
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(c) e (d) h0

(e)

Figure 12. Assessment of a masonry arch: (a) Axial force diagram, (b) Bending moment dia-
gram, (c) Curve of pressure, (d) Values of h0-in grey- along the curve of pressure, (e) Collapse
mechanism.
European Journal of Environmental and Civil Engineering 19

hmin/2, necessary to resist bending moments. If at each side of the curve of pressures,
is drawn h0/2, we obtain the position of h0, at any point of the arch when the cross
section yields (Figure 12(d)). If those values of h0 lie wholly into the arch, the arch
is safe enough. If those values of h0 attain the arch boundaries in a number enough
of points to transform the arch into a mechanism, it has been found the yield estate
of the structure. Finally, if they lie outside the arch boundaries, the arch cannot
counteract the bending moments and collapses.

(11) Appliance of the lower bound theorem

To draw the curve of pressure implies knowing the real values of the clamping
moments and of the thrust. That seems a difficult question for a non elastic structure.
The ultimate state of the structure can only be found, by trial and error, if a curve of
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

pressure is found for which the values of h0 reaches the arch boundaries in a sufficient
number of points.
However, as Heyman has shown in Heyman (1966) applying the lower bound
theorem of limit analysis, if a curve of pressure is found that corresponds to the
equilibrium state of the arch and to the equilibrium with loads, and h0 lies entirely
into the arch, the arch is safe enough. Lower bound theorem establishes that the col-
lapse load must be greater than that supposed in the above analysis. In any case,
for an equilibrium state, differences between the different bending moment and nor-
mal force diagrams that accomplish the above conditions are very small. The same
can be said for the values of clamping moments and of the thrust. In the following,
we will use simply possible bending moment and normal force diagrams, that is
anyone corresponding to a curve of pressure for which h0 fits into the arch, accom-
plish the equilibrium state supposed, and that fulfils (2), without questioning if it
corresponds to the exact real case.

(12) Line of thrust

From the XIX century on to our days, it has been very usual to state the stability of
masonry arches using lines of thrust, instead of curves of pressure. A line of thrust is
really one of the funicular curves of the load, that is, the axis of one of the arches that
support the load only by means of axial forces, without any bending. Its ordinates, mea-
sured from the springing line of the arch, y0f can be deduced from the first equation of
(2). If for y = y0f , M = 0, then,
Mv  H  y0f ¼ 0, and therefore: y0f ¼ Mv
H
Deriving the above equation, we obtain the slope of the funicular:
dy0f
dx ¼ tan bf  tan a ¼ H , and substituting this value in the third of (2) equations, and
Tv

making β = βf, we obtain the axial force acting at the funicular: Nf ¼  cosHb
f

Also, if the first of (2) equation is divided by H, it is obtained:


0 0 0
H ¼ H  y ¼ yf  y , that is, the difference between the ordinates of the funicular
M Mv

and those of the arch axis is the relationship between bending moment and thrust. This
is the reason because we have found that the lines of thrust obtained by Heyman for fly-
ing buttresses correspond to bending moment diagrams.
Lines of thrust are therefore less accurate than curves of pressure, because they do
not establish the real value of the eccentricity M
N.
20 V. Quintas
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 13. Line of thrust and funicular arch within a flying buttress.

It would seem – at least for vertical loads – that the use of lines of thrust is safer
since H is always smaller than N. However, this would happen if M H is measured in the
perpendicular direction to the arch axis, not in the vertical direction, as is done when
tracing a funicular. The eccentricity of the funicular NMf ¼ M H cos bf can be obtained from
the line of thrust (Quintas, 1997), but it is always smaller than that of the arch, M N as
Nf ≥ N.
Anyway, lower bound theorem – and common sense – has an appliance for lines of
thrust. If a line of thrust is found inside the arch, with the depth necessary to resist by
means of compressive stresses the loads of the arch (Figure 13), it has been found
inside the structure an another structure capable to withdraw the arch loads without
transforming itself into a mechanism – the “stress distribution” of lower bound theorem
(Nielsen, 1984) – and therefore the structure is safe enough. This use has as a disadvan-
tage that the behaviour of the structure obtained within the real structure can be very
different to that of the real structure. For instance, if it is supposed that the line of thrust
(see Figure 4) of a straight arch corresponds to the real state of the structure, it can be
deduced that the reactions at both sides act downwards. As we have seen, the reactions
can act downwards, at a horizontal line or upwards depending on the value of the thrust,
being bending moment diagram, and therefore the shape of the line of thrust
unchanged.

(14) Stone lintels

Lines of thrust and curves of pressure can be very similar as happens in the case of
Figure 13, or very different as happens with arches submitted to inclined punctual loads
(Quintas, 1997). Anyway, there exists a structure in which the form of the line of thrust
is exactly the same than that of the curve of pressure: a horizontal straight arch
(Figure 14), what can be designated as a stone lintel or, in words of Heyman, a
“horizontal plate-bande.” In this case, if the structure has an active thrust H acting at
each support, N = H, by simple equilibrium of forces, and is constant, and therefore
H ¼ N , the line of thrust is identical to the curve of pressure.
M M

Following the current strength of materials assumptions, the only thrust that can
exist is an active one, as happens with pre-stressed beams and as has been established
European Journal of Environmental and Civil Engineering 21

20 KN/m

25 T 25 T
L/2

h
Θ
Θ

LcosΘ/2
hsinΘ
L´/2

10.1 mT
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

10.1 mT 10.1 mT

25 T

Figure 14. from up to down: collapse mechanism, bending moment, normal force and curve of
pressure of a stone lintel submitted to its dead load.

above. However, if, when performing the deformation as a yielded beam, the real
rotations are taken into account, it can be found that the final length of the stone lintel
(Figure 14) is: L0 ¼ L cos h þ 2h sin h, if L is the length of the stone lintel, h is its height
and θ the rotation at the middle of the span. This means that for a range of rotations
and for stone lintels with low slenderness Lh, the length of the deformed structure, L′, is
greater than L. Therefore, the lower corners of the dowels at the supports push against
the buttressing, producing at the same time a clamping negative bending moment and a
passive thrust. It can then be deduced that if the stone lintel is clamped between two
stiff walls or buttresses, it can work, as described in Figure 14, with a natural pre-
stressing. The line of thrust is identical to the curve of pressure, and the joints of the
dowels can be traced following lines perpendicular to those curves, in the direction of
no shear force. The depth of the funicular arch is also the projection perpendicular to its
22 V. Quintas

4T

11 T 11 T

4 mT

4 mT 4 mT
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

11 T

Figure 15. from up to down: collapse mechanism, bending moment, normal force and curve of
pressure of a stone lintel submitted to a punctual load at its mid-span.

axis of h0. The case of Figure 15, a stone lintel without dead loads and submitted only
to a punctual load, is important for the equilibrium of upper flying buttresses as will be
seen in the following.

Flying buttresses

(15) Evolution of flying buttresses. The first flyers of Amiens Cathedral

The first flying buttresses were simply rampant arches with their axes forming a quarter
of circle, acting as curved struts in order to counterbalance vault thrusts, as that of Fig-
ure 13(a). Very soon it was realised that – as it has be seen in Section 1 – the best strut
is the straight one, since it not produces a passive thrust and it can accommodate itself
to resist a wide range of active thrusts produced by the vaults or the wind forces. The
problem was that a straight strut cannot support its dead load unless an active thrust is
acting at its springing, and this was not guaranteed by the structural behaviour of the
rest of the structure. It was then decided to support the straight strut by a rampant arch
through an infill of masonry (Figure 14(a)), and this is the classical design of flying but-
tresses (Viollet le Duc, 1997). This design was improved replacing the masonry infill by
a row of arches, as in the first flying buttresses of Amiens (Figure 14(b)), or simply by
a tracery as in Saint Urbain of Troyes (Figure 14(c)).
The design of the first Amiens flying buttresses seems especially perfect; the straight
strut balances the wind forces acting at the roof. The circular arch counterbalances at
European Journal of Environmental and Civil Engineering 23
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 16. Evolution of flying buttresses, from (6).


24 V. Quintas

their springing the thrust of the vaults by means of its passive thrust. In fact, Viollet-le-
Duc took this design as a model when rebuilding the flying buttresses of Evreux Cathe-
dral (Leniaud, 1994) during its restoration (Figure 16).

(16) The second flyers of Amiens Cathedral. Double flying buttresses

However, some years later, the whole buttressing system – with the exception of
that remaining of the choir – was replaced by a new one. This became the standard
buttressing system of gothic vaults: the double flying buttress (Figure 17). In the place
of the straight strut, it was built a curved rampant arch supporting a straight strut, the
upper flying buttress, and below, without any infilling connection, it was built another
rampant arch, the lower flying buttress with its straight strut. The role of this lower
flying buttress is clear; it continues producing a passive thrust that counterbalances the
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

thrust of the vault and that of the transverse arches. The role of the upper flying but-
tress seems especially obscure. It is not only to counteract wind forces, since the best
design was that of the first flying buttresses: if wind does not act, the straight strut
produces no thrust, when wind acts it can resist nearly any horizontal force. Instead
of this, the upper flying buttress produces always a passive thrust at the façade wall,
in the place where no structure exists to counteract that force. The only possible equi-
librium mechanism is that composed by inclined struts included into the depth of the
façade wall transferring inclined forces on to the vaults. The wall works horizontally
in a similar way of the stone lintel supporting one punctual load described in point
14. In any case, the upper flying buttress produces a compressive force at the rib
vaults. It is tempting to relate this structural action to Sabouret cracks (Heyman,
1983). These cracks appear at the crown of the rib vaults trying to separate the vault
from the façade wall. In fact Amiens vaults show plenty of them, as sketched in Fig-
ure 17(c). In this way, the upper flying buttress should be placed to close or to stop
the propagation of Sabouret cracks.
In any case, the fact that the thrust of the upper flying buttress acts against a wall
could lead the gothic builders to reduce the thrust by placing a column at the upper sup-
port. In this way, they forced the arch to work in its motionless state and, therefore, to
reduce its thrust. A possible bending moment diagram corresponding to that state (Fig-
ure 17(a)) needs more depth than that provided by the arch ring, and therefore it is nec-
essary to take into account part of the infilling of the arches. It is interesting
(Figure 17(c)) to note that the moment produced by the two eccentric lower vertical
reactions of the two arches at the buttress are nearly exactly balanced by the load of the
pinnacle placed eccentrically at the other side of the buttress, as founded previously by
means of photo elastic analysis by Mark (1992). If the buttress is analysed as a column
submitted to its dead load, the thrusts of the flyers and the eccentric reactions of flyers
and the pinnacle, it can be found that it accomplish Young’s criterion, that is, it is
extremely stiff, it works entirely in compression, and therefore the cracking progression
described in Ochsendorf et al. (2004), de Lorenzis et al. (2012a) and de Lorenzis et al.
(2012b), is very far from its beginning. This also happens in all the buttresses studied in
this work.

(17) Beauvais Cathedral

Beauvais Cathedral was constructed at the same time than Amiens, but with a delay
of some 15 years. This allowed Beauvais builders to profit the experience of the Amiens
European Journal of Environmental and Civil Engineering 25

construction. In this way from the beginning, the flying buttresses were of the Amiens
second type, that is, double flying buttresses, with the difference that the upper flyer
was placed pushing directly against the upper part of the façade wall (Figure 18).

+2.2 mT
-3.7 mT
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

+1.68 mT

+1.8mT

-2.8mT

+1.5mT

5T

4.7 T

13.4 T

13.1 T

Figure 17. Second flying buttresses of Amiens Cathedral: (a) Bending moments, (b) Normal
forces, (c) Curve of pressure and actions against the structure.
26 V. Quintas

5.92 T

0,7 T

26.5 T
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

5.47 T

5.92 T
0.22T

5.47 T

12.23 T

12.14 T

Figure 17. (Continued).

The upper part of this wall worked, therefore, as a row of the stone lintels of point
14 (Figure 15) working horizontally and transferring the thrust to the rib vaults crowns
(Figure 19). Once more it is tempting to suppose that Beauvais builders were trying to
avoid Sabouret cracks by producing a compression at the region in which they begin to
propagate. Anyway, as seen in point 14, the façade wall can transfer the thrust of the
upper flyer to the crown of rib vaults if a family of struts able to resist the thrust com-
ponents can be included into the wall section (Figure 19), as is the case. Another prob-
lem is if this action produces an inacceptable horizontal displacement at the façade wall.
Beauvais vaults were the greatest constructed at its time: buttresses were separated 9 m.
That displacement would be inacceptable not only because of aesthetics, but also
because a horizontal displacement at the rampant arches will force them to work at the
“beam state” and to reduce its thrust and its resistance. In fact, the former buttressing
system of Evreux Cathedral, that was nearly identical to that of Beauvais (7), was, as
said, replaced by Viollet-le-Duc by the first buttressing system of Amiens because the
upper part of the façade wall showed “disorders” in words of Lasteyrie (1926). All this
leaded the Beauvais builders to try to make the upper flyer thrust as low as possible. In
this way, they reduced the span of rampant arches dividing it into two by means of an
intermediate pier (see Figure 18(c)), placed in “porte a faux” over part of the aisle
European Journal of Environmental and Civil Engineering 27

vaults span. Also, they placed a column at the upper springing of the upper arch,
forcing it to work in motionless state and reducing even more its thrust. The result is
that the thrust of the upper flying buttress is extremely low, in the order of half of that
of Amiens. With such a low thrust, it is necessary to take into account part of the
infilling between the ring and the straight strut to resist bending moments. The lower
flying buttress arch, on the contrary, was forced to work with its natural thrust by mak-
ing a great vertical joint between the upper spring and the groin of the rib vault. In this
way, its thrust nearly duplicates that of the upper arch, and it is necessary only the arch
ring to resist bending moments in most of the arch length.
It could be imagined that the collapse of Beauvais vaults was produced because the
low compressive force produced by the upper flying buttress was not enough to stop the
propagation of Sabouret cracks. In fact, following Taupin (2001), the study of the roof
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

+1.08 mT
-0.79 mT

+1.41mT
-0.87 mT

+1,16 mT

+0.84 mT

+2,42 mT

-1.19 mT

2.5 T

8.7 T

4.4T

10.9 T

Figure 18. Flying buttresses of Beauvais Cathedral: (a) Bending moments, (b) Normal forces,
(c) Curve of pressure and actions against the structur.
28 V. Quintas
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 18. (Continued).

wooden framework shows that only collapsed part of the transversal vaults, remaining
the rest of the structure undamaged. The repairs only involved doubling the transversal
vaults transforming them into sexpartite vaults, with a much less need of compressive
forces, and leaving unchanged the rest of the structure.

(18) The repairs during the XIV century. Notre Dame of Paris

Being the above true or not, the fact is that, from the Beauvais collapse and along
the XIV century, many of the former buttressing systems were deeply modified. In
Chartres (Viollet le Duc, 1997), a new upper flyer was added to the former one that was
of the primitive type, that is, a rampant arch counteracting the vault thrust at its supports
(see Figure 20(a)). In Mantes (Radding & Clark, 1992), the flying buttresses were
European Journal of Environmental and Civil Engineering 29
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 19. The transfer of the flying buttress thrusts to the vault crowns.

moved on to the top of the façade wall, leaving the vaults springing without any
buttressing system (Figure 20(b)).
Perhaps the most remarkable – and well known – of these repairs was that of Notre
Dame of Paris (Viollet le Duc, 1997).The whole of the old buttressing system
(Figure 21) was entirely replaced by a great upper flying buttress spanning 12 m, more
than two times Beauvais span, pushing directly against the top of the façade wall
(Figure 21). A great vertical joint was placed at the upper springing, in order to make
the rampant arch work with its natural thrust.
The result is a thrust of 20 tones, more than six times that of the upper flyer of
Beauvais. The span of the rampant arch is so great – in fact is the greatest flying but-
tress ever built – that, even working with its natural thrust, a curve of pressure that fits
into the arch ring cannot be found.
As in Beauvais, the façade wall must transfer that thrust to the vault crowns. How-
ever, this problem was in this case of little importance: not only are the vaults much
smaller than that of Beauvais, but also they are of the sexpartite type: the transversal
vaults are duplicated and, therefore, the distance between vault crowns is divided into
two (Figure 22).

(19) The perfect flying buttresses. Narbonne

Following Viollet le Duc (1997), at the end of the XIII century, gothic builders
completely mastered the design of flying buttresses. Three cathedrals – according to
30 V. Quintas
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 20. Former and current state of: (a) Chartres, (b) Mantes.
European Journal of Environmental and Civil Engineering 31
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

Figure 21. Former and current state of Notre Dame of Paris.

Viollet-le-Duc-show a perfect design of the buttressing system: Limoges, Clermont-Fer-


rand and Narbonne. None of them show any movement at the arches or the buttresses.
In words of Viollet-le-Duc, they continue staying up in a perfectly vertical position. It
would be of interest to study the buttressing system of Narbonne, perhaps that of the
three cathedrals showing the best construction.
The two rampant arches (see Figure 23) are very sloping. Instead of forming a
quarter of a circle as usual, they finish at an angle of 72°. As in Beauvais, the thrust is
minimised by placing a column at the upper springing of each arch. The result is that
although they have a greatest span than those of Beauvais they have nearly the same
thrust of Beauvais lower flyer. As shown in point 3 – and as happens in any arch – a
highest crown produces a smaller thrust for the same span. As in the upper flying but-
tress of Beauvais, it is impossible to find a curve of pressure that fits into the arch ring;
it is necessary to take into account as structure part of the infilling between the arch ring
and the strut. The low thrust must explain the low movements of the walls and the
buttresses observed by Viollet-le-Duc. Narbonne builders learned how to obtain low
thrusts using the geometry of arches instead of reducing their span by means of a “porte
32 V. Quintas

20.09 T
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

20.09 T

35.9 T

Figure 22. Curve of pressure and actions against the structure of the Notre Dame of Paris flying
buttresses

à faux” as in Beauvais, and this means indeed an improvement in the flying buttresses
design.
The upper flying buttress produces against the vaults a thrust that nearly duplicates
that of the upper flyer of Beauvais. As Narbonne vaults show a perfect maintenance –
no cracks can be observed at their surface – it could be supposed that the upper flyer
produces at the vaults the necessary compression. Narbonne has smaller vaults than
Beauvais; therefore it could be supposed that Beauvais vaults needed a greater thrust
and therefore greater arches in its buttressing system.
European Journal of Environmental and Civil Engineering 33

(a)
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(b)

Figure 23. Flying buttresses of Narbonne Cathedral: (a) Bending moments, (b) Normal forces,
(c) Curve of pressure and actions against the structure.
34 V. Quintas

(c)

5.2T

2.1T

5.4T

1.85T
5.2T
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

5.4T

13.8T

13.7T

Figure 23. (Continued).

(20) Mallorca Cathedral

Many Architecture Historians point the construction of Narbonne Cathedral as the


beginning of the gothic building school developed in the old Kingdom of Aragon, in
Spain, during the XIV century. One of its greatest achievements is Mallorca Cathedral.
Its vaults are as great as those of Beauvais and its nave has nearly the same highness
(Mark, 1992). The design of the buttressing system is, however, completely different (see
Figure 24). Instead of making the rampant arches more sloping as in Narbonne, they are
extremely flat. As seen, this increases the thrust. Moreover, the arches have a greater
span than those of Narbonne, and a vertical joint, instead of the columns of Narbonne.
The flyers working with its natural thrust and with a low crown produce high thrusts.
The upper flyer, a passive thrust of 18 tones, similar to that of Paris, but the lower flyer
pushes with 25 tons, the same of Paris. These great thrust results, in the lower flying but-
tress, into great axial force acting at the arch. This allows finding a curve of pressure that
fits into the arch ring. If we sum to the passive thrust of the lower arch ring, the active
thrust that can withdraw the lower straight strut, it can be found that the lower flying but-
tress can support a thrust of 96 tons produced by the vaults. The great passive thrust of
the upper flyer could have moved inwards the upper part of the façade wall, and so the
upper springing of the upper rampant arch, leading it to work in the beam state, as can
be seen in the deformation of some of those arches (Roca, 2001).
An explanation on the need of such heavy thrusts could be found in the Catalonian
way of designing rib vaults. All of them, but specially those of Mallorca Cathedral,
European Journal of Environmental and Civil Engineering 35

14.8 T
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

25.2 T

14.8 T

25.2 T

31.2T

34.9T

Figure 24. Curve of pressure and actions against the structure of the Mallorca Cathedral flying
buttresses analysis of flying buttresses.
36 V. Quintas

(Roca, 2001) have heavy dead loads acting at the ribs intersection, that is, at the key-
stone of the rib vault. As can be found by membrane theory, great loads acting at the
key of rib vaults produce overall compressive stresses at the shell and consequently con-
tinuous thrust acting at the transverse boundaries of the shell and the façade wall. But
all this would be the object of another work (Figure 24).

Conclusions

(1) A rampant arch – and any kind of arch – has several yield patterns, or collapse
mechanisms, depending on the boundary conditions.
(2) These boundary conditions depend on the capability of the arch springings to
move on one or another direction.
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

(3) Although lines of thrust can be used to assess the stability of a masonry arch,
the correct way of analysis is the use of curves of pressure.
(4) The only possible explanation of the use of upper flying buttresses is to induce
a compressive force at the upper part of rib vaults.

Disclosure statement
No potential conflict of interest was reported by the author.

References
Courbon, J. (1964). Resistance des Matériaux (Vol. 1, Deuxième édition). Paris: Dunod.
de Lorenzis, L., Dimitri, R., & Ochsendorf, J. (2012a). Structural study of masonry buttresses:
The trapezoidal form. ICE Proceedings Structures and Buildings, 165, 483–498.
de Lorenzis, L., Dimitri, R., & Ochsendorf, J. (2012b). Structural study of masonry buttresses:
The stepped form. ICE Proceedings Structures and Buildings, 165, 499–521.
Heyman, J. (1966). The stone skeleton. International Journal of Solids and Structures, 2, 249–279
Heyman, J. (1983). Chronic defects in masonry vaults: Sabouret’s cracks. Monumentum, 26,
131–141.
Heyman, J. (1992). Leaning towers. Meccanica, 27, 153–159.
Lasteyrie, R. de. (1926). L′Architecture Religieuse en France à L′Époque Gothique [The
Religious Architecture in France during the Gothic Age]. In Auguste Picard (Ed.), Paris.
Leniaud, J.-M. (1994). Viollet-le-Duc ou le Délires du Système [Viollet-le-Duc or the System
Deliriums]. Paris: Mengès.
Livesley, R. K. (1978). Limit analysis of structures formed by rigid blocks. International Journal
for Numerical Methods in Engineering, 12, 1853–1871.
Mark, R. (1992). Experiments in gothic structure. Cambridge, MA: M.I.T. Press.
Nielsen, M. P. (1984). Limit analysis and concrete plasticity. Englewood Cliffs, NJ: Prentice Hall.
Ochsendorf, J. A., Hernando, J. I., & Huerta, S. (2004). Collapse of masonry buttresses. Journal
of Architectural Engineering, ASCE, 10 88–97.
Quintas, V. (1997). On lines of thrust and stability of masonry arches. In S. Sanchez Beitia , &
C. A. Brebbia (Eds.), Structural studies, repairs and maintenance of historical buildings
pp 168–177. Southampton: Computational Mechanics Publications.
Radding, C., & Clark, W. (1992). Medieval architecture, medieval learning: Builders and masters
in the age of Romanesque and Gothic. New Haven, CT: Yale University Press.
Roca, P. (2001). Studies in the construction of Gothic Cathedrals. Guimaraes: Historical
Constructions.
Taupin, J. L. (2001). La mémoire mécanique des Cathédrales. Réfléchissements sur les deforma-
tions[The Mechanical Memory of Cathedrals, Reflections on its deformations]. First European
Congress on Restoration of Gothic Cathedrals, Vitoria.
European Journal of Environmental and Civil Engineering 37

Tralli, A., Alessandri, C., & Milani, G. (2014). Computational methods for masonry vaults: A
review of recent results. The Open Civil Engineering Journal, 8, 272–287.
Viollet le Duc, E. (1997). Dictionnaire de l′architecture médiévale [Medieval Architecture
Dictionary]. In Facsimile of the 1854 Edition. Poitiers: Bibliothèque de L′Image.
Downloaded by [Valentín Quintas Ripoll] at 10:05 05 February 2016

View publication stats

You might also like