Buried Structures Static and Dynamic Strength Bulson 1984

@seismicisolation
@seismicisolation
Buried Structures
Static and Dynamic Strength
@seismicisolation
@seismicisolation
E D I T O R ’S FO REWORD
D ram a tic in n o v atio n s an d developm ents have occurred in civil an d stru ctu ral
engineering in recent years. D ifficulties o f analysis w hich ap p eared in surm ountable
only tw enty years ago have largely disappeared w ith the advent o f the m ainfram e
c o m p u ter a n d the finite elem ents m e th o d ; new g eneration m icrocom puters now
increasingly provide such analyses w ith great convenience an d econom y. The engineer
to d a y has m ore tim e to devise new form s o f con stru ctio n , to im prove design details,
an d to allow fo r p h en o m en a an d d a ta w hich were previously overlooked o r
ap p ro x im ated . M u ch o f this new expertise has been used to im prove the design o f ships
an d aircraft, offshore platfo rm s, subw ay systems, high-rise tow ers an d buildings, and
m an y o th e r form s o f c o n stru ctio n previously designed by rules-of-thum b an d simple
codes o f practice. T here is now m uch m ore intern atio n alism in engineering to o , w ith
design m eth o d s an d codes becom ing m ore standardized, an d large com puters
providing technical literatu re a n d p a te n t in fo rm atio n from all over the w orld. T here is
a need fo r these advances to be presented to an in tern atio n al audience by leading
engineers o f in tern atio n al re p u te ; this is the p u rpose o f the new Civil E ngineering
Series by C h a p m a n an d H all.
T he th ird of th e new series is by D r P.S. B ulson w ho is head of a defence research
establishm ent specializing in m ilitary engineering a t C hristchurch, E ngland, an d a
visiting p ro fesso r in th e Civil E ngineering d ep artm en t o f S o u th am p to n U niversity. H e
has w orked fo r the B ritish M inistry o f D efence since 1953, follow ing p o stg rad u ate
studies a t the U niversity o f B ristol an d service as an officer in the R oyal Engineers.
T h o u g h interested in all aspects o f m ilitary engineering, he has personally specialized
in stru c tu ra l stability, pneum atic stru ctu res an d u n d erg ro u n d structures. H e is the
a u th o r o f m any technical reports and papers, and has already written books entitled
S tability o f Flat P lates an d (as co -au th o r) Background to Buckling. H e has c o n tribu ted
to o th e r b o o k s on stru ctu ral stability.
D r Bulson’s new book is entitled Buried Structures: Static and Dynamic Strength,
w hich covers u n d erg ro u n d structures co n stru cted by a ‘cu t and fill* m eth o d rath er
th a n by tunnelling. T h o u g h m ost of th e research area is directed to w ard s the o ptim um
design of defence installations, pipelines and dom estic nuclear shelters, sim ilar
conclusions an d recom m endations apply to b uried structures constru cted for any
p urpose. T he b o o k will therefore be o f considerable interest to m ost civil an d stru ctu ral
engineers, p articu larly those designing covered tunnels, conduits an d defence w orks
u n d er static a n d dynam ic loads.
E. L ig h tfo o t
O xford
O T H E R T IT L E S IN T H IS S E R IE S
Theory and Design o f Steel Structures
G. Ballio an d F.M . M azzolani
Probabilistic Methods in Structural Engineering
G . A ugusti, A. B aratta an d F. C asciati
@seismicisolation
@seismicisolation
Buried Structures
Static and Dynamic Strength
P. S. BULSON
Deputy Chief Scientific Officer,
Ministry of Defence and
Visiting Professor of Civil Engineering,
University of Southampton
Taylor & Francis

Taylor & Francis Group
LONDON AND NEW YORK
@seismicisolation
@seismicisolation
First published 1985
by Taylor & Francis
2 Park Square, Milton Park,
Abingdon, Oxon, 0 1 4 4RN
Published in the USA
by Taylor & Francis
270 Madison Ave,
New York NY 10016
Transferred to Digital Printing 2006
© 1985 P.S. Bulson
ISB N 0 412 21560 8
All rights reserved. No part of this book may be reprinted,

or reproduced or utilized in any form or by any electronic,
mechanical or other means, now known or hereafter
invented, including photocopying and recording, or in
any information storage and retrieval system, without
permission in writing from the publisher.
British Library Cataloguing in Publication Data
Bulson, P.S.
Buried Structures. Taylor & Francis
1. Underground construction
I. Title
624.1’9 TA712
ISBN 0-412-21560-8
Library of Congress Cataloging in Publication Data
Bulson, P.S., 1925-

Buried structures.
Bibliography: p.
Includes indexes.
1. Underground construction. 2. Structural design.
3. Structures, Theory of. I. Title.
TA712.B85 1985 624.1’9 84-7761
ISBN 0-412-21560-8
Publisher’s Note
The publisher has gone to great lengths to ensure the quality of this reprint
but points out that some imperfections in the original may be apparent
@seismicisolation
@seismicisolation
Contents
P r e fa c e ix
N o ta tio n xi
1 Introduction 1
1.1 Early history 1
1.2 C ontem porary structures 2
1.3 The future 3
1.4 Purpose of the book 4
2 General principles 6
2.1 Arching 6
2.2 Elastic stress distribution 15
2.2.1 Stress distribution around a circular cavity 15
2.2.2 Stress distribution around non-circular cavities 20
2.2.3 Two circular cavities in close proxim ity 25
2.2.4 Thick-cylinder theory 27
2.2.5 Concentrated free surface loads 29
2.2.6 Stress distribution around a spherical cavity 30
2.2.7 Stress distribution above a yielding trapdoor 32
2.3 Properties of soils 35
2.3.1 Static properties of granular soils 36
2.3.2 Static properties of cohesive soils 39
2.3.3 Static properties of rock 41
2.3.4 Dynamic properties of soils 41
2.3.5 Wave propagation in soils 45
3 Thick-walled pipes under static loads 52

3.1 Ditch-type excavations 53
3.2 Positive projecting conduits 57
3.3 Negative projecting conduits 61
3.4 Wide trenches and multiple pipes 65
3.5 Submerged pipes 67
3.6 Concentrated free surface loads 68
@seismicisolation
@seismicisolation
vi Contents
3.7 O ther m ethods of load analysis 71
3.8 C onduit beddings 74
3.9 Full-scale experiments 75
3.10 M odel tests 82
4 Thin-walled pipes under static loads 89

4.1 Deflection analysis (traditional methods) 90
4.2 Deflection analysis (recent developments) 97
4.2.1 Closed-form elastic solution 97
4.2.2 Finite element program 100
4.3 Buckling under static loads 104
4.3.1 Elastic analysis 104
4.3.2 Ultim ate strength related to flexibility 109
4.3.3 Ultim ate strength related to cover depth 110
4.3.4 Experimental studies 114
4.4 Design rules 121
4.4.1 US C orrugated Steel Pipe Institute guidelines 121
4.4.2 US Naval Civil Engineering Laboratory reports 124
4.4.3 C lR IA recom m endations 126
4.4.4 U K military design codes 127
5 Non-circular pipes, closed cylinders and shells under static loads 134
5.1 Thin-walled square sections 134
5.2 Elliptic sections 143
5.3 Pipe arches 147
5.4 Arches 152
5.5 Thick-walled arches 156
5.6 Thick-walled closed cylinders 158
5.7 Thin-walled closed cylinders 160
5.8 Vertical capsules 162
5.9 Thin-walled domes 164
5.10 Thin-walled spherical shells 168
6 Structures under dynamic loads 173

6.1 Tests at the University of Illinois, USA 173
6.2 Tests at the US Waterways Experiment Station 176
6.3 Tests at the US Air Force W eapons Laboratory 180
6.4 Tests in the U K 181
6.5 Dynamic analysis 189
6.6 U nderground concrete structures under localized
explosive loads 193
6.7 U nderground concrete structures under surface
pressures from nuclear explosions 196
@seismicisolation
@seismicisolation
Contents vii
6.8 Structural design of domestic nuclear shelters 198
6.9 Backpacking 200
6.10 M ilitary shelters 202
6.11 Burster slabs 206
7 Loads, strength and safety 213

7.1 Rail and road tunnels 213
7.2 Large water pipelines 214
7.3 C orrugated steel pipes 216
7.4 U nderground blast shelters 218
7.5 Conclusions 220
A u th o r in d e x 223
S u b je c t in d e x 226
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Preface
As a schoolboy I frequently journeyed to the Dorset coast through the road

tunnel at Beaminster, built by Lang in the 1830s. Later, when I first became a
student of engineering, the walk to lectures took me through the derelict
surface workings of the great D olcoath tin mine near Camborne in Cornwall.
It never entered my mind that I would one day write on the subject of
underground structures.
I became involved in the subject through a defence interest in the behaviour
of thin-walled buried structures under static and dynamic surface loading, a
subject not closely connected with the design of masonry tunnels or mine
workings, but nevertheless relevant to the general field of soil-structure
interaction. Recently I have detected an increasing civil engineering interest in
the problem, and I have therefore attem pted to summarize the available
analysis and test work for the benefit o f engineers and scientists coming to the
subject for the first time.
I have acknowledged sources of information where they appear in the text,
and I am indebted to the U.S. Army Standardization Office in London, and
the U.S. Defence Nuclear Agency in W ashington, for their help in obtaining
clearance to quote from U.S. Technical Reports. There are inevitable gaps in
the presentation because a good deal of inform ation from defence research
sources still carries a security classification and cannot be published in open
literature. Because some of the experimental work was carried out before the
days of m etrication, readers are asked to accept a mixture o f f.p.s. and SI units
in the text. However, a conversion table is provided at the end o f the N otation
section on p. xvi.
I wish to record my thanks to Miss Joyce Carter, who typed the manuscript,
and to M r Phillip Read of Chapm an and Hall, who waited so patiently for the
final draft. Above all, I express my gratitude to M r J. Ellis, former director of
the M ilitary Vehicles and Engineering Establishment, Ministry of Defence, for
allowing me the facilities to complete the work, and for his encouragement.
P.S. Bulson
Christchurch, 1984
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Notation
All symbols are defined in the text where they first occur. The symbols listed
below are those that appear repeatedly, or are of greatest interest.
Lower-case letters
a distance between circular holes
vertical semi-axis of elliptic arch
radius of cylindrical surface
b width of underground structure
horizontal semi-axis of elliptic arch
c cohesion of soil
d diameter
displacement
dc diametrical shortening
/ stress
fc design stress, ultim ate compressive strength
/ yield or proof stress
h m inor axis of ellipse
equivalent wall thickness
k soil constant
m odulus of passive resistance
coefficient of subgrade reaction
k0 coefficient of earth pressure at rest
kc constant of soil reaction for clay
kL coefficient of elastic soil reaction
km m odulus of soil reaction
coefficient of soil reaction
m odulus of subgrade reaction
k ml’
i9 k mz, m om ent coefficients
/Cpi, kp2, kp3 axial force coefficients
kT ratio between horizontal and vertical earth pressure
k constant of soil reaction for sand
L spring constant for soil
@seismicisolation
@seismicisolation
N otation
length of underground structure
volume of voids per unit volume of soil
percentage void content
pressure
internal pressure
atm ospheric pressure
static collapse pressure
uniform radial pressure
allowable surface pressure
collapse pressure in air
elastic critical radial pressure
external pressure
interface pressure
peak incident pressure
value of ps at deep covers
peak reflected pressure
surface pressure to cause collapse
peak instantaneous transverse pressure
free field pressure
free field stress
overpressure
polar coordinate
distance from centre of circle
radius
projection ratio
settlement ratio
ultim ate unit resistance
shearing resistance of soil
sz remote state of stress
thickness
time
distance of water table above crown
duration of positive phase
pore pressure
dilational seismic velocity
radial displacement
tangential displacement
impulse velocity
velocity of shock front
weight of explosive charge
radial deflection
, coordinate axes
@seismicisolation
@seismicisolation
Notation xiii
Capital letters
A arching factor
footing area
thrust area per unit length
Ag geometry factor
As plan area
B width of underground structure
Bc overall conduit width
Bd ditch width
Cc load coefficient (positive projection)
Cd load coefficient (narrow trench)
D diam eter of underground structure
D' flexural rigidity of pipe wall
Dc diam eter shortening
E m odulus of elasticity
E' specific stiffness
E* m odulus of equivalent pipe
Ec elastic m odulus of clay
Es elastic m odulus of sand
elastic m odulus of soil
F blast load
concentrated load
line load
G shear modulus
Gs secant shear m odulus
H depth of cover
H total impulse
I second m om ent of area
impulse
I0 impulse to produce given level of damage
L side depth of underground structure
span
path length of projectile
Lk relative stiffness
M bending mom ent
Ml loading modulus of soil
AM l additional m om ent
Mp ultim ate m om ent (midspan)
Ms secant m odulus of soil
Mt average tangent m odulus for soil
Mu ultimate moment (support)
unloading modulus of soil
@seismicisolation
@seismicisolation
Notation
thrust in tunnel lining
flexibility num ber
vertical penetration
force
radius
stiffness-geometry factor
rise of pipe arch
angular linear penetration
m om ent reduction factor
stiffness of elastic medium
span of arch
settlement
perimeter of structure
periodic time
strain energy
yield of explosion
concentrated surface load
total weight of bomb
vertical load per unit length
equivalent weight
horizontal deflection
vertical deflection
;ek letters
semi-angle of arch
semi-angle of bedding
peak stress attenuation factor
depth coefficient
soil-structure interaction factor
angle between vector and vertical
burial factor
soil density
partial safety factor
y2, y3 partial safety factors
bedding factor
drained density
hyperbolic shear strain
saturated density
deflection/time factor
density of water
deflection
central deflections
@seismicisolation
@seismicisolation
lateral strain in soil
direct strain
hardening strain of backpacking
safety factor
angle of obliquity
coefficient of ground reaction
ductility ratio
Poisson’s ratio
Poisson’s ratio for soil
mass density
radius of curvature
breaching range
stress
ring compressive stress
tangential compressive stress
compressive stress to cause collapse
ultim ate strength
critical elastic buckling stress
largest principal stress
longitudinal stress
compressive yield stress
radial stress
ultimate soil bearing strength
vertical soil stress
hydrostatic stress
stress in x-direction
overburden pressure
tangential stress
tangential compressive stress
shear stress (polar coordinates)
shear stress (cartesian coordinates)
angle of shearing resistance
angular velocity
@seismicisolation
@seismicisolation
xvi Notation
Conversion factors
P ro p erty SI U nits f.p.s. U nits
L ength 1.0 m m 0.0394 in.

1.0m 3.28 f t
A rea 1.0 m m 2 0.00155 in .2
1.0 m 2 10.76 ft2
M ass 1.0 kg 2.2051b
D ensity l.O k g m -3 0.0624 lb ft" 3
F o rce 1 .0 N 0.225 lb f
1.0 k N 225 lb f
Stress 1.0 k N m -2 0.145 lb in ." 2
15.44 N m m " 2 l.O to n fin .-2
@seismicisolation
@seismicisolation
CHAPTER ONE
Introduction
1.1 Early history

F or m any thousands of years, the main subterranean activities of man were
cave dwelling, mining and tunnelling. Mining was often carried out at
considerable depths in ancient rocks, and the cavities excavated for this
purpose were generally stable, with some form of structural lining being used
as a means of preventing local falls of the roof or sides. Many mining galleries
could be left unlined, just as the galleries of burrowing animals are unlined,
and m any only needed local roof props.
Tunnelling, on the other hand, was often carried out at shallow depths in
younger geological formations, and was used for water supply, drainage
or military fortifications. The lining was needed to m aintain the integrity
of the cavity for conveyance purposes, and was designed as a permanent
structure, resisting local loads by using brick or masonry in the form of
vaulted arches. The tunnels of Babylon, Athens and Rome were built in
this way, and there were few design changes throughout ancient history.
Even in the Middle Ages, the substructures of our cathedrals and castles
m ade extensive use of the m asonry vault.
Between the seventeenth and nineteenth centuries, there were notable
advances in the techniques of tunnelling, owing to the introduction of
gunpowder and dynamite for blasting, and of hydraulic and pneumatic
drills for rapid excavation. These advances coincided with a sharp increase
in the need for traffic tunnels (highway, railway, navigation and subway)
and conveyance tunnels (water supply, drainage, sewage, hydroelectricity),
particularly in Europe. Tunnels were normally constructed without disturb
ing the surface, using what became known as ‘classical’ methods, where
tem porary timber elements in a variety of configurations were employed to
support the heavy linings during erection. Later, Brunei invented the shield
method, a moving metal casing driven in advance, to support the surround
ing earth or rock without the need for timbers.
Lining materials changed from brick and m asonry to concrete, reinforced
concrete, cast iron and steel, particularly in the construction of highway and
subway tunnels in the early part of this century. M any shallow-depth ‘m etro’
tunnels were constructed by the ‘cut-and-cover’ method, particularly if the
@seismicisolation
@seismicisolation
2 Buried Structures
soil was poor and saturated by ground water. There was a need for all these
structures to withstand the long-term degeneracy associated with a soil and
water environment, uneven bedding and fluctuating loads. They were
therefore rigid in construction, inherently strong and robust. Modes of
failure were similar to those exhibited by m asonry arches - the formation of
hinges at the springings and crown of the arch or vault, or inward failure of
the lower sides due to lateral pressure followed by upward collapse of the
floor. Structural analysis supporting the design was mainly concerned with
establishing levels of loading and predicting failure of the heavy cross
sections. M any designers still employed thrust line theory to check the
stability of the lining in the way it was used in the seventeenth century to
design domes and vaults.
1.2 Contemporary structures

At the end of the nineteenth century, mass-produced corrugated steel
sheeting became available to world markets, and its use in all fields of the
construction industry grew rapidly. Corrugated metal pipe was soon
developed and used for culverts; it was shop fabricated into a variety of
cross-sectional shapes (round, elliptic, pipe arch and underpass), and as
experience grew diameters were increased to 2.5 m and above. This was
probably the first use of thin-walled flexible linings for subterranean struc
tures, and gave rise to a good deal of laboratory and field research in the
early part of this century towards a rational design method. The idea of
using soil-structure interaction as a means of supporting the loads on
flexible pipes and tunnels began to be formulated as a result of this research.
Heavier plating was developed, with larger corrugations, capable of field
assembly into culvert and underpass shapes, and diameters of over 6 m were
successfully constructed. It was soon clear, however, that in addition to the
problems associated with elastic deflections, and the mobilization of the
resistance of the surrounding soil, there could be a danger of instability
under compressive forces of the relatively thin walls of these structures - a
condition that had not been met in classical tunnel design. Research on this
aspect was sponsored in the USA by the American Iron and Steel Institute
in the late 1960s.
Meanwhile, during the Second W orld War, it became necessary to design
underground shelters as a protection against blast effects from high
explosives, for both civil and military purposes. The economic thin-walled
lining was used extensively in corrugated or stiffened form, and considerable
knowledge was accumulated on the behaviour of this type of structure under
dynamic loading. After the war, the military use was extended to take
account of the blast effects of nuclear devices on subterranean military
@seismicisolation
@seismicisolation
Introduction 3
installations of varying types, and more research was undertaken. The load
ing differed from that associated with culverts and underpasses, in that
the pressure due to the soil cover was augmented by a surface blast pressure,
dynamic in character. The setting of some of the structures was no longer
horizontal - some thin-walled tubular constructions were set vertically to
act as subterranean missile-firing silos.
1.3 The future

At the time of writing, even larger buried culverts are being used, particularly
in N orth America, and some have been instrum ented to study their perfor
mance. Arch-shaped culverts with spans greater than 15 m have been the
subject of field experiments to measure soil arching, displacement and
deformation, and earth pressures in adjoining areas. Cover depths of more
than 13 m have been employed.
The large-diameter flexible pipe is also required for the development
of Britain’s water resources, particularly for the proposed inter-regional
grid to transfer bulk water supplies quickly between m ajor strategic centres.
In a survey of the problem, particular attention has been given to the analysis
of reinforced and unreinforced plastic pipes with diameters up to 5 m,
including the placing and com paction of the backfill.
In the future, one foresees increasing interest in the design and construction
of structures under the sea bed, in conjunction with deep-sea mining and
energy exploration. The loading on a structure is then a large hydrostatic
pressure due to the depth of water, superimposed on that due to a relatively
shallow covering of sea-bed m a te r ia l- a problem not unlike that of the
blast-resistant buried structure. Few codes of practice are likely to exist, so it
is im portant that designers have a good understanding of the fundamental
behaviour, with particular regard to stability, limit states of deflection and
deformation, and safety.
The distinction between rigid and flexible structures may also become
less obvious as time goes by. Taking underground pipes as an example, a
rigid pipe of concrete or cast iron will norm ally fail by bending of the wall,
and a flexible pipe (plastic, corrugated or sheet steel) by buckling. But the
distinction is not an ideal one because most structures have some degree of
flexibility, and for unconventional designs it may be better to consider the
whole range of behaviour in terms of slenderness, rather as we do for struts,
plates and shells. Further, unlike many above-ground structures, the modes
of failure are influenced by the properties of the surrounding soil, and the
loading is influenced by deformation in a non-linear fashion. The nature of
the backfill is also im portant for structures emplaced by the cut-and-cover
m ethod.
@seismicisolation
@seismicisolation
4 Buried Structures
1.4 Purpose of the book

The m ore conventional underground structures have been dealt with
comprehensively in the literature. The analysis and design of thick-walled
tunnels has been the subject of two im portant b o oks: The Art o f Tunnelling
by Szechy [1.1] and Statik des Tunnel- und Stollenbaues by Kastner [1.2].
In neither case, however, is the connection between theory and practice
fully explored. Underground excavations in rock have been dealt with by
Hoek and Brown [1.3], and Megaw and Bartlett have written on the plan
ning, design and construction of tunnels [1.4,1.5]. Buried pipelines have been
examined in detail by Clarke [1.6] formerly of the U K Building Research
Establishment, and in Report 78 of the U K C onstruction Industry Research
and Inform ation Association [1.7]. The analysis of soil arching over buried
structures has been set down in detail by Terzaghi in his book on Theoretical
Soil Mechanics [1.8], and by Lambe and W hitman in their publication of
1969 [1.9].
As stated in the preface, however, it is not the author’s intention to deal
with the subject of mining tunnels or rail and road tunnels, but to concentrate
on underground structures of a less perm anent character in which stability
under dynamic and static surface loading, and resistance to localized
explosions, are a feature. There is a very large store of knowledge, analytical
and experimental, available from engineering and scientific journals,
research reports, conference proceedings and theses, throughout the world,
that apply to m any of the less frequently met problems of underground
structures.
The main purpose of this book is to bring together what the author
knows of this store of knowledge in as concise a way as possible, and in
doing so to strike a balance between the needs of the progressive designer
and those of the research worker. Problems for which there is no acknow
ledged code of practice, such as the design of sub-sea-bed structures, can best
be approached by applying fundamental reasoning based on existing
knowledge. The research student or engineer about to em bark on a project
is helped by a readable summary of the subject, based on recent work.
The book begins with general principles, before going on to deal with a
num ber of basic structural forms, describing in turn the analysis, experi
m ental results and (if available) generally accepted design rules. Some of the
forms are well known - the long cylinder (pipeline), or long arch (culvert) -
but some are less widely used - the dome, spherical shell, or vertical cylinder,
for example. All may be im portant during the next half-century.
The problem of dynamic loading is pursued in Chapter 6, with particular
reference to the resistance of ‘hardened’ underground structures to nuclear
and localized explosions. The large test facilities in the USA and elsewhere
are described. M uch of the work on the behaviour of buried structures under
@seismicisolation
@seismicisolation
Introduction 5
shock conditions has a close link with the analysis of structures under
earthquake loads, but the book does not extend into this area of the subject.
The final chapter surveys the problem of loading and safety in tunnels,
pipelines and underground construction in a general way. Attention is
drawn to the need for a unified approach to a subject that is becoming more
im portant as the num ber of projects in the world involving major under
ground construction grows.
References
1.1 Szechy, K. (1973) The A rt o f Tunnelling, A kadem iaikiado, Budapest.
1.2 K astner, H. (1962) Statik des Tunnel- und Stollenbaues, Springer-Verlag, Berlin.
1.3 H oek, E. an d Brow n, E.T. (1980) Underground Excavations in Rock, Institute
o f M ining an d M etallurgy, L ondon.
1.4 M egaw, T .M . and B artlett, J.V. (1981) Tunnels: Planning, Design and Construction,
V olum e 1, Ellis H o rw o o d, Chichester.
1.5 M egaw , T .M . and B artlett, J.V. (1982) Tunnels: Planning, Design and Construction,
V olum e 2, Ellis H o rw o o d , Chichester.
1.6 C larke, N .W .B. (1968) Buried Pipelines, M aclaren, L ondon.
1.7 C om pston, D .G ., et al. (1978) Design and Construction o f Buried Thin- Walled Pipes,
C IR IA R ep o rt 78, C onstruction In dustry R esearch and Inform ation A ssociation,
L ondon.
1.8 T erzaghi, K. (1943) Theoretical Soil Mechanics, Wiley, N ew Y ork.
1.9 L am be, G .W . and W hitm an, R.V. (1969) Soil Mechanics, Wiley, N ew Y ork.
@seismicisolation
@seismicisolation
CHAPTER TWO
General principles
Before specific types of structures are considered, it will be useful to devote

a chapter to a review of fundamental aspects of the general problem. Theories
for analysing stress distributions in soils when buried structures or inclusions
are present are examined, together with a summary of test results.
The properties of soils under static and dynamic loading conditions
are discussed in the light of the considerable am ount of experimental work
carried out in the USA in recent times. The attenuation of dynamic surface
pressure with depth and the general problem of wave propagation in soils
have been examined in connection with the design of underground pro
tective structures to resist surface explosions, and this work is summarized.
2.1 Arching
An inclusion, in the form of an underground structure, in a uniform soil
field will cause a redistribution of the ‘free field’ stresses and deflections.
The nature of this redistribution influences the load that reaches the structure,
whether it is due to weight of the soil above the structure (the ‘overburden’)
or to a surface loading. The proportion of the superimposed load that reaches
the structure is governed by the characteristics of the soil, the geometry and
stiffness of the structure, and whether the surface loading is static or dynamic.
Because this redistribution in many practical cases results in a decrease
in loading over the deflecting or ‘yielding’ areas of a structure, and an increase
over adjoining rigid and stationary parts, the transfer of pressure has become
known as ‘arching’. It is particularly applicable to structures that deflect
enough under the action of the superimposed load to mobilize the shearing
resistance of the soil immediately adjoining the deflecting part. Theories
of arching that take account in various ways of the frictional resistance
along surfaces of sliding, or the plastic flow of soil located above the yielding
area, have been summarized by Terzaghi [2.1], and the reader is recom
mended to consult this work.
In m any instances, however, the state of stress in a soil-structure system
is not associated with sliding or plastic flow. Particularly in the early stages
of loading, the stress distribution can be assessed accurately by assuming
@seismicisolation
@seismicisolation
General principles 1
that the soil is perfectly elastic. This is especially true if the geometry of the
structure is such that large deflections do not take place, and if the structure
remains relatively undistorted until a catastrophic failure, such as buckling,
occurs. The application of the classic theory of elasticity to stress distribution
problems in underground structures, leading to the evaluation of elastic
stress concentrations in their vicinity, has also been summarized in later
chapters of ref. 2.1. Of particular im portance is the deviation of stress
distribution in the vicinity of vertical shafts and horizontal tunnels in semi
infinite elastic solids with a horizontal surface.
There have been m any experimental studies to examine stress distribution
and arching, the most famous being conducted by Terzaghi [2.2] in 1936
using a deflecting trapdoor in the base of a soil bin. He found that the
pressure acting on a long trapdoor was independent of the state of stress in
the soil located m ore than two or three door widths above the door. The
experiments were concerned with a plane strain condition, with only two
plane surfaces of sliding. In a three-dimensional situation, for example
a circular door, the equivalent distance is one to one-and-a-half diameters.
The results of one of Terzaghi’s experiments are given in Fig. 2.1. The
width of the long trapdoor is b, the depth of cover is H and the vertical soil
stress at a horizontal section at any depth below the surface is crv. The
vertical stress at z = H if there were no arching (i.e. the hydrostatic stress)
would be <rvh. It can be seen that for z/b > 2.5 there is no relief of vertical
Figure 2.1 Terzaghi’s tra p d o o r experim ents (n o te : z m easured upw ards from trapdoor).
@seismicisolation
@seismicisolation
8 Buried Structures
stress due to arching, but that when z = H, immediately over the trapdoor,
<rv is less than 10% of <rvh.
The form of the experimental curve can be predicted theoretically if it is
assumed that the surfaces of sliding that occur when the door is deflected
downwards are vertical. This arching analysis is due to Janssen, who assumed
also that the vertical pressure on the yielding element, as shown in Fig. 2.2,
is equal to the difference between the pressure due to the weight of soil
above the element and the frictional resistance along the sides of the element.
The shearing resistance is determined from the familiar equation
s = c + <7tan </>, (2 . 1)
where c is the cohesion (i.e. shearing resistance per unit area if compressive
stress is zero) and </>is the angle of shearing resistance.
The ratio between horizontal and vertical pressure is an empirical con
stant kr (not to be confused with subgrade m odulus or the constant of
horizontal soil reaction). The vertical stress on a horizontal section at any
depth z below the surface is <rv, and the norm al stress on the vertical sliding
surface is therefore kroy. Resolving for vertical equilibrium on a unit length
of the section, we obtain
by dz = b(crv + d<7v) —boy + 2c dz + 2k o y dz tan </>, (2.2)
where y is the unit weight of soil. If the uniform surcharge or surface loading
is q, then solving this equation gives
b( y - 2c/b)r 2z
exp I —kr tan </> + q exp - kr tan (2.3)
<Ty 2k r tan <f> |_
F or a cohesionless soil, such as dry sand, c = 0, and the vertical pressure on a
yielding strip at very large depths (o ) approaches the value
by
(2.4)
2kr tan </>
@seismicisolation
@seismicisolation
F urther experiments by Terzaghi, in which he found the variation in the
force required to extract flat metal tapes from positions in the soil fill above
a yielding trapdoor, suggested that k T increased from unity immediately
over the yielding surface to 1.5 at an elevation of b above the centreline of
the surface. At elevations greater than 2.5b, the movement of the trapdoor
did not alter the state of stress in the fill at these elevations.
If we take a typical value of </>= 30°, and consider the area just above the
trapdoor, where kr « 1, equation (2.4) becomes
by
(2.5)
<Jvo° = L16’
so that the pressure on the trapdoor at very deep covers is equal to the
hydrostatic pressure due to a depth of soil of about 0.9b acting on the door.
The form of the experimental curve in Fig. 2.1 can be explained if it is
considered that any soil beyond a distance of 2.5b above the door acts as a
surcharge, and that for the purposes of the analysis can be represented as
such. If we then apply equation (2.3), taking z = 0 when H = 2.5b, and taking
q to be the hydrostatic pressure due to the weight of soil above this elevation,
a theoretical curve similar to that shown in Fig. 2.3 can be deduced. This is
very like the experimental relationship in Fig. 2.1.
The negative exponential indices in equation (2.3) imply ‘active’ arching,
i.e. the condition resulting from the downward movement of the trapdoor
or the downward deflection of the flat roof element of an underground
structure. If the movement is upwards, tending to compress the fill, arching
is ‘passive’, and the indices are positive. Passive arching could result if the
buried structures were much stiffer than the surrounding medium, but for
thin-walled flexible structures arching is usually active.
The analysis described so far has been limited to a ‘plane’ condition,
i.e. a section of an infinitely long trapdoor or flat roof has been considered.
0\,/yb
z/b
Figure 2.3 T heoretical arching curve (n o te: z m easured dow nw ards from surface).
@seismicisolation
@seismicisolation
10 Buried Structures
This is a special case of a more general analysis for a yielding element having
finite length /, as well as finite breadth b. Van H orn [2.3] has shown that the
more general form of equation (2.3) is
L(y — 2c/L)
a = 1 —exp( —2kr tan J + <7exP ^ —2/cr tan J,
2kr tan </>
(2.6)
where L = bl/(b + /)• N ote that when l-> oq , L -►b, and that for circular or
square trapdoors or roof elements, L = b/2.
A series of experiments was undertaken in 1964-65 by M cNulty [2.4]
to investigate active and passive arching in sands, using a circular trapdoor
m ounted flush with the bottom of a circular soil container. The influence
of large surface pressures was investigated by subjecting the sand surface
r 6 x
| +S
k £ -H
(d/a) x io o o
Figure 2.4 M cN ulty’s experiments.
@seismicisolation
@seismicisolation
to air pressure. It was possible to vary the depth of cover, the trapdoor
diam eter and the properties of the soils. Test results were presented as
arching curves showing the relationship between the ratio of the average
pressure on the door to the surface air pressure and the ratio between
deflection and diam eter of the door. Several curves are summarized in Fig. 2.4,
which shows that the arching ratio falls off quickly to zero or some finite
limiting value for active curves, but that the approach to an ultimate value
for passive arching requires much larger movements of the trapdoor.
A useful study of m uch existing experimental arching data for plates
and cylinders in a granular soil has been conducted by Gill [2.5]. He pointed
out that, when the data were plotted as a function of a geometry-stiffness
factor, it was possible to form an empirical equation connecting this factor
and an arching factor. Referring to Fig. 2.5, suppose that the ‘free field’
pressure at the level of the deflecting element (i.e. the pressure at that level
if there were no inclusion) is pv, and that the actual interface pressure is pr,
then the portion of the load transferred away from the structure is 1 —p jp y.
This is known as the ‘arching’ A, and the relationship between this factor
and the geometry-stiffness factor R takes the form
(2.7)
where A 0 and n are experimentally determined constants, and R is given by
(2.8)
where A g is the geometry factor, M s is the secant modulus of the soil from a
confined compression test and S is the relative deflection between structure
and free field.
F or convenience, Gill writes
(2.9)
Arch ing, A = 1- p. /p^

H
m
UlO in-J
n— 18 in
Figure 2.5 G ill’s circular alum inium slabs.
@seismicisolation
@seismicisolation
so that equation (2.7) becomes
/ a Y~ a
( ‘- t J = e ~”n (2-,0)
Gill points out that the physical dimensions of the structure in relation
to the depth of burial would affect arching, and that the plan area is also
im portant. The non-dimensional term that he suggests to account for this
is SH /A s, where S is the perimeter of the structure in plan, As is the plan
area and H is the depth of cover. N ote that 2A s/S is equivalent for rec
tangular structures to Van H orn’s factor L, defined after equation (2.6).
Gill also considers the im portance of the stiffness of the soil in relation
to the stiffness of the structural response to loading, and he therefore suggests
the ratio p.J(S/b), where p{ is the average unit load reaching the structure,
and S/b is the ratio of displacement of the structure to its m ajor dimension.
This represents the strain-strain characteristic of the soil immediately
adjacent to the structure, and it seems logical that this should be compared
with the secant m odulus of the soil, M s .
Thus, Gill’s complete stiffness-geometry factor is {SH/As) («5Ms/bpj), so
that Ag in equation (2.9) is S H /A sb for horizontal cylinders. If the value 0.87
is assigned to A 0 and 0.135 to n, the curve compares closely with the results
of experiments on circular and rectangular (in plan) aluminium slabs
supported by peripheral footings and subjected to surface overpressure in a
soil test tank. These experiments are discussed in detail in a report by Gill
and True [2.6].
N ote the similarity between the arching equation (2.7) and the exponential
terms in equations (2.3) and (2.6). In fact, if kr = 1 and </>= 30° in these
equations, then
- , 1 - * 0.87.
2kr tan (p
Fig. 2.6, taken from ref. 2.6, shows the relationship between equation (2.7)
and the test results for first loadings on the circular slabs.
Further data on arching have been established by Gill and Allgood [2.7]
and Allgood et al. [2.8] as the result of static and dynamic loading tests on
small horizontally set buried arches. These will be discussed later in the
book, as will the results of tests on buried domes and vertical cylinders.
F or a further discussion on the fundamental equations of Terzaghi, the reader
should see a report by M ason [2.9], which examines the effect of the com
pressibility of the structure on the ‘sliding prism’ theory.
The horizontal plane above an inclusion at which the settlement under
load is similar to that of the adjacent free soil field is the plane of equal
settlement. All soil above the plane can be treated as a superimposed load
on a soil surface coinciding with the plane.
@seismicisolation
@seismicisolation
Stiffness -geom etry factor, R

Figure 2.6 G ill’s static arching experim ents : circular (O ) and rectangular ( □ ) structures.
The vertical distance between this plane and the upper surface of the
buried structure can be estimated by considering the limiting conditions of
the volume of soil immediately above the structure. For an inclusion of
width b and infinite length, we can use equation (2.4) for dry, granular soil,
so that the depth of soil between the plane and the structure, H 0, is given by
H by
°y 2kr tan </>’
or
. (2 . 11)
b 2k, tan 4> ' ’
The use of this equation in design calculations will be discussed in later
chapters.
@seismicisolation
@seismicisolation
Arching experiments in cohesive soils are few in number. Perhaps the
most useful tests have been reported by Jester [2.10], who conducted tests
in a buckshot clay having a cohesive strength of 15 pounds per square inch
(psi) at 26% water content and an angle of friction of zero. He concluded
that active and passive arching could be present in cohesive soils, and that
it occurs within a distance of one diameter above an inclusion. He underlined
the im portance of the problem of creep (discussed later with regard to
dynamic and static tests on thin-walled cylinders in clay).
Jester also noted two differing classes of behaviour. At lower loads,
the behaviour of the system is governed by the relative stiffness of soil
and inclusion. At higher loads, the limitation is the bearing capacity of the
soil underneath the inclusion.
We have so far discussed the effect of arching on the loading over the
upper surfaces of yielding structures, but if the underground structure is
in the form of a horizontal tunnel, then it is necessary to know the forces
on the side walls and the loads that cause upwards heaving of the base.
These have been examined by Terzaghi [2.11] for tunnels rectangular in
section. He points out that for sand the soil adjoining the side walls subsides
because of the yield inwards of the lateral support. An active Rankine state of
the adjoining soil results, with surfaces of sliding descending at an angle
45° + 0/2 to the horizontal, as shown in Fig. 2.7. The inclined boundaries
of the zone of subsidence therefore rise from the lower corners of the tunnel
at this angle, and at the level of the roof the width of the yielding strip is B,
given by
B = b + 2D tan(45° - 0/2), (2.12)
n a
t
7 (D+H)
Vertical pressure
on dd
Figure 2.7 R ectangular tunnel in com pacted sand (Terzaghi).
@seismicisolation
@seismicisolation
where D is the side depth of the tunnel. The vertical stress across the width
B is therefore found from equation (2.3) with B replacing b. F or a material
with c = 0, equation (2.3) then becomes
* = 2 [ 1 - exp ( - 4> f ) ] + q e x p ( - kt tan * f ) . (2.13)
On each side of the tunnel, this pressure acts like a surcharge on the top
surface of the wedge of subsidence. The total lateral force on the side walls
can then be calculated by Coulom b’s theory, treating the sides as retaining
walls. The total norm al com ponent pAn on one side wall is given by
jyD 2k A
PAn = ^ T— (2. 14)
sin a
where a is the angle between the sides and base of the tunnel (90° for the
tunnel shape considered above), and kA is the coefficient of active earth
pressure. The unit norm al pressure at depth z is given by
Pa. = ? A (2.15)
(since sin 90° = 1).
The above analysis does not take into account the effect of premature
buckling of the side walls that can occur in thin-walled rectangular tunnels.
This problem is discussed later.
The vertical pressure on a horizontal section through the tunnel will
take the form shown in Fig. 2.7, and this can produce a heaving pressure on
the base of the tunnel, where the vertical pressure is absent. The action of
this pressure is similar to the base failure of a vertical bank, and the reader
is referred to a discussion of this problem in ref. 2.11.
2.2 Elastic stress distribution

It was indicated earlier that at low loads, and for certain structural geome
tries, the stress distribution in a dry granular soil might be analysed by
considering the soil as an elastic medium rather than by examining shearing
failure stresses. It is often stated that, if the safety margin of a soil mass with
respect to plastic flow failure is about 3, the assum ption of elasticity is a
valid one. The key to the assum ption is the extent to which the stress-strain
relationships of the soil can be considered linear during the earlier stages of
loading.
2.2.1 Stress distribution around a circular cavity

The stress distribution in the vicinity of a cylindrical unlined cavity under
a state of elastic equilibrium was examined in the 1920s and 1930s by Schmidt
@seismicisolation
@seismicisolation
Figure 2.8 Axes of reference for M indlin’s analysis.
[2.12], Yamaguti [2.13] and Anzo [2.14]. However, the most im portant
contribution of that era was due to M indlin [2.15], who investigated the
problem of a horizontal cylindrical hole of circular cross section in a semi-
infinite elastic solid, under the action of gravitational forces. He did not
consider the action of a pressure or surcharge norm al to the surface in
addition to the weight of the elastic medium.
M indlin took the axes of reference shown in Fig. 2.8, and assumed that,
because the length of the cavity is long in com parison with its diameter, the
problem could be treated as one of plane strain, i.e. the stresses are everywhere
parallel to the xy-plane and independent of z, and that the resulting dis
placements parallel to z, along the length of the tunnel, can be allowed.
This enables two-dimensional solutions of the fourth-order differential
equations to be employed, because the applied stress field is assumed not
to vary along the length of the cavity.
M indlin also considered three states of stress that could exist in the
semi-infinite solid before the boring of the cavity (Table 2.1). These are
Table 2.1 Three states of stress th at can exist in a semi-infinite solid
State A Each elem ent is in a state of isotropic pressure due to the

(hydrostatic) weight of the m aterial, so th a t there will be a linear distribution
of norm al pressure, ox = — yy along the boundary
State B Initially, long before the tunnel was bored, the m aterial was
(lateral) com pletely restrained from lateral displacem ent. It can be
show n th at this gives a linear d istribution of pressure,
ax = [ — v/(l —v)]yy along the b o undary
State C C om plete absence of lateral stress, so th a t a x = 0

(unidirectional)
@seismicisolation
@seismicisolation
(a) (b )
crX =0
u tL .-,
(c)
nmnmk
Figure 2.9 B oundary conditions for the semi-infinite solid (Mindlin). (a) Case A,
hydrostatic. k0 = 1; (b) Case B, no lateral deform ation, k0 = v/(l — v ); (c) Case C, no
lateral restraint, k0 = 0.
Figure 2.10 Values of oJay and oJ<Jy (O bert and Duvall).
@seismicisolation
@seismicisolation
shown in Fig. 2.9 in the form of conditions along vertical boundaries normal
to the x-axis, situated a great distance from the cavity. These three initial
states correspond to values of the coefficient of earth pressure at rest, k0,
of k0 = 1, k0 = v/(l — v) and k0 = 0. N ote how the second of these states
depends critically on the value assumed for v. When v = ^, kQ when
v = 3 >^o = 2 >and when v = j , k 0 = 1.
The elastic solution is greatly simplified if a bipolar system of coordinates
is used, as originally applied by Jeffery [2.16]. The results of the analysis for
a single circular hole can then be presented in the form of radial stress <rr,
tangential stress oe, and shear stress xr0, where 9 is the polar coordinate and
9 = 0° represents the horizontal axis. The exact solution for a hole far enough
from the surface to enable any alteration in stress due to the removal of
material from the hole to be ignored is given in Fig. 2.10. The solution for
k0 = 0 is similar to that for a hole in an infinite plate under a unidirectional
stress field. The general solution is
(2.16)
where r 0 is the hole radius, and r is distance from hole centre.

Notice that the stresses are independent of the elastic constants and the
numerical value of hole radius. Fig. 2.10 gives values of o J o y at 9 = 0 for
various values of r/r0. It can be seen that the tangential stress concentration
at 9 = 0° is a m aximum at the edge of the hole, and decreases quickly with
distance from the edge, so that, at values of r/r0 > 3, ae approaches oy. The
radial stress concentration changes from zero at the edge of the hole to a
value almost equal to the applied stress in that direction at r/r0 = 4.
The distribution of ae around the edge of the hole is shown in Fig. 2.11
for the range 9 = 0 to n. F or k Q= 0, the stress concentration varies from + 3
at 9 = 0 to - 1 at 9 = tc/2. Thus, for a compressive stress applied in the
y-direction, a zone of tension exists along the boundary between 9 = n/3
and 2n/3. When k 0 = 1, i.e. ax = ay, the stress concentration is 2 at all points
on the boundary of the hole. For a more detailed discussion of these results, see
a recent book by Obert and Duvall [2.17].
The distribution of od for depths of burial of the cavity that are not great
enough to ignore the effect of removing material from the original elastic
m edium has been given by M indlin in the form of a relationship between
o J ly R and (H + R)/R for the highest point (crown) of the tunnel. This is
@seismicisolation
@seismicisolation
Figure 2.11 D istrib u tio n of ag under a com pressive stress in the ^-direction (O bert and
Duvall).
shown in Fig. 2.12 for k0 = 1, and from it the following conclusions can be
d raw n :
1. The disturbing effect of the upper free boundary (surface) is not felt when
{H + R)/R > 2.0 (or H /R > 1.0), if v = 0.5.
2. F or (H + R)/R > 2.0, stresses <rd decrease linearly with depth.
M ore recently, the stress distribution around a tunnel has been examined
by means of the finite element method of analysis, but as this work has been
aimed mainly at the buried culvert problem, further discussion will be
left for C hapter 4.
Figure 2.12 V ariation of a0 with depth of burial.
@seismicisolation
@seismicisolation
2.2.2. S tr e ss d istrib u tio n a ro u n d n o n -circu la r c a v itie s
A solution for the stress distribution in a uniformly loaded elastic plate

containing a small hole of a generalized shape was obtained by Greenspan
[2.18]. He expressed the coordinates of the boundary of the hole in the form
x = p cos /? + r cos 3/?, y = q sin /? —r sin 3/?, (2.17)
where p, q and r are parameters, and /? is an angle. By adjustment of the
values p, q and r, a variety of hole shapes can be obtained - circles, ellipses,
approximate rectangles with rounded comers - all having symmetry
about the x- and y-axes. If b and h are the overall dimensions in the x- and
y-directions, then
b = (p cos 0 + r cos 3P)p =0 — (p cos 0 + r cos 30)p =lso ,
so that
b = 2(p + r). (2.18)
Similarly,
h = (q sin /? —r sin 3P)p =90 —(g sin /? —r sin 3/?)'/?= 270 ’
so that,
h = 2(q + r). (2.19)
G reenspan’s solution for the tangential stresses at the boundary of the
hole is
[ ( p 2 + 6 rq) sin2/? + (q 2 + 6 rp) cos2/? - 6r(p + q) cos2 2/? + 9 r 2]<r0
= K + a )(p2 sin2 ft + q2 cos2 fi - 9r2) - r (p + q)2 sin 20

x y xy p + q + 2r
(2.20)
where ax and oy are states of stress at points remote from the hole, and xxy
is the constant shearing stress.
The results of a num ber of applications of this solution have been given
by O bert and Duvall [2.17], and three of these are shown in Figs 2.13, 2.14
and 2.15. Some useful conclusions can be drawn from these diagrams:
1. F or an applied stress parallel to one axis of an elliptic hole, the stress
concentration at the end of that axis is approxim ately — 1 (as we found
earlier for circular holes).
@seismicisolation
@seismicisolation
Figure 2.13 B oundary stresses for elliptic opening (b/h = 2) (O bert and Duvall).
Figure 2.14 B oundary stresses for square with rounded corners - sides parallel to axes
{b/h = 1) (O bert and Duvall).
@seismicisolation
@seismicisolation
Figure 2.15 B oundary stresses for square with rounded corners - diagonals parallel to
axes {b/h = 1) (O bert and Duvall).
2. For an applied stress parallel to the m inor axis, the stress concentration
at the end of the m ajor axis increases as the ratio (major axis)/(minor axis)
increases, and as the radius of curvature at the end of the major axis
decreases.
3. F or an applied stress parallel to the m ajor axis, the stress concentration
at the end of the m inor axis decreases as the ratio (major axis)/(minor axis)
increases.
The accuracy of this analysis was improved by Brock [2.19], who added
more terms to the functions in equation (2.17) to enable rectangular openings
with rounded corners to be mapped less approximately. His equations were
x = (A + B) cos p + C cos 3P + D cos 5p + E cos Ip + ... (2 21)
y = (A — B) sin p — C sin 3p — D sin 5P — E cos Ip — ...
Using these, he produced results for the stress distribution on the boundary
of square holes as a function of the radius of curvature of the corner, as
shown in Fig. 2.16. The results for a circular opening are shown on the same
@seismicisolation
@seismicisolation
ft (d e g)
Figure 2.16 B oundary stresses for squares with rounded corners (Brock).
axes. N ote that the maximum stress concentration for r/h = f is about 2.8,
which is less than the m aximum for a circle of 3.
A further study by Heller et al. [2.20] extended this work to account for
rectangles with variable h/b as well as r/b. The maximum stress concen
trations for a num ber of values of these param eters are given in Fig. 2.17.
Using these and earlier results, Obert and Duvall [2.17] produced a very
useful com parison between maximum stress concentration and shape for
three different types of opening: ellipse, ovaloid and rectangle. This is shown
in Fig. 2.18.
Stress concentrations around non-circular holes have also been discussed
in a book by Savin [2.21], who compares experimental data with theoretical
results, and illustrates the close agreement. The isostatics, or lines of principal
stress, for a square hole under a uniaxial stress, found experimentally by
@seismicisolation
@seismicisolation
O 0*1 0-2 0-3 0-4 0-5
r/b
Figure 2.17 M axim um stress concentrations for rectangular openings w ith rounded
corners. U niaxial stress field, ay (Heller et al.) (after O b ert an d Duvall).
Figure 2.18 M axim um stress concentration for different shape openings - uniaxial
stress field, ay (O bert and Duvall).
@seismicisolation
@seismicisolation
Coker and Filon [2.22], are given in the book. The stress concentration at
the centre of the sides parallel to the stress field is shown to be 1.4, which
com pares closely with the theoretical value of 1.47. It is of interest to note
that the hole has virtually no influence on the stress field at a distance greater
than about twice the side length from the centre of the hole.
2.2.3 Two circular cavities in close proximity

The stress distribution near a line of holes in a plate subjected to an applied
stress in a direction norm al to the line joining their centres has been analysed
by Howland [2.23], and his results in the form of stress concentration
factors for the case when the distance between the edges of the holes is
equal to their diameters is shown in Fig. 2.19.
The figure gives the stress concentration o j s at points around the
boundary of one hole, where, as before, sy is the state of stress in the direction
at a point remote from the hole. Com parison with the results for a single hole
in an infinite plate (Fig. 2.10) shows that the maximum stress concentration
has increased from 3 to 3.26, which is a relatively small increase. Fig. 2.19
also shows the ratio oy/sy along the line joining the centres of the holes.
At the centre, the ratio is about 1.6, reducing virtually linearly along the
y-axis through the centre, until at a distance y/r0 = 3 from the centre the
concentration is 1.0. This suggests that a line of holes, spaced one diameter
Figure 2.19 Stress co ncentration for two circular holes. Applied stress norm al to line
of centres. C entres of holes tw o diam eters ap art (Howland).
@seismicisolation
@seismicisolation
apart (two diameters centre to centre) have no influence on the stress field
at a distance greater than about 1.5 x hole diam eter from the m idpoint
of the line joining the hole centres.
The influence of decreasing the distance between hole centres has been
examined by Ling [2.24], with reference to two circular holes in a uniaxial
or biaxial stress field. The values of stress concentration a e/ s y at the horizontal
diam eter (6 = 0, n) for various com binations of sx and sy are given in Fig.
2.20. The most interesting curve to us is for sy ^ 0, sx — 0, i.e. applied stress
in a direction norm al to the line joining the hole centres. Note that when
a/D = 1.0 (where a is the distance between the edges of the holes), the concen
tration is 3.07, rather less than that resulting from an infinite row of holes
(3.26). F or values of a/D < 0.5, there is a rapid increase in stress concentration,
which becomes 3.87 when a/D — 0. However, at this stage the average stress
in the pillar between the holes has increased to infinity, so the concentration
factor is of academic interest. In general, small values of a/D should be
avoided unless special precautions are taken. F or values of a/D > 1.0, the
stress concentration is virtually equal to that for a single hole in an infinite
plate.
4 0
30
2 -5
2-0 1 « 1 1
O 0-5 1 1-5 2
a/D
Figure 2.20 Influence of decreasing distance between hole centres (Ling).
@seismicisolation
@seismicisolation
2.2.4 Thick-cylinder theory

It is sometimes convenient to consider the stress field around a hole
enclosed by a cylinder of elastic material, when the outer surface of the
cylinder is acted on by a uniform radial pressure. The general solution
for a thick-walled cylinder having an internal radius r0, an external radius
rex, and subjected to an internal pressure of p0 and external pressure of
pex(Fig. 2.21) is
(2.22)
N ote that the sum of or + cQis constant and independent of r. For a thick
cylinder subjected to external pressure only, we have
(2.23)
Figure 2.21 Thick-w alled cylinder under internal and external pressure.
@seismicisolation
@seismicisolation
re x / r o
Figure 2.22 V ariation of m axim um tangential stress with rex/r 0 for thick-walled
cylinder.
When r = rQ, or = 0, and

2 r e2xpr ex
(2.24)
If rex -►oo, ae -*■ — 2pex, which is the same stress concentration that we noted
earlier for the biaxial stress field a x = o y . The variation of the ratio o O'j pr ex as
the ratio rex/rQincreases from unity to 5 is shown in Fig. 2.22, which indicates
that the effect of the value of rex is hardly felt when reJ r 0 > 2.0. N ote that this
conclusion is similar to that drawn in Section 2.2.1 about the disturbing
effect of a free boundary, which suggests that thick-cylinder theory gives a
good approxim ation to the analysis of a hole in a biaxial stress field near a
free boundary where the distance from the boundary to the hole centre is
equal to the external radius, rex, of the thick cylinder.
It can also be shown (see for example the book by Wang [2.25]) that the
radial displacement (u) at any point on the inner surface of the elastic cylinder
is given by
(2.25)
so the change in diam eter is twice this value. It is of interest to note that,
combining equations (2.25) and (2.24), we obtain
(2.26)
E
which is not surprising, since u/r0 is the diametral and therefore the circum
ferential strain around the inner hole.
@seismicisolation
@seismicisolation
-F
Thickness A
Figure 2.23 C oncentrated line load on the b o u n d ary of a semi-infinite plate.
2.2.5 Concentrated free surface loads

If a semi-infinite plate of thickness t is acted on by a concentrated line load
F, acting norm al to the plate and uniformly distributed across the thickness,
it is convenient to replace the concentrated force by an equivalent stress
distribution acting on a small cylindrical surface of radius a, as shown in
Fig. 2.23.
If the stress distribution in a radial sense is (<7r)r=a, where
_ — 2F sin 9
y r’r=a nta ' '
then the vertical com ponent is F and the horizontal com ponent is zero.
Analysis shows that, with this treatm ent, the radial stress at a distance r from
the point of application is given by
a (2.28)
so that or is inversely proportional to radius.

If the concentrated load occupies a defined area, and can be thought of as
a uniform stress/d istrib u te d over an area of length 2a and width t, then by
using the results of the concentrated load analysis it can be shown that the
stress distributions along the y-axis (see Fig. 2.24), expressed as the ratio
oy/ f for various values of y/a, take the form shown in the figure. At larger
@seismicisolation
@seismicisolation
Figure 2.24 Stress distrib u tio n along y-axis for a uniform stress / over a length 2a on
edge of elastic half-space.
values of y/a, the stress distributions are very similar to those for a point load.
In general, at any point in the elastic half-space subtending an angle /?
with the ends of the distributed load, we have
a = — —(ft — sin /?),

n
c = - t ( P + sinp), (2.29)
y 71
x =0.
These equations are all derived from the original analytical work on point
loads on a semi-infinite solid by Boussinesq [2.26], reported in some detail
in ref. 2.11.
2.2.6 Stress distribution around a spherical cavity

A theoretical solution for the stress distribution around a spherical cavity in
a unidirectional stress field was first obtained by Southwell and Gough
[2.27] in 1926. D istributions around ellipsoidal cavities were examined by
Sadowsky and Sternberg [2.28, 2.29] in 1947 and 1949. Subsequently a
sum m ary was given by Terzaghi and Richart [2.30] in 1952.
It is inappropriate to give details of the analysis here, but the results in
term s of stress concentration factors are interesting. The stress distribution
around a ‘spherical cavity’ in a uniformly distributed applied stress field sz,
acting in the z-direction at a large distance from the cavity, is shown in
Fig. 2.25. The distribution is dependent on Poisson’s ratio v, and although
@seismicisolation
@seismicisolation
1 1 1 1 1 1 s,
z
Figure 2.25 Spherical cavity notation.
changes in the ratio do not have a marked effect on the tangential stress ae on
boundaries denoted by z = 0 (the equatorial plane), they exert a large in
fluence on the values of oe at the top of the cavity, z = r0, where r0 is the cavity
radius.
The m aximum stress concentration, occurring at the boundary of the
cavity in the equatorial plane, is o J sz = 2.0 (when v = 0.2), rising to 2.045
when v = 0.3 and 2.167 when v = 0.5. We saw earlier that the corresponding
concentration for a circular tunnel is 3.0.
At the top of the cavity, the concentration, as for a circular tunnel, is
negative, and has the values a j s z = —0.5 (when v = 0.2), increasing num eri
cally to —0.682 when v = 0.3, —0.900 when v = 0.4, and — 1.167 when
v = 0.5. The concentrations for v = 0.2 are shown in Fig. 2.26.
2*0
r/r „
1-5
A 1-0
0 -5
r /r ,l
(a) (b )
Figure 2.26 Stress co n cen tratio n s w hen v = 0.2 (spherical cavity), (a) 0 = 0, to p of
cavity; (b) 6 = n/2, b o u n d a ry at eq u ato rial plane.
@seismicisolation
@seismicisolation
In general,
(5 - 10v)r2
on = 1 4 - lOv + 21- j ) sin2 0
1 4 - lOv
(9 - 15v)r2 12^o
+ (2.30)
so that, when 6 = 0, r = r0 (the top of the cavity),

oe 3 + 15v
(2.31)
sZ 14 — 1 0 v ’
and when 6 = tc/2, r = r0 (the equatorial plane),

2 7 - 15v
(2.32)
1 4 - 10v'
By superimposing the results, the concentrations for biaxial and triaxial

stress fields can be found. The triaxial field, corresponding to a hydrostatic
stress condition ( —s) gives the simple result
1 -3
s \ r6
(2.33)
2r
W hen r = rQ, along the inside of the cavity, a j s = 0, and o j s = —\ .
After discussing the results for ellipsoidal cavities and comparing them
with circular tunnels and spherical cavities, Obert and Duvall [2.17] came
to the conclusion that for engineering purposes the maximum stress concen
trations for single, isolated openings can be conservatively estimated by using
the concentrations for holes in plates, where the cross section of the hole is
similar to that of the three-dimensional opening under consideration.
2.2.7 Stress distribution above a yielding trapdoor

The ‘arching’ analysis of Section 2.1, based on the theory of plastic or sliding
flow of the soil, has been supplemented in recent years by an examination in
terms of elastic stress distribution, by Finn [2.31] and Chelepati [2.32, 2.33].
These solutions assume a rigid, horizontal boundary in the soil, two-
dimensional boundary conditions and artificially controlled displacements of
the strip. Finn limited his analysis to an infinite depth of soil, but Chelepati
investigated the m ore practical situation of a soil field of finite depth, with a
horizontal free surface subjected to a high overpressure. He argued that if
there were no displacement of the horizontal strip, the pressure p transm itted
@seismicisolation
@seismicisolation
to the strip is given by
p = q + yH, (2.34)
where q is the overpressure, H is the depth of cover and y is the density of the
soil, but that if the yielding strip were depressed by an am ount d, the total
force on the strip would be relieved by a quantity equal to the tensile forces
induced by the displacement d. Assuming the principle of superposition, the
arching is in fact equal to the total tensile force on the strip, but the analysis
must take account of the presence of very high stresses towards the edges,
owing to the discontinuous displacements there. Since the soil is not capable
of transm itting tensile stresses, these do not contribute to the arching. This is
explained in Fig. 2.27: Fig. 2.27(a) represents a uniform compression on the
unyielded strip, according to equation (2.34); Fig. 2.27(b) is the tensile stress
distribution due to displacement d alone (with infinite stress at the disconti
nuity); and Fig. 2.27(c) is the resulting stress distribution (shown by the
shaded area). N ote that when the tensile stresses reach the value p, the region
beyond this is assumed to be in a state of tension and not effective. The
am ount of arching is therefore represented by the area shaded in Fig. 2.27(a)
minus the area shaded in Fig. 2.27(c).
Taking the axes of reference at the m idpoint of the strip, and the total
width of the strip as 2b, Chelepati computes the value of the tensile stress
oy in terms of the param eters x/b, d/H and Poisson’s ratio v, where x is the
Tensile stress
infinite at
discontinuity
(a) (b)
ii
(c)
Figure 2.27 Stress d istribution above yielding tra p d o o r (Chelepati). (a) uniform
com pression, p = q + y H ; (b) tensile stress d istribution due to displacem ent d; (c) result
ing stress distrib u tio n (compression).
@seismicisolation
@seismicisolation
x/b
Figure 2.28 Pressure d istribution on strip, v = 0.33 (Chelepati).
distance along the strip from its mid-point. N ote that d/H indicates the
uniform strain in the elastic medium if the strip extended to infinity in both
directions, and (dE/H) is the corresponding stress.
The results show that v has little effect on the distribution except at low
depths of cover given by b/H — 1.00. F or most practical cases, i.e. b/H < 0.8,
the differences are marginal. The distribution of tensile stresses given by the
param eter oyH/dE for v = 0.33 is shown in Fig. 2.28. The analysis next
located the value of x/b at which the resultant pressure becomes negative,
and using this the total arching, expressed as a percentage of the total load
(2pb) on an unyielding strip, can be computed. Again, the results are largely
independent of v when b/H < 0.8, and are given for v = 0.33 in Fig. 2.29.
To com pare the change of arching with depth of cover for a given deflection,
let us examine the results for a condition expressed by pH/dE = 10. Then, the
relationship between arching and 2b/H is as shown in Fig. 2.30. As we are
@seismicisolation
@seismicisolation
pH/ dE
Figure 2.29 R elationship betw een percentage arching and pH/dE, v = 0.33 (Chelepati).
2 b/H
Figure 2.30 R elationship betw een arching and 2b/H (Chelepati).
considering a purely elastic medium, there will be no ‘plane of equal settle

m ent’, and the results are expressed from 2b/H = 0.1 to 2.0.
2.3 Properties of soils

The properties of soils form a separate subject, and when properties under
dynamic as well as static loading conditions are considered, it becomes a
@seismicisolation
@seismicisolation
subject of great complexity. The intention here is to discuss a num ber of
fundamental aspects of soil behaviour that have a direct bearing on our
discussions on the strength, stability and soil-structure interaction properties
of buried structures.
There are several types of environm ent: undisturbed granular soils, clays
and rocks, and ‘placed’ granular soil - often a compacted backfill around an
underground structure placed in an excavated hole or trench. Allgood [2.34]
has m ade a useful survey of information on material properties essential to
the design of buried structures, in which he makes the im portant observation
that clays are not suitable for soil structure systems that have to resist high
loading. The higher m odular and shear strengths of granular soils, together
with their relative unsusceptibility to plastic flow and creep under load, make
them desirable for construction or backfilling, particularly if they have
sufficient m oisture to give them some cohesive strength.
2.3.1 Static properties o f granular soils

The natural properties of a soil are norm ally expressed as angularity, grain
size distribution, mass density, water content and void ratio. These give rise
to three ‘intrinsic’ properties - cohesion, angle of friction and pore pressure -
which are independent of the size of the soil mass or its boundary conditions,
but which govern the shear strength. These have been defined in a report by
W hitm an [2.35].
The static ‘m aterial’ properties are those given by laboratory tests on soil
specimens, i.e. the confined compression test leading to a stress-strain
relationship, from which the secant confined compression modulus M s and
the unloading modulus can be obtained, and also the triaxial shear test,
which gives the shear stress-strain relationship and volume change curves.
These allow the shear strength and shear m odulus G to be derived.
For static application, the soil stress-strain curve for initial, not repeated,
loadings is of m ajor interest, and a typical curve from a confined compression
test is shown in Fig. 2.31. The relationship between M s and vertical stress
<rv for an unsaturated, medium sand, well compacted,has been approxim ated
by Luscher [2.36] as
M s ^ 1000<rv°-8, (2.35)
and an alternative, linear relationship has been given for ‘Frenchman Flat’
soil in the USA (a hard, friable tan silt) by Hendron and Davisson [2.37] in
the form
M s ^ 4 0 0 0 + 3<7v, (2.36)
where M s and <rv are expressed in pounds per square inch (psi). This approxi
m ation covers the range <rv = 500 to 5000 psi.
@seismicisolation
@seismicisolation
Figure 2.31 Typical curve from confined com pression test.
These are merely examples to indicate the type of simple relationship that
can be used for design purposes when accurate soil data are not available.
For a particular design problem, it is im portant to have an accurate analysis
of the soil and the results of confined compression tests, if available.
From a knowledge of M s it is possible to obtain a value of E, related to a
given pressure, which can be used in an elastic analysis, by means of the
equation
(1 + u)(l - 2v)
(2.37)
E= i - V -
where v is Poisson’s ratio. N ote that when v = f , E = § M s .

Typical shear stress-strain and volume change curves from triaxial tests
are shown in Fig. 2.32 (taken from ref. 2.35). Successive load cycles produce
similar stress-strain relationships, but there is generally a continuing volume
increase. In the absence of a triaxial test, it is possible to use equation (2.1)
to obtain the shear strength, i.e.
s = c + a tan </>, (2.38)
if the values of c and 0 are known from other tests. A more accurate version
of this equation, to allow for the effects of pore water pressure, has been given
by Terzaghi:
s = c + (o —w)tan </>, (2.39)
where u = ul + u2— u3, and iq is the initial pore pressure,u2 is the excessive
pore pressure developed during shear at constant water content, and u3 is
@seismicisolation
@seismicisolation
Decrease possible for soft soils
Shear strain
Figure 2.32 Shear stre ss-strain and volum e change curves (W hitman).
the excessive pore pressure reduced by consolidation during loading.

Equation (2.38), however, is thought to be sufficiently accurate for most
design purposes.
The shear m odulus G is the slope of the shear stress-strain curve at the
stress under consideration.
M any of the experiments carried out by the author, and described else
where in the book, used a sand having the typical characteristics shown in
Table 2.2.
The value for E is given for low stress conditions, i.e. the value of Es
Table 2.2 Typical characteristics of sand
Unified Soil Classification G ro u p Symbol SP

(poorly graded sand, m edium grain size, with no fines)
Particle size distribution

Fine gravel (6.00-2.00 mm)
C oarse sand (2.00-0.60 mm)
M edium sand (0.60-0.20 mm)
Fine sand (0.20-0.06 mm)
Silt and clay (0.06 mm) 1%
C om position characteristics (BS 1377:1967, test 13)
(V ibrating ham m er com paction)
O ptim um m oisture content 10%
M axim um dry density 112 lb/ft3
E 5 x 103 to 104 psi
@seismicisolation
@seismicisolation
corresponding to relatively low overpressures, of the order of 30 to 100 psi,
because most of the test results give critical overpressures of this order of
magnitude.
2.3.2 Static properties o f cohesive soils

Although clays and similar soils are not ideal media in which to bury struc
tures, there is sometimes no alternative, and account must be taken of their
properties.
Clays are often specified in terms of specific gravity, liquid and plastic
limits, m oisture content, density (wet and dry) and shear strength c (see
equation (2.1)). F or the purposes of analysis, it is convenient to assume that
clay behaves as an elastic medium, and Cooling and Skempton [2.38] have
proposed that the relationship between E and c can be expressed approxi
mately as
E = 140c. (2.40)
This is not an ideal way of treating the action of clays, because their behaviour
under a superimposed stress is time-dependent, but the usefulness of the
simple relationship is dem onstrated in the chapter on buried thin-walled
cylinders.
Experiments have been carried out by the author in a high-plasticity clay
and a loam (low-plasticity clay) with the characteristics shown in Tables 2.3
and 2.4. Using equation (2.40) the ‘equivalent’ values of E are:
clay £ = 1 1 2 0 psi,
loam E = 2800 psi.
Table 2.3 Typical characteristics of clay
Unified soil C lassification G ro u p Symbol C H

(inorganic clay of high plasticity), specific gravity 2.74
A tterbury limits
Liquid limit 60
Plastic limit 22
Plastic index 38
C om position characteristics (BS 1377: 1967, test 12)

(Heavy com paction)
O ptim um m oisture content 30.4%
M ean dry density 92 lb/ft3
M ean wet density 120 lb/ft3
Shear strength 8 psi
@seismicisolation
@seismicisolation
Table 2.4 Typical characteristics of loam
Unified Soil Classification G ro u p Symbol CL

(inorganic clay of low plasticity), specific gravity 2.70
A tterbury limits
Liquid limit 36
Plastic limit 21
Plastic index 15
C om position characteristics (BS 1377: 1967, test 12)

(Heavy com paction)
O ptim um m oisture content 19%
M ean dry density 100 lb/ft3
M ean wet density 119 lb/ft3
Shear strength 20 psi
Table 2.5 Typical characteristics of rocks
Granite
C om pressive strength max. 42.6 x 103 psi
min. 23.0 x 103 psi
range 25-35 psi for > 50% of data
Static elastic m odulus max 11.0 x 106 psi

min 2.5 x 106 psi
Limestone
Com pressive strength max. 37.6 x 103 psi
min. 5.3 x 103 psi
range 20-3 0 psi for > 50% of data
Static elastic m odulus max. 12.0 x 106 psi
min. 4.2 x 106 psi
Sandstone
C om pressive sterngth max. 34.1 x 103 psi
min. 4.8 x 103 psi

min. 1.4 x 106 psi
Marlstone
Com pressive strength max. 28.2 x 103 psi
min. 10.4 x 103 psi
range 10-20 psi for > 50% of data
min. 0.6 x 106 psi
@seismicisolation
@seismicisolation
2.3.3 Static properties o f rock

The properties of rock from intact samples are not strictly relevant to the
behaviour of rock in the vicinity of a flexible underground structure, but have
been included here for reference. The engineer is more likely to be concerned
with the mass properties of jointed rock.
The static mechanical properties of rock are normally obtained from
uniaxial (unconfined) compressive, tensile, shear and flexural tests; and from
triaxial compressive and shear tests. The properties of some common rocks
have been discussed by O bert and Duvall [2.17], and typical values taken
from their work are given in Table 2.5. A typical uniaxial compressive
stress-strain curve for a granite (in this instance a Colorado Granite) is
shown in Fig. 2.33. N ote the slight S shape to the loading curve, which is
typical, although there are some cases of a virtually linear elastic relationship
(Georgia G ranite for example). It is possible, if there are microscopic cracks
in rock, for the stress-strain relationship to be similar to that for confined
sand, as shown in Fig. 2.34, which refers to a limestone specimen. In general,
however, it is reasonable to assume linear elastic behaviour for most rocks.
(Figs 2.33 and 2.34 are taken from ref. 2.17.)
2.3.4 Dynamic properties o f soils

The effect of strain rate on the properties of dry granular soils is not marked.
The compressive properties are virtually unaffected, and the angle of friction
(p is rarely changed by more than 10 or 15%. F or wet granular materials, the
Compressive strain (microinches per inch)

Figure 2.33 S tress-strain curve for C olorado G ran ite (O bert and Duvall).
@seismicisolation
@seismicisolation
in
a
«/>
5
8
2!
a
E
o
O 500 1000 1500 2000

Compressive strain (microinches per inch)
Figure 2.34 S tress-strain curves for lim estone (O bert and Duvall).
excess pore-water pressure resulting from consolidation is affected by strain

rate, but by and large the shear resistance is not. The differences in pore-
water pressure at high strain rates can increase the static values of physical
properties by a value of 2 in moist sands and also in cohesive soils.
It is known, too, that short-duration surface pressures decay rapidly with
depth of soil, particularly pressures due to high explosives. The attenuation
of blast pulses in this way is discussed in the section on loading.
H endron and Davisson [2.37] have com pared the static and dynamic
constrained moduli for ‘Frenchm an Flat’ soil, found on the US Nevada Test
Site, which has a cohesion of 8.5 psi, an angle of internal friction of 34.9° and
0 10 20 30 40x10'
Constrained secant moduli; static tests (p s i)
Figure 2.35 C om parison of static, rapid and dynam ic m oduli (H endron and Davisson).
@seismicisolation
@seismicisolation
a dry density in the range 84-87 lb /ft3. They concluded that, for short
rise-time dynamic loading (2-3 ms), the dynamic constrained m odulus is
about twice the static modulus, but that, for rise-times of 100 ms or longer,
the m odulus is virtually equal to the static m odulus for an undrained sample.
This is shown in Fig. 2.35, taken from their report, which compares the
constrained secant moduli M s from static tests with the moduli from rapid
tests (rise-time > 100 ms) and dynamic tests (rise-time 2-3 ms). The secant
m oduli-axial stress relationships are shown in Fig. 2.36. In the range of axial
stress from 2000 to 10 000 psi, the dynamic results can be represented by the
approxim ation
M s = 13 000 + 4(7, (2.41)
where M s and o are expressed in pounds per square inch.
The dynamic shearing properties of compacted clay have been examined
by Olson and Parola [2.39], who conducted over 100 triaxial compression
tests on clay com pacted at water contents ranging from 9% dry of optimum
to 3% on the wet side. The specimens were subjected to continuing pressures
ranging from 10 psi to 1000 psi, and were loaded to failure in times ranging
from 2 ms to 1 h. A change in failure time from 100 min to 10 min only
resulted in a marginal increase in compressive strength, whereas a change
from 60 ms to 6 ms resulted in an 18% increase. The secant modulus at 1%
axial strain increased steadily at the rate of 15% per decade reduction in
failure time.
x103
Axial stress, cr ( p s i )
Figure 2.36 R elationship between static, rapid and dynam ic secant m oduli and axial
stress (H en d ro n an d Devisson).
@seismicisolation
@seismicisolation
0-6
6 ms 6 0 ms 600 ms 6s 60 s 600 s 6000 s
Time to failure
Figure 2.37 R elationship between secant m odulus and tim e to failure for clay (Olson
and Parola).
The results are summarized in Fig. 2.37, which relates the secant modulus
at 1% axial strain for a given failure time, M s(f), with the same modulus for a
failure time of 1 min, M s( 1 min), for a range of values of time to failure. The
curve is a single line drawn through the scatter band of results for water
contents varying between 6.5 and 13.5%, and for confining pressures of 100
and 1000 psi.
E
JO
U 1ms 10 ms 100ms Is 10 s 100 s 1000 s 10 00 0 s

Time to failure
Figure 2.38 R elationship between com pressive strength and tim e to failure for clay
(O lson an d Parola).
@seismicisolation
@seismicisolation
Fig. 2.38 shows the scatter band of results relating the compressive strength
for a given failure time t with the strength for a failure time of 1 min for a
range of values of time to failure.
O ther investigations of dynamic properties have been made by Olson and
Kane [2.40], Casagrande and Shannon [2.41] and Richardson and W hitman
[2.42],
2.3.5 Wave propagation in soils

U nderground structures used as protective installations against surface blast
pressures that result from very large explosions have to withstand loads due
to airblast-induced ground m otions in the superseismic region. Because
of the dynamic character of the loading, there is an attenuation of vertical
stress with depth that is a function of the energy absorption characteristics
of the soil. It has been shown that the attenuation of peak vertical stress due
to a large air burst with depth of soil is influenced significantly by hysteresis
in the soil, and consequently the variation of peak vertical stress with depth
is a function of the type of soil, as well as of the yield of the explosion, the
overpressure and the depth of soil.
There are two types of air-blast-induced ground motion. In those where
the velocity v of the shock front (see Fig. 2.39) exceeds the dilational seismic
velocity u of the soil, the sloping shock front is then made up of vertical
pressures that propagate through the ground, because v > u. If v = w, the
ground shock front becomes almost vertical near the surface. If v < u, the
shock front in the ground ‘outruns’ the front in the air, and reaches the
underground structure first. Typical seismic velocities for soils and rocks
are given in Table 2.6.
Now, the air shock velocity varies with the overpressure q. For example,
an overpressure of 200 psi travels at 4000 ft/s, 100 psi at 3000 ft/s and 50
Figure 2.39 W ave p ro p ag atio n (N ew m ark): u is the seismic velocity of soil, v is the
shock front velocity.
@seismicisolation
@seismicisolation
Table 2.6 Typical seismic velocities u
Loose and dry gran u lar soils 600-3300 ft/s

Clay and wet soils 2500-6300 ft/s
C oarse and com pact soils 3000-8500 ft/s
L im esto n e-ch alk 7000-21 000 ft/s
Jointed granite 8000-15 000 ft/s
V olcanic rocks 10 000-22 600 ft/s
Table 2.7 Typical values of overpressure q
Alluvium less th an 40 psi

G ravel (dry) 10-100 psi
G ravel (wet) 40-500 psi
Sand 100-500 psi
S andstone 500-2000 psi
Lim estone 1500 psi an d above
G ran ite 3000 psi and above
psi at 2200 ft/s, so that for pressures above 200 psi, the seismic velocity u
will be exceeded in loose and dry soils and in some clays and compacted
soils. This will give a sloping shock front condition, and the result will be a
difference in phasing of the loads reaching the structure, which will receive
loads from the surrounding soil after it receives loads from the overpressure.
N ote that the lower the overpressure, the slower v becomes, and there is
m ore chance of v being less than u, with consequent outrunning. An indication
of the critical values of q is given in Table 2.7.
F or values of q less than these, outrunning conditions are possible. These
figures are taken from an address by Newm ark [2.43], who also discusses
the attenuation of the pressure wave with depth, as indicated in Fig. 2.40.
A typical pressure-duration curve for an explosion is shown at the surface,
with a fast rise-time followed by an exponential decay. However, when the
wave front has reached a depth z 1, the curve has a longer rise-time, lower
peak and longer decay. At larger values of z the relationship takes a much
sm oother form, with a long rise-time and a somewhat lengthened duration
(at depth z2 for example). F or most pressures the impulse, or area under the
pressure-duration curve, remains fairly constant, even though the shape
changes. It has also been established by experiment that the rise-time at
any depth is about half the time of transit of the shock front to the point
considered. The changes in the shape of the pressure pulse are a direct
result of the stress-strain characteristics of the soil in unidirectional com
pression.
@seismicisolation
@seismicisolation
A simplified empirical equation, useful for design purposes, has been

given by Newm ark and Haltiwanger [2.44], and it was shown to give good
agreement with test results obtained by Newm ark and Hall [2.45]. The
relationship between peak vertical stress az, at a depth z, and the peak
overpressure on the surface q, is given by
cTz = ccq, (2.42)
where a is the peak stress attenuation factor. The factor is defined as
« = ■. A , - , (2-43)
1 + z /^ w
Figure 2.41 B ilinear stre ss-strain relationship (H endron and Auld).
@seismicisolation
@seismicisolation
where
/lO O p siY 'V W Y /3

i » - 230ft( " J ( tm t ) • <2“ »
and W = yield of the explosion, or ‘weapon’, in megatons (MT). Further
analytical work by H endron and Auld [2.46] suggested that equation (2.44)
was only applicable to a soil having a strain recovery ratio of j , and a pro
pagation velocity of the peak stress for initial loading vL of 622 ft/s. If the
unidirectional stress-strain character of a soil can be generalized by the
bilinear hysteresis-type curve shown in Fig. 2.41, where M L is the loading
modulus, and M v the unloading modulus, then the propagation velocity
vL is given by
where p is the mass density of the material.

The strain recovery ratio is defined as M L/M U? and might well lie in the
range 0.3 to 0.7. Therefore, to take account of a greater range of conditions,
H endron and Auld have proposed the equation
and the solutions to equation (2.43) for various values of M L/M U and q,
are given for W — 1 MT in Figs 2.42 and 2.43. They also showed that for a
Attenuation factor, a
Figure 2.42 R elationship between attenu atio n factor a and depth when q = 100 psi
(H endron and Auld): vL = 622 ft/s, W = 1 M T.
@seismicisolation
@seismicisolation
Attenuation factor, ac
0-2 0-4 0-6 0-8 10
Figure 2.43 R elationship between attenu atio n factor a and depth when q = 300 psi
(H endron and Auld): vL = 622 ft/s, W = 1 M T.
given peak overpressure and strain recovery ratio, a will scale as

z
cc oc (2.47)
(W )il3v.
F or a further discussion of wave propagation in soils and a review of elemen

tary wave theory, see Allgood [2.34].
References
2.1 T erzaghi, K. (1943) Theoretical Soil Mechanics, Wiley, New Y ork, p. 66.
2.2 T erzaghi, K. (1936) Stress d istribution in dry and in saturated sand above a yielding
tra p d o o r. Proc. 1st Int. Congr. on Soil Mechanics, C am bridge, M a, vol. 1, p. 307.
2.3 V an H o rn , A .D . (1963) A Study o f Loads on Underground Structures, p art III,
Iow a Engineering Experim ent Station.
2.4 M cN ulty, J.W . (1965) An Experimental Study o f Arching in Sand, U S W ate r
w ays E xperim ent S tation, Tech. Rep. 1-674.
2.5 Gill, H .L. (1967) Active Arching o f Sand During Dynamic Loading, US N aval
Civil E ngineering Lab., Rep. T R 541.
2.6 G ill, H .L. and T rue, D .G . (1966) Active Arching o f Sand During Static Loading,
US N aval Civil Engineering L ab., Tech. N ote N-759.
2.7 Gill, H .L. and A llgood, J.R . (1964) Static Loading o f Small Buried Arches, US
N aval Civil Engineering L ab., Tech. Rep. R-278.
2.8 A llgood, J.R ., et al. (1963) Blast Loading o f Small Buried Arches, US N aval Civil
E ngineering Lab., Tech. Rep. R-216.
2.9 M ason, H .G . (1965) Effects o f Structural Compressibility on Active and Passive
Arching in Soil-Structure Interaction, D A SA Rep. 1718.
@seismicisolation
@seismicisolation
2.10 Jester, G .E . (1970) An Experimental Investigation o f Soil-Structure Interaction
in a Cohesive Soil, US A rm y W aterw ays Experim ent Station, Tech. Rep. N -70-7.
2.11 Terzaghi, K. (1943) Theoretical Soil Mechanics, W iley, New Y ork, p. 194.
2.12 Schm idt, H. (1926) Statische Probleme des Tunnel undDrucksfollenbanes, Springer,
Berlin (English tran slatio n : US B ureau o f R eclam ation, Tech. M em o 262, 1931).
2.13 Y am aguti, S. (1929) O n the stresses aro u n d a horizontal circular hole in a gravitat
ing elastic solid. J. Civ. Eng. Soc. Jpn., 15, 291.
2.14 A nzo, Z. (1937) Technol. Rep. Kynshu Univ., Jpn., 12 (3), June.
2.15 M indlin, R .D . (1939) Stress distribution aro u n d a tunnel. Proc. ASCE, A pr., 619.
2.16 Jeffery, G .B. (1921) Plane stress and plane strain in bipolar co-ordinates. Phil.
Trans. R. Soc. A, 221, 265.
2.17 O bert, L. and D uvall, W .E. (1967) Rock Mechanics and Design o f Structures in
Rock, W iley, L ondon.
2.18 G reenspan, M. (1944) Effect o f a sm all hole on the stresses in uniform ly loaded
plate. Q. J. Appl. M ath., 2, 60.
2.19 B rock, J.S. Analytical Determination o f the Stresses Around Square Holes
with Rounded Corners, D avid T aylor M odel Basin, Rep. 1149.
2.20 H eller, S.R ., Brock, J.S. and Bart, R. (1958) The stresses around a rectangular
opening w ith rounded corners in a uniform ly loaded plate. Trans. 3rd US Congr.
on Applied Mechanics, A IM E .
2.21 Savin, G .N . (1961) Stress Concentration Around Holes, Pergam on Press, New
Y ork.
2.22 C oker, E .G . and Filon, L .N .G . (1931) A Treatise on Photo-elasticity, Cam bridge
U niv. Press.
2.23 H ow land, R .C .J. (1934) Stresses in a plate containing an infinite row o f holes.
Proc. Camb. Phil. Soc., 30, 471.
2.24 Ling, C hih-B ing (1948) On the stresses in a plate containing two circular holes.
J. Appl. Phys., 19, Jan., 77.
2.25 W ang, C.T. (1953) Applied Elasticity, M cG raw -H ill, N ew Y ork.
2.26 Boussinesq, J. (1885) Application des Potentiels a FEtude de cFEquilibre et du
Mouvement des Solides Elastiques, G authier-V illars, Paris.
2.27 Southw ell, R.V. and G ough, H .J. (1926) On the co ncentration o f stress in the
n eighbourhood o f a sm all spherical flaw. Phil. Mag., p. 71.
2.28 Sadow sky, M .A . and Sternberg, E. (1947)7. Appl. Mech., Trans. ASM E , 69, A-191.
2.29 Sadow sky, M .A . and Sternberg, E. (1949) J. Appl. Mech., Trans. ASM E , June,
149.
2.30 T erzaghi, K. and R ichart, F.E . (1952) Stresses in rock aro u n d cavities. Geotech
nique - Int. J. Soil Mech., 3 (2), June, 57.
2.31 Finn, W .D . (1963) J. Soil Mech. Found. Div., ASCE, 89, SM 5, Sept., 39.
2.32 C helepati, C.V. (1965) Arching in Soil due to the Deflection o f a Rigid Horizontal
Strip, US N aval Civil Engineering L ab., Tech. N ote N-591.
2.33 C helepati, C.V. (1964) A rching in soil due to the deflection o f a horizontal strip.
Proc. Symp. on Soil-Structure Interaction, Univ. o f A rizona, Tucson.
2.34 A llgood, J. (1972) Summary o f Soil-Structure Interaction, US N aval Civil Engi
neering L ab., Tech. Rep. R771.
2.35 W hitm an, R.V. (1970) The Response o f Soils to Dynamic Loadings, US A rm y
W aterw ays Experim ent S tation, C o n tract Rep. 3-26.
@seismicisolation
@seismicisolation
2.36 L uscher, U. (1966) Buckling o f so il-su rro u n d ed tubes. / . Soil Mech. Found. Div.,
ASCE , 92, SM 6, N ov., 211.
2.37 H en d ro n , A .J. and D avisson, M .T. (1964) Static and dynam ic constrained
m oduli o f F renchm an F lat soils. Proc. Symp. on Soil-Structure Interaction, Univ.
o f A rizona, Tucson.
2.38 C ooling, L.F. and Skem pton, A .W . (1942) J. Inst. Civ. Eng., 17, 251.
2.39 O lson, R .E. and P arola, J.F . (1967) D ynam ic shearing properties o f com pacted
clay. Proc. Symp. on Wave Propagation and Dynamic Properties o f Earth Materials,
U niv. o f N ew M exico, A lbuquerque.
2.40 O lson, R .E. an d K ane, H. (1965) D ynam ic shearing properties o f com pacted clay
at high pressures. Proc. 6th Int. Conf. on Soil Mechanics and Foundations, vol. 1.
2.41 C asagrande, A. and S hannon, W .L. (1949) Strength o f soils under dynam ic loads.
Trans. ASCE, 114.
2.42 R ichardson, A .M . and W hitm an, R.V. (1963) Effect o f strain-rate u pon undrained
shear resistance o f a satu rated rem oulded fat clay. Geotechnique, 13.
2.43 N ew m ark, N .M . (1964) O pening address. Proc. Symp. on Soil-Structure Inter
action, U niv. o f A rizona, Tucson.
2.44 N ew m ark, N .M . and H altiw anger, J.D . (1962) Air Force Design Manual: Princi
ples and Practices fo r Design o f Hardened Structures, U S A F Special W eapons
C enter, K irtlan d A ir F orce Base, Rep. A FSW C-TD R -62-138.
2.45 N ew m ark, N .M . and H all, W .J. (1959) Preliminary Design Methods fo r Under
ground Protective Structures, U S A F Special W eapons C enter, K irtland A ir Force
Base, Rep. A FSW C-TR -60-5.
2.46 H en d ro n , A .J. and A uld, H .E. (1978) The effect o f soil properties on the a tten u a
tion o f air-blast induced g round m otions. Proc. Symp. on Wave Propagation and
Dynamic Properties o f Earth M aterials, U niv. o f N ew Mexico, A lbuquerque.
Further reading
B arjansky, A. (1944) D isto rtio n o f B ousinnesq field by circular hole. Q. J. Appl. Math.,
2.
D ar, S.M . and Bates, R .C . (1974) Stress analysis o f hollow cylindrical inclusions.
J. Geotech. Eng. Div., ASCE, 100, G T2, Feb.
Einstein, H .H . and Schw artz, C.W . (1979) Simplified analysis for tunnel supports.
J. Geotech. Eng. Div., ASCE , 105, G T4, A pr.
G etzler, S., G ellert, M . and E itan, R. (1970) Analysis o f arching pressures in ideal
elastic soil. J. Soil Mech. Found. Div., ASCE, 96, SM4, July.
K ennedy, T .C. an d Lindberg, H .E. (1978) M odel tests for plastic response o f lined
tunnels. J. Eng. Mech. Div., ASCE, 104, EM 2, A pr., 399.
M u rth a, R .N . (1973) Arching in Soils with Cohesion and Intergranular Friction, US
N aval Civil E ngineering Lab., Tech. Rep. R 793.
@seismicisolation
@seismicisolation
CHAPTER THREE
Thick-walled pipes
under static loads
Perhaps the simplest underground structure is the buried pipe conduit of

relatively small dimensions, used as a passageway for fluids, gases or various
types of cables. The walls of these pipes have in the past been made thick
enough for the structure to be considered as rigid, so that failure under high
surface loading is associated with bending of the cross section, rather than by
buckling or compression. In general, pipes made from concrete and cast
iron are termed ‘rigid’, whereas thin-walled pipes made from plastic, steel
sheet or corrugated metal sheet are considered ‘flexible’, and are more likely
to fail by buckling of the pipe wall. There is, of course, a transition between
rigid and flexible construction, and this will be discussed later. This chapter
will concentrate on the existing analysis and design rules for thick-walled
pipes and conduits, which are governed to a large extent by the way the
pipe is bedded in the soil, by the properties of the soil and by the method of
backfill.
M ost codes of practice for the design and laying of rigid pipes have made
use of extensive experimental and analytical studies carried out originally
by M arston, who began work on the problem at the Iowa State University,
Ames, USA, in the early years of this century, and who published an authori
tative paper on the theory of loads on pipes in ditches and tests of cement
and clay drain tile and sewer pipes in 1913 (M arston and Anderson [3.1]).
A series of bulletins was issued from then on, but perhaps the most useful
was a review of theories of external loads on closed conduits, published in
1930 (M arston [3.2]). The work was carried forward by Spangler, a former
student of M arston, who wrote a most comprehensive review in 1948 [3.3].
M arston discussed the loads on rigid conduits in terms of the shearing
forces developed in the prism of soil above the structure, along the lines
discussed in Section 2.1, making use of the following soil properties: density
y,the ratio between active lateral and vertical pressure kr, cohesion c, angle of
internal friction 0, and the stress-strain characteristics of the soil. He also
took account of the relative settlement of the soil above and below the
structure and the soil adjacent to the structure, the vertical deflection of the
@seismicisolation
@seismicisolation
Thick-walled pipes under static loads 53
Table 3.1 The three m ain classes of installation
D itch conduits These are buried in n arro w ditches excavated in

undisturbed soil
Positive projecting These are installed in shallow bedding with the top
conduits projecting above the natural ground surface, then covered
with a soil backfill
N egative projecting These are installed in shallow ditches with the to p below
conduits the natu ral ground surface, then covered with a soil
backfill
top surface of the structure with respect to the base, and the compression in
the soil prisms above and adjacent to the structure. He considered three
main classes of installation, as shown in Table 3.1.
In surveying the work on thick-walled pipes and conduits, we will begin
by considering the loads on the conduit due to each type of installation in
turn.
3.1 Ditch-type excavations

M arston applied the arching analysis of Janssen, replacing the trapdoor
width by the width of the ditch Bd, so that, from equation (2.3), taking the
cohesion as zero, we obtain
(3.1)
(see Fig. 3.1). M arston argued that cohesion could be neglected, because
considerable time must elapse after backfilling before effective cohesion
can be developed between the backfill soil and the adjacent, undisturbed
soil; furthermore, rainfall or some other action could occur which would
eliminate any cohesion that had developed. N ote then that the above theory
is based on the assum ption of a purely frictional fill, and that the shear
stresses along the vertical sides of the ditch are proportional to the mean
value of the vertical stresses at any section.
M arston took the value of kr to be equal to the Rankine value for active
pressure:
frequently expressed as
kr = tan 2 (45° — </>/2). (3.3)
@seismicisolation
@seismicisolation
Figure 3.1 M arsto n ’s analysis of ditch-type excavations.
The maximum value that /cr tan</> in equation (3.1) can reach for a good
sand fill is 0.19, and corresponds to </>= 31°. For a soft clay fill, (j) is often
taken as 8°, and kr tan </>= 0.11.
The value of y used in design is normally taken as the saturated density
(i.e. all voids filled with water); but if the backfill is made from particularly
coarse gravel that is free-draining, the void water is not retained, and the
drained density is used (the material wet but drained). If the saturated density
is y , the drained density is yd and the density of water is yw, then
y, = 7d + "yw- <3-4>
where
volume of voids
unit volume of soil
Table 3.2 Range of kr tan </> values
kr tan </> Soil type
0.19 G ran u lar m aterials w ithout cohesion

0.165 Sand and gravel
0.15 S aturated topsoil
0.13 O rd inary clay
0.11 Saturated clay
@seismicisolation
@seismicisolation
Cd
Figure 3.2 R elationship between H/B and Cd (trench fill) (Clarke).
The relationship between H /B d and Cd, where

= 1 - exp[ - 2kr tan </>(tf/B d)]
d 2kr tan (j)
has been given by Clarke [3.4] for a num ber of values of kr tan 0 (see Table
3.2), and is shown in Fig. 3.2.
It has been shown by Schlick [3.5] that, for trenches without parallel
sides, the width Bd should be measured at the top of the pipe, and this is
illustrated for one or two examples in Fig. 3.3. F or the ‘V’ trench, or wide
trench with a subtrench, note that the vertical surfaces of sliding are wholly
within the backfill, so the angle of internal friction will depend on the sliding
Figure 3.3 T renches w ithout parallel sides, show ing m easurem ent of Bd.
@seismicisolation
@seismicisolation
<f> (d e g)
Figure 3.4 C om parison of proposals for kT(Larsen).
H/Bd
Figure 3.5 Vertical pressure at crow n level on a ditch conduit, when </> = 30°, for sand
fill (Larsen).
@seismicisolation
@seismicisolation
between vertical faces of the backfill, not between the backfill and the vertical
face of the undisturbed soil.
The accuracy of M arston’s theory was examined by Wetzorke [3.6] in
1960, and he proposed that a better value for kr would be 0.5 for loose fill
and 1.0 for dense fill. He suggested that kr = 0.5 should be used for sand and
clay fill, while the minimum value of kTtan </> of 0.11 should be retained for
saturated clay. F or frictional fill, this proposal means a noticeable decrease
in <7v from that predicted by M arston, as shown in Figs 3.4 and 3.5 taken from
a thesis by Larsen [3.7].
A further study of the value of kr was undertaken by Christensen [3.8]
in 1967, who proposed a value of
which leads to the values of <rv given in Fig. 3.5.
3.2 Positive projecting conduits

When a thick-walled pipe is laid with the top projecting above ground and
then covered with a soil fill, the conditions are as illustrated in Fig. 3.6. The
soil adjacent to the soil prism immediately above the structure is now
filling material, and a consideration of the relative settlements is necessary
in order to determine the magnitude of the shearing forces.
The plane of equal settlement (see equation (2.11)) is now important,
and so is the ‘critical’ p la n e - a horizontal plane passing through the top
surface of the conduit when the level of backfill is at that surface, and before
the rem ainder of the fill is placed.
In discussing relative settlement, it is im portant to recognize the three
possibilities shown in Fig. 3.6. These are:
Natural ground level

Complete projection Neutral Complete ditch
Figure 3.6 Positive projecting conduits (complete).
@seismicisolation
@seismicisolation
Top of embankment
/ \
No Plane of equal
friction settlement
Natural ground level

Incomplere projection Incomplete ditch
Figure 3.7 Positive projecting conduits (incomplete).
1. The settlement of the top of the structure is less than that of the critical
plane, resulting in downward shearing forces in the soil prism. This
‘projection’ condition will therefore result in a greater pressure on the
top of the pipe than would be applied by the weight of the soil prism alone.
2. The settlement of the top of the structure is equal to that of the critical
plane, so that no shearing forces are developed.
3. The settlement of the top of the structure is greater than that of the
critical plane, resulting in upward shearing forces on the soil prism. This
‘ditch’ condition relieves the pressure on the top of the pipe.
Further, the plane of equal settlement influences the behaviour of the soil —
structure system. If it is located between the top of the pipe and the surface,
the shearing forces only extend upwards as far as this plane, giving an
incomplete projection condition, or an incomplete ditch condition. If
there is no plane of equal settlement, the shearing forces extend to the
surface, giving complete projection or ditch conditions. This is illustrated
in Fig. 3.7.
(a) Complete positive projection condition

Applying the arching analysis of Janssen, and taking account of the reversal
in direction of the shearing forces, leads to an expression similar to equation
(3.1), but with a reversal in sign. Thus, taking cohesion as zero,
(3.6)
where Bc is the overall width of the conduit.
@seismicisolation
@seismicisolation
(b) Complete ditch condition
This condition is identical to that given by equation (3.1), but with Bd replaced
by B giving
<7 = 1 —exp [ — kt tan <p —

2kr tan </>
+ qexp( — kr tan <j> (3.7)
(c) Incomplete positive projection condition

The equation given for complete projection holds, but with the depth of
cover taken equal to H e, the distance from the top of the pipe to the plane of
equal settlement, and an allowance made for the extra surcharge due to the
weight of soil in the vertical distance (H — H ). Equation (3.6) then becomes
2H
o = exp( + k r tan <j> )_ 1
2kr tan 4>
2H
+ [q + ( H - H J y] exp ( + kr tan 4> (3.8)
(d) Incomplete ditch condition

Applying similar reasoning, we obtain
Bj
— expl —kx tan 4>
2kr tan <f> B
2H
+ [q + (H - ffe)y] exp - kr tan </>__ e (3.9)
B
It is now necessary to evaluate H e by equating the total settlements of
the ‘interior’ and ‘exterior’ prisms of soil below the plane of equal settlement.
If it is assumed that the effective width of each of the ‘exterior’ prisms of soil
is equal to the total overall width of the conduit, as illustrated in Fig. 3.8,
then the total load at the critical plane is 3HyBc per unit length of the pipe.
But the load due to the soil immediately above the pipe is ov£ c, so that the
sum of the loads on the exterior prisms is
3HyBc oV Bc . (3.10)
acting on a total width 2B c.
The settlements are also shown in Fig. 3.8, where S{ is the settlement
downwards of the conduit invert, Sg is the settlement downwards of the
original surface of the natural ground, Sm is the extra settlement of the
critical plane adjacent to the conduit, due to the action of the overlying fill,
@seismicisolation
@seismicisolation
Plane of equal settlement
Inferior Exterior
soil prism soil prism
* S+S S+d
m. g f. C
- -I ; --------- — Before loading
Critical plane
After loading
y- 4" Before loading

Natural ground
J After loading
Figure 3.8 N o tatio n for incom plete ditch condition.
and dc is the vertical diametrical shortening of the conduit. Then, the relative
settlement of the critical plane to that of the conduit is, from Fig. 3.8,
(Sm + Ss) - ( S , + de), (3.11)

and if this is divided by Sm, we get a measure of the relative deformation
of the interior and exterior prisms of soil, known as the ‘settlement ratio’
r sd ’ th U S
(Sm + S ) - ( S f + d.)
(3.12)
The deform ation of the critical plane at the crown of the conduit, relative
to its deform ation in the adjacent soil, is governed by the am ount that the
top of the conduit projects above the surface of the ground - and the ratio
of this projection to the conduit width Bc is known as the projection ratio rp.
Then the projection, as shown in Fig. 3.8, is rpBc.
Using the ratios rsd, rn and equation (3.10), it can be shown that, for c = 0,
ex p [ + 2/cr tan 0 (H e/Bc)] - 1 f ± H H
@seismicisolation
@seismicisolation
Incomplete ditch condition
, ^ ,
% / 7 bc
Figure 3.9 R elationship betw een H /B c and aJ yB e for positive projecting conduits.
The upper signs are valid for the incomplete positive projection condition,
the lower signs for the incomplete ditch condition. Equation (3.13) can be
solved numerically to obtain values of H e for given values of H /B c, rsdrp and
kr tan </>. If H e > H, the complete positive projection or ditch condition
exists, and the pressure os on the structure is found from equation (3.6) or
(3.7). If H e < H , incomplete conditions exist, and <rv is found from equation
(3.8) or (3.9). Using this method, the relationship between H /B c and
o J y B c has been computed by Spangler [3.3] for a range of values of rsdrp,
when k T tan </>= 0.19 (granular materials without cohesion). His results are
given in Fig. 3.9, and it is generally thought that figures based on k Ttan
</>= 0.19 can be reasonably chosen as safe working values for most projecting
conduits. The relationship between H /B c and <rJyBc has also been given by
Clarke [3.4] for k T tan </>= 0.13 (ordinary clay with minimum favourable
friction), as shown in Fig. 3.9. He points out that the major portions of the
curves are linear, and to facilitate use he gives simple empirical relation
ships between H /B c and o J y B c for a range of values of rsdrp (see p. 247 of ref.
3.4).
Spangler [3.9] recommended values of rsd for design purposes, as the
result of field m easurements of settlement ratios of various highway culverts
in the USA. These have been summarized by Clarke on p. 219 of ref. 3.4.
3.3 Negative projecting conduits

If a thick-walled pipe is laid in a trench dug in the natural ground before it is
covered with a soil fill, as shown in Fig. 3.10, it is possible to combine the
@seismicisolation
@seismicisolation
Top of embankment-
advantages of load reduction of a ditch-type construction, with the ease of

installation of projecting structures. A trench having a width slightly greater
than B c is excavated,’ so that the crown of the conduit is below natural
ground level, and after placing the conduit the fill is built up to a level well
above the level of the original ground surface. This method usually results in
the pressure on the conduit due to the prism of soil above it being less than the
hydrostatic pressure due to the height of the prism, and the main reason
for this is that the soil prism can now settle more than the adjacent effective
soil, producing vertical shearing forces directed upwards.
The settlement situation is shown in Fig. 3.10. For negative projecting
structures the ‘critical’ plane passes through the top of the ‘subtrench’,
not through the top surface of the conduit, and the width of the interior
prism is Bd, the ditch width, not Bc. As before, the plane of equal settlement
influences the behaviour of the soil-structure system. If there is no plane of
equal settlement, the shearing forces extend to the top of the backfill, and this
results in the complete ditch condition; if it lies between the top of the pipe
and the surface, we have the incomplete ditch condition.
(a) Complete ditch condition

This condition is identical to that given by equation (3.1); thus, for c = 0/
we have
@seismicisolation
@seismicisolation
(b) Incomplete ditch condition
The equation for the complete ditch condition holds, but with the depth of
cover taken as H e, and an allowance made for the extra surcharge due to the
weight of soil in the vertical distance (H — H e). Equation (3.14) then becomes
o M
v 2krtan <p 'd / J
(3.15)
As before, it is now necessary to evaluate H e, taking account of the relative

settlements of the ‘interior’ and ‘exterior’ prisms of soil below the plane of
equal settlement. By a similar analysis to that leading to equations (3.11),
(3.12) and (3.15), it can be shown that, for c = 0,
(3.16)
where H' is the distance from the natural ground surface to the top of the
embankm ent, and H'e is the distance from the natural ground surface to the
plane of equal settlement (Fig. 3.10). The projection ratio F is the ratio
between the am ount that the top of the conduit is below the natural ground
surface and the trench width, so that the ‘negative’ projection, as shown in
Fig. 3.10, is rpBd.
Equation (3.16) can be solved numerically to obtain values of H e for
given values of H /B d, rsd, rfp and kr tan </>, remembering that
H' = H — rp'B d ’ (3.17)
and
(3.18)
If H e > H, the complete ditch condition exists, and <rv is found from equation
(3.14). If H e < H, the incomplete ditch condition exists, and <rv is found from
equation (3.15). Using this procedure, the relationship between H /B d and
aJyB d has been com puted by Spangler [3.10] for various values of rsd and
F , when kr tan (p = 0.13. It is generally agreed that figures based on kr tan (p =
0.13 are safe for negative projecting conduits under ordinary backfills.
Spangler’s findings have been summarized by Clarke in ref. 3.4 (pp. 248-51),
@seismicisolation
@seismicisolation
<rv/ y B d
Figure 3.11 R elationship between H /B d and o JyB d for negative projecting conduits
(incom plete ditch condition).
who also gives empirical linear relationships. A typical chart is given in

Fig. 3.11 for kr tan (j) = 0.13, and r'p = 1.0 (i.e. the crown of the conduit is set
one trench width below the natural ground surface).
(c) Conduits with backpacking

Negative projection construction assures a reduction in load on the thick-
walled conduit, but installation can be difficult. To ease construction
difficulties and retain the advantages of the negative projection condition,
the ‘imperfect’ or ‘induced’ ditch method of construction was proposed by
M arston (not to be confused with the ‘incomplete’ ditch condition of the
previous two sections). In this method the conduit is set with its invert at
natural ground level, and a thoroughly compacted fill is built up to a level
about one or two diameters above the top of the conduit. At the end of this
stage a part of the completed fill immediately over the conduit, and equal to
its width, is excavated, and ‘backpacked’ with loose soil or other highly
compressive material, as shown in Fig. 3.12. The embankm ent is then
completed in the usual way.
The settlement of the soil prism immediately above the conduit will now be
greater than that of the adjacent soil, and as a result the shearing forces on the
prism walls will be directed upwards, thus reducing the load on the structure.
Spangler [3.10] considered the backpacked trench as a special case of
negative projection construction, with the substitution of the conduit width
@seismicisolation
@seismicisolation
Top of em bankm ent
Normal fill
Trench dug offer compaction,

Coi filled with compressible material
Natural ground
Figure 3.12 C onduit with ‘backpacking’.
Bc for the trench width B d in equations (3.14) to (3.16). It is interesting to note

that this is a compromise between the positive and negative projecting
conditions, in that the conduit is installed in the former, and loaded in the
latter.
3.4 Wide trenches and multiple pipes

The discussion of ditch-type excavations has so far been limited to geometries
in which the width Bd, measured at the top of the pipe, has not greatly
exceeded the pipe width Bc. Schlick [3.5] has indicated that the formula
for total load,
where Cd is the load coefficient for the narrow trench condition, does not
apply as Bd is increased indefinitely, but the total load reaches a limiting
value, given by
where Cc is the load coefficient for the positive projection condition. The
trench width at which the two values are equal, i.e. when yB2 d Cd = yB2Cc,
is known as the ‘transition’ width from the trench to the positive projection
(embankment) condition.
Thus, the transition width (Bd\ is given by
(3.19)
where, from equation (3.1),
1 —exp( —kr tan 4>(2H/Bd))
Cd
2kr tan 0
(3.20)
@seismicisolation
@seismicisolation
-i 1
Figure 3.13 T hree pipes in the same ditch (Clarke).
and from equation (3.6)

exp(kr tan </>(2H /B c)) - 1
2kr tan </>
When £ d < (£d)t, the trench is said to be narrow. When Bd > (Bd)t, it is
said to be wide. N ote that the transition width is a function of the depth of
cover H , and becomes smaller as the cover depth increases.
This conception of ‘narrow ’ and ‘wide’ trenches can be used to discuss
the loads imposed on several parallel conduits, buried at close spacing in
the same trench, and examples of the method of analysis have been given by
Clarke [3.4]. An illustration of the type of arrangem ent he considered is
given in Fig. 3.13, which shows three pipes, A, B and C, in the same trench.
If the distances from the centre of pipes A and B to the edge of the excavation
are thought of as ‘half-ditches’, then these can be conveniently expressed as
(Bd)A/2 and (£d)B/2 respectively.
If (Bd)A and (Bd)B do not exceed the transition widths for pipes A and C,
then these pipes can be considered to be under the loading actions associated
with the narrow trench condition. If (£c)c + 2a (where a is the distance
between the inside edges of the pipes) is greater than the transition width for
pipe C, then pipe C can be considered to be under a loading action associated
with the wide trench condition. Then the total loads on the three pipes a re :
pipe A
l o a d = f [ C d><(B^ + C y(B c)J],
pipe B
load = % C dy(Bd)2
E + Ccy(Bc)2Bl (3.22)
pipe C
load = Cj (Bq) |.
@seismicisolation
@seismicisolation
If (Bc)c + 2a is less than the transition width, the prisms of soil over the
areas adjacent to or between the pipes will not settle in relation to the prisms
over the pipes, and the loads will b e :
pipe A
load = i{ C dy(Bd£ + t f r [(Bc)A + a]} ,
pipe B
load = ±{Cdy(Bd)2
B + Hy[(Bc)B + *] }, (3.23)
pipe C
load = Hy[(Bc)c + a].

If the distance between the pipes is very small, the three pipes can be assumed
to share equally a ditch loading calculated for the entire width of the ditch,
Bd, Thus,
load on each pipe = \ C dyB 2d . (3.24)
If doubt exists in the m ind of the designer, a good procedure is to calculate
the load on a pipe by each of the three m ethods given above, and take the
largest of the figures for design purposes.
3.5 Submerged pipes

W hen a pipe or conduit is submerged in a mixture of backfill and water,
and if the pipe is empty, there will be an uplift of ^nB 2yw per unit length,
acting on the pipe, where yw is the density of water. The load due to the soil
backfill will therefore be reduced by this amount, so that
load per unit length due to soil backfill = H B c(y —yw) —^ n B 2yw. (3.25)
If this load is less than the dead weight per unit length of the empty conduit,
it will want to float, and steps must be taken to anchor it down when empty.
If the backfill is submerged in water to a height of H w above the crown of
the pipe (see Fig. 3.14), the density of the submerged fill will be y —yw, and
there will be a reduced load on the pipe due to buoyancy. For the narrow
ditch condition, and when H w < H, the value of oy at the crown of the pipe
(including the effects of an overpressure q) is:
^=i ^ [ 1-eXp(-k'tan^^)]+«CXp(-k'tan^^)
2
(3.26)
@seismicisolation
@seismicisolation
— fid — ^ Ground level
/Full Density y
friction
H
Water level
“f
Hw ✓Reduced Submerged
friction density
,_L_
n
Figure 3.14 Subm erged pipe in a narrow ditch.
3.6 Concentrated free surface loads

It has been custom ary to examine the effects of concentrated free surface
loads on thick-walled pipes by using the elastic theory of Boussinesq, as
described in Section 2.2.5. M arston [3.2] proposed this method in 1930, after
Spangler et al. [3.11] had conducted tests on concentrated loads in the 1920s.
If the concentrated load is over the centreline of the pipe, the stress distri
bution on a horizontal plane through the crown of the pipe (cry in equation
(2.29)) is shown in Fig. 3.15. At an arbitrary point N in this plane, vertical
coordinate z (measured from the surface downwards), horizontal radial
coordinate r, where r is the distance between N and a vertical axis through
the point of application of the load F,
3F
a* = ^ i co&5P’ <3-27>
where ft is the angle between the vector O N and the vertical axis through O.
Figure 3.15 D istribution of stress due to point load on the surface (Boussinesq).
@seismicisolation
@seismicisolation
Table 3.3 R elationship between I and r/z
r/z /a r/z K r/z /a
0 0.477 1.0 0.084 2.0 0.0085

0.1 0.466 1.1 0.066 2.2 0.0058
0.2 0.433 1.2 0.051 2.4 0.0040
0.3 0.385 1.3 0.040 2.6 0.0029
0.4 0.329 1.4 0.032 2.8 0.0021
0.5 0.273 1.5 0.025 3.0 0.0015
0.6 0.221 1.6 0.020 3.2 0.0011
0.7 0.176 1.7 0.016 3.4 0.0009
0.8 0.139 1.8 0.013 3.6 0.0007
0.9 0.108 1.9 0.0105 3.9 0.0005
This is one of the well-known Boussinesq equations. The direction of the

largest principal stress cTj at N passes through O, and the stress intensity is
3F
<7. = - — ~cos3/?. (3.28)
1 2nzl
When Poisson’s ratio v = 0.5, the other two principal stresses are zero, and the
load F then produces a unidirectional state of stress. Equation (3.27) can be
rewritten as
where / is an ‘influence value’ depending on /?, or the ratio r/z. The relation
ship between / and r /z was com puted for a range of values of r /z by Gilboy
[3.12] in 1933. His results are summarized in Table 3.3.
Equation (3.29) can be used to compute the vertical normal stress oz at
point N, when the surface load is applied over a rectangular area centred
on the vertical through N ; or, conversely, tne distributed load over a rect
angular area at depth z below a concentrated surface load F located on a
vertical through N, the centre of the area (see Fig. 3.16).
The vertical norm al stress oz at point N is found by dividing the total
area into four rectangular sections, and then finding the influence of a
surface load covering each of these sections in turn on the value of oz at a
point located at depth z below one corner. Using equation (3.29), Newmark
obtained by integration the relationship between Act, the stress due to any
one of the four areas, and the uniformly distributed surface load q in terms of
B /z and L /z , where B and L are the side dimensions of the sub-area under
consideration. His results can be presented in graphical form in a way due to
@seismicisolation
@seismicisolation
Acr/q
O 0 0 5 0-100-15 0-20 0-25
2
z/B
4
6
8
10
4 o-s stress at/V z

due Co q on
shaded area
cr s A Act N
Figure 3.16 Vertical stress due to uniform ly distributed surface pressure over a rectangu
lar area (Terzaghi).
Steinbrenner [3.13], as shown in Fig. 3.16, which relates So/q and z/B for
various values of L/B.
To find the value of oz at point N, the ratios B/z and L/z are computed
for each of the four subrectangles in turn, and the corresponding value of
Aa/q is determined from Fig. 3.16. The total value of oz is the sum of the
four values of A<r found in this way.
It is interesting to compare the value of az at various values of z/B below
the centre of a square area IB x 2 B, with values of oz at the same depth,
under a surface point load F — 4B2q acting at the centre of the area. This is
shown in Fig. 3.17, indicating that the differences are small when z/B > 4.
The analysis described above was used by Spangler and Hennessey
[3.14] in 1946 to find the distributed load on a rectangular area located
in the plane of the pipe crown, with one side equal to the width Bc of the
<?z /q
0-5 1-0
Surface pressure q
6 over square area 2 8 x 2 8
Figure 3.17 C om parison of az for a point load, and for the same load distributed over a
square area.
@seismicisolation
@seismicisolation
LWf / F
Figure 3.18 C oncentrated surface load coefficients for conduits with L = 3 ft.
pipe, when a concentrated load F was applied at the surface vertically

above the centre of the pipe.
If the load reaching the underground conduit is expressed as WF per unit
length, then the total load on a length L is L W F, and the ratio L W f /F is a
measure of the relief of load due to the presence of the soil. This ratio can be
deduced from Fig. 3.16, and its value for a length of conduit of 3 ft, having
various widths Bc, has been given in graphical form by Spangler [3.3]
for cover depths up to 10 ft. The relationship is shown in Fig. 3.18.
3.7 Other methods of load analysis

M arston’s theory is of great practical value, because it takes into account a
variety of trench and em bankm ent conditions and soil properties, and makes
allowance (by the use of load factors - discussed later) for different pipe
bedding conditions. It is possible, however, to treat much more idealized
thick-walled pipes and conduits by applying the classical theory of linear
elasticity, and this is useful for attem pting to predict horizontal as well as
vertical loads on the structure and for giving information about the effect
of im portant parameters. Finite element m ethods have proved very useful in
recent years for examining simplified examples.
An alternative to the shear plane analysis was first put forward in 1937
by Voellmy [3.15], who proposed that the total vertical load on a projecting
conduit was equal to the weight of the soil mass shown shaded in Fig. 3.19,
where H e is the depth of the crown of the pipe below the plane of equal
settlement, and the angle of inclination </> of the inclined planes is equal to
the angle of friction of the cohesionless soil. The resultant of the stresses in
@seismicisolation
@seismicisolation
Figure 3.19 Voellmy’s analysis.
the inclined planes is therefore horizontal, and does not contribute to the
vertical load. The load was assumed to be acting over a width nD/4, as shown
in the figure. The value of H e can be found by the method given earlier in this
chapter, but the reader should be aware of an alternative analysis due to
W astlund and Eggwertz [3.16].
In 1961, a m ethod of analysis for rigid projecting conduits was proposed by
Pruska [3.17], which was simplified by taking a linear elastic medium with
v = 0. The vertical load per unit length on the crown of the conduit was taken
as the weight per unit length of the soil prism immediately above it (yHBc)
plus the force per unit length P, as shown in Fig. 3.20, where P is the force
intensity required to produce a plane of equal settlement coinciding with the
free surface. N ote that the effect of P is added, because a rigid projecting
conduit attracts more load than that due simply to the prism of soil.
@seismicisolation
@seismicisolation
The settlement of the soil prism above the pipe is \ y H 2/E, and the settle
ment of the soil adjacent to the pipe is \y(H + B c)2/ E ; so that the addition of
force P should produce an additional settlement equal to the difference
between these expressions:
ty{H + Bc)2/ E - \ y H 2IE. (3.30)
It can be shown that the additional settlement corresponding to P is
P 1 'H
(a + sin a) d/z, (3.31)
^ B C~E o
where
_ sin \ B h )
“ (BJ 2 )2 + h2' (132)
so that
h n B [{H + B ) 2 - H 2]
P = -2- ,-----"• (3-33)
Jo (a -f sin a) dh
One of the m ost useful finite element analyses of the problem is due to
Leonhardt [3.18]. He analysed a buried concrete pipe of known stiffness
surrounded by a linear elastic-plastic soil, obeying Coulom b’s failure
criterion. His results are presented in the form of radial and tangential
pressure distributions for a variety of pipe installations, and two examples are
shown in Fig. 3.21. The distribution in Fig. 3.21(a), where the figures are in
N /cm 2, is for a pipe 1 m in diameter, buried a depth of 5 m in sand having a
12-8 N/cm 2 13-7 N/cm2
Figure 3.21 L eo n h ard t’s pressure distribution, (a) Pipe diam eter lm , buried 5m in sand,
invert bedded on sand; (b) pipe diam eter lm , buried 5m in sand, invert bedded on rock.
@seismicisolation
@seismicisolation
density 19.6 kN /m 3 and an angle of friction (j) of 35°. That in Fig. 3.21(b)
is for similar dimensions and soil properties, but the invert is bedded on a
slab of rock. As might be expected, the main difference is the level of radial
and tangential stresses in the vicinity of the invert, and there is a clear
discontinuity at the point where the rock bedding begins.
In 1974, Anand [3.19] used the finite element method to analyse the
behaviour of shallow buried rigid pipes under various types of surface
pressure. He chose a i m diameter concrete pipe (diameter/thickness
ratio = 40) buried under 1 m of linearly elastic soil, with a further 1 m below
the pipe and then a rigid base. He showed that the radial and tangential
pressure distribution was not affected much by the soil modulus, and that
loads calculated by the Boussinesq m ethod gave good agreement at the
crown - but were at variance elsewhere. He investigated ‘backpacking’ by a
soft m aterial all around the pipe, and showed that this could reduce radial
pressure by at least 25% and tangential pressure by 50%.
O ther finite elements studies have been conducted by Krizek et al. [3.20],
and T rott and G aunt [3.21] in connection with full-scale field tests on steel
and concrete buried pipes. Their results are surveyed in a later section that
deals with experimental work. A finite element analysis of thick-walled
ditch conduits has also been reported by Nayak et al. [3.22], producing
results similar to those of Leonhardt. A com puter program for the analysis
and design of buried pipe culverts, code-named CANDE, has been developed
by Allgood et al. [3.23], and is described later with respect to thin-walled
tubular construction.
3.8 Conduit beddings

It is clear from the analysis of loads described above that the method by
which ditch conduits are bedded at the base of the trench has an influence
on the loading actions as well as on the strength of pipe. Before going on to
discuss full-scale experiments and design procedures, it will be useful to
survey briefly the classifications of bedding norm ally employed in codes of
practice. Spangler [3.3] dealt with this aspect in his 1948 paper, and recently
a useful review of U K procedures has been written by Young [3.24]. He
makes the point that conduits should not rest on blocks or similar hard
packings.
(a) Concrete cradle bedding

This method guarantees maximum strength, and minimizes the disturbance
by future adjacent excavations. A cross section of the arrangement proposed
by Young is given in Fig. 3.22(a).
@seismicisolation
@seismicisolation
Main backfill Main backfill Main backfill

Compacted Compacted
^Vi mm «...
soil (min.) 5011 Compacted
soil
Concrete cradle Selected granular Trimmed trench

material (compacted)) bottom
(a) (b) (c)
Figure 3.22 Conduit beddings, (a) Maximum strength; (b) first class; (c) class D.
(b) First-class bedding

The conduit is set on a fine granular material carefully shaped to fit the
lower part of the circumference, and extending to a width of at least 60%
of the width of the conduit. The remainder is entirely surrounded by a
tam ped backfill of granular material. Young recommends that the bedding
should give support over the lower 180° where possible, as shown in Fig.
3.22(b). Spangler’s ‘ordinary’ bedding is very similar, but the soil foundation
extends to a width of 50% of the conduit width, and the tamped backfill
only extends 0.5 ft above the crown.
(c) Class D bedding

This standard, given by Young, consisted of merely trimming the trench
bottom to fit the profile of the conduit over a short length of its circumference,
as illustrated in Fig. 3.22(c). Com paction of the backfill adjacent to the
conduit is carried out by hand, or by light plate vibrator.
3.9 Full-scale experiments

M arston’s theories were based on the results of full-scale tests, and in ref. 3.3
Spangler compares the results of these tests with the theories for ditch and
projecting conduits. Fig. 3.23, taken from his paper, shows close agreement
between the ditch theory and tests on 18 inch and 36 inch diameter rigid
pipes in 2.24 and 4.15 ft ditches respectively. Fig. 3.24, from the same paper,
indicates good agreement for loads on projecting conduits in the ‘imperfect
ditch’ or ‘incomplete projection’ condition. Spangler [3.25] also recorded the
behaviour of a rigid conduit of diam eter 112 cm under 4.6 m of sandy soil
backfill over a period of more than 20 years. Radial earth pressures were
@seismicisolation
@seismicisolation
Load on pipe (kips/ft)

Figure 3.23 M easured loads on ditch conduits (Spangler).
Load on pipe (kips/ft) Load on pipe (kip s/ft)

Figure 3.24 M easured loads on projecting conduits (Spangler).
m easured by pressure ribbons, and the distributions at the beginning of this

period for concrete and cast iron pipes are shown in Fig. 3.25. The stress
pattern closely follows the predictions of finite element theory, and the
weighed loads were about 1.45 times as great as the load due to the weight
of m aterial directly over the conduits. Over the period of the test, no sub
stantial change in load occurred, although there were measurable settlements,
and Spangler recommended the use of a settlement ratio rsd = 0.7 in the
design of rigid conduits in ‘ordinary earth’, and rsd = 1.0 in rock or un
yielding material, where this ratio is defined as in equation (3.12).
@seismicisolation
@seismicisolation
Height of fill 4- 6 m of sandy soil
Concrete pipe Cast iron pipe

Figure 3.25 M easurem ents of radial pressure d istribution (M arston and Spangler).
In 1929 Braune et al. [3.26] reported on tests carried out between 1924
and 1927 at the University of N orth Carolina, USA, on a variety of pipes,
rigid and flexible. The results for a thick-walled concrete pipe, taken from
their work, are given in Fig. 3.26, indicating a virtually linear relationship
between load and depth of cover for a concrete pipe 51 cm in diameter with a
sand backfill. The weighed vertical load on the pipe was about 1.50 times as
great as the load due to soil prism about it - agreeing closely with M arston’s
Depfh of cover ( m )
Figure 3.26 Vertical load on buried pipes, from tests by Braune et al.
@seismicisolation
@seismicisolation
findings. The results given in Fig. 3.26 were for a pipe projecting 100% from
a concrete bed.
After the establishment of the M arston-Spangler theory, and its confirma
tion by test results, there was a relative lull in full-scale research activity
until the 1960s. Since then, however, a num ber of investigations have been
sponsored by the US Bureau of Reclamation, the American Concrete Pipe
Association, the M inistry of Housing and Local Governm ent (UK) and the
Transport and Road Research Laboratory (UK). The first of these, reported
by Pettibone and Howard [3.27, 3.28], were laboratory tests, but on a large
enough scale to be thought of as more than model tests. They investigated
the behaviour of 61 cm diameter concrete pipes in sandy clay and sand
backfills, with uniform and concentrated surface loads applied by a structural
testing m achine; the tests were carried out in the ditch condition, and, as far
as possible within the limitations of the apparatus, in the embankment
condition. Their most interesting observation is summarized in Fig. 3.27,
which shows the distribution of radial earth pressure for the em bankment
condition (Fig. 3.27(a)), and the ditch condition (Fig. 3.27(b)). Lateral
pressure increases for condition (a) are about 40% of the increase in the
applied surcharge pressure, whereas for condition (b) the increase in lateral
pressure was virtually zero.
The tests sponsored by the M inistry of Housing (UK) were reported by
H ainsw orth et al. [3.29] in 1967. Rigid clay, cast iron and concrete pipes were
installed at the bottom of narrow trenches and the strains in the pipe walls
were m easured as the backfill was emplaced. The clay pipes were 30.5 cm
diameter, laid in a 65 cm wide trench, and the cast iron and concrete pipes
Figure 3.27 E xperim ents by Pettibone and H ow ard, (a) E m bankm ent condition,
diam eter 61 cm, load applied over 213 cm with 91 cm of clay cover; (b) ditch condition,
diam eter 61 cm, load applied over 109 cm w ith 91 cm of sand cover; surface pressure 69
k N /m 2 for (a) and (b).
@seismicisolation
@seismicisolation
were 91.5 cm diameter, laid in a 150 cm wide trench. In all tests the final
depth of cover, after com paction of the backfill, was 120 cm.
At the early stages of backfilling, the loads on the pipe were noticeably
less than the prediction of the M arston theory, but after a period of time and
the application of transient heavy surface loads of the type associated with
wheeled traffic, the comparison between theory and experiment was found
to be close for the cast iron and concrete pipes.
F urther confirmation of the ratio (weight of soil prism)/(load on conduit)
as 1/1.4 was obtained by full-scale tests on three concrete pipelines in Sweden
by Janson [3.30] in 1965. His tests, however, suggested that the Boussinesq
m ethod of dealing with local surface loads gave somewhat unconservative
figures for rigid pipes.
The American Concrete Pipe Association has carried out a long series
of field tests on reinforced concrete culverts in Ohio and California, reported
in ref. 3.20, and m ore comprehensively in a recent paper by Krizek and
M cQuade [3.31]. The Ohio tests, carried out at the Transportation Research
Center, East Liberty, consisted of instrum ented lengths of pipe of 153 cm
Figure 3.28 East L iberty test installations (K rizek and M cQ uade).
@seismicisolation
@seismicisolation
diam eter buried in the em bankm ent and ditch conditions under a cover of
7.65 m. All param eters were measured at regular intervals for a period of
more than three years, during which time changes were very small (maximum
differences were of the order of 10%). The em bankm ent soil was a compacted
silty clay, and the ditch was excavated in 3.7 m of very fine sand over 5.5 m of
sandy clayey silt. The ditch itself was excavated with sloping sides, as shown
in Fig. 3.28, taken from ref. 3.31. The installation conformed with Highway
Specifications, and the strength of the pipe was determined by the M arston-
Spangler theory (Spangler [3.32]).
N orm al interface stresses were measured and compared with theoretical
results obtained from a plane strain finite element model. Plane strain
was assumed on the basis of experimentally measured longitudinal displace
ments, which were negligible. The com parison is shown in Fig. 3.29. Vertical
strains in the adjacent soil and pipe diam eter changes were also measured.
Radial earth pressures were found to increase in proportion to the height of
fill, and their distribution in Fig. 3.29 suggests that, although great care was
taken to bed one culvert on a shaped seating, it was virtually supported along
a line. Behaviour during loading and unloading of an applied surface
pressure suggested that during this process the system responded elastically.
The California tests were carried out at M ountain-house Creek, on a 73 m
long reinforced concrete pipe, 2.1 m diameter, 200 mm thick, in the em bank
ment condition under a fill of 41.5 m. Various bedding and backfilling
configurations were employed in a series of six zones along the test length
of the pipe. The results were originally reported by Davis et al. [3.33] in
Height of cover 25 ft
Embankment Trench
Figure 3.29 D istribution of radial stress for East Liberty installations (Krizek and
M cQ uade).
@seismicisolation
@seismicisolation
1974, and, in com paring these with the finite element analysis, Krizek
showed that there was good agreement between calculated and measured
diam eter changes. The four lengths of pipe that were laid on shaped bedding,
or on sand in trench, attracted full hydrostatic pressure due to the weight
of soil above the section. The two lengths bedded on polystyrene or line
bearing on hard earth showed a noticeable relief in loading due to arching -
due to settlement into the polystyrene or to local crushing of the invert
along the contact line with the hard earth.
The idea of ‘cast-in-place’ concrete pipelines was first investigated in the
USA in the 1950s, because of cheapness of installation and virtually ideal
bedding conditions. Load testing of this type of culvert was reported by
Johnson and Hess [3.34] in 1963, and more recently the finite element
analysis of an idealized model has been undertaken by Bush and Chak
[3.35]. The load tests were carried out on the cross section shown in Fig.
3.30, and measurem ents were taken of diam eter change and soil pressures
when a 48 inch internal diam eter pipe, under a range of backfills, was sub
jected to an increasing surface load. This load was applied by the near
wheels of a truck, and increased incrementally by loading the truck with
oil drum s filled with steel balls. Visual observations indicated no distress
to the culvert even under the most extreme loading conditions, which was
a load of double the legal axle load in Arizona at the time (18 000 lb), and
a cover of only 6 m of granular backfill.
Trucks were also used by Page [3.36] to examine the effect of vehicle
speed on the response of concrete conduits of 18 inch and 36 inch diameter,
passing under a road and buried at depths of 4 ft and 9 ft below the surface.
Im pact factor (or ratio of dynamic to static strains and deflections of the
pipe) was found to increase linearly with speed, and was independent of
pipe size, cover depth, backfill material or type of bedding for the conduits
tested. M any of the vehicles tested gave impact factors at 30 mph vehicle
speed in excess of 3.0, although the design factor normally used at the time
was 2.0.
Compacted
granular Height of backfill
backfill varied between
6 and 24 inch
_lCast-in -place’
concrete pipe
Figure 3.30 ‘C ast-in-place’ pipe (Johnson and Hess).
@seismicisolation
@seismicisolation
3.10 Model tests
M odel tests on long, thick-walled cylinders have been carried out mainly
in connection with the design of strategic underground structures to with
stand air explosions. A useful experimental program me was reported by
Allgood and H errm ann [3.37], who examined reinforced concrete model
cylinders buried in dry sand with their longitudinal axis parallel to the surface.
The cylinders were 9 inch outside diam eter with a single layer of reinforce
ment at midsection. Twelve cylinders were tested, with varying thickness,
reinforcement and cover depth. The thinner ones were flexible, and the
results will not be discussed until Chapter 4; but the thick-walled cylinders,
with diam eter/thickness ratios of the order of 30, failed by ‘caving’ in the
direction of the applied load. Because the surface pressures were so high
in relation to the gravity loading of the backfill, the model material could
be identical to the full-scale material, and Ultracal concrete, which closely
parallels the behaviour of Portland Cement concrete, was used to construct
the specimens.
The thick-walled cylinder deflections, as might be expected, were very
small at low surface loads, but loss of stiffness due to the formation of
longitudinal cracks resulted in thick-walled cylinders deflecting more than
thin-walled cylinders at higher loads. There were also large variations in
the thrust around the circumference of the thick cylinders, and they ex
perienced larger moments.
A laboratory study to examine the behaviour of thick-walled cylinders
buried in sand under high static surface load was reported by N orm an and
Prendergast [3.38] in 1973. It was undertaken to obtain more information
about the efficient design of thick-walled cylindrical structures to withstand
very high overpressures at shallow burial depths, and a series of 6 inch
outside-diam eter steel cylinders were tested in a horizontally laid position
in dense, dry sand. Static loading tests were conducted for each of nine
cylinders at cover depths of 3, 6 and 9 inches. M easurements were made of
hoop strain, vertical diam eter change and associated ‘free field’ stress.
The geometry of the test specimens is shown in Fig. 3.31. The length of
the central test section was 12 inch; the closed end caps were 8 inch long
U— 8 — dl-* 12 H i- — ® — *j
inch | inch | inch
^ inch gap 1 inch gap
Figure 3.31 Thick-w alled steel cylinders (N orm an and Prendergast).
@seismicisolation
@seismicisolation
3000
.c
u
c
2000
in § Wall thickness (inch)

D
y- 1000
— i------1-------- 1-------1-------1____ » i
o 200 400 600 800
Surface overpressure (p s i)
Figure 3.32 E xperim ental values of th ru st at springline (N orm an and Prendergast).
constructed so that the cylindrical section of the end cap was structurally
separate from the central test section. This ensured that the ends of the
test section were essentially representative of a free boundary, since there
was no transfer of axial load. Strains were m easured circumferentially at
the m idpoint of the test section, and relative deflections of crown and invert
were recorded. The surface load was applied by air pressure though an
impervious membrance, and a typical loading rate was 130 psi/min.
It was found that the thrusts developed in the walls of the cylinders were
in general a linear function of the applied surface pressure, with the maximum
and m inim um values recorded at the springline and crown respectively.
Typical relationships are shown in Fig. 3.32, for 6 inch diam eter cylinders
having thicknesses of J and f inch respectively.
The values of bending moments recorded at the crown, invert and spring-
line (the ends of the horizontal diameter) were not linearly related to surface
pressure, but increased at a slower rate as the surface pressure increased
during the early stages of loading, as indicated in the typical experimental
plot shown in Fig. 3.33. The bending moments were in good agreement with
the results of an analysis of a thick-walled cylinder subjected to a uniform
vertical pressure of p and a uniformly distributed horizontal pressure k0p,
where k0 is taken as the coefficient of earth pressure at rest. When the cylinder
is so stiff that its horizontal deflection at the springline can be neglected,
it is reasonable to assume this value for k0.
The m om ents per unit length, M e, in terms of the polar coordinates
R and 0, where 0 is m easured anticlockwise from the springline, are given by
the expression
(3.34)
(see Fig. 3.34).
@seismicisolation
@seismicisolation
Surface overpressure (p si)

Figure 3.33 E xperim ental values of m om ent 30° below springline (N orm an and
Prendergast).
Figure 3.34 Thick-w alled cylinder loading (H endron et al).
Clearly, if A:0 = 1 there are no moments in the cylinder wall, and hydro
static loading is then represented. W ith k0 taking a typical value of 0.4,
the m om ent at the springline becomes
M 0 = - 0 .1 5 p R 2. (3.35)
If the cylinder is flexible enough to deflect horizontally, k0 will increase
in value, and the greater will be the tendency to produce a uniform radial
pressure distribution, with a consequent reduction in bending moment.
The tests also indicated that, for the high-stiffness cylinders tested, the
influence of burial depth was insignificant, and could not be defined with
@seismicisolation
@seismicisolation
Figure 3.35 Design m om ents for thick-w alled tunnel liners (N ew m ark et al.).
sufficient clarity to w arrant its inclusion as a design parameter. On this

basis, Newm ark et al. [3.39] have proposed the design curves shown in
Fig. 3.35, linking the ‘m om ent coefficient’ M /p R 2 with the ‘normalized
pressure’ p R 2/ E l for all depths of burial greater than D/8, where D is the
cylinder diameter. These curves are thought to be conservative.
References
3.1 M arsto n , A. and A nderson, A.O. (1913) Iow a Engineering Experim ent Station,
Bull. no. 31.
3.2 M arston, A. (1930) Iow a Engineering Experim ent S tation, Bull. no. 96.
3.3 Spangler, M .G . (1948) U nderground c o n d u its-a n appraisal o f m odern research.
Trans. A S C E , 113, 316.
3.4 C larke, N .W .B. (1968) Buried Pipelines, M aclaren, L ondon.
3.5 Schlick, W .J. (1932) Loads on Pipe in Wide Ditches, Iow a Engineering Experim ent
S tation, Bull. no. 108.
3.6 W etzorke, M. (1960) Uber die Bruchsicherheit von Rohrleitungen in parallel wandi-
gen Graben, Vevoffentlich. Inst. Siedlungs W asserw irtsch., Tech. H ochschule,
H annover, 5.
3.7 L arsen, H. (1977) E arth pressure aro u n d buried pipes. P hD thesis, Univ. of
C am bridge.
3.8 C hristensen, N .H . (1967) R igid Pipes in Sym m etrical and Unsymmetrical Trenches,
D anish G eotechnical Institute, Bull. no. 24.
@seismicisolation
@seismicisolation
3.9 Spangler, M .G . (1950) Iow a Engineering Experim ent S tation, Bull. no. 170.
3.10 Spangler, M .G . (1950) A theory o f loads on negative projecting conduits. Proc.
US Highw. Res. Board, 30.
3.11 Spangler, M .G ., et al. (1926) Iow a Engineering Experim ent S tation, Bull. no. 79.
3.12 G ilboy, G . (1933) Influence tables for solution o f Boussinesq equations. Proc.
A S C E , 59, 781.
3.13 Steinbrenner, W. (1937) Wasserwirtsch. Technik, 4, 27.
3.14 Spangler, M .G . and H ennessey, (1946) A m ethod o f com puting live loads tran s
m itted to u n d erground conduits. Proc. US Highw. Res. Board.
3.15 Voellm y, A. (1937) Eingebelfete Rohre, Eidgenossischen Tech. H ockschule,
Zurich.
3.16 W astlund, G. and Eggwertz, S. (1949) Dimensionering au betongror, Svenska
K om m unaltekniska Foreningens H andlingar, no. 6.
3.17 P ruska, M .L. (1961) Ann. Inst. Tech. Batim. Trav. Publics, M ar./A pr.
3.18 L eo n h ard t, G. (1973) Strasse Briicke Tunnel, 25 (12), Dec., 324.
3.19 A nand, S.C. (1974) Stress distributions aro u n d shallow buried rigid pipes. J.
Struct. Div., A S C E , 100, ST1, Jan., 161.
3.20 K rizek, R .J., et al. (1974) Field perform ance o f reinforced concrete pipe. Transp.
Res. Rec., no. 517, p. 30.
3.21 T ro tt, J.J. an d G au n t, J. (1972) Experimental Work on Large Steel Pipeline at
Kirtling, T ra n sp o rt and R oad Research Lab., Rep. LR472.
3.22 N ayak, G .C ., et al. (1976) F inite elem ent analysis o f ditch conduits. Proc. Int.
Symp. on Soil-Structure Interaction, Univ. o f R oorkee, India.
3.23 A llgood, J.R ., et al. (1976) C AN D E: A Modern Approach fo r the Structural
Design and Analysis o f Buried Culverts, US N aval Civil Engineering Lab.
3.24 Y oung, O .C. (1977) The Structural Design and Laying o f Small Underground
Drains o f R igid M aterials, T ran sp o rt and R oad Research Lab., Suppl. Rep. 303.
3.25 Spangler, M .G . (1973) Long-tim e m easurem ents o f loads on three pipe culverts.
Highw. Res. Rec., no. 443.
3.26 B raune, G .M ., et al. (1929) E arth pressure experim ents on culvert pipe. Public
Roads, 10 (9), N ov., 153.
3.27 P ettibone, H .C . and H ow ard, A .K . (1966) Laboratory Investigation o f Soil
Pressures on Concrete Pipe - Progress Report No. 2, US B ureau o f R eclam ation,
Rep. no. EM-718.
3.28 P ettibone, H .C . and H ow ard, A .K . (1967) D istribution o f soil pressures on
concrete pipes. J. Pipeline Div., ASCE, 93, PL2, July.
3.29 H ainsw orth, I.H ., et al. (1967) Pilot Experiments to Determine the Loads Causing
Failure o f Sewer Pipes Under Roads, Institu tio n o f Civil Engineers, L ondon.
3.30 Janson, J.E . (1965) Undersdknung avseende belastninger pa styva ror i mark,
K ungl.vag-och V attenbyggnadsstynelsen, Stockholm .
3.31 K rizek, R .J. an d M cQ uade, P.V. (1978) B ehaviour o f buried concrete pipe.
J. Geotech. Eng. Div., ASCE, 104, G T7, July.
3.32 Spangler, M .G . (1960) Soil Engineering, 2nd edn, In ternation al T extbook Co.,
Scranton, PA, p. 396.
3.33 D avis, R .E. et al. (1974) C oncrete pipe culvert behaviour, P arts 1 and 2. J. Struct.
Div., ASCE, 100, ST3, M ar., 599.
@seismicisolation
@seismicisolation
3.34 Jo hnson, A .M . and Hess, J.D . (1963) L oad testing o f no-joint, cast-in-place
concrete pipe. US Highw. Res. Rec., no. 30, p. 20.
3.35 Bush, R .Y . an d C hak, J.S. (1978) S tructural perform ance o f cast-in-place pipe.
J. Transp. Eng. Div., ASCE, 104, TE5, Sept., 605.
3.36 Page, J. (1966) Impact Tests on Pipes Buried Under Roads, R oad Research Lab.,
Rep. R RL-35.
3.37 A llgood, J.R . and H errm an n , H .G . (1969) Static Behaviour o f Buried Reinforced
Concrete M odel Cylinders, US N aval Civil Engineering Lab., Tech. Rep. R606.
3.38 N o rm an , D .C . and Prendergast, J.D . (1973) Behaviour o f S tiff Cylinders Buried
in Sand Under Static Loading, US W aterw ays Experim ent Station, Tech. Rep.
N-73-1.
3.39 N ew m ark, N .M ., et al. (1968) Design o f Cylindrical Reinforced Concrete Tunnel
Liners to Resist Air Overpressures, US N aval Civil Engineering Lab., C o ntract
R ep. C F 68.010.
Further reading
A hm ed, S., M cM ickle, R.W . an d Brassow , C.L. (1981) S o il-p ip e in teraction and
pipeline design. J. Transp. Eng. Div., ASCE, 107, TE1, Jan., 45.
A tkinson, J.H . and P otts, D .M . (1976) Subsidence Above Shallow Circular Tunnels in
Soft-ground, U niv. o f C am bridge, Tech. Rep. C U E D /C -Soils, T R 27.
A zar, J.J., B om ba, J.G . and R andolph, V.G. (1975) Pipeline stress analysis. J. Transp.
Eng. Div., ASCE, 101, TE1, Feb., 163.
Brown, C.B. (1967) Rigid culverts under high fills. J. Struct. Div., ASCE, 93, ST5, Oct.,
195.
D avis, R .E ., Bacher, A .E. and O berm uller, J.C . (1974) Structural Behaviour o f a Con
crete Pipe Culvert, C alifornia Business & T ran sp o rtatio n Agency, D ept o f Public
W orks, Rep. R & D 4 - 7 1 .
H ann a, M .M . (1963) Stresses in circular pipes buried in open cuts. Proc. Inst. Civ.
Eng., p ap er 6629.
H avell, R .F . and Keeney, C. (1976) L oads on buried rigid c o n d u it - a ten year study.
J. Water Pollut. Control Fed., 4 8(8), A ug., 1988.
H oeg, K. (1968) Stresses against u n derground structural cylinders. J. Soil Mech. Found.
Div., ASCE , 94, SM 4, July, 833.
Jam es, R .G . an d L arsen, H. (1975) Centrifugal M odel Tests on Buried R igid Pipes,
Univ. o f C am bridge, Tech. Rep. C U E D /C -Soils, T R 21.
K ovach, E .M . (1982) Protection o f pipelines subject to surcharge loads. J. Transp. Eng.
Div., ASCE , 108, TE6, N ov.
L eonards, G .A . (1962) Culverts and Conduits, Foundation Engineering, M cG raw -H ill,
N ew Y ork.
N ath , P. (1977) Finite Element Analysis o f a Large Diameter Buried Steel Pipeline. T ran s
p o rt and R o ad R esearch L ab., Rep. 778.
Page, J. (1966) Impact Tests on Pipes Buried Under Roads, T ran sp o rt and R oad R esearch
L ab., Rep. 35.
Parm elee, R .A . and C orotis, R.B. (1972) The Iow a deflection form ula: an appraisal.
@seismicisolation
@seismicisolation
P earson, F .H . (1977) Beam behaviour o f buried rigid pipelines. J. Environ. Eng. Div.,
ASCE, 103, EE5, Oct.
R ichards, R. and A graw al, J.S. (1974) Stresses on shallow circular pipe by transform ed
section. J. Geotech. Eng. Div., ASCE, 100, G T6, June, 637.
R ude, L.C. (1982) M easured perform ance o f a lab o rato ry culvert. J. Geotech. Eng. Div.,
A SCE , 108, G T12, Dec.
R ude, L.C. (1983) L oad reductions on buried rigid pipe. J. Transp. Eng. Div., ASCE,
109, TE1.
@seismicisolation
@seismicisolation
CHAPTER FOUR
Thin-walled pipes
under static loads
Thin-walled metal pipes, often m anufactured from corrugated sheeting,

have been used in the construction of culverts and conduits since the early
years of this century. These flexible pipes of relatively lightweight, low-
strength construction, can support backfills of considerable height by using
the passive pressures induced in the surrounding soil as the sides try to
move outwards against this soil. The high-deformation characteristics of
thin-walled pipes mobilize the passive pressures to great advantage, whereas
we have already seen that thick-walled, rigid pipes only have the active
lateral pressure of the surrounding earth to augment their load-carrying
capacity.
Analysis of the behaviour of thin-walled conduits must take account of
two requirem ents: the need to know the horizontal deflections at the spring-
line, and the need to know the ultim ate strength of the pipe in terms of
buckling of the wall - not in terms of the bending strength of the wall. The
problem of horizontal deflection was first examined in depth by Spangler
[4.1], following a good deal of test work on flexible conduits by Iowa State
College and the US Bureau of Public Roads in the 1920s and 1930s. Field
m easurem ents were m ade on conduits of widely varying flexibility, including
thin-walled steel and corrugated pipes, and the broad conclusions from
tests under a 35 ft high embankment, undertaken by the American Railway
Engineering Association, were as follows:
1. The vertical pressures due to backfill on flexible conduits were about
one-third of those due to similar backfills on rigid conduits.
2. The vertical pressures were about one-half of the weight of the column
of soil (or ‘soil prism’) above the conduits.
3. Pressures at the springing were equal to or greater than the pressures
of the crown or invert.
Details of these tests are given in ref. 4.2. As a result of these and other
conclusions, Spangler realized that any attem pt to rationalize the structural
analysis should consider a m ethod for determining pipe deflections under
@seismicisolation
@seismicisolation
specified conditions of load and installation, and we will deal first with
this aspect.
4.1 Deflection analysis (traditional methods)

It is best to start by examining the deflection of thin rings under simple
loading systems, using as a base the equation of equilibrium for a ring under
conditions of flexure. Thus, from Fig. 4.1, we have
M
(4.1)
Tr
where p is the radius of curvature of the elastic curve of the loaded ring,
R is the mean radius of the unloaded ring, 1/p — l/R is the change in curvature
due to the load, M is the bending m oment at any point on the ring, and
I is the second mom ent of area of ring cross section.
C onsideration of the relationship between radial displacement w and
radius of curvature leads to the final differential equation
d 2w MR
+ w= (4.2)
~eT
where 6 is the angular displacement from the invert of any point, having a
radial deflection w.
Consider next a ring of radius R compressed by two forces P acting along
a diameter. Denoting the bending moment at A and B by M 0, we find that
the m om ent M at any cross section is
M = M, cos 0). (4.3)
Figure 4.1 T hin ring under com pression by diam etrically opposed forces.
@seismicisolation
@seismicisolation
Thin-walled pipes under static loads 91
and equation (4.2) becomes
d 2w M nR 2 PR'
+ w= — (1 —cos 6). (4.4)
d62 El 2EI
Solving this equation gives the expression
P R 3( (4.5)
w= cos 6 4- 6 sin 6 —
4E l
from which the radial deflections at any point can be found.
The lengthening of diam eter AB is
P R 3/ 4
(4.6)
2 E I\n ~
and the shortening of diam eter CD is
PR3/ 8
(4.7)
4 E f[n ~ n
The variation in mom ent around the ring is given in Fig. 4.2. The expression
for the m om ent at C and D is
M c = M d = 0.318 PR, (4.8)
and at A and B it is
M a = M b = - 0 .1 8 2 P P (4.9)
The above analysis uses small-deflection theory assumptions about the
radius of curvature and is strictly valid only as long as the deflections are
less than 3% of the ring diameter. F or larger deflections the relationship
between P and w becomes non-linear, and this has been shown in experiments
on thin-walled pipes of 36, 42 and 60 inch diameter, reported by Spangler
[4.1].
0*182
120 150 180

6 (d e g )
Figure 4.2 V ariation in m om ent a ro u n d the ring in Fig. 4.1 (Spangler).
@seismicisolation
@seismicisolation
WC /2 R
kU X/2)
Subgrade
Wc / 2 f l c o s a
Figure 4.3 D istrib utio n of pressure on a flexible culvert (Spangler).
If the wall of the pipe is corrugated, it is necessary to use an ‘equivalent’

value for I in equation (4.5). F or sheet having corrugations ^ inch deep
spaced 2f inch centre to centre, a formula developed by Jensen at Iowa can
be used. Thus,
7 = 0.029 251 0.001 5012 + 0.104 25 V3 - 0.002 2514. (4.10)
A close approxim ation to this, often used, is
I = 7/30. (4.11)
The results of two experiments, in 1927 and 1934, in which the behaviour
of 42 inch diam eter corrugated pipe was observed (including the measure
m ent of horizontal earth pressures) led Spangler to propose that the stress
distribution around a flexible pipe should take the form shown in Fig. 4.3.
The assum ptions may be summarized a s :
1. The vertical load is distributed uniformly over the width of the pipe.
2. The vertical reaction is distributed uniformly over the width of the bedding
of the pipe.
3. The horizontal pressure is distributed parabolically over the middle
100°, and the maximum unit pressure is equal to the modulus of passive
resistance of the fill m aterial multiplied by one-half the horizontal de
flection of the pipe (AX/2).
Using this, he developed mathem atical expressions for moments, thrusts
and deflections, and found that the horizontal deflection (A 27) was given by
yBWCR :
AX = (4.12)
E1 + 0.061 k R * ’
@seismicisolation
@seismicisolation
where the bedding factor yB is given by
yB = 0.5 sin a —0.082 sin2 a + 0.080a/sin a
—0.16 sin a (n — a) —0.04 sin 2a/sin a
+ 0.318 cos a - 2 0 8 , (4.13)
k is the m odulus of passive resistance, or ‘coefficient of subgrade reaction’

of the soil, which has a non-linear relationship with stress, a is one-half
of the bedding angle (see Fig. 4.3), and W c = vertical load per unit length
of the pipe.
From equation (4.13) the relationship between yB and a is:
a(deg) 0 15 22.5 30 45 60 90
yB 0.11 0.108 0.105 0.102 0.096 0.090 0.083
as shown in Fig. 4.4.

Later experiments by Spangler confirmed the assumptions of uniform
vertical load distribution and of parabolic distribution of horizontal pressure
for a = 45°. The total horizontal deflections agreed closely with equation
(4.12). It was noted that the test pipes continued to deflect vertically long
after the com pletion of the backfill - in some instances for many years after
the trials. The data for a 42 inch diam eter culvert under a 15 ft backfill
were:
Tim e (years) 0 0.5 4.5 7.25 9.75

Vertical deflection (inch) 1.43 1.78 2.44 2.59 2.62
30 60 90
Half bedding angle, oc (d e g )
Figure 4.4 R elationship between yB and a.
@seismicisolation
@seismicisolation
Because of this, a second factor, yT, known as the ‘deflection/time factor’
is often incorporated in equation (4.12), so that
AX = W rK * 3 (414)
E l + 0.061 k R K
An alternative proposal is to accept that the m odulus of passive resistance,
k, is also dependent on time. If the vertical load is due to a pressure p acting
over the width of the pipe, then.
Wc = 2pR. (4.15)
Using this equation, and taking a = 90°, yT = 1, gives
and since, for most thin-walled pipes, the stiffness term in the denom inator
(El) is small com pared with the term due to passive resistance of the soil
(0.061 /cP4), a good approxim ation is
A X = 2.1 p/h. (4.17)
This approxim ation has been used by M eyerhof [4.3] and it implies that,
for a given constant standard of soil com paction and bedding, total deflec
tion is proportional to the average vertical pressure on the crown. Note
that this is independent of the elastic properties of the pipe.
A problem arising in the application of equation (4.14) or (4.17) is that
k is not related to a standard soil property readily measured in the laboratory.
A m ore readily available property is the coefficient of lateral earth pressure,
or ‘constant of horizontal soil reaction’, and M eyerhof has therefore proposed
the approxim ate relationships:
clays
k = kJL5R, (4.18)
dry and moist sands
k = kt H/1.5R, (4.19)
submerged sands
k = k H /3 R , (4.20)
where kc is the constant of horizontal soil reaction for clay, ks is the constant
of horizontal soil reaction for dry and moist sand, and H is the cover depth.
N ote that kc and ks have different dimensions. The former, in imperial units,
is expressed in pounds per square inch, and the latter in pounds per cubic
inch. Typical values of kc and ks, proposed by Meyerhof, are given in Table
4.1.
@seismicisolation
@seismicisolation
T able 4.1 Typical values of kg and kc
kc (clay) stiff 500-1000 psi

very stiff 1000-2000 psi
hard > 2000 psi
ks (sand) loose 1.5-4.0 pci

com pact 4.0-12.0 pci
dense > 12.0 pci
If the moduli of deformation of sand and clay are Es and Ec respectively,

then, approxim ately
K = Ec, (4.21)
K = EJH , (4.22)
so that equation (4.17) becomes
dry and moist sand
A X = 4 pR /E s, (4.23)
clay
A X = 4 p R /E c. (4.24)
Returning to equation (4.16), and substituting for k, we can write the equation
in a different way. Thus, for sands,
“ *o-("(!£ ) ( i +0.J e,i»/h ) «4251

a variation proposed by Spangler some time after the publication of his
original paper.
A further variation was proposed by Barnard [4.4], in connection with
studies carried out on behalf of the American W ater W orks Association
Com mittee and given in their design manual. He regarded the pressure
exerted by the thin-walled pipe on the side fill as similar to the pressure on
a footing strip having a width proportional to the pipe diameter, and assumed
that the ‘horizontal settlement’ would be equal to the shortening of a
uniformly stressed length of soil, equal to 1.25D. Then, for a given value
of the lateral strain in the soil, es,
A X /D = 2 x 1.25 x es = 2.5es, (4.26)
or
A X = 5esR. (4.27)
@seismicisolation
@seismicisolation
N oting that es = p/Es, we get
A X = 5pR/Es, (4.28)
which gives a value for A X about 25% greater than the approxim ate equation
(4.23).
Returning to equation (4.14), an often-used rearrangem ent is
AX = — ’ (4.29)
E l / R 3 + 0.061 E'
where E’ = kR. This was suggested by W atkins and Spangler [4.5], who
pointed out that kR was dimensionally correct, and similar to the compressive
modulus of elasticity of s o i l - a pipe-soil interaction m odulus in fact.
Further, E I / R 3 can be regarded as a ‘ring stiffness’ factor, because it is the
product of the elastic m odulus of the pipe wall material and the moment
of inertia of unit length of the cross section ( = t 3/12). Thus,
A x = _________ loading param eter____________________ ^
ring stiffness factor + soil stiffness factor
Tests by the US Bureau of Reclamation, reported by Howard [4.6], give
values for Ef in pounds per square inch as in Table 4.2.
These figures were representative of calculations based on measurements
in 113 field installations in which initial deflections were measured after
construction; pipe deflection was measured between the time of placing
the soil to the top of the pipe and completion of backfilling; the horizontal
deflection AX was measured in most instances. If AX was not measured,
but the vertical deflection A Y was known, then it was assumed that
AX = 0.913 A 7, which is the ratio given by equations (4.6) and (4.7). Howard
Table 4.2 Values for E (psi)
Relative density (Proctor)
D um ped <40% 40-70% > 70%
Fine-grained soils with

less th an 25% coarse
grained particles 50 200 400 1000
F ine-grained soils with

m ore th an 25% coarse
grained particles 100 400 1000 2000
C oarse-grained soils 200 1000 2000 3000
C rushed rock 1000 3000 3000 3000
@seismicisolation
@seismicisolation
Table 4.3 A ccuracy of predictions for deflections
D um ped and slightly com pacted over 90% within ± 2%

M oderately com pacted over 90% within ± 1%
H igh com paction over 80% within ± 0.5%
showed that, by using the figures in Table 4.2, the deflections could be
predicted quite closely; as shown in Table 4.3.
F or a further discussion of the relative merits of the traditional Spangler
and B arnard approaches, the reader should consult ref. 4.7.
4.2 Deflection analysis (recent developments)

The semi-empirical methods surveyed above are based on parameters
that are defined from test results, so care should be taken in employing
the solutions to a wide range of problems. In an attem pt to generalize the
analysis, interest has turned towards the treatm ent of the soil as an elastic
continuum , interacting with the circular conduit as an elastic shell, along
the lines discussed in Section 2.2.1. Two methods of solution have been
considered, a closed-form elastic solution and a finite element program.
4.2.1 Closed-form elastic solution

This has been examined in detail by Burns [4.8] and Burns and Richard
[4.9], who provided an ‘exact’ solution for an elastic circular conduit encased
in an isotropic, homogeneous infinite elastic medium, with a uniformly
distributed pressure acting on horizontal planes at an infinite distance.
The conduit-m edium interface was treated as (a) a frictionless boundary,
with only tangential forces transm itted, and (b) as a bonded boundary,
with tangential and norm al forces transm itted. The solutions give radial
and tangential soil pressures on the conduit, radial and tangential displace
ments of the conduit wall, and moments and thrusts in the wall.
Burns took the values of M s and k0 defined in Chapter 2, i.e.
E( 1 — v)
M s = - ------------ ■— > 4.31)
s (l+ v )(l-2 v )
and defined two new constants related to k0. These were
(4.33)
@seismicisolation
@seismicisolation
and
c = ^ \ - k 0) = x- f x 2v
2 \ 1 —v J (4.34)
Two further dimensionless param eters related to M s were defined as:
EA 6E l
where A is the thrust area per unit length of conduit.

Burns and Richard expressed stresses and displacements in the elastic
medium in terms of a stress function, using plane strain relationships, and
also used the displacement equations for circular cylindrical shells. They
showed that, for equilibrium and compatibility of displacements, the radial
displacements of the pipe w are given b y :
bonded interface
WR
w = [BU( 1 - a0) - CV( 1 + a2 - 2b2) cos 20], (4.36)
Ms
where
a0 = C(2BU - 1)/B(1 + 2 C U \
a2 = [CV(2BCU - C ) + C2U/2 - B~\/d, (4.37)
b2 = [ CV(B + 2BCU) - BCU/2 - B~]/d,
d = CV(\ + B + 2BCU) + CU{1 + C/2) + C + 1;
frictionless interface
WR
w = — [ B C / ( l - a o) - f C F ( l + 3a2 - 4 h 2)cos20], (4.38)
Ms
where
a0 = C(2BU - l)/B(\ + 2C U \
a2 = (4B C V + C)/d,
(4.39)
b2 = (4B C V - B)/d,
d = 4B C V + C + 2.
The interaction between the conduit and the surrounding medium
occurs mainly within a distance of 1.5 conduit diameters from the conduit
center, as we found earlier in Chapter 2, and the conduit-soil system can
@seismicisolation
@seismicisolation
therefore have finite boundaries. The solution is therefore applicable to cover
depths H , from the suface to the crown of the conduit, greater than 2r, and
should not be used for H <r.
The tangential displacements v of the pipe corresponding to equations
(4.36) and (4.38) are:
bonded interface
WRl
v = - — -(1 - a 2 + 2b2 C/B) sin 20, (4.40)
Ms 2
where a2 and b2 are defined in equation (4.37);

frictionless interface
WR C V
v = - y — j i l + 3a2 “ 4h2) sin 20, (4.41)
where a2 and b2 are defined in equation (4.39).

C onduit displacements and the variation of displacements in the soil
with distance from the pipe wall are shown in Figs 4.5,4.6 and 4.7, assembled
from ref. 4.9. The displacements are given in terms of w M J W R and v M J W R ,
and their attenuation is shown to be quite rapid along the horizontal through
the centre of the pipe.
Figure 4.5 C on d u it displacem ents and atten u atio n of soil displacem ents for k0 = 0.5,
U = 0.1, V — 100, bonded interface (Burns and Richard).
@seismicisolation
@seismicisolation
wms / wr
0-53
2R vMs / W R
Figure 4.6 C o nduit displacem ents and atten u atio n of soil displacem ents for k0 = 0.5,
U = 0.1, V = 100, frictionless interface (Burns and Richard).
0-46
0-5
2H WR
Figure 4.7 C o nduit displacem ents and atten u atio n of soil displacem ents for kQ= 0.5,
U = 0.1, V — 3, bonded interface (Burns and Richard).
4.2.2 Finite element program

This has been fully described in a report by K atona et a l [4.10]. Their
program, named C ANDE (Culvert ANalysis and DEsign), uses a static
form ulation derived from ‘virtual work’ principles, and three basic element
types: a quadrilateral element for the soil, an interface element for the
conduit-soil junction, and a bending-thrust element for the conduit.
@seismicisolation
@seismicisolation
The quadrilateral element is a non-conforming element developed by

H errm ann [4.11], and associated with it are three types of material behaviour:
(a) linear elastic, (b) incremental elastic, where the elastic moduli are depen
dent on overpressure, and (c) a variable m odulus model. The basic mesh
employs 86 quadrilateral elements and 10 bending-thrust elements for the
conduit, shown in Fig. 4.8.
(a) linear elastic soil model

Although a soil rarely behaves as a linear elastic medium, such an approxi
m ation often provides a reasonable representation, particularly during the
early stages of a design when the effect of varying geometry and stiffness
@seismicisolation
@seismicisolation
param eters is under scrutiny. The linear form (for planestrain conditions) is
[C n 0 re
c 12 ^22 0 (4.42)
a> r 0 0 c 33
where
c » = c 22 = M s,
C 12 = Msk0’ (4.43)
r — M s( 1- - * o ) .
33
2(1 - v)
(b) Incremental elastic soil model

This is the application of the linear model in a series of steps, assuming that
soil stiffness increases with overpressure (valid if the soil is in confined com
pression). A typical relationship between overburden pressure oy and axial
strain ey is shown in Fig. 4.9, which gives a relationship between M s and oy
of the type shown in Fig. 4.10. Increments of stress Aoy are related to incre
ments of strain Aey, by
Aoy = M TAey, (4.44)
where M T is the average tangent modulus.
@seismicisolation
@seismicisolation
Figure 4.10 T ypical relationship between secant m odulus and overburden pressure
(K ato n a et al).
(c) Variable modulus model

A variable m odulus model due to H ardin [4.12] was used in the CANDE
program. This model relates the secant shear m odulus Gs to the maximum
value of shear m odulus Gmax and the hyperbolic shear strain yh, in the form
G = ^ max (4 45)
s l + y h’ 1 j
where yh is dependent on the ratio of shear strain y to a ‘reference’ shear

strain.
K atona et a l reported that the predictions of the CANDE program for
vertical and horizontal diametrical deflections are in good agreement with
experimental data for 10 ft and 7 ft diameter conduits, the former of corrugat
ed steel, the latter of reinforced concrete.
A finite element program has also been evolved by Robinson [4.13], using
the mesh shown in Fig. 4.11to analyse the effect of burial depth under static
loading. A com puter code known as SAP4, a Structural Analysis Program
for static and dynamic response of linear systems, described by Bathe et a l
in ref. 4.14 was used for the analysis. The mesh boundaries were located at
two diameters from the edge of the cylinder, and the boundary conditions
were chosen to simulate a roller bearing surface at the vertical and lower
horizontal edges. All other grid points were unrestrained. The tunnel liner
was modelled by 24 interconnecting bending-thrust elements, which infers
full bonding and no slip between the structure and the soil. The stress analysis
followed closely a treatm ent by Hoeg [4.15], which is similar to that described
above in relation to the work of Burns and Richard, but is not restricted to a
uniaxial strain condition.
@seismicisolation
@seismicisolation
Figure 4.11 Basic m esh of SAP4 static analysis system.
4.3 Buckling under static loads
4.3.1 Elastic analysis

Experiments show that when thin-walled cylindrical tubes are buried in well
com pacted soils, with a depth of cover from crown to surface of more than one
radius, the chances of excessive deformation during the loading process are
very slight. The cross section of the tubes remains sensibly circular until
catastrophic failure by buckling of the curved wall occurs. This can take place
in the elastic range of wall compression stresses if the diameter/thickness
ratio {D/t) is very high, or in the plastic range at lower values of D/t. At
interm ediate values, there is a transition zone in which there is an interaction
between the two collapse modes.
@seismicisolation
@seismicisolation
Figure 4.12 Thin-w alled circular tube under radial pressure in an elastic medium.
If a thin-walled circular tube of infinite length is surrounded by a well

com pacted m edium with elastic properties, and is subjected to uniform radial
pressure, as shown in Fig. 4.12(a), the tangential compressive stress ac is
found from
pR pD
(4.46)
The upper limit to the value of ac is the crushing of the thin wall in compres
sion due to breakdow n of material. This is only reached at very low values of
D/t. W hen D/t has a high value, it is necessary to calculate the elastic critical
radial pressure pcT to cause buckling of the wall. In general, for thin-walled
long tubes, it can be shown that
(n1 - 1)EI
(4.47)
p" (i >2 ) R 3 ’
a well known formula, the derivation of which is given, for example in

Tim oshenko and Gere [4.16], p. 292.
Here n is the num ber of full waves formed around the circumference at
buckling, and v is Poisson’s ratio. Since I r/1 2 , per unit length of tube,
equation (4.47) can be rewritten as
@seismicisolation
@seismicisolation
A thin-walled tube under uniform radial pressure, but not supported in an
elastic medium, will buckle into an oval shape with n = 2, so that
3EI
P“ = (1 - v 2)K3' (4'49)
If the tube is constrained by an elastic medium or some other external effect
to buckle with large numbers of full waves, then n is large compared with
unity, and it is sufficiently accurate to write equation (4.47) as
n2EI
p" = ( i ^ W (450)
W hen we analyse the effect on critical radial pressure of a soil medium, it is
first assumed that the soil provides elastic radial support, as shown in
Fig. 4.12(b),where the spring constant is kz (in units of pressure per unit
radial deflection). Cheney [4.17], has derived the expression
El k R
(^ -D -^ + r r ^ "> 2, (4.51)
lR2 + {n2 - 1)’
where the buckling mode in terms of numbers of full circumferential waves
was given by
■-[■♦(Syr
Substituting for n in (4.51) gives the approxim ate expression
£ / \ 1/2
P„ = 2 ( V rtJ • (4.53)
which underestim ates the critical pressure by less than 10% for n > 5, and
by lessthan 1% for n > 10. In the application of this method, the m ajor
problem is to relate the value of kz to a m easurable soilproperty. Suppose we
assume that the strain in the surrounding soil is equal to the radial deflection
divided by the tube radius, then
kz = E J R , (4.54)
where Es is the soil modulus obtained from triaxial tests at the density and
stress applicable to the design conditions.
We can then write (4.53) as
Pcr = 2 ( £ s^ j . (4.55)
or
pcr - 2(8 E E ') 1'2, (4.56)
where E' is the specific stiffness of the tube, and is equal to E l/D 2.
@seismicisolation
@seismicisolation
At about the time Cheney was deriving this expression,Meyerhof and
Baikie [4.18] were evaluating pcr by modifying the theory of buckling of flat
plates on an elastic foundation. Using the m odulus of subgrade reaction /cm,
with units of pressure per unit radial deflection, they found that
( « + l)2 E l (1 - v 2) k R
1 - v 2 K3 + (n + l)2 - l ’ ( *
which, for larger values of n, becomes
„ 4 k™EI Y '2 (4-58)

p« Z\ ( 1 - V2)R 2) '
In assessing the values of km, they proposed the conservative relationship
km = EJl.5R, (4.59)
and using this, equation (4.58) becomes
Es E l \ l/2
1.5(1 - v2)~R-
F urther work by Luscher and Hoeg [4.19] in 1964 gave, for the higher
buckling modes,
E lY 2
P„
Pcr =
= 2^ ( K
VL ^ J , (4.61)
and
k R 3\ 114
-in -) • (462)
where kL is the coefficient of elastic soil reaction (in units of pressure per unit
radial strain). N ote that if kL is taken equal to Es, equation (4.61) becomes
EI_ y /2
PCr = 2 E — \ , (4.63)
which is identical to equation (4.55).

If the soil is taken as a true elastic medium, with a m odulus Es and Poisson’s
ratio vs, Forrestal and H errm ann [4.20] found in 1965 that for a long cylin
drical shell in which there was no bond between the shell and the elastic
medium (i.e. no shear stresses present)
(n2 - 1)£ / E,
= \ ■ (4-64)
(1 - v2)K3 + (1 + v j ( l —2vJ(n + l)n
At high values of n this becomes
@seismicisolation
@seismicisolation
Later Duns [4.21] used the Donnell shell stability equation to arrive at a
similar relationship. F or the unbonded condition he found that, when n ^ 7,
there was an approxim ate relationship between n, kL and Es in the form
nE(
- - '- f t . (4.66)
80 - v-)
W hen there was full bonding the value of kL was increased fourfold, so that
(4OTI
At high values of n, taking v = vs and neglecting bonding,
2n2E I . R ( 12E \ 1/3
= tr z v w ' where n= 2 t{-r) ■ (468)
Thus, for a given large num ber of full circumferential waves, the value of pcT
is double that given in (4.50) for buckling in free air.
Papers by H abib and Phong [4.22] and Sonntag [4.23] include a further
discussion of buckling in a soil-surrounded medium, and the reader should
see these.
All the theories discussed so far assume that when buckling occurs the
elastic medium stays in contact with the wall of the cylinder. Experiments
have shown that this is not always the case, and that the half-waves in the
cylinder wall which deflect inwards part com pany with the medium as they
do so. This condition was analysed by Chelepati [4.24], who chose an elastic
foundation with linear spring characteristics in compression and zero resist
ance in tension. He found that
{n2 - l)§ + w h -y n>z (469)

Com paring this with equation (4.51) shows that the loss of contact introduces
a factor 2 in the denom inator of the second part of the expression for pcr.
The approxim ate lower bound solution of equation (4.69) is
» “ Y“
and when this is com pared with equation (4.53) the loss of contact is seen to
reduce pcr in the ratio l/y /2.
The most commonly used equations are (4.55) or (4.56), and these are
recommended in a num ber of design practice documents, for example, by
Com pston et a l [4.25] in CIRIA Report 78. Accurate tests by Allgood and
Ciani [4.26] and by Bulson [4.27] have suggested that these equations depart
from the general trend of test data for deeply buried cylinders, a fact noted
by Cheney [4.28].
@seismicisolation
@seismicisolation
The design equation (4.55) can be written in terms of the thin shell property
I = E 3/ 12, and the thin sheet property E -> E/( 1 — v2), as follows:
(4.71)
The theory presented by Cheney [4.28] yields an expression for pcr that gives
closer agreement with the trend of results. Thus
pcr = 0.54£2/3£ 1/3(t/K). (4.72)
It is of interest that the two-thirds power has links with the buckling of
straight, flat strips on an elastic foundation. In fact, if the critical axial stress
for a thin flat sheet supported by an elastic medium is replaced by pcrR/t,
it can be shown that
pcr = 0.52E^13E l/3(t/R). (4.73)
Thus, for larger values of n, the cylinder wall behaves like a flat strip supported
by an elastic medium.
4.3.2 Ultimate strength related to flexibility

When the calculated elastic critical value of pcr gives rise to tangential com
pressive stresses in the wall of the cylinder that exceed the limit of propor
tionality of the stress-strain relationship for the cylinder material, the
ultim ate strength, or buckling strength, will fall below pcr. This is analogous
to the strength of a simple strut, which falls below the Euler buckling stress
as the compressive stress approaches the plastic range of the material.
The value of compressive stress ac that causes collapse of the lining can
be linked to the flexibility of the cylinder by a buckling curve similar in form
to well known strut curves. This is illustrated in Fig. 4.13, which shows the
relationship between <rc and a flexibility term (1/E)(R/t)2. At low values of
Flexibility, ^ ( j ) 2
Figure 4.13 Buckling curve for radially com pressed cylinders.
@seismicisolation
@seismicisolation
flexibility, oc approaches the compressive yield stress of the material op.
F or m aterials having no sharply defined yield, <7p could be taken as the 1 or
2% proof stress. At high values of flexibility, <
jc approaches the critical elastic
buckling stress oCT. At interm ediate values there is a transition between these
extremes, and W atkins [4.29] has proposed that this can be represented by a
formula similar to the Rankine formula for struts, i.e.
(4.74)
This is sometimes expressed as
The upper limiting value of <rc, viz. op, is often known as ‘ring compression
crushing’ such as would occur if the cylinder were enclosed by a medium
having an infinite m odulus (Es = oc). The ring compressive strength of
thin-walled cylindrical conduits that do not buckle has been examined by
White and Layer [4.30].
Equation (4.74) leads to rather low values of oc at flexibilities corresponding
to the junction of the curves oc = op and ocr = pcr R/t. At this point oc is
g J 2 . A m ore realistic interpretation would be to use a generalized version
of equation (4.74), in the form
(4.76)
By judicious variation of a and b the transition curve can be brought nearer

to the two basic curves, although the limiting conditions still hold. This
relationship is speculative, because there have been few, if any, well conducted
experimental program mes to study the true form of the curve. If a = b = 2,
for example, the relationship between oc and g c t at the transition point is
g c = G j y J l . If a = b = 3, the relationship becomes g c = gJ Z ^ / 2 , and so on.
4.3.3 Ultimate strength related to cover depth

In most practical designs, the radial pressure around a thin-walled circular
conduit is not uniform. W hen the pressure is due to the dead weight of earth
cover, there will be a pressure gradient across the vertical diameter of the
pipe, and, if this gradient is a significant fraction of the mean pressure, it
cannot be ignored. If the surface of the backfill is subjected to a high static
overpressure, the depth of the crown of the tube below the surface will have a
m arked effect on the level of pressure to cause collapse. This is particularly
true when the cover depth is less than one diameter, a condition that has been
the subject of much research and testing.
@seismicisolation
@seismicisolation
Surface
* N
\
/
/
Figure 4.14 A nnular ring of soil.
Luscher [4.31] used the concept of a thin cylindrical tube surrounded

by an annular ring of soil, as shown in Fig. 4.14. The radius of the annular
ring is R + H, where H is the cover depth over the crown of the tube. The
term H /R influences the value of kL in equation (4.61), and a method of
calculating kL is to examine the modulus of resistance of the soil ring, under
conditions of plane strain, to a uniform pressure applied radially inside the
cavity that would be formed if the metal tube were removed. Luscher used
the elastic theory of thick cylinders to establish the equation
{ l - [ R / ( R + fl)]2}£s
(4.77)
L ( l + v’s ) { l + [ R / ( R + t f ) ] 2 ( l - 2 v s ) } '
This relationship is given in Fig. 4.15 for vs = 0.3 and 0.5. N ote that when
vs = 0.5,
(4.78)
0 0-2 0 -4 0-6 0-8 1-0

R/(R+«)
Figure 4.15 V ariation of the coefficient of elastic soil reaction.
@seismicisolation
@seismicisolation
Figure 4.16 Thick elastic cylinder analysis.
and that when H = 0, kL = 0. When H is large com pared to R, kL -+^ Es . The

value of kL does not alter much at cover depths greater than H = 2R, but
there is a lot to be gained in terms of buckling strength if the tube is buried
so that H /R is at least 1.5.
The relationship between collapse pressure and depth of cover has also
been examined by Bulson [4.32], using the concept of a ring of soil, radius
(R + H \ as shown in Fig. 4.16. From the theory of thick elastic cylinders
(4.79)
where b = R H, p l is the uniform radial pressure on the outside of the

soil ring, p2 is the pressure exerted between the tube and the inside of the
soil cylinder, and is the tangential compressive stress at the inner
surface of the cylinder. N ote that changing the sign of the first term leads to an
expression for the radial compressive stress at the inner surface of the soil
cylinder. Thus,
(4.80)
where (eR)R is the radial compressive stress at the inner surface of the soil
cylinder. The tangential compressive strain at the inner surface (£R) is found
from the relationship
(4.81)
where vs is Poisson’s ratio for the soil, and Es is the elastic modulus of the
soil.
@seismicisolation
@seismicisolation
Now,
14.82)
where E is the elastic m odulus of the material of the tube, and if we substitute
for sR, and (<7^ in equation (4.81), then
P,
p2 2 ^
I 5'
E.R +. — R
i-L = I f (1 + V) + _I1:
Et
2 f| ,1 _ v - E R
i:—
(R + H)2 \ s Et
(4.83)
W hen H /R is large, the last term approaches zero, and in addition, E R / E t

is a small number. If we neglect EJR/Et, then at deep covers (H -*■ oo),
Pl = W 1 + Vs>’ (4 -8 4 >
and at shallow covers (H -►0),
P , = P 2. (4-85)
If the surface pressure ps is so high that pressure due to the dead weight of the
fill can be neglected, then, by elastic theory,
Pi = 2(1 — vs)ps. (4.86)
This equation assumes that no lateral expansion can occur and that the tube
acts as a rigid inclusion. Then, at deep covers, from (4.84)
= P i (1 + ,v.) (487)
s 4 ( 1 - vs) v
As the surface pressure is increased, p2 will eventually reach a value equal to
the collapse stress of the tube. If the R/t ratio forthe tube is sufficiently high
for buckling to occur in the elastic range, then at failure p2 = pcr. Further,
if the value of ps to produce this condition at deep covers is pmax, equation
(4.87) can be rewritten as
p = gcr(1 + V ^ (4 8 g)
ax 4 (1 - vs) '
At interm ediate values of cover
oc 1 - f — — I . (4.89)
R +H
Using equation (4.88) in conjunction with equation (4.68), and taking vs = 0.3,
we can establish the relationship
Pm„ = 0.12£s2/3£ 1/3(t/R), (4.90)
which is about half the value for p that would be derived from equation
@seismicisolation
@seismicisolation
(4.73). F or buckling in the elastic range, equation (4.90) agrees well with the
lower scatterband of a large num ber of experimental results, as will be shown
later.
If tube buckling does not occur in the elastic range, then the value of pcr
will need to be modified, using equation (4.76), after employing the relation
ship <rcr = pcrR/t. Thus,
where
(4.91)
and a and b are defined in the code of practice to give a conservative re
presentation of the transition curve.
4.3.4 Experimental studies

If cylindrical tubes are buried with a very small depth of cover (i.e. H < R /2),
and the surface is subjected to a static overpressure, tests show that collapse
usually results from a caving-in of the crown, as shown in Fig. 4.17(a). At
deeper covers, however, collapse in a well compacted sandy soil usually
follows buckling of the thin wall, shown in Fig. 4.17(c). At intermediate depths
a com bination of these two modes can take place (Fig. 4.17(b)). The collapse
sequence at deeper covers is shown in Fig. 4.18. As surface pressure is applied,
the crown of the tube deflects downwards and the sides very slightly out
wards; the invert has virtually zero deflection. A typical plot of radial deflec
tion during this early loading phase is shown in Fig. 4.19, and it can be shown
Surface
Well compacted
granular soil
©
(a )
©
(b )
O
(c)
Figure 4.17 C ollapse m odes of thin-w alled buried cylinders, (a) Shallow cover; (b)
interm ediate cover; (c) deep cover.
@seismicisolation
@seismicisolation
o o
(a) (b)
Q
(c) (d)
Figure 4.18 C ollapse sequence at deep covers, (a) O riginal shape; (b) lower rim buckles;
(c) one buckle g ro w s; (d) final shape.
Figure 4.19 Typical radial deflection plot.
by calculation, using the bending theory for rings, that the ratio between
total vertical deflection and total horizontal deflection is about what one
would expect if the pressure were uniform around the tube.
As the surface pressure is increased, the thin wall of the tube around the
lower half of the rim buckles into a num ber of elastic half-waves, and subse
quently one of these buckles grows large. This action causes the vertical
diam eter of the tube to contract, followed by inward collapse of the crown,
as indicated in the final drawing of Fig. 4.18. The m ajority of static test
program mes have used model-scale cylinders, usually burying them with a
horizontal orientation in com pacted sand or clay, contained in a closed tank.
The tank has usually been circular or square in plan, with a removable bonnet,
and pressure has been applied by compressed air or by water, acting through
a diaphragm onto the soil surface. All the available studies up to July 1965
@seismicisolation
@seismicisolation
E I/R 3
Figure 4.20 C om parison of tests and theory (Dorris).
were summarized in a report by Dorris [4.33], who also described his own
experiments. He com pared test results with the theories of Cheney [4.17]
and Forrestal and H errm ann [4.20], as shown in Fig. 4.20.
The author conducted a series of model tests in the late 1960s, in the rig
shown in Fig. 4.21, details of which are given in Bulscn [4.34]. It measured
5 ft square in plan, 4 ft deep, and was filled with soil in 3 inch layers. Each
layer was compacted to give a consistent density. When the soil level reached
the bottom edge of two inspection ports, in opposite vertical sides of the
container, the tube under test was placed horizontally on the soil and in line
Lid Water pressure

n / / n
i □ □ [ £
777S7/7777W//////////////////A
□ □ □ □
Te st* s p e c im e n ' u
1 1 1 | 1 - ! —
S o li
Container, 1-5 m square in plan

Figure 4.21 Static test rig.
@seismicisolation
@seismicisolation
with the ports. The tube ran the whole width of the testing tank between the
ports and was sliced transversely into sections to eliminate overall longitu
dinal bending. The length of the tubes was thought to be sufficiently great to
eliminate any arching action in the longitudinal direction between the ends
of the tube.
Further layers of soil were then laid and compacted until the desired cover
was reached. In order to preserve accurately the shape of the tube during this
phase, it was necessary to support it internally with wood sections, and these
were removed after compaction. A pair of soft rubber diaphragms, edged by
a steel frame, was then placed on the soil surface, and above this a restraining
structure was connected via links to the side of the tank. As water was pumped
into the space between the diaphragms, a uniform pressure was applied
over the surface of the soil, measured by a gauge in the water supply system.
The m aximum working pressure was 100 psi. By limiting the diameter of
cylindrical tubes under test to about 4 inches, an order of m agnitude less than
the rig dimensions, it was assumed that the side walls of the tank had no
effect on the action of the specimens under load.
The first series of tests explored a range of cover depths, two types of tube
material (steel and brass) and two types of soil (well compacted sand and
remoulded clay), and the results are shown in Figs 4.22 to 4.25. The lower
scatterband of the results agrees well with the form of equation (4.89).
Applying equation (4.90) suggests that in the same type of soil, under
similar conditions of compaction, and for tubes having similar radii,
(4.92)
[p
which gives a ratio of 1.4 for the tubes under test. From Figs 4.22 to 4.25, the
experimental ratio is also 1.4.
Equation (4.90) also suggests that, for tubes m anufactured from similar
0 2 4 6 8 10 12
Cover depth, H (inch)
Figure 4.22 Steel cylinders in sand.
@seismicisolation
@seismicisolation
Figure 4.23 Brass cylinders in sand.

Figure 4.24 Steel cylinders in soft clay.

Figure 4.25 Brass cylinders in soft clay.
@seismicisolation
@seismicisolation
~in 2 0 r
a Scatter of test results

4) fl=2inch t - 0*0312 inch
o

Figure 4.26 Polyester resin/slate cylinders in soft clay.
material, and having similar radii,
(4.93)
which gives a ratio of 4.5 for the tubes under test. From Figs 4.22 to 4.25, the
experimental ratio is 4. The test results therefore strongly supported the
analysis leading to equation (4.89) and (4.90).
F urther experiments were conducted in clay using j ^ inch thick polyester
resin/slate tubes of 4 inch diameter, and are reported in ref. 4.35. The material
was chosen because of its relatively brittle nature, and the influence of this
lack of ductility on buckling and ultim ate strength was examined in the tests.
The results are shown in Fig. 4.26. Taking the modulus of elasticity of the
polyester resin/slate as 6 x 105 psi, and again using equation (4.90), gives
(4.94)
As Figs 4.24 and 4.26 indicate, the experimental ratio at deep covers lies in
the range 1.0 to 1.4.
Using the same test rig, experiments were carried out by Allison [4.36] to
investigate tube behaviour when discontinuities were produced in the soil by
the introduction of rigid interfaces. The results gave an experimental guide
to the minimum dimensions of test tanks to ensure that the walls have no
influence on collapse pressure. Brass tubes, 4 inch diameter, 0.0050 inch
thick, were used, and heavy, inflexible stone slabs provided the adjustable
boundaries.
In one series of tests the tubes were buried under a 9 inch cover and the rigid
slab was set horizontally in the soil below the tube. The distance of the slab
below the tube invert was varied successively from 0 to 9 inches. The results
given in Fig. 4.27 show that the proximity of the invert to the rigid base had
no measurable effect on collapse pressure. In another series the tubes were
@seismicisolation
@seismicisolation
9 <nch
4 irtch
60 7
Rigid slab
40
Scatter of test result’s-

sa 20
a
o
2 4 6 8 10
Distance of slab from in v e rf (inch )
Figure 4.27 Effect of rigid slab below the invert.
again buried 9 inches deep, and one rigid slab was set 3 inches below the
invert. Two vertical slabs were then brought progressively nearer to the
specimen from each side. The results given in Fig. 4.28 indicate that wall
proximity begins to have an effect at distances below 1.5 x diameter from the
side of the tube, and this compares favourably with the stress analysis of
holes in a semi-infinite isotropic sheet, discussed in Chapter 2. Two similar
holes placed four diameters apart, i.e. 2(D/2 + 1.5D), have no influence on
each other.
The tests so far described were concerned with the effect of a high surface
pressure acting on the surface of the soil, when this produces a much more
severe loading on the pipe than the dead weight of the backfill. However,
many practical problems concern pipes or conduits under external loading
due to backfill and fluctuations in the ground-water table only. Tests have
been conducted by English and Schofield [4.37] using a centrifuge to model
the effect of gravity stresses in the backfill. English [4.38] analysed the onset
of instability for a buried cylinder with ground water rising slowly in the
permeable sand around it, and more recently Valsangkar and Britto [4.39]
have conducted centrifuge experiments to investigate the combined action
of the self-weight of the backfill and of surface loading. The reader should
also see their earlier report [4.40].
Larger-scale tests on flexible pipes have been sponsored by the US Bureau
of Reclamation, as reported by Howard [4.41] and Pettibone and Howard
[4.42]. Three equally spaced diameters were selected, 18, 24 and 30 inches
@seismicisolation
@seismicisolation
Rigid slab-1 9 inch
Distance varies
from O ta 9 inch
3 inch
O 2C)I--------1-------1------1------ 1------ 1
° 0 2 4 6 8 10
Distance of vertical slabs
from side of cylinder (inch)
Figure 4.28 Effect of vertical rigid slabs alongside cylinder.
(46, 51 and 76 cm), and wall thickness of 7, 10 and 14 gauge were used. The
pipes were buried in a 7 ft deep container with a plan area of 6 ft x 7 ft, and
their longitudinal axes were set 4 ft below the soil surface, as described in
ref. 4.42. A large universal testing machine applied surface loads, which were
increased until either the pipe failed or a pressure of 100 psi (the design limit)
was reached.
All pipes deflected in a similar way until a 10 or 15% diameter change was
reached, but thereafter two patterns emerged, depending on pipe flexibility.
The stiffer pipes deformed elliptically and failed after the formation of plastic
hinges at the ends of the horizontal diameter. The more flexible pipes deform
ed rectangularly, and four plastic hinges formed at 60° and 70° either side of
the vertical diameter. The hinge formation was followed by inelastic buckling
of the top of the pipe.
The proximity of the container sides to the test pipes influenced the results,
but the testing team felt that the relationships established could be correlated
and refined with the results of field trials on similar specimens.
4.4 Design rules
4.4.1 US Corrugated Steel Pipe Institute guidelines

The work of Spangler, White and M eyerhof has formed the basis of design
rules for flexible soil-steel pipe structures in the USA, and much of their
@seismicisolation
@seismicisolation
work was sponsored by the Corrugated Steel Pipe Institute and by the
ARM CO International Corporation. A useful handbook, covering strength,
durability, subdrain design, earth control and installation, is published by
A RM CO [4.2]. The Corrugated Steel Pipe Institute commissioned
M cCavour [4.43] to prepare a design guide in the late 1960s, and the first
part, dealing with circular pipes under a minimum fill height of one diameter,
was issued in M ay 1966. The second part, issued in December 1968, extended
the design procedure to circular pipes under a minimum fill height of only
one-quarter of the diameter.
The C orrugated Steel Pipe Institute guide emphasized the importance of
analysing the earth fill and subgrade as well as the strength of the structure.
The following checks were proposed:
1. Classification and evaluation of soil type.
2. D eterm ination of fill soil properties (above and around the structure).
3. D eterm ination of subgrade soil properties (below the structure).
4. Structural analysis of the pipe:
(a) circumferential ring stress,
(b) elastic buckling stress,
(c) axial stress,
(d) joint strength,
(e) longitudinal ring stress,
(f) circumferential deflection.
The last check is vital, because failures of thin-walled corrugated pipes have
been due mainly to excessive circumferential ring deflection, and the cause
has usually been insufficient com paction of the backfill in critical areas. The
economic balance is between the cost of excavation and soil compaction on
the one hand, and the cost of the pipe material on the other.
The recommended procedure is to classify the soil type by using the Unified
Soil Classification System, and to conduct triaxial compression tests on
com pacted soil samples, m aintaining a steady lateral hydrostatic pressure as
the axial pressure is increased to failure. From this test the modulus of soil
deform ation Es is obtained, and the m odulus of soil reaction km is found from
equation (4.59) or, more accurately, from the equation
The relationship between km and the depth of cover is then calculated on the
lines proposed by Luscher [4.31], see equation (4.78), so that
(4.96)
@seismicisolation
@seismicisolation
This procedure was recommended by Meyerhof [4.44], based on the work of
Luscher and Hoeg [4.19]. An alternative m ethod is to use the value of k from
equations (4.18) to (4.20), when a rough estimate is required and test data are
not available; having found this value, equation (4.96) is again used to take
account of shallow cover depths.
Circumferential ring stress is next evaluated, and the work of Allgood
[4.45], Donnellan [4.46] and Dorris [4.33] is used to produce the following
formulae:
ring thrust, c = pR, (4.97)
where the earth pressure p is the sum of that created by the dead weight of the
soil and the live loads on the surface, i.e.
P = P d + Pl - (4-98)
It is further recommended that a proposal due to White [4.47] should be
used to find pD. He suggested that the pressure should be equal to the product
of soil density y and the average height of fill above the circular pipe, where
the average height is approxim ately H + 0.225R. This gives
pD = y(H + 0225R). (4.99)
To determine pL, Boussinesq’s equation is used (see Section 2.2.5), in the
well known three-dimensional form
3W z3
PL = ^ r , (4.100)
where W isthe concentrated surface load, r isthe horizontal distance from

this load to the point at which the stress is applied to the structure, and z is the
vertical distance between this point and the surface.
If A is the cross-sectional area of the corrugated steel conduit per unit
length of horizontal projection, the axial stress oc is given by
° c = pR/A. (4.101)
The value of crc to produce elastic buckling of the pipewall (crcr) is found from
equation (4.58). Thus,
R 2/ k \ 1/2
a = p — = — [ ------------------- , (4-102)
cr cr A A \ ( l - v 2) J K ’
where km is given by equation (4.96). To take account of the inelastic buckling
range, equation (4.75) is used. M cCavour proposes a safety factor of 2 on
buckling strength and 3 on the shear or bearing failure of the bolted joint.
Longitudinal ring stress ax due to longitudinal deflection of the pipe after
consolidation and com paction of the soil is predicted from the analysis of a
@seismicisolation
@seismicisolation
circular ring supporting a triangular load over the entire length of the pipe.
This is given as
W REAy
G\ = — j2 , (4.103)
where I and Ay are defined in Fig. 4.29.

Deflection formulae based on the work of Spangler and Meyerhof are
recommended. The horizontal deflection is taken from equation (4.16), and
the total vertical diametral shortening Dc from a formula due to Meyerhof
[4.44]:
0A61pR4
° c = EI + 0.061 {I - [R/(R + H)~\2)kR*' (4104)
If the vertical diam eter decreases by about 20% it is normally assumed that
the pipe is in a state of incipient failure, and that any further vertical load
will cause collapse. Measurements of deflections in practical installations
have shown that the average vertical deflection of all culverts is about 2.5%
of the diameter. M cCavour concludes by giving useful design examples and
by discussing m aterials and design factors of safety.
4.4.2 US Naval Civil Engineering Laboratory reports

An extensive program me of theoretical and experimental work, extending
over many years under the guidance of J.R. Allgood, culminated in the publi
cation of a Summary o f Soil-Structure Interaction in 1972 [4.48]. This includ
ed design recom m endations for horizontally buried cylindrical pipes, and
established boundaries to distinguish flexible, intermediate and stiff cylinders,
as follows:
flexible
@seismicisolation
@seismicisolation
interm ediate
(4.105)
stiff
where M s is the effective secant confined compression modulus at a stress

equal to the applied load.
Cylinder failure connotes: wall crushing or excessive cracking, buckling,
separation and rupture of seams, excessive deformation and deflection that
impairs the functional performance. The calculation of deflection
recommended in the report is based on the work of Hoeg [4.15], as simplified
by Allgood [4.49, 4.50], and gives the following expression for diametral
deflection at the crown :
Ay _ pv 14F + 1
(4.106)
~ D ~ E s 3 (2 F + 3 )’
where
and py is the effective free field stress at the elevation of the crown. For flexible
cylinders where F > 1, equation (4.106) reduces to
This assumes a Poisson’s ratio of an at-rest coefficient of lateral earth

pressure of y, and no interface shear (i.e. full slippage) between the cylinder
wall and the surrounding soil. In practice, full slippage is rarely attained, and
Allgood proposes the amended relationship
Since p J M s is proportional to the vertical soil strain, this equation illustrates

the obvious fact that in order to keep deflections small, the soil must be well
compacted. Allgood recommends the use of equation (4.56) for computing
the elastic buckling stress, with Chelepati’s amendment for loss of contact
during inward buckling, and Luscher’s equation (4.77) to take account of the
change in soil m odulus with depth. After some m anipulation he obtains the
@seismicisolation
@seismicisolation
equation
(4.109)
where kL is defined in equation (4.77), and
Allgood also discusses the use of backpacking to reduce the pressure

reaching the thin-walled pipe. For backpacking to be effective it must yield
to transfer a larger percentage of the applied load to the soil arch, and the
H/D ratio must be large enough to permit formation of such a soil arch.
Further, if a certain minimum stress exists under the soil arch to maintain
its integrity, the peak pressure on the structure will be the yield stress of the
backpacking, as long as the buckling resistance of the pipe is greater than
this yield stress. Allgood calculates the required thickness of the backpacking
tL from the formula
(4.110)
where H' is the distance from the top of the backpacking to the invert of the
cylinder, es is the average strain in free field over vertical diameter of the
cylinder.
ec is the average strain over vertical diam eter of cylinder,
S is the perimeter of structure, is the plan area of structure, d0 is the

depth of cover from surface to backpacking (see equation (2.8)), and ehL is
the hardening strain of the backpacking.
Allgood [4.48] gives a practical example to illustrate the design method
procedure including the effect of backpacking. The factor of safety against
wall buckling is 3.0.
4.4.3 C IR IA recommendations
The CIRIA report [4.7] contains a critical review of design procedures
and proposes a design method very similar to that in the US Corrugated
Steel Pipe Institute proposals. The pressure due to dead weight of soil is
@seismicisolation
@seismicisolation
given as
pD = y{H + 0.107D), (4.111)
which is very similar to that given by equation (4.99), and the pressure due to
live loads is calculated from the Boussinesq equation (4.100).
The maximum pressure due to ground water (pw) is found from the
equation
Pw = y J , (4.112)
where yw is the specific weight of water, and h is height of water table above
the pipe invert. An allowance can be made for a reduction of apparent
density of soil resulting from submergence, when the pipes are set in perme
able submerged soils. Tem porary lowering of the ground-water level during
installation must also be considered, and an allowance should be made
for possible ‘vacuum’ conditions in the pipe. The buckling pressure of the
pipe wall is found from equation (4.56), and the effect of plastic range buck
ling is taken into consideration by employing equation (4.75), with a safety
factor of 3.0.
Deflections are calculated according to equation (4.29), with yT = 1 and
yB = 0.083 (assuming a half-bedding angle of 90°). Wc is expressed as the
change in vertical pressure to cause a diam etral change A X /D , and is denoted
as <5p, so that equation (4.29) becomes
**= aQ83^ - (4113)

D 8 F ' + 0.061 E 9 v '
where E" = E I/D 3 and E = £ .
It is also necessary to check that the pipe is suitable for handling and
installation without undue care or bracing. The minimum pipe stiffness
required decreases as pipe diam eter increases, and the requirement is that
E"D > 4000 N/m. (4.114)
The report also discusses the use of the Southwell plot for predicting failure
from a knowledge of the initial deviations from the true shape of the pipe.
4.4.4 UK military design codes

The author has recommended design procedures after several years of
reseafch at the M ilitary Vehicles and Engineering Establishment, Christ
church, UK, carried out on behalf of the Scientific Advisory Branch of the
Home Office. The procedures are summarized in ref. 4.27. The research was
mainly concerned with underground tubular structures loaded statically by
@seismicisolation
@seismicisolation
the application of surface pressures far greater than pressures due to the
dead weight of the soil. The latter was neglected in most of the design re
commendations.
The surface pressure to cause collapse of a thin-walled buried tube of
circular cross section, when the tube is buried under a very large cover depth,
is calculated using equation (4.90). Thus
pmax = 0 A 2 E ^ E l/3(t/R), (4.115)
a formula based on the theoretical work of Bulson [4.32], Duns [4.21] and
Cheney [4.28], and on the experimental results of the first-named.
The surface pressure ps to cause collapse at a finite cover depth H is found
from
R
Ps=P r 1 (4.116)
R +H
and this formula gives a relationship between ps and pmax that not only agrees
with the author’s tests on buried tubes, but also agreed closely with the
observations of Gill [4.51], who plotted all available arching data for plates
and cylinders in granular soil.
The allowable surface pressure pa is related to ps by the equation
>/Pa = Ps> (4.117)
where rj is the safety factor. For steel and aluminium tubes under static
loading, the following values are proposed:
Large cylindrical tubes, hum an habitatio n t] = 3.0

Cylindrical tubes for pipelines or conduits r] = 2.0
F or G R P pipes it is proposed that the above values become 3.75 and 2.5
respectively.
When there is no surface pressure present, this code suggests that the soil
pressure should be calculated for an imaginary surface 3R above the crown
of the cylinder, and that this can be taken as a surface pressure on a pipe
having a burial depth of 3R. If the soil density is y,
ps = y ( H - 3 R ) . (4.118)
This value can be com pared with the value calculated from equation (4.116)
using 3R instead of H , so that
Ps = r e P (4.119)
The formulae recommended above differ from those in the design proce
dures discussed in Sections 4.4.1, 4.4.2 and 4.4.3, which were based on the
@seismicisolation
@seismicisolation
work of Luscher [4.31] and M eyerhof [4.44]. However, when the two
m ethods are applied to the same design problem, very similar solutions are
derived. This is because the procedures in Sections 4.4.1, 4.4.2 and 4.4.3 are
used with equation (4.75), whereas the design formulae from Section 4.4.4 are
used on the basis that factors a and b in equation (4.76) have high values.
This is justified because the relationship between ps and pz in equation (4.87)
is based on a very conservative assum ption about the state of elastic stresses
in the medium.
References
4.1 Spangler, M .G . (1937) Structural design o f flexible pipe culverts. 17th Annual
M eeting, Highway Research Board.
4.2 A R M C O In tern atio n al C o rp o ratio n (1958) Handbook o f Drainage and Con
struction Products, A R M C O .
4.3 M eyerhof, G .G . and Baikie, L .D . (1963) S trength o f steel culvert sheets bearing
against com pacted sand backfill. 42nd Annual M eeting, Highway Research Board.
4.4 B arnard, R .E. (1957) Design and deflection co ntrol o f buried steel pipes su p p o rt
ing earth loads and live loads. 60th Annual M eeting, A STM .
4.5 W atkins, R .K . and Spangler, M .G . (1958) Some characteristics o f the m odulus o f
passive resistance o f soil: a study o f similitude. Highw. Res. Board Proc., 37, 576.
4.6 H ow ard, A .K . (1977) M odulus o f soil reaction values for buried flexible pipeline.
J. Geotech. Eng. Div., ASCE, 103, G T1, Jan., 33.
4.7 C om pston, D .G ., et al. (1978) Design and Construction o f Buried Thin-Walled
Pipes, C IR IA R ep o rt 78, C onstruction Industry R esearch and Inform ation
A ssociation, L ondon.
4.8 Burns, J.Q . (1965) An analysis o f circular cylindrical shells em bedded in elastic
media. P hD Thesis, Univ. o f A rizona, Tucson.
4.9 Burns, J.Q . and R ichard, R .M . (1964) A tten u atio n o f stresses for buried cylinders.
4.10 K ato n a , M .G . et al. (1976) C A N D E - A M odern Approach fo r the Structural
Design and Analysis o f Buried Culverts, U S N av al Civil E ngineering Lab.
4.11 H errm an n , L .R. (1972) User's Manual fo r Plane Strain Incremental Construction
Program, Univ. o f C alifornia, Davis.
4.12 H ard in , B.O. (1973) Effects o f Strain Amplitude on the Shear Modulus o f Solids, US
A ir Force W eapons Lab., K irtland A ir Force Base, Tech. Rep. A FW L-TR -72-201.
4.13 R obinson, R .R . (1978) Theoretical Investigation o f Loads on Buried Structure, US
A ir Force W eapons L ab., K irtland A ir Force Base, Tech. Rep. A FW L-TR-78-6.
4.14 Bathe, K ., W ilson, E.L. and Peterson, F.E. (1973) Univ. o f C alifornia, Berkeley,
Rep. E E R C 73-11.
4.15 Hoeg, K. (1966) Pressure Distribution on Underground Structural Cylinders, US
A ir Force W eapons L ab., M IT, Tech. Rep. TR-65-98.
4.16 T im oshenko, S. P. and G ere, J.M . (1961) Theory o f Elastic Stability, M cG raw -H ill,
New Y ork, p. 292.
4.17 Cheney, J.A . (1963) J. Eng. Mech. Div., ASCE, 89, EM 5, Oct.
@seismicisolation
@seismicisolation
4.18 M eyerhof, G .G . and Baikie, L .D . (1963) see ref. 4.3.
4.19 Luscher, U. and H oeg, K. (1964) The Interaction Between a Structural Tube and the
Surrounding Soil. US A ir Force W eapons Lab., K irtland A ir Force Base, Rep.
R T D TDR-63-3109.
4.20 F o rrestal, M .J. and H errm ann, G. (1965) Buckling o f a long cylindrical shell
su rrounded by an elastic m edium . Int. J. Solids Struct., 1, 297.
4.21 D uns, C.S. (1966) The Elastic Critical Load o f a Cylindrical Shell Embedded in an
Elastic Medium, Univ. o f S outham pton, Rep. CE/10/66.
4.22 H abib, P. and P hong, L.M . (1966) Ann. Inst. Tech. Batim. Trav. Publics, no. 218,
Feb.
4.23 Sonntag, G . (1966) Die Stabilitat dunnwardoger Rohve in Notrasionslosen Kantinun-
on (Rockmechanics and Engineering Geology), Springer Verlag, Berlin, vol. IV/3.
4.24 C helepati, C.V. (1966) Critical Pressures fo r Radially Supported Cylinders, US
N aval Civil Engineering Lab., Tech. N ote N-773.
4.25 C om pston, D .G ., et al. (1978) see ref. 4.7.
4.26 A llgood, J.R . and C iani, J.B. (1968) The influence o f soil m odulus on the behaviour
o f cylinders in sand. Highw. Res. Rec., no. 249, p. 1.
4.27 Bulson, P.S. (1972) Thin-Walled Tubular Structures Under Soil Cover and Surface
Pressure, M ilitary Vehicles and Engineering E stablishm ent, Rep. 7515/18.
4.28 Cheney, J.A . (1976) Buckling o f the Thin-Walled Cylindrical Shell in Soil, C am
bridge U niv. Engineering D ept, Rep. C U E D /C -S oils T R 26.
4.29 W atkins, R .K . (1964) Structural design trends in buried flexible conduits. Proc.
Symp. on Soil-Structure Interaction, Univ. o f A rizona, Tucson.
4.30 W hite, H .L. and Layer, J.P. (1960) The corrugated m etal conduit as a com pression
ring. Proc. Highw. Res. Board, 39, 389.
4.31 Luscher, U. (1966) Buckling o f soil-surrounded tubes. J. Soil Mech. Found. Div.,
ASCE , 92, SM 6, N ov., 211.
4.32 Bulson, P.S. (1966) Stability o f Buried Tubes Under Static and Dynamic Over
pressure. Part One: Circular Tubes in Compacted Sand, M ilitary Engineering
E xperim ental E stablishm ent, Res. Rep. RES 47.5/7.
4.33 D orris, A .F . (1965) Response o f Horizontally Orientated Buried Cylinders to
Static and Dynamic Loading, US A rm y Engineer W aterw ays Experim ent Station,
Tech. Rep. 1-682.
4.34 Bulson, P.S. (1964) Buried tubes under surface pressure. Proc. Symp. on S o il-
Structure Interaction, Univ. o f A rizona, T ucson, p. 211.
4.35 Bulson, P.S. (1969) Static and Dynamic Overpressure on Brittle Circular Tubes in
Clay, M ilitary Engineering Experim ental E stablishm ent, Res. Rep. 47.5/15.
4.36 A llison, C.J. (1967) An Experimental Study o f the Strength o f Circular Tubes in a
Confined Medium, M ilitary Engineering E xperim ental E stablishm ent, Res. Rep.
47.5/8.
4.37 English, R .J. and Schofield, A .N . (1974) A technique o f inspection and testing o f
thin-w alled buried structures. Geotechnique, 24 (1).
4.38 English, R .J. (1973) Centrifuge m odel testing o f buried flexible structures. PhD
thesis, M anchester Univ.
4.39 V alsangkar, A .J. an d B ritto, A .M . (1979) Centrifuge Tests o f Flexible Circular
Pipes Subjected to Surface Loading, T ran sp o rt and R oad R esearch L ab., Suppl.
Rep. 530.
@seismicisolation
@seismicisolation
4.40 V alsangkar, A .J. and B ritto, A .M . (1978) The Validity o f Ring Compression
Theory in the Design o f Flexible Buried Pipes, T ran sp o rt and R oad Research Lab.,
R ep. SR 440.
4.41 H ow ard, A .K . (1968) Laboratory Load Tests on Buried Flexible Pipe, Progress
Report No. 1, US B ureau o f R eclam ation, Rep. EM-763.
4.42 P ettibone, M .C . and H ow ard, A .K . (1966) Laboratory Investigation o f Soil
Pressures on Concrete Pipe - Progress Report No. 2,US B ureau o f R eclam ation,
Rep. EM -718.
4.43 M cC avour, T.C . (1968) Composite Design fo r S o il-S teel Structures, C orrugated
Steel Pipe Institu te, Tech. Bull. no. 205.
4.44 M eyerhof, G .G . (1966) C om posite design o f shallow -buried steel structures.
Canadian Good Roads Association, Annual Convention.
4.45 A llgood, J.R . (1964) The behaviour o f shallow buried cylinders. Proc. Symp. on
S oil-Structure Interaction, Univ. o f A rizona, T ucson, p. 189.
4.46 D onnellan, B.A. (1964) The response o f buried cylinders to quasi-static over
pressures. Proc. Symp. on Soil-Structure Interaction, Univ. o f A rizona, Tucson,
p. 449.
4.47 W hite, H .L . A Rational Approach to Soil Pressures Surrounding Flexible M etal
Structures, A R M C O .
4.48 A llgood, J.R ., et al. (1972) Summary o f Soil-Structure Interaction, US N aval
Civil Engineering L ab., Tech. Rep. R771.
4.49 A llgood, J.R . (1971) Structures in soil un d er high loads. J. Soil Mech. Found. Div.,
A SCE , 97, SM 3, M ar.
4.50 A llgood, J.R . (1972) Balanced design and finite elem ent analysis o f culverts.
51st Annual Meeting, Highway Research Board.
4.51 G ill, H .L. (1967) A ctive Arching o f Sand During Dynamic Loading; Results o f an
Experimental Program and Development o f Analytical Procedure, US N aval Civil
Engineering Lab., Tech. Rep. R541.
Further reading
A bel, J.F ., N asir, G .A . and M ark, R. (1977) Stresses and Deflections in Soil-Structure
System s Formed by Long-span Flexible Pipe, P rinceton U niv., D ept o f Civil E ngineer
ing, Res. Rep. 77-SM-13.
B arker, H .A . (1977) The Design o f Flexible G RP Pipes, Inst. M ech. Eng., paper C237.
B rockenbrough, R .L. (1963) The Influence o f Wall Stiffness on the Design o f Corrugated
M etal Culverts, A pplied R esearch Lab., U nited States Steel, Tech. Rep. on Project
51. 12-400(3).
B rockenbrough, R .L. (1964) Influence o f wall stiffness on corrugated m etal culvert
design. Highw. Res. Rec., no. 56, p. 71.
B rody, O. (1979) E stim ating supporting strength o f flexible pipes. J. Transp. Eng. Div.,
A SCE, 105, TE4, July, 473.
B row n, C.B., G reen, D .R . and Pawsey, S. (1968) Flexible culverts under high fills. J.
Struct. Div., A SCE , 94, ST4, A pr., 905.
Butler, B.E. (1972) S tructural design practice o f pipe culverts. Highw. Res. Rec., no.
413.
@seismicisolation
@seismicisolation
C ates, W .H . (1964) Design o f flexible steel pipe under external loads. J. Pipeline Div.,
ASCE, 90, PL1, Jan., 21.
C helepati, C.V. and A llgood, J.R . (1972) Buckling o f cylinders in a confining medium.
Cheney, J.A . (1977) On the Buckling o f Buried Structures, Univ. o f Cam bridge, Tech.
Rep. C U E D /C -S oils, T R 41.
C onstan tin o, C .J. and Vey, E. (1969) R esponse o f buried cylinders encased in foam.
J. Soil Mech. Found. Div., ASCE, 95, SM5, Sept.
D aD ep p o , D .A . (1965) Inelastic Buckling o f an Elastically Supported Buried Cylinder,
U niv. o f D etro it, R esearch Institute o f Science and Engineering, Project 462.
D oyle, J.M . and C hu, S.L. (1968) Plastic design o f flexible conduits. J. Struct. Div.,
A SCE , 94, ST8, A ug., 1935.
G reen, W .E. (1971) D eterm ination o f stress distribution aro u n d circular tunnels using
conform al m apping technique. Highw. Res. Rec., no. 345, p. 111.
H ow ard, A .K . (1977) M odulus o f soil reaction values for buried flexible pipe. J. Geotech.
Eng. Div., ASCE, 103, G T1, Jan., 33.
M eyerhof, G .G . (1968) Some problem s in the design o f shallow -buried steel structures.
Proc. Canadian Structural Engineering C on f, Univ. o f T oronto.
M eyerhof, G .G . and Fisher, C.L. (1963) C om posite design o f underground steel
structures. Eng. J. Can., Sept.
N ielson, F .D . (1972) D esign o f circular so il-cu lv ert systems. Highw. Res. Rec., no. 413.
Schwinn, K .H . (1967) U ber den Einfluss eines diinnw andigen, im Boden verlagten
R ohres a u f das T ragverhalten des Bodens. D issertation D17, Univ. o f D arm stadt.
Sevin, E., Shenkm an, S. and W elch, R .E. (1961) Ground Shock Isolation o f Buried
Structures, US A ir Force Special W eapons C enter, Rep. A FSW C-TR-61-51.
Shrock, B.J. (1978) In stallation o f fibreglass pipe. J. Transp. Eng. Div., ASCE, 104,
TE6, N ov.
Sonntag, G . (1965) U ntersuchung elastischer R ohre in K ohasionslosen K ontinum .
Die Bautechnik, 42 (8).
Sonntag, G . (1966) Stabilitat des elastisch gebetten R ohres un ter A ussendruck. Forsch.
Ing-Wes., 32 (6).
S onntag, G. (1967) B em erkungen zur Frage der Biegebeanspruchung und des Beulens
diinnw andiger T unnelauskleidungen in nachgiebiger Bettung. Die Bautechnik, 44
(9), Sept.
Sonntag, G. (1969) Einfluss von A bweichungen von idealen L astannahm en a u f die
Stand festigkeit diinnw andiger gebetter R ohre. Die Bautechnik, 46 (8), Aug.
V alentine, H .E. (1964) S tructural perform ance and load reaction patterns o f flexible
alum inium culvert. Highw. Res. Rec., no. 56, p. 47.
V alliappan, S., M atsuzaki, K. and R ajasekar, H .L. (1977) N on-linear stress analysis
o f buried pipes. Proc. Int. Symp. on Soil-Structure Interaction, Univ. o f R oorkee,
India.
V alsangkar, A .J., B ritto, A .M . and G unn, M .J. (1981) A pplication o f the Southwell
plot m ethod to the inspection and testing o f buried flexible pipes. Proc. Inst. Civ. Eng.,
p t 2, 71, M ar.
W atkins, R .K . (1967) Pipeline econom y th rough design o f backfill. J. Pipeline Div.,
ASCE, 93, PL3, N ov., 45.
@seismicisolation
@seismicisolation
W atkins, R .K . (1975) Buried Structures, Foundation Engineering, Van N ostrand
R einhold, N ew Y ork.
W atkins, R .K ., G havam i, M. and L onghurst, G .R . (1968) M inim um cover for buried
flexible conduits. J. Pipeline Div., ASCE, 94, PL1, O ct., 155.
W atkins, R .K . and M oser, A .P. (1972) N ew design m ethod for buried corrugated
steel pipe. Civ. Eng., ASCE, June, p. 61.
@seismicisolation
@seismicisolation
CHAPTER FIVE
Non-circular pipes,
closed cylinders and shells
under static loads
5.1 Thin-walled square sections

The behaviour of square metal tubes, buried with a relatively shallow soil
cover and subjected to a Static overpressure on the surface, was examined in
a series of model tests supervised by the author during the late 1970s.
Although less structurally efficient than the circular tubes discussed in
earlier chapters, there are a num ber of advantages in the on-site construction
of tunnels m ade from an assembly of flat sheets.
In a thin-walled circular tunnel, the application of surface loads at deep
cover results in virtually no inward deflection of the curved wall before
buckling occurs. The sides of square tunnels, however, deflect inwardly from
the start of loading, causing soil arching to take place with a consequent
redistribution of interface stresses. At deep covers, deflections of the roof and
sides are similar in magnitude until plastic hinges form and immediately
following this the side walls buckle inwards and collapse, as indicated in
Fig. 5.1. It is generally true for tubes of constant wall thickness that at covers
deeper than half the side depth, buckling of the side walls takes place when
roof deflections reach a critical value.
At very shallow cover depths, the application of surface pressure produces
large downward deflections of the roof, and very small inward deflections of
the sides. After the formation of plastic hinges at the junction of the roof and
sides, collapse occurs when one side buckles inwards. Because the lateral
pressures are not active, this collapse is delayed, and takes place only after
exceedingly large roof deflections. This behaviour is shown in Fig. 5.2, which
relates surface pressure with the central roof deflection for an 8 inch square
steel tube, 0.024 inch thick, buried in well compacted sand. At covers greater
than 6 inch, collapse occurred when the roof central deflection reached
0.8 inch. At zero cover, and 3 inch cover, the central deflection reached
1.9 inch.
@seismicisolation
@seismicisolation
Non-circular pipes, closed cylinders and shells 135
Figure 5.1 Stages in collapse of thin-w alled square tubes at deep covers, (a) Original
shape; (b) inw ard deflection; (c) inw ard buckling of vertical sides.
Central roof deflection (inch)

Figure 5.2 Deflections of thin-w alled square tubes in com pacted sand.
Results in well compacted sand for 4 inch square tubes m anufactured from
mild steel and aluminium alloy are summarized in Figs 5.3 and 5.4 and
indicate a similar pattern of behaviour. It is worth noting that deflection
m easurements give a good guide to pressure distribution around the tubes at
various cover depths. If a thin-walled square tube is subjected to a pressure
p 1 on one pair of sides, and a pressure p2 on the other pair, and the deflection
@seismicisolation
@seismicisolation
Central roof deflech'on (inch)

Figure 5.3 Deflections of 4 inch square thin-w alled mild steel tubes in com pacted sand.

Figure 5.4 Deflections of 4 inch square thin-w alled alum inium tubes in com pacted
sand.
in the direction of the p2 pressure is zero, then it can be shown quite simply
that
P2/P i = f- (5.D
If the deflection in the direction of p2 is one-third of that in the direction of
p 1, then
pjPi=fi- <5-2)
@seismicisolation
@seismicisolation
♦MlPs
T
H/b
Figure 5.5 V ariation of pressure and deflection ratios with H/b.
If the deflections are equal, = p2 of course. It can be seen that the deflection
ratio is more sensitive than the pressure ratio. From the inspection of many
test results, we can draw an approxim ation in the form of Fig. 5.5 in which
<5X and S2 are the central deflections of the roof and sides respectively. In
these model tests, the dead weight of the compacted sand above the roof was
neglected, but for larger-scale work it could not be ignored, and would reduce
the surface pressure to cause collapse.
When all the experimental relationships between surface pressure to cause
collapse ps and depth of cover H were brought together, it was concluded that
ps increased linearly with H, and that the experimental relationship was
ps = p0(l+ocH/b), H/b >0.5, (5.3)
where p0 is the overpressure to cause plastic hinges at zero cover, and b is the
length of the side of the square section.
The value of p0 is governed by the tube geometry and material properties.
Since collapse takes place when three plastic hinges have formed, at the centre
and both sides of the roof, we can write
p0 = i x y ym \ (5.4)
where / is the yield or proof stress of the material.
When H/b < 0.5, collapse occurs when the side wall buckles, and since the
@seismicisolation
@seismicisolation
experiments showed that the wall is effectively fixed over a length 3b/4, we
can write
b 4n2E t3
(5.5)
Ps2 ~ 12(3fc/4)3'
Thus,
(5.6)
This expression is independent of H/b, and the graph relating ps and H/b is a
horizontal line parallel to the H/b axis. Fig. 5.6 suggests that for design
purposes a = 5 for compacted sand. Thus,
Ps = 4 x 3/y(f/i>)2(l + 5H/b\ (5.7)

or
(5.8)
whichever is the larger.
The form of this relationship is shown in Fig. 5.6.
An upper limit to ps, which will be designated pmax, is reached when the
cover depth is sufficient to produce uniform pressure on all faces of the tube.
This usually occurs when H = 2b, so from equation (5.3)
Pmax=Po<1 + 5 X 2 )= l l p 0 . (5.9)
At depths of cover greater than 2b, pmax remains constant at this level, as
shown in Fig. 5.6. N ote that pmax includes the pressure due to the dead weight
of compacted sand, where this cannot be neglected in comparison with the
surface pressure.
12 Ps ** V P0 ~ Pmox
P = P (1 + 5 H/b)
s 0
Zone of test results, b = 4 inch
2 - /
O 0-5 10 1-5 2*0 2-5

H/b
Figure 5.6 Flexible square tubes in com pacted sand under static loading.
@seismicisolation
@seismicisolation
H/ b
Figure 5.7 Flexible square tubes in soft clay under static loading.
F urther tests were carried out on similar tubes buried in a soft, remoulded
clay, and the relationship between ps and H/b is shown in Fig. 5.7. A noticea
ble difference between the behaviour in soft clay and in sand is that in the
former environm ent the formation of plastic hinges at the centre and edges
of the roof does not propagate collapse at any cover depth. This is because
lateral pressures on the sides of the tubes are never active, even at low covers,
and the soft clay adjacent to the side walls does not follow up the inward
deflections of the wall.
Thus, p0 is given by equation (5.6), and the relationship between ps and
H/b remains linear until pmax is reached when H/b = 2. The value of a,
however, is much reduced from that for compacted sand, and in tests on 4
inch square tubes, 0.01 inch thick, m anufactured from mild steel sheet, it was
0.6. Equations (5.3) and (5.4) then become
ps = j j n 2E(t/b)} (l + 0.6 H/b), (5.10)
and
Pm»* = P o ( l + 0 - 6 x 2 ) = 2.2po . (5.11)

G raphs relating ps and were approxim ately linear at lower cover depths,
but for H/b = 0.75 and above the relationship was markedly non-linear, as
shown in Fig. 5.8.
Thin-walled metallic tubes collapse after buckling with large deflections
and surface folds, but rarely does the metal rupture or tear. To examine the
influence on behaviour when the thin material forming the tubes had a
brittle nature, a series of 4 inch square thin-walled specimens were manu-
@seismicisolation
@seismicisolation

Figure 5.8 Deflection of 4 inch square thin-w alled tubes in soft clay.
factured from a mixture of polyester resin and slate powder, which produced
a wall texture similar to bakelite. The tubes were ^ inch thick, and were
tested in the same apparatus used for the experiments with steel and alumi
nium tubes. Because of the lack of ductility, the tubes failed after catastophic
cracking of the roof along the lines that the plastic hinges formed in the
metallic specimens. The results are given in Fig. 5.9 for polyester resin/slate
’v/> 6 0
a
P = 11 p^
max ,0
o
\
^ 40
o
2!
D
l/>
O) Ps = pQ (1 + 5 H / b )
<D
20
b = 4 in ch , t - 0 -0 6 2 5 inch
4 6 8 10
C o ver dephh (inch)
Figure 5.9 B rittle polyester resin/slate square tubes in com pacted sand under static
loading.
@seismicisolation
@seismicisolation
tubes in com pacted sand. The linear relationship of equation (5.3) still
applies, but over the whole range of H/b values. As before, a = 5. Thus,
ps = p0{ \ + 5 H / b \ (5.12)
and
UPo-
Tests were also made in a soft remoulded clay medium, and although the
linear relationship between ps and H/b was still evident, the value of a was
much greater than was m easured in the tests on metal tubes. The experiments
gave
ps = p0(l+ 2 .8H /b), (5.13)
and pmax was reached when H/b = 1.5, so that
P ^ = 5-2Po- (5.14)
The tests also showed that when pmax was reached there was a change in
collapse mode from roof failure to floor failure.
In an earlier series of tests on metallic tubes, the author examined experi
mentally the effect on collapse strength of progressively increasing the corner
80
4>
&
i 60
o
u
o
£
D
</> 40
U )
8
a I
8 I
o
l_ 20 It
D
I/) /
_ •_
•••
1-
0 1 2 3 4
Corner radius (inch)
Figure 5.10 Effect of corner radius on collapse pressure of a steel tube.
@seismicisolation
@seismicisolation
radius of thin-walled square tubes. Tests were m ade on square tubes of 8 inch
side, in which corner radii were increased in the following sequence: 0, 1, 2,
3, 3.5 and 4 inches. The cover depth was held constant at 3 inches, and the
wall thickness at 0.015 inch.
The results are shown in Fig. 5.10 in terms of collapse pressure and corner
radius. The square tubes buckled and collapsed at a low overpressure, and it
was not until the corner radius reached 3 inches that there was any significant
increase in collapse pressure. This coincided with a change in buckling mode
from the inward buckling of predom inantly flat sections to snap through
buckling of predom inantly curved sections. Fig. 5.11 suggests a design curve
that approxim ates to the results of the tests; it assumes no increase in collapse
strength as the corner radius changes from 0 to 36/8, but thereafter a sharp
linear increase to the strength associated with a true circular section (corner
radius = 6/2).
A problem that arises when the results of model studies are used as a basis
for judging the behaviour of full-scale structures is whether there are in
fluences other than gravity effects, such as grain size of the compacted soil.
The results of tests on two square tubes, differing in geometric scale by 30 to 8,
throw some light on this question.
A square steel tube, 8 inch side, 0.0165 inch thick, was buried to a depth
of 3 inches and loaded statically. The initial roof deflection due to the dead
weight of the compacted sand backfill was negligible. The relationship
between ps and 6 J b was recorded.
A second steel tube, 30 inch side, 0.062 inch thick, was buried to a depth of
11.25 inches and loaded statically in a large 24 ft static test rig by Bulson [5.1].
The surrounding sand was compacted in layers by a vibrating roller, and its
Figure 5.11 A pproxim ate design relationship for tubes with corner radii.
@seismicisolation
@seismicisolation
S ,/b
Figure 5.12 Tests on square tubes in com pacted sand at two scales.
average density after com paction was 104 lb /ft3, which agreed closely with
densities m easured in the small test rig. The initial pressure due to the dead
weight of the sand was 0.65 psi, which caused an initial deflection of 2.1 inch.
Fig. 5.12 shows the relationship between ps and d j b for the early stages of
loading and the close agreement suggests that in this range of tube dimen
sions the grain size of the sand is not an influential factor.
5.2 Elliptic sections

Large elliptic culverts have been used in Canada, and two of these systems
have been instrum ented to determine the distribution of soil pressure and the
nature of structural deformations. The culverts at Kettle Creek, near
St Thomas, O ntario, were 37 ft wide along their m ajor axes, and 27 ft deep
along m inor axes, and were constructed from corrugated steel sheeting.
Fig. 5.13 shows that they were covered by a granular and cohesive fill
extending to a depth of 26 ft above the crown of each culvert. The instru
m entation and the measurements obtained have been described in papers by
Selig [5.2] and by Selig et al. [5.3], and have been the subject of a report
prepared for the Federal Highway Administration, Washington, by the US
Naval Civil Engineering L aboratory at P ort Hueneme, California [5.4].
M easurem ents were also made on a 27 ft span elliptic corrugated steel
culvert at Thunder Bay, O ntario. The rise, or m inor axis length, was 16 ft,
but because the depth of cover over the crown was only 3.5 ft, the designers
specified a reinforced concrete slab to be placed on top of the soil cover to
distribute vehicle loads, as shown in Fig. 5.14. Earth pressure distribution
was m easured along the bottom of the slab, and the deflection of the slab
relative to the crown was recorded.
@seismicisolation
@seismicisolation
Stream bed
Invert’ Side view

Figure 5.13 K ettle Creek culverts (Selig).
Slab, 1 ft thick
r 3 5 ft
Excavation line
Figure 5.14 T hunder Bay culvert (Selig).
In both installations the superimposed load was due to the backfill and
the load from vehicles traversing the top of the embankment, and the total
loads were at no time sufficiently large to cause wall buckling of the elliptic
pipe. At Kettle Creek changes in strain with time were measured directly
over the crown and over the soil column to one side of the culvert, and showed
that the largest vertical compression was located over the soil column. The
smallest compression was over the crown, indicating an arching pattern, as
shown in Fig. 5.15. The pattern remained constant over a period of more than
three years, but the strains increased significantly from the original values.
An elastic analysis of the stress distribution around an elliptic pipe, using
the finite element method, was carried out by Abel et al. [5.5]. The configura
tions and loadings used are shown in Fig. 5.16. The ‘deep’ pipe is buried to a
depth equal to half the length of the m ajor axis of the ellipse, and is subjected
@seismicisolation
@seismicisolation
Horizontal distance from crown (tf)

Figure 5.15 Vertical com pression at K ettle Creek (Selig).
t NDeep buried
2a/3
a /3
i Jl
Shallow buried
Figure 5.16 Pipe configuration for elastic analysis (Abel et al.
to a uniformly distributed surface loading. The ‘shallow’ pipe is representative

of an early stage during construction, when the depth of cover is one-sixth of
the m ajor axis length, and the surface is subjected to a concentrated line load.
This line load corresponds to wheel or track loads from construction
equipment.
F our conditions were analysed: (a) an unlined elliptic tunnel; (b) a lined
tunnel; (c) a lined tunnel with thrust beams or wedges placed at the springing
@seismicisolation
@seismicisolation
Table 5.1 D im ensions of elliptic structure
W idth across m ajor axis 36 ft (11 m)

Pipe wall, 5 gauge I = 1.52 in4/ft,
A — 3.32 in 2/ft
O verpressure (deep pipe) 3600 lb/ft2, representing
30 ft (9 m) of soil
C oncentrated load (shallow pipe) 10 000 lb/ft (150 kN /m )
Steel pipe m odulus, E 210 N /m m 2
of the arched upper perimeter; and (d) a lined tunnel with compressible
underbedding. The finite element analysis assumed perfect bonding between
the elastic medium and the tunnel lining. The dimensions of the structure
were as shown in Table 5.1.
The equivalent constant-thickness, uncorrugated pipe that would interact
with the soil in a similar m anner to the corrugated pipe can be found from
the equations
EA = E*bh
(5.15)
E l = E*bh*/\2
where E* is the modulus of equivalent pipe material, b is the width of equiva
lent section in the axial direction, and h is the equivalent pipe wall thickness.
These equations give:
h = (12 I/A )112,
b = AE/hE*. (516)
If b is taken as 1 ft, the equivalent values are h — 2.4 inch, E* = 24 N /m m 2.
Photoelastic models based on the transform ed pipe section properties were
manufactured, and the finite element analysis was carried out using the
dimensions and material properties of the photoelastic models.
An earlier study by Doderer [5.6] had attem pted to review the importance
of param eters in terms of their effect on soil stresses. Using a dimensional
analysis of an elastic half-space subjected to surface forces he concluded as
follows:
D epth of cover, pipe radius and thickness highly significant

R atio of Y oung’s m oduli for pipe and soil less significant
Poisson’s ratio of pipe and soil least significant
Thus, although the elastic properties of the soil are often difficult to measure,
this is not necessarily a severe restriction on the use of the finite element
@seismicisolation
@seismicisolation
Circumferential position
Botl-om Side Top^
CD
c
o Free field
-120
Figure 5.17 T angential stress d istribution (Abel et al).
analysis. Fig. 5.17 shows the distribution of tangential (hoop) stress in the
lining of the deeply buried pipe, and indicates the arching action in the soil
and the significant reduction in hoop stress if the pipe-soil interface is
bonded. N either the thrust beams nor the soft underbedding had any
noticeable effect on the interface stresses.
To calculate the elastic buckling stress for an elliptic pipe it is sufficiently
accurate to use equation (4.55) for circular pipes, with the radius R replaced
by the radius of the arched upper portion of the pipe (R f). Thus,
(5.17)
where p'cr is the critical radial pressure over the arched upper portion of the
pipe.
This approach has been used in a report by M ott et al., who proposed that
the ‘ovalling’ of an initially circular thin-walled pipe under the action of
surface pressure would lead to a reduction in the critical radial pressure if the
radius of the upper portion of the pipe were substituted for R in equation
(4.55). This links deform ation and stability in a useful way when a pipe is
subjected to large initial ‘out-of-roundness’ due to poor bedding (see Gumbel
and Wilson in ‘F urther Reading’).
5.3 Pipe arches

The behaviour of a 35 ft (11 m) span aluminium pipe arch, constructed in
Greenbrier, West Virginia, USA, has been reported by Ducan [5.7]. During
the design it has become clear that mom ents and deflections could be reduced
by using tension members between the crown and invert during backfilling,
and measurem ents were made during construction to verify the accuracy of
the analysis. The backfill material was well graded limestone gravel, with an
@seismicisolation
@seismicisolation
Surface
Aluminium
plate
---- 3 4 -6 ft -
•
Tension
T 1*5 - x 1 0 3 ••I [.--Tests
/
bar force 1-0 •/ \.
(,b/fr) 0-5 V Theory
i
\/
\
rs_
O - 8 - 4 0 +4
Cover depth (ft)
Figure 5.18 Tension b ar forces at G reenbrier (Duncan).
average relative com paction of 95%. The com parison between measured and
calculated tension bar forces are shown in Fig. 5.18.
Duncan concluded from his finite element analysis and tests that the
dividing line between ‘shallow cover’ and ‘deep cover’ is reached for pipe
arches when the cover depth to the crown approached a quarter of the major
axis span. W hen the cover depth is less than this, moments should be evaluat
ed. Axial forces in this shallow cover condition are given b y :
P = kpl7S2 + kp2yH
p2 S + kp3L L, (5.18)
where P is the maximum axial force in the culvert wall, y is the unit weight of
backfill, kpl is the coefficient for axial force due to backfill up to the crown,
kp2 is the coefficient for axial force due to the cover over the crown, kp3 is the
coefficient for axial force due to live load (Ll ), H is the cover depth, and S is
the span.
The values of /cpl and kp2, which vary with the rise/span ratio of the arch,
and kp3, which varies with cover depth/span ratio, are shown in Fig. 5.19.
These are approxim ate relationships, suitable for design calculations. The
m aximum bending m om ent (iVfB) due to backfill loads is given by
M B = R B(kmly S > - k m2yS2H), (5.19)
where kml and km2 are m om ent coefficients that vary with flexibility. P ro
posed design relationships are given in Fig. 5.20. Flexibility is denoted by the
‘flexibility num ber’, defined by Nielson [5.8] and Selig [5.9] as N f , where
N { = EsS 3/EI. (5.20)
Here Es is the secant m odulus of the backfill soil, E is Young’s modulus for
the culvert material, and / is the mom ent of inertia of the culvert wall per unit
length. The units must be chosen to give a non-dimensional form to N f.
@seismicisolation
@seismicisolation
0-2 0-4 0-6
Rise/span ratio, R/ S
Rise/span ratio, R/ S
/fp3 = 1-25 - H / S
0 -5 1-0
Cover depth/span ratio, H/ S
Figure 5.19 Values of axial force coefficients for pipe arches (Duncan).
k
ml = 0 -0 0 4 6 -0 0 0 1 log 10 N.t
k = 0 -0 0 0 9
/
ml
Flexibility number, A/f

0-002
k = 0 -0 0 1 8 -0 -0 0 0 4 log10 A/f
rm2
0-001
k*x\2 =° '00032
/
~ 102 103 104 105

Flexibility number, Nf
Figure 5.20 M om ent coefficients for pipe arches (Duncan).
@seismicisolation
@seismicisolation
1-2 r
1-0-
fl.B
0-8 -
0-6-
0 4 ' ---------1-------- 1---------•---------1

0- 2 0-3 0-4 0-5 06
R /S
Figure 5.21 M om ent reduction factor for pipe arches (Duncan).
R b in equation (5.19) is the moment reduction factor for backfill moments,

governed by the rise/span ratio, as shown in Fig. 5.21. The determ ination of
Es for initial design purposes has been given in terms of the depth of cover at
the quarter point of the arch by Duncan, for various types of backfill, and is
shown in Fig. 5.22. These approxim ate curves have been drawn from the
results of a large num ber of finite element analyses.
The finite element technique was also used to calculate bending moments
due to live loads on the surface. Live loads over the crown cause downward
flexure of the crown, and live loads over the quarter point cause inward
Percentage of maximum
dry density^
CVJ
O 10 20 30 40
H + R I 2 (ft)
Figure 5.22 Pipe arches: values of Es used by D uncan.
@seismicisolation
@seismicisolation
H/S
Figure 5.23 R eduction factor for live-load m om ents on pipe arches (Duncan).
flexure of the quarter point. At shallow cover depths, with the live load over
the quarter point, the effects of live load and backfill combine to produce a
moment at the quarter point larger than at any other position. The addi
tional m om ent at the quarter point (AML) is given by the formula
= R hkm3S(LL), (5.21)
where R L is the reduction factor depending on N { and H /S, and km3 is the live
load m om ent coefficient, which also varies with N f .
The relationship between R L and H /S is shown in Fig. 5.23, and that
between km3 and N f in Fig. 5.24. Both these graphs are taken from Duncan’s
paper, and may be used for design purposes.
F urther inform ation on long-span underground pipe arches has been
given by W ong and Duncan [5.10] and by Duncan [5.11] - see also papers
by Allgood and Takahashi [5.12] and Sonntag [5.13].
Figure 5.24 Live-load m om ent coefficients for pipe arches (Duncan).
@seismicisolation
@seismicisolation
5.4 Arches
Arches of the type shown in Fig. 5.25 are a more complex design problem,
because in addition to the type of checks given in the previous section for
thrust and bending moment, the behaviour of the footings is important.
These can experience large downward ‘punching’ deflections with respect to
the floor slab, and if the floor slab is integral with the arch large moments can
be developed near the springings. An early program me of research on the
buckling of underground flexible arches was undertaken by Meyerhof [5.14],
who showed that at low values of flexural rigidity of the arch wall the failure
mode was by buckling at the crown. These arches had a sufficiently large
central rise to ensure that buckling did not occur by shallow arch snap-
through. At high values of flexural rigidity the characteristic failure was
yielding of the arch material near the supports.
M eyerhof found that large model corrugated steel circular arches of
constant section (radius 12 inch, span 23.4 inch, and a semi-angle a of 77°,
as shown in Fig. 5.25) failed due to sudden and catastrophic buckling of the
crown for small depths of cover, and local yielding near the springing for
depths of cover greater than one-sixth of the radius. These arches were
buried in dense sand. Smaller semicircular arches (radius 9 inch, a = 90°),
in clay, failed by crown buckling, at small cover depths ( < 2 inch), and by
local buckling of the springing for greater depths.
It can be shown that for a circular thin-walled arch with a semi-angle a
and hinged feet, the hydrostatic buckling pressure is equal to that for a
complete ring, but the buckling mode is represented by un/lv., where n > 2.
Substituting this into equation (4.57), and replacing (n + 1) by n for larger
num bers of full waves, the buckling stress ocr becomes
2
El (1 - v2)k R 4
1 + r/ o ^ (5.22)
cr (1 - v2)R2A [(nn/2a)2 - ! ]£ /
h
'Dense
;sand
U---------------- 23 4 inch---------------
Figure 5.25 M odel corrugated steel flexible arch (Meyerhof).
@seismicisolation
@seismicisolation
where A is the cross-sectional area of arch material, I is the second moment

of area of arch material,7 and k m is the coefficient of soil reaction.
If the circular arch has fixed feet, it can be shown that the buckling mode
changes to n(n + a)/2a, where n ^ 2 and a is a constant that increases from
0.857 for a < 30° to 2.0 for a = 90°.
M eyerhof shows that the buckling stress may be written in the same form
as equation (4.58), and that, in fact, the relationship holds very closely for
ratios of R / L k ^ u/ ol for arches with hinged feet, and R /L k ^ 1.57i/a for arches
with fixed feet. Lk is the relative stiffness of the arch and the soil, and is given
by
(5.23)
Equation (4.58) can be recast as
(5.24)
with the upper limit of oci as oy, the yield or proof stress of the material. Since
this equation is independent of radius and feet fixity, it can be used for arches
of other shapes, such as the three-centred arch, horseshoe arches or pipe
arches, at higher modes of buckling.
The buckling stress for an arch under shallow cover can be approximately
found by multiplying the expression in equation (5.24) by a reduction factor
of {h/2R)l/1, where h is the average depth of cover over the central 90° crown
segment (0.1 ^ h/2R ^ 1). The minimum reduction factor is 0.12 for zero
cover over the crown.
The ultim ate strength oc can be found by equation (4.76) when allowance
is made for initial eccentricities and imperfections.
Road surface
44 ft Structural zone
sand and gravel
/ 1
Thrust beam
I L — 50 ft
j-— 51 f r — J
Figure 5.26 Arch culvert on the Vieux Comptoir river (Lefebvre et al.).
@seismicisolation
@seismicisolation
M eyerhof showed that the above analysis gave good agreement with a
series of tests on model circular and three-centred arches in clay and dense
sand.
M easurements of displacements, earth pressures and wall stresses were
carried out by Lefebvre et a l [5.15] on a full-scale arch culvert built over the
Vieux C om ptoir river in Canada. The culvert span was 51 ft (15.5 m) and it
was covered by a 44 ft (13.4 m) high embankment. The rise of the arch was
26 ft (7.9 m) over its foundations, as shown in Fig. 5.26. The corrugated steel
wall was 6.86 mm thick, with 5.1 cm x 15.2 cm corrugations.
During emplacement and com paction of the backfill, the lateral curved
walls were pushed about 5 cm inwards, and the crown was raised about 10 cm.
When the fill was deep enough to cover the structure the trend of deformation
was reversed, until the final position of the crown was 5 cm upwards.
The earth pressure immediately above the crown was only 25% of that due
to the overburden, showing considerable arching. At elevations of 7 ft (2.1 m)
and 30 ft (9.0 m) above the crown, the pressure was 70% of that due to the
overburden, as indicated in Fig. 5.27. Note that at elevations about 25 ft
(7.6 m) the full overburden pressure was acting on each section, and that this
depth is approxim ately equal to the rise of the arch.
Static loading tests were carried out on model semicircular arches of 15
inches radius in the early 1960s by Gill and Allgood [5.16]. The arches were
30 inches long, 0.0478 inch thick, and the cover depth was 6 inches. They
were buried in the soil pit of the US Naval Civil Engineering Laboratory
blast simulator, and surface pressures up to 25 psi were applied statically by
means of a pressurized pneumatic bag made of neoprene. Two widths of
footing were chosen, 1.20 inch and 2.40 inch, and this variation was found
Figure 5.27 C om parison of m easured vertical pressure and overburden pressure

(Lefebvre et al).
@seismicisolation
@seismicisolation
to have a large effect on the test results. During backfilling the largest
measured m om ent of the arch with narrow footings was at the crown. For
the wide footing arch it occurred at a point 15° from the vertical. The narrow
footings deflected downwards about 50% more than the wide footings, and
the soil arching was consequently greater (60% for narrow footings, 40%
for wide footings, remaining virtually constant during the increase in over
pressure from zero to 25 psi).
The thrusts and moments due to backfilling were high, up to 40% of their
values under the 25 psi surface load. Typical deflection patterns are shown
in Fig. 5.28. The arches did not fail due to instability, and at 25 psi surface
load the stresses were small com pared to the yield stress of the material.
Buckling of the arch crown was, however, investigated experimentally in
1970 by M unn et a l [5.17], who performed tests on 16 circular arches with
18 inch outside diameters. Wall thicknesses, depths of cover and footing
areas were varied. Two thicknesses were chosen, 0.036 inch for the thin-
walled arch tests and 0.119 inch for the thick-walled specimens. Pressure was
applied to the soil surface by air pressure through a thin rubber bag, and the
depths of burial were one-quarter, one-half and three-quarters of the arch
radius with well compacted sand.
Three modes of failure were observed: buckling of the crown; punching
of the footings into the supporting soil; and buckling of the arch haunches.
Thrusts were uniform throughout the structural shell. For failure to occur,
the thrust m ust be great enough to cause the footings to punch into the soil,
or plastic hinges must form at the three points of maximum moment. Buckl
ing of a 120° arch under uniform lateral pressure q is given by
qcr = 6 J 5 (E I/R 3)p, (5.25)
where ft is a factor to account for soil-structure interaction.
Static overpressure
m t H K I I H l H
6 inch
Figure 5.28 D eflection p attern for arch with wide footings (Gill and Allgood).
@seismicisolation
@seismicisolation
Figure 5.29 Idealized failure m ode (M unn et al).
The critical pressure for bearing failure of the footings is

q„ = (auA /R )P , (5.26)
where A is the footing area, <ru is the ultimate soil bearing strength deter
mined from static plate bearing tests, and p f is a factor accounting for depth of
burial of the footing.
If the two modes of failure are to occur simultaneously,
P 6.15EI
(5.27)
The dimensionless num ber p / p is an index indicating the type of failure, and
the test results suggest that when p / p > 0.175 the failure mode is by punching
of the footings. W hen p / p < 0.175, the arch fails by crown buckling. Fig. 5.29,
taken from the report, shows the combined structural response and the
idealized failure mode of the crown.
5.5 Thick-walled arches

Unreinforced thick-walled concrete arches were tested statically in dry sand
by Meyer and Flathau [5.18], as part of a program me of static and dynamic
laboratory tests. The objectives of the study were to investigate (a) distribu
tion of strains, thrusts and moments, (b) arching actions of the soil, (c)
deflections at the crown, and (d) the effect of cover depth.
@seismicisolation
@seismicisolation
All the arches were 12 inch outside diameter, and varied in thickness with
dimensions of j , 1 and 2 inch. The depths of burial ranged from zero to 12
inches above the crown, and a burial depth of 2 inches was chosen for the
final maximum overpressure loading for each structure. The arches were 24
inches long, to minimize end-wall effects on response at the centre section.
The metal end walls were rigidly attached to the base slab but not to the
arch ring. The arches were buried in a 4 ft diam eter test chamber, and the soil
surface was loaded by gas or water under pressure, through neoprene
diaphragm.
The sequence of structural damage as the surface load increased was:
1. Cracking of base slab reducing degree of fixity of the ends of the arch.
2. Cracking at the intrados due to tensile bending strains at the crown.
3. Spalling due to compressive strains at the intrados near the springlines.
4. Collapse by flexural instability at the crown after the formation of three
hinges.
M easured crown deflections for a burial depth of 2 inches are shown in
Fig. 5.30, which also shows the form of the arches. Only the \ inch thick
arch collapsed, at an overpressure of 150 psi, after passing through the
stages listed above. The 1 inch thick arch reached the spalling stage, and
was clearly very close to collapse at an overpressure of 379 psi.
It was concluded that pmax, the static overpressure to produce failure
with 2 inches of cover, is at least 50% greater than the pressure to
cause collapse under hydrostatic loading directly onto the outside of the
O 004 0 0 8 0-12 0-16

D e fle c tio n (in c h )
Figure 5.30 E xperim ental crow n deflections of thick-w alled concrete arches (M eyer
and Flathau).
@seismicisolation
@seismicisolation
158 B u ried Structures
arch. Thus
(5.28)
where / c is the ultim ate compressive strength of the concrete (psi).
It was recommended that this approxim ate scale could be applied to
concrete arches having a range of stiffness, E I / R 3, from 80 to 8000 psi.
5.6 Thick-walled closed cylinders

The previous sections have been concerned with the behaviour of tubes
sufficiently long to ensure that the end conditions of support or restraint do
not influence the behaviour of the central areas which can then be treated as
two-dimensional stress problems. However, there are many types of practical
underground structure in which the length of the tubular section is only one
or two times the diameter, and the ends of the tube are closed by flat or domed
sheets. These include horizontally or vertically set capsules, suitable for
protective shelters against nuclear blast or conventional explosive weapons,
and hardened vertical silos used to protect long-range rockets. Other struc
tures falling into this category are domes and spherical shells.
A num ber of experimental studies were carried out in the USA in the early
1960s, and of these the most notable were due to M arino and Riley [5.19].
They applied static pressures to the surface of a bin of O ttaw a Sand, in which
horizontally set thick-walled steel tubes were buried. The tubes were all
5 inch diameter, with an overall length of 15 inches, including domed
hemispherical ends of 2.5 inch radius. This m eant that the uniform central
section, circular in form, was 10 inches long, as shown in Fig. 5.31. The
cylinders were buried to a depth (surface to crown) of 10 inches, and three
diam eter/thickness ratios were employed: 40, 80 and 160. They were m anu
factured from cold drawn seamless tubes.
M arino and Riley measured diam eter changes at the central transverse
plane of each shell as the static soil overpressure was increased. Longitudinal
and hoop m em brane forces and bending moments were computed. It was
found that in all the static experiments the load transm itted to the cylinder
Cylinder
L
Hemispherical
nemi
ends
2-5 inch
10 inch
Figure 5.31 Thick-w alled closed steel cylinders tested by M arino and Riley.
@seismicisolation
@seismicisolation
Surface pressure (p s i)
Figure 5.32 R elationship betw een surface pressure and force at springing (M arino
an d Riley).
in term s of the hoop mem brane force at the springing of the central transverse
cross section was about 90% of the force calculated by multiplying the applied
overpressure by the projected area of one-half the model. The am ount of
‘arching’ was relatively small, as indicated in Fig. 5.32. However, when the
specimen having D/t = 40 was subsequently loaded dynamically (see ref. 5.19),
the load on the cylinder was greater than the value of the overpressure
multiplied by the projected area, almost double in fact. This suggests that the
rapid (7 ms) application of surface pressure does not allow sufficient time for
the developing of arching action. In fact, the soil-structure system behaves as
an elastic medium with an inclusion of high rigidity, and, using equation
(4.87), one might expect this result.
A second study was reported some years later by Gates and Takahashi
[5.20], who carried out tests on model steel tanks fabricated from stainless-
steel sheets. The tests, reported in 1971, were made in the small blast-load
generator at the US Waterways Experiment Station (described in Section 6.2).
Each model consisted of a 12 inch long cylindrical section, with 12 inch
diam eter hemispherical end caps, buried 12 inches below the surface in
Cook’s Bayou Sand. The tank wall thickness was 0.024 inch, giving a D/t
ratio of 500. The models that were set horizontally were first half-filled with
liquid silicone, to represent a fuel storage tank, and then internally pressurized
to 100 psi (see Fig. 5.33). The external pressure on the soil surface was then
increased to 150 psi, and at this stage the internal pressure was slowly
@seismicisolation
@seismicisolation
160 B uried Structures
Static surface pressure
t t m u m
Compacted sand
12 inch
Figure 5.33 Static tests on m odel fuel storage tanks (G ates and Takahashi).
released. The surface pressure was then increased to a higher value and the
cycle repeated until the model tank collapsed. The purpose of this sequence
was to provide all identifiable com binations of internal and external pressure
in as few cycles as possible. The tank buckled at 498 psi surface pressure and
13 psi internal pressure, due to the stresses in the wall reaching yield. This
was well above the value predicted for unpressurized circular pipe sections
in soil.
5.7 Thin-walled closed cylinders

The author investigated the influence of cover depth and soil properties on
the static collapse pressures of horizontally set, short, closed cylinders, and
reported his results in 1971 (Bulson [5.21]). The cylinders were 4 inch
diameter, made from very thin brass and steel shim, and the soils were
com pacted sand, soft clay and a loam. The upper and lower scatter bound
aries of all test results indicated an approxim ately linear relationship between
ps/pa and H/D, where p&is surface pressure to cause collapse, pa is pressure to
cause collapse of the cylinder in air, H is the cover depth to the crown of the
specimen, and D is the diameter of the ‘s pecimen. The test rig has been
described in C hapter 4 (Section 4.3.4).
The specimens were 5.75 inches long, closed with a flat end disc of the
same thickness as the curved wall. In order to measure the collapse pressure,
each specimen was fitted with a short length of copper tube, soldered to one
of the end discs, and a long length of fine flexible nylon tube led from this,
though the soil, to a water manometer. The sudden buckling and collapse of
the cylinder wall produced an instantaneous reduction in volume, and the
@seismicisolation
@seismicisolation
H/D
Figure 5.34 Tests on thin-w alled closed cylinders (Bulson).
resulting increase in internal pressure gave rise to a sudden jum p in the

water level.
The relationship between p j p a and H/D is plotted for sand and clay in
Fig. 5.34. The results lie within a small scatterband for both types of soil
medium, suggesting that p j p a is not greatly influenced by the large difference
in elastic m odulus between steel and brass. A considerable increase in
collapse pressure is attainable by burial in compacted sand, ranging from
pjp* * 10 when H/D = 0.5, to ps/pa ~ 35 when H/D = 3.0. In clay, however,
the range is much lower, the lower boundary of the scatter increasing from
p j p a ~ 3 when H/D = 0.5 to p j p a « 8 when H/D = 4. Sand gives a greater
protection than soft, remoulded clay, the strength ratio being about 4:1.
Tests in soft clay are time-dependent.
Using the formulae for buckling for thin-walled cylinders given in Chapter
4, e.g. equation (4.90), we infer that the ratio of collapse pressures at deep
covers is
H r n - f F\TE,<
Pm«<clay> Tclay>/
T )” «52’»
and taking £ s(sand) = 10 000 psi and £ s(clay) =1120 psi, this gives a theoreti
cal ratio of 4.5. This compares reasonably with the experimental value of 4.
If we assume that £ s(loam) is 2800 psi, then equation (5.29) yields the relation
ship
Prnax<l0am) = , ^
Pmax(day)
If this ratio is applied to the scatter boundaries for clay from Fig. 5.34, we'
@seismicisolation
@seismicisolation
30r
1-85 x clay scatterband
O 1 2 3
H/D
Figure 5.35 Tests on thin-w alled closed cylinders in loam y soil (Bulson).
arrive at fig. 5.35. It can be seen that this ratio satisfactorily predicts the
range of results for p j p a in loam.
The test results also indicated that the design procedure recommended
for long thin-walled tubes in soil can be satisfactorily applied to closed
tubes having a length/diam eter ratio of 1.5, providing the closing discs
have the same thickness as the cylinder wall.
5.8 Vertical capsules

There have been a num ber of useful experimental studies on the behaviour
of closed cylinders set vertically under soil cover. Gates and Takahashi [5.20]
tested their model steel fuel storage containers in this mode, with 12 inch
Small blast load

generator, 48 Inch dia.
Static surface pressure
Compacted sand
Figure 5.36 Static tests on vertical capsules holding fuel (G ates and Takahashi).
@seismicisolation
@seismicisolation
cover of compacted sand above the centre of the upper hemispherical end,
as shown in Fig. 5.36. The vertical orientation was chosen because on certain
sites, with low water tables, the full-size structures (10-12 ft in diameter)
can be buried in vertically drilled holes more economically than if they are
set horizontally.
The loading cycles took the same form as those described in Section 5.6,
and it was found that the tank buckled under a surface pressure of 316 psi,
while internally pressurized to 100 psi. It was found that, in contrast to the
horizontally buried tank, the vertically orientated specimens behaved as
stiff inclusions, producing ‘negative’ arching. Taking the arching factor A
as (1 —p j p v\ where pe is the vertical soil pressure at the interface of the
tank crown and pv is the uniform vertical pressure in the free field at the
crown elevation, the arching factor was negative for vertical capsules and
positive for horizontal capsules, under static loads.
Vertically orientated thick-walled cylinders with both thin-walled and
rigid end closing discs were tested in the early 1960s by Constantino and
Longinow [5.22]. The experimental rig is shown in Fig. 5.37, and consisted
of a 3 ft diam eter and 3 ft deep pressure vessel, filled with well compacted
dry O ttaw a Sand. The dimensions of the structural model were 6 inch
diameter, 6 inches deep, and the basic configuration was a thick outer
cylinder and base plate supporting a flexible roof disc. This disc could
be changed so that experiments could be made on a range of flexibilities,
defined by roof thicknesses of and \ inch. The last of these was
considered to represent a ‘rigid’ roof. The models were buried at known
depths below the surface, and the central vertical displacement of the discs
were m easured as the surface pressure was increased.
A typical result for central deflection in terms of the depth/diam eter
ratio for a given flexibility is shown in Fig. 5.38. Note that for depths of
burial greater than one diam eter there is little change in the relationship
Pressure inlet
Roof disc
Structural
model
Compacted Pressure
dry sand V vessel
1
i i
3 ft dia. *|
Figure 5.37 The test rig of C onstantino and Longinow.
@seismicisolation
@seismicisolation
Cenlral deflection (inch)

Figure 5.38 Typical test results of C onstantino and Longinow.
between the surface pressure to produce a central deflection of 0.03 inch

and the depth/diam eter ratio for the disc having a flexibility (El) of 1054 lb.
inch. N ote the similarity between this figure and the relationship between
collapse pressure and cover depth for thin-walled square tubes in Section 5.1.
The results suggest that there is nothing to be gained in terms of strength by
burying vertical capsules of this type more than 1.5 diameters deep. The
ratio of the surface pressure to cause a given deflection at 1.5 diameters
with that to cause the same deflection at zero cover was about 9:1. The load
transm itted to the vertical walls of the model at deeper cover was about
half that due to the surface overpressure acting over the plan area of the
model, and the pressure distribution across the roof plate became less and
less uniform with plate deflection and with burial depth.
The load concentration on vertically orientated cylindrical capsules was
examined by Abbott [5.23]. He concluded that load concentration is a
function of the length/diam eter ratio of the cylinder, and that it is possible to
experience loads on the upper end more than 25 times the load at the same
level in the free field. Allgood [5.24] proposes that this undesirable condition
can be overcome by backpacking, discussed in a later chapter.
5.9 Thin-walled domes

An extensive series of model tests were carried out on statically loaded
buried domes by W hitm an et al. [5.25] in the early 1960s. The loading
@seismicisolation
@seismicisolation
60 inch dia.
A ir bag
Steel rings,
10 inch deep
Figure 5.39 The test rig of W hitm an et al.
system is shown in Fig. 5.39, and consisted of a 5 ft diameter soil container

having a m aximum depth of 3 ft 6 inch and walls made from 10 inch deep
steel rings. A uniform static pressure was applied to the sand surface from a
self-contained loading frame through a rubber bag. The bag was fabricated
from black neoprene by vulcanizing together two discs and a long strip; it
was 4 inches thick, and fitted neatly within the upper cover plate. Pressure
was applied to the bag from nitrogen bottles, to a maximum working pressure
of 300 psi. The bases of all domes tested were 12 inch diameter, and it was
considered that this was small enough in com parison with the bin diameter
to minimize boundary effects on the behaviour of the domes. The sand
backfill was placed by a ‘showering’ system that produced an ‘in-place’
density of about 102 lb/ft3. A num ber of plate bearing tests were carried out
to check the uniformity of the dense sand mass, and the standard deviation
was found to be less than 4% from the mean. The density of a loose sand
mass produced by a simple ‘pouring’ system was 90 lb/ft3, so the increase
in density by using a flexible hose, a mesh wire screen and the ‘showering’
m ethod was about 13%.
The domes used in the tests are shown in Fig. 5.40. They were m anu
factured by spinning flat sheets of aluminium 6061/0 alloy to produce the
segment of a sphere, radius 9 inches, central angle 85°, thickness 0.024 inch.
The buckling pressure, found by subjecting one dome to direct water pressure,
was about 25 psi, a figure associated with the snap-buckling of domes,
which agreed satisfactorily with results obtained by Kloppel and Jungbluth
[5.26]. N ote that this is a much lower pressure than would be given by the
classical linear buckling theory for spherical shells. The ‘classical’ buckling
pressure is 97 psi.
@seismicisolation
@seismicisolation
-12 inch did.
0*024 inch
85°-
Figure 5.40 D om e dim ensions for tests by W hitm an et al.
During the tests, the vertical movements of the crown of the dome were
recorded, as was the total vertical force reaching the dome. The applied
surface pressure was increased in increments of 10 psi until failure was
approached. This developed along the springline when the average vertical
stress against the dome reached about 70 psi. The surface pressure required
to cause this stress varied with sand density and the flexibility of the dome
support. After this yielding occurred, it was possible to increase the surface
load without causing complete collapse.
The strains within the sand increased as the square root of the applied
pressure, which confirmed the conclusions of Chaplin [5.27]. The sand
masses were essentially in a state of one-dimensional strain, where the stress/
strain ratio represents a constrained modulus. The overpressure on the sand
surface to cause failure was similar for burial depths of 0.4 and 1.5 times the
dome span, and the relationships between this pressure, the vertical pressure
on the dome and the density of the sand are given in Table 5.2. (The ‘rigid’
support was provided by setting the load cells directly onto a concrete floor;
the ‘flexible’ support was provided by introducing a layer of cork between
the cell and the floor.)
Even in cases where it is not possible to provide a flexible foundation,
the pressures to cause failure were almost three times as great as the theoreti
cal elastic snap-buckling pressure. Further, collapse was by bending failure
Table 5.2 R elationships for a dom e
Vertical pressure D ensity F o undation

applied surface pressure (lb/ft3) condition
1 90
1 104 rigid
0.3 104 flexible
@seismicisolation
@seismicisolation
near the support, and was not the result of elastic buckling. The difficulty
of analysing the collapse behaviour of shallow spherical caps is well known,
and has been summarized by Allen and Bulson [5.28]. F or design purposes,
the strength of the domes used in the above experiments would be taken as
97/4 « 24 psi under hydrostatic pressure, so that the experimental collapse
pressure of 70 psi shows that the presence of well compacted sand had a
significant effect on the mode of failure. In assessing the strength of domes in
a practical installation, it would probably be conservative to assume (a)
a rigid foundation, (b) densely compacted soil, and (c) that the static surface
pressure to cause buckling would be half the hydrostatic pressure to cause
buckling of a deep spherical shell according to the classical theory. To
adopt this approxim ation, the depth of cover over the crown should be at
least 0.4 times the dom e span. Collapse pressures in soils other than well-
com pacted sand could be evaluated by using the relationship given in
equation (5.9).
An experimental study of buried domes was reported by Getzler and
Lupu [5.29] in 1969. The domes were m anufactured from duralum in and
copper, and two spherical radii were chosen, 40.40 and 24.55 cm. The base
radius was constant at 30 cm, and the thicknesses of the domes after spinning
were 0.27,0.37 and 0.44 mm. The loading device, shown in Fig. 5.41, consisted
of a vertical duralum in tube, 36 cm diameter, 48 cm long, closed at both
ends with tie-bolted rigid lids. Three types of loading were used: hydrostatic;
combined hydrostatic and pneum atic loading, by means of an inflatable
rubber cushion overlying the surface of compacted sand; and pneumatic
^Com pressed air inlet

Y////77777/777777. Tzzzzzzzzzzzzm
Rubber cushion Tie rod

^Tubular container'.-y
’Compacted sandy.*v
; 'ry Dome - •
L°acl cell
&
36 cm dia
Figure 5.41 The test rig of G etzler and Lupu.
@seismicisolation
@seismicisolation
loading on the sand surface, with the thickness of the sand layer over the
crown of the dome varying from 3 to 38 cm. Because the base diameter of
each dom e was only slightly smaller than the diameter of the container,
the results are only applicable to the study of the restraining effect of the soil,
and cannot be used to draw conclusions about dome behaviour in an in
finitely large soil field. However, the results with duralum in domes showed
that, at depths of soil cover about equal to the radius of each dome base,
the surface pressure to cause buckling or yielding collapse was about double
the experimental buckling pressure of uncovered domes under hydrostatic
pressure.
5.10 Thin-walled spherical shells

Tests were carried out by Bulson [5.30], in 1970, on thin-walled spherical
shells m anufactured from a plastic (polybutylchloride) sprayed internally
with alum inium powder. They were pressed in the form of two hemispheres,
as shown in Fig. 5.42, and joined along the external circumferential lip by
high-frequency welding. There was a slight variation in shim thickness from
specimen to specimen, and about a 10% variation in E. No attem pt was
m ade to check for true sphericality. Five nominal diameters were used:
60, 80, 100, 125 and 200 mm; the corresponding average shim thickness was
in the range 0.0060 to 0.0090 inch for specimens between 60 and 125 mm.
diameter, and in the range 0.0104 to 0.0126 inch for the 200 mm diameter
specimens. The mean value of E was 4.03 x 105 psi. The test rig has been
described in Section 4.3.4.
In order to measure the surface pressure to cause buckling, each specimen
was fitted with a fine flexible nylon tube, 0.125 inch outside diameter, which
was inserted in a small hole in the shell wall. This tube was taken through
the com pacted soil to the outside of the rig, and connected to a water m ano
meter. The sudden snap-buckling of the shell wall produced an instant
Figure 5.42 Thin-w alled spherical shells (Bulson).
@seismicisolation
@seismicisolation
Table 5.3 Average test soil properties
Sand density 104 lb/ft3

m oisture content 1-3.5%
Clay m ean wet density 120 lb/ft3

m oisture content 30.4%
m ean dry density 92 lb/ft3
reduction in volume, and the resulting surge in pressure gave rise to a sudden
jum p in the water level.
The spheres were bedded, and the soil compacted in a similar m anner
to the tests described in Section 4.3.4. The average soil properties were as
shown in Table 5.3.
The tests have been reported in full in ref. 5.30, and are shown in terms
of the param eters ps/pa and H/D in Fig. 5.43, where ps is the surface pressure
to cause collapse, pa is the external pressure to cause collapse in air, H is the
depth of cover, and D is the diam eter of the sphere. In sand the lower scatter-
band rises from ps/pa = 1 .5 at H/D = 0.5 to ps/pa = 4 at H/D = 3. In clay
there was virtually no increase in collapse strength over the range of covers
tested, and this suggests that the soft remoulded clay used in the tests trans
mits a true hydrostatic pressure, with no arching, at all depths.
H/ D
Figure 5.43 Tests on thin-w alled spherical shells (Bulson).
@seismicisolation
@seismicisolation
In the absence of a theory for the buckling of spherical shells in an elastic
medium, it is only possible to draw broad conclusions about the effect of
soil m odulus on collapse pressure. In broad terms, ps (sand)/ps (clay) = 4,
which is similar to the result for closed cylinders. It is therefore reasonable
to assume that this ratio is governed by a similar relationship to that found
for cylinders, i.e. (E J E c)2/3.
The collapse pressures in air were between one-quarter and three-quarters
of the classical elastic buckling pressure, given by
\2E(tlRf,
so a conservative design formula for well compacted sand would b e :
ps = 0.3 E{t/R)2 (I + H/D). (5.30)
References
5.1 Bulson, P.S. (1964) Buried tubes under surface pressure. Proc. Symp. on S o il-
Structure Interaction, U niv. o f A rizona, T ucson, p .2 1 1.
5.2 Selig, E.T. (1975) Instrumentation o f Large Buried Culverts; Performance M onitor
ing fo r Geotechnical Construction, A STM STP 514, A m erican Society for Testing
an d M aterials.
5.3 Selig, E .T., et al. (1978) Long-span buried structure design and construction.
J. Geotech. Eng. Div., ASCE, 104, G T7, July.
5.4 Selig, E .T., et al. (1977) Review o f the Design and Construction o f Long Span
Corrugated M etal, Buried Culverts, US N aval Civil Engineering L ab., Rep.
F H W A -R D -7 7 -131.
5.5 A bel, J.F ., M ark, R. and R ichards, R. (1973) Stresses aro u n d flexible elliptic
pipes. J. Soil Mech. Found. Div., ASCE, 99, SM 7, July.
5.6 D oderer, E.S. (1970) Elastic interaction fields for buried pipe. P hD Thesis,
Princeton Univ.
5.7 D uncan, J.M . (1979) B ehaviour and design o f long-span m etal culverts. J. Geotech.
Eng. Div., ASCE, 105, G T3, M ar.
5.8 N ielson, F .D . (1972) Experim ental studies in so il-stru c tu re interaction. Highw.
Res. Rec., no. 413.
5.9 Selig, E.T. (1972) Subsurface so il-stru ctu re interaction: a synopsis. Highw. Res.
Rec., no. 413.
5.10 W ong, K.S. and D uncan, J.M . (1976) Summary o f Field Instrumentation Study
on a 35 f t Span Aluminum Culvert in Greenbrier County, West Virginia, R eport to
H ighw ay Products D ivision, K aiser A lum inum and Chem ical Sales Inc., O akland,
CA.
5.11 D uncan, J.M . (1975) Design Studies fo r a 35 f t Span Aluminum Culvert fo r Green
brier County, West Virginia, R ep o rt to H ighw ay P roducts D ivision, K aiser
A lum inum and C hem ical Sales Inc., O akland, CA.
5.12 A llgood, J.R . and T akahashi, S.K. (1972) Balanced design and finite elem ent
analysis o f culverts. Highw. Res. Rec., no. 413.
@seismicisolation
@seismicisolation
5.13 S onntag, G . (1968) M odel studies o f corrugated tubes in an elastic medium.
Der Bauingenieur, 43 (7).
5.14 M eyerhof, G .G . (1970) Some research on u n d erground flexible arches. Proc.
6th World Highway Conf., M ontreal.
5.15 Lefebvre, G ., et al. (1976) M easurem ents o f soil arching above a large diam eter
flexible culvert. Can. Geotech. J., 13 (1), Feb.
5.16 G ill, H .L . and A llgood, J.R . (1964) Static Loading o f Small Buried Arches, US
N aval Civil E ngineering L ab., Rep. R278.
5.17 M un n , J.F ., C arre, G .L . and K ennedy, T.E. (1970) Failure o f Footing-Supported
Buried Steel Arches Loaded Statically, US A rm y W aterw ays E xperim ent Station,
Misc. P ap er N-70-2.
5.18 M eyer, G .D . a n d F la th a u , W .J. (1967) S tatic and Dynamic Laboratory Tests o f
Unreinforced Concrete Fixed-End Arches Buried in D ry Sand, US W aterways
E xperim ent Station, Tech. Rep. 1-758.
5.19 M arin o , R .L. an d Riley, W .F . (1964) R esponse o f buried structural m odels to
static an d dynam ic overpressures. Proc. Symp. on Soil-Structure Interaction,
U niv. o f A rizona, T ucson, p.464.
5.20 G ates, W .E. and T akahashi, S.K. (1971) Static and Dynamic Tests o f M odel
Pressurised Underground Fuel Storage Containers, U S N aval Civil Engineering
L ab., Tech. Rep. R728.
5.21 Bulson, P.S. (1971) C ollapse pressures o f thin-w alled cylinders in soil. Int. J.
Mech. Sci., 13, 431.
5.22 C o n stan tin o , C. J. and Longinow , A. (1964) E xperim ents on circular cylinders w ith
flexible ro o f plates buried in sand. Proc. Symp. on Soil-Structure Interaction,
U niv. o f A rizona, T ucson, p. 423.
5.23 A b b o tt, P.A . (1966) Nonlinear Static Arching fo r Vertically Buried Prismatic
Structures, U S A ir Force W eapons Lab., Tech. Rep. A FW L-TR -160.
5.24 O delio, R .J. and A llgood, J.R . (1970) Response o f Buried Capsules in the High
Overpressure Region, US N aval Civil Engineering Lab., Tech. Rep. R668.
5.25 W hitm an, R.V., G etzler, Z. and H oeg, K. (1962) Static Tests Upon Thin Domes
Buried in Sand, M IT , D ept o f Civil Engineering, Rep. R 6 2 -4 1 .
5.26 K loppel, K. and Jungbluth, O. (1953) Beitrag zum D urchschlagproblem dunn-
w andiger K ugelsohalen. Der Stahlban, 22.
5.27 C haplin, T .K . (1961) Some m echanical properties o f gran u lar m aterials at low
strains. Proc. M idland Soil Mech. Found. Eng. Soc., 4.
5.28 Allen, H .G . and Bulson, P.S. (1980) Background to Buckling, M cG raw -H ill, New
Y ork.
5.29 G etzler, Z. an d L upu, L. (1969) E xperim ental study o f buckling o f buried domes.
J. Soil Mech. Found. Div., ASCE, 95, SM 2, M ar.
5.30 Bulson, P.S. (1970) Static Overpressures on Spherical and Cylindrical Shells
Buried in Sand Clay, M ilitary Vehicles and Engineering Establishm ent, Rep.
RES 47.5/17.
Further reading
B akht, B. (1981) S o il-steel structure response to live loads. J. Geotech. Eng. Div., ASCE,
107, G T6, June, 779.
@seismicisolation
@seismicisolation
D em m in, J. (1966) Field verification o f ring com pression co nduit design. Highw. Res.
Rec., no. 116, p.36.
Fleck, H ., Spang, J. and Sonntag, G. (1980) Beitrag zur statischen B erechnung von
T unnel-auskleidungen. Die Bautechnik, 57.
G um bel, J.E . and W ilson, J. (1981) Interactive design o f buried flexible p ip e s - a fresh
ap p ro ach from basic principles. Ground Eng., M ay.
Linger, D .A . and F ernandez, P. (1968) Soil pressure d istribution on buried structures.
M eyerhof, G .G . (1970) Some research on u n derground flexible arches. Proc. Conf. on
Subway Construction, B udapest.
Sonntag, G . (1968) A unlichkeitsm echanik langsgewellter R ohre in nachgiebiger
B ettung. Der Bauingenieur, 43 (7).
@seismicisolation
@seismicisolation
CHAPTER SIX
Structures
dynamic loads
The behaviour of simple underground structures under dynamic surface

pressure has received a good deal of attention in connection with the design
of underground structures to withstand the effects of nuclear or high-
explosive blast. The construction of blast simulators in US Army, Air Force
and Naval Laboratories during the 1950s and 1960s enabled a num ber of
experimental studies to be carried out. Later the author used horizontal
and vertical shock tubes in the U K to check on the behaviour of thin-walled
cylinders in sand and clay. These studies were followed by theoretical
work leading to design rules, but most of the design information still comes
directly from test results.
6.1 Tests at the University of Illinois, USA

The dynamic testing apparatus was designed and constructed by Egger
[6.1], and later modified by Sinnamon et a l [6.2]. It is shown diagrammati-
cally in Fig. 6.1, and permits simulation of blast loading from a nuclear
weapon by means of a high-pressure gas system. The peak surface pressure
was 500 psi uniformly over the surface of the soil.
The steel container was 26 J inch high and 23 J inch, diameter and could
be filled with com pacted soil, in which the test structure was buried. The
surface of the soil was protected from gas penetration by means of a ^ inch
thick neoprene diaphragm, and above this was a baffle and grid to help
achieve uniform pressure distribution. Above the grid was a transition
expansion chamber, connecting the soil container to a 12 inch diameter
slide valve chamber. The grid consisted of a steel plate, \ inch thick, containing
925 \ inch diam eter holes.
To operate the dynamic device it was necessary to charge the external
compression chamber and the auxiliary chambers that cause the piston to
travel upwards. As it did so, the gas in the compression chamber vented
into the expansion chamber which rose very rapidly to a peak pressure.
Using helium, a pressure rise-time of 3 ms could be achieved, and with
@seismicisolation
@seismicisolation
Figure 6.1 Blast load sim ulator at the U niversity of Illinois (Hanley).
Time (m s )
Figure 6.2 Pressure profile for ‘rapid’ loading in Illinois rig.
@seismicisolation
@seismicisolation
Structures under dynamic loads 175
heavier, but less expensive, nitrogen a rise-time of 13 ms was possible. A hole
bored on the inside of the transition ring that held the grid permitted the
installation of a pressure transducer which measured gas pressure as a
function of time. A typical overpressure-tim e relationship is shown in
Fig. 6.2 for nitrogen, which produces ‘rapid’ rather than ‘instantaneous’
loading.
This apparatus was used by Dorris [6.3] to test horizontally buried
cylinders, and by Hanley [6.4] to test vertically buried cylinders. The latter’s
results are discussed later in this chapter, but ref. 6.4 contains a num ber of
good photographs of the elements of the rig. Dorris tested the following
thick-walled alum inium cylinders under static and rapid loading conditions
at the University of Illinois: diameter, D = 3.5 inch; thickness, t = 0.065 and
0.022 inch; length, / = 10.5 inch; soil, Sangamon Sand; cover depths, H = 0,
D/8, D/4, D/2 and 3D/4. He measured hoop strain on the inside and outside
of the cylinders by means of metal-film strain gauges which were connected
to a bank of amplifiers and to a 12-channel, direct-write recording oscillo
graph. In the dynamic tests, plane strain values were recorded, and from them
m om ents and thrusts at certain cross sections were calculated.
All of the cylinders that failed did so by a catastrophic snap-through
(caving) of the crown. Because of the low D/t ratios, there were no cases of
local wall buckling. Typical results for thrust at the springline (at each end
of the horizontal diameter), as indicated in the figure, are shown in Fig. 6.3.
N ote that the values of thrust for a given surface overpressures are about
20% greater for rapid loading than for static loading, and that the over-
600
O 50 100 150 200 250 300 350 400

Surface pressure ( p s i )
Figure 6.3 Springline thru sts under static and rapid loading (Dorris).
@seismicisolation
@seismicisolation
pressure to cause collapse is also about 20% higher. The vertical diameter
changes in the static tests decreased as depth of burial increased, and the
following table gives these changes at a surface overpressures of 400 p si:
C over dep th , H D/S D/4 D/2

Vertical diam eter decrease (m) 0.062 0.048 0.028
The above results were obtained from cylinders manufactured from D6061-0
alloy, having a proof stress of 7600 psi. The test series was repeated with
cylinders m ade from D5052-0 alloy, with 0.2% proof stress of 12 700 psi.
A similar increase in thrust values for rapid loading (about 20%) was observed.
The third group of cylinders, having a wall thickness of 0.22 inch, was
m anufactured from D6061-T6 alloy with a proof stress of 12 700 psi. The
cylinders were only one-twentieth as stiff as those used in the first two series,
and the pressures to induce caving of the crown were lower by a factor of 2 or
3. The difference between the effects of rapid and static loading was more
m arked than for the stiffer cylinders. The values of thrust for a given surface
overpressure were more than 50% greater for rapid loading than for static
loading.
6.2 Tests at the US Waterways Experiment Station

The small blast load generator (SBLG) at the US Waterways Experiment
Station (WES) was designed and constructed by Boynton Associated [6.5],
and is shown diagrammatically in Fig. 6.4. The steel container was 48j
inch diameter, and was constructed from a num ber of rings, so that the overall
height could be varied. After filling with soil and burying the test model,
7 ^ B„ „ . ( v 5£s
^ S o il surface
— Steel container
I"1 ...................—... —
1 ,
| Approx. 4 fr did. |
Figure 6.4 E xternal view of small blast load generator, W aterw ays Experim ent Station.
@seismicisolation
@seismicisolation
Time (ms)
Figure 6.5 P re ssu re -d u ra tio n curve for dynam ic loading in small blast load generator.
the surface of the soil was covered by a rubber diaphragm to prevent gas
and water penetration. Tests could be carried out with a rigid concrete base
or with an ‘infinite’ base consisting of a column of sand, 4 ft diameter, extend
ing 9 ft below the floor level. A ‘bonnet’ bolted onto the top of the container
could be used to apply static loading by filling it with water or gas that
was subsequently pressurized. F or dynamic loading, prim acord was de
tonated in two firing tubes that were incorporated in the bonnet, to give the
dynamic overpressure duration relationship shown in Fig. 6.5, having an
initial rise-time of 0.30 ms.
Dorris [6.3] used this rig to repeat some of the tests on horizontal cylinders
that were performed at the University of Illinois, and then tested similar
cylinders buried in Buckshot Clay. The repeat tests were made in Cook’s
Bayou Sand, which was slightly stiffer than the Sangamon Sand used at
Illinois. There was a variation in response, particularly in the springline
thrusts and diam eter changes, which were somewhat greater in the WES
experiments. Dorris felt this was due to the differing methods of emplacement.
The Sangamon Sand was vibrated and rodded in, whereas the Cook’s Bayou
Sand was sprinkled into place. The difference between static and dynamic
measurements of thrust for a range of cover depths are shown by comparison
of the scatterbands in Fig. 6.6.
The tests on Buckshot Clay were carried out for static as well as dynamic
loading. The static tests indicated higher bending m oments and larger
diam eter changes than would be expected in sand, whereas the values of
thrust were similar. The im portant conclusion was that, as in the case of
@seismicisolation
@seismicisolation
Surface pressure
2 5 0 psi
600r
-C
u
.E 50 0
2
* 400
D
v.
300
200
180 270 0 90 180
Cross-seclion location, 9 (deg)
Figure 6.6 S catterbands of static and dynam ic tests on horizontal cylinders in C ook’s
Bayou Sand (Dorris).
the sand tests, the static failure pressures served as a good basis for estimating
dynamic failure pressures, particularly for 3.5 inch diameter cylinders buried
at depths of D/4 and D/2. In general, the dynamic thrusts for the same peak
overpressure were about 20% higher than the static thrusts.
Albritton et al. [6.6] tested 6 inch diameter cylinders manufactured
from seamless steel tubing in the same rig. The wall thickness was 0.12 inch,
giving a D/t value of 50. The central test section, 12 inches long, was isolated
from the closed ends to eliminate any effects of an abrupt ending of the section
on the surrounding stress field. This was accomplished by closing the ends
with 6 inch long caps, as shown in Fig. 6.7, supported by § inch diameter
steel rods. The caps were separated from the test section by pliable gaskets
m anufactured from silicon rubber. The cylinders were tested at burial depths
0, 6 and 12 inches in Cook’s Bayou Sand, having a test density of 110 lb/ft3
and an angle of internal friction of 42°. At cover depths of 6 and 12 inches, the
loadings were both static and dynamic in nature. Albritton drew the following
conclusions from the tests:
Gasket 6 inch
Support rods
I 1 1
------- --
1 1 1
I
i_ _ y . 6 inch dia.
1 i
End cap I Test section

L — 12 inch
Figure 6.7 Steel cylindrical test specimen (A lbritton et al.).
@seismicisolation
@seismicisolation
1. The dynamic strains are 20-40% higher than strains at an equivalent
static load.
2. The cylinder responses at the 12 and 6 inch depths of burial were very
similar.
3. H oop strains were not a linear function of applied overpressure. Tensile
strains tended to become constant for pressures greater than 100 psi
and at least up to 350 psi.
A further series of tests carried out jointly by the previous two authors,
Dorris and A lbritton [6.7], used a similar experimental technique to examine
the response of a 4 inch diam eter cylinder, 18 inches long, manufactured
from inch thick steel tube. This represented protection conduits used in
the USA to encase com m unication cables leading to im portant defence
installations. It was felt that, in the absence of knowledge on the structural
behaviour of these conduits, excessively large factors of safety had been
used. The objectives of the study were to examine the response to dynamic
loading typical of a nuclear blast overpressure of 1000 psi at several depths
of cover in two types of soil media.
The experiments were conducted at depths of 2,4 and 16 inches in a dense,
dry sand, and at 8 inch depth in a stiff clay. As before, the sand was Cook’s
Bayou, with a density of 108.5 lb/ft3, and the clay had a wet density of 120
lb/ft3 with a water content of 25%. The results in sand indicated that the
m aximum hoop strains in the tube walls under dynamic loading were about
25% greater than the strains at the equivalent static loads. The conclusions
were as follows:
1. The conduit buried in sand was subjected to 466 psi static overpressure
and 295 psi dynamic overpressure without significant yielding or structural
failure.
2. The conduit buried at two diameters in clay collapsed at static pressures
of 348 and 428 psi. The strains in the conduit were much larger in clay
than in sand at com parable pressures, and in both soils strain was a
non-linear function of overpressure.
3. The conduit should be able to withstand dynamic surface overpressures
of 1000 psi if the following criteria are met:
(a) burial at depths greater than three times the diam eter;
(b) burial in a material approxim ating a dense, dry sand; and
(c) the loads exclude axial loads on the conduit, longitudinal bending
moments, and concentrations due to joints or abrupt changes in the
surrounding environment.
Experimental results were com pared with theories of elastic response in a

short report by W alker et al. [6.8], published in April 1966.
A much larger blast load generator (BLG) was commissioned at WES
@seismicisolation
@seismicisolation
4 8 tr
/ \
Concrete reaction
structure
30 ft Firing tubes"
Test cham ber W T

23 ft dia. 13ft
1 Ground
Rails
22-5 ft I Foundations
l_i
Figure 6.8 Large blast load generator at the W aterw ays E xperim ent Station.
in the mid-1960s, taking the form of a central firing station and test chambers
as shown in Fig. 6.8. The former is a massive, post-tensioned prestressed
concrete reaction structure, designed to resist the dynamic loads generated
in the test chambers. It consists of a 48 ft x 28 ft x 28.5 ft deep slab, with a
13 ft deep tunnel cut through it. Into this tunnel fits a circular test chamber,
about 23 ft in diameter, that contains the soil medium and the structure
under test. The chamber is made by stacking three rings and adding a ring
containing firing tubes. Pressures from 30 to 500 psi, with rise-times of
2 -4 ms, can be reproduced in the generator, and static loads up to 1000 psi
can be sustained. A baffle system is used to ensure that the pressure applied
to the soil surface is uniformly applied. The BLG has been used for tests
on slabs and arches but most of the test work at WES has used the small
generator. O ther agencies have also used this facility, notably the Naval
Civil Engineering Laboratory, who tested model pressurized underground
fuel storage containers.
6.3 Tests at the US Air Force Weapons Laboratory

An alternative method of applying dynamic pressures to model structures
in soil was constructed at the US Air Force W eapons Laboratory (AFWL),
K irtland Air Force Base, New Mexico, in the late 1960s. This used the concept
of a vertical shock tube, 2 ft diameter, 40 ft long, set over a cylindrical soil
bin 48 inches high, 22 inches in diameter, and is described in a report by
A bbott [6.9]. When primacord, distributed over the top 10 ft of the tube,
was detonated, the remaining 30 ft of the tube became an expansion chamber,
allowing the shock compression wave to develop into a clean and plane
@seismicisolation
@seismicisolation
wave before striking the surface of the soil. The maximum charge was
4.57 lb (32 000 grains). D uring the calibration tests, measurements were made
of the incident pressures on the soil surface and on the peak reflected pressures
at the surface due to the shock wave reflection at the bottom of the soil
bin. The peak reflected pressures were generally about five times as great
as the incident pressure, and this agreed closely with the calculated reflected
pressure ratio assuming a perfectly rigid reflecting surface. Thus,
(6.1)
where pr is the theoretical peak value of reflected pressure, p. is the peak

value of incident pressure, and p0 is atmospheric pressure. The relation
between the reflected pressure p and time agreed closely with the Friedlander
equation
P = PrU - (/ (o)e" '0» (6.2)
where t is time, and t0 is duration of positive phase.

It was clear that a very short time existed between the incident and reflected
waves in the soil bin, and the rig was only used for studies in which these
early stress reflections did not present a problem.
6.4 Tests in the UK

The author carried out a series of tests on model thin-walled cylinders in a
vertical shock tube facility not unlike that described in the previous section.
It was built at the Atomic W eapons Research Establishment (AWRE),
Foulness, and is fully described in a report by Clare [6.10]. A 5 ft square,
4 ft deep soil bin was placed centrally under a 40 ft long, 2 ft diameter shock
tube. The tube was fired by means of a coiled charge of Cordtex at the top,
which gave a blast pressure-duration curve at the soil surface of the Fried
lander form as shown in Fig. 6.9. The peak pressure could be varied by
changing the length of Cordtex.
The test specimens were set in exactly the same way as in the static tests
described in the third and subsequent paragraphs of Section 4.3.4 above,
with the sand and clay under similar conditions of moisture and compaction.
High-speed cine-film was taken, through the tunnels, of each collapse
sequence. Tube deflections were indicated by gluing a serrated strip around
the internal circumference of the tube, halfway along its length. The inspection
ports at the ends of the tunnels were covered by \ inch thick perspex to
prevent sand blowing out of the rig and enveloping the camera.
The firing procedure was first to set the camera working by electrically
operated remote control from a blastproof building a short distance from
the rig, and a few milliseconds later the camera operated an electrical charge-
@seismicisolation
@seismicisolation
Duration (m s )
Figure 6.9 P re ssu re -d u ra tio n curve at the base of the shock tube (Clare).
firing circuit. The m om ent of firing was recorded by a flash on the film from
a photoflood bulb operated by the firing circuit. The camera speed was
3000 frames per second, with a running time of about 3 s, including accelera
tion and deceleration. The time interval from firing to complete decay of the
blast pressure was about 20 ms (0.02 s).
Pressure gauges of the piezoelectric type were set in the wall of the shock
tube near its m outh. These were calibrated in relation to gauges set in a
plate m ounted across the bottom of the tube, so it was possible to measure
accurately the blast pressure on the soil surface.
Cover depths were chosen in the range H = 0 to 2D, where D is the tube
diameter, and at each depth specimens were subjected to two or three in
creasing values of peak pressure, chosen to give a no collapse - appearance
of buckles - complete failure sequence. This enabled a reasonably accurate
assessment of the peak blast pressure to cause complete collapse to be made.
F or example, for 4 inch diameter thin-walled tubes buried at a cover depth
of 1.5 inch, the 10.2 psi peak had no perm anent effect, the 24 psi peak caused
the appearance of rim buckles, and 31.3 psi resulted in complete failure.
Complete failure would therefore first occur between 24 and 31.3 psi peak
pressure.
Fig. 6.10 shows how the pattern of behaviour varied with cover depth
for brass tubes 4 inch diameter, 0.005 inch thick in compacted sand. It
suggests a zone in which complete collapse might first be expected to occur.
The ordinate is the peak dynamic overpressure, but on the same axis the
lower scatter boundary of the static collapse pressures, taken from tests
described in C hapter 4, is also shown. Within the limits of experimental
accuracy there seems to be a close link between static collapse pressure
and peak dynamic pressure, and the response of the tubes to dynamic
pressures was so fast that no account needed to be taken of blast impulse or
strain energy absorbed during the collapse process. Photographs were
@seismicisolation
@seismicisolation
40
\n
a Com plete failure
30
a C ollapse zone
8 Rim buckling, no collapse
B 20
3
lA
O Low er scatter boundary, static tests
E
c 10
>
TJ N o deform ation
o
0)
Q. _______I________ I________ I
0 2 4 6 8
Depth of cover (inch)
Figure 6.10 D ynam ic tests on 4 inch diam eter brass tubes in com pacted sand (Bulson).
Fram e num ber
Figure 6.11 F ram es from a high-speed film of collapse of thin-w alled circular tube
(diagram m atic) (Bulson).
taken from the high-speed film, indicating stages in the collapse, and are
shown in diagram m atic form in Fig. 6.11.
The tests were repeated, using both brass and steel tubes, in a remoulded
clay, and the results are given in Figs 6.12 and 6.13. An im portant feature
was the similarity in dynamic collapse pressures for the brass tubes in both
sand and clay, whereas we have already seen that the static collapse pressures
in sand are about four times as large as in clay. In dynamic conditions clay
seems to offer similar support to the specimens as compacted sand, and the
time-conscious behaviour of soft clay in static or quasi-static loading states,
which leads to lower collapse strengths, seems to have no effect when the
pressure is applied virtually instantaneously.
@seismicisolation
@seismicisolation
Depfh of cover (inch)

Figure 6.12 D ynam ic tests on 4 inch diam eter brass tubes in soft clay (Bulson).
Depth of cover (inch)

Figure 6.13 D ynam ic tests on 4 inch diam eter steel tubes in soft clay (Bulson).
The fact that peak dynamic pressures to cause collapse are similar to
the static pressure suggests that the time to collapse of the soil-specimen
system is about equal to the duration of the positive loading phase, in these
experiments about 20 ms. Also, the total impulsive load (the area under the
pressure-duration curve for the positive phase) is equal to that required to
cause collapse.
@seismicisolation
@seismicisolation
If the peak instantaneous vertical pressure is pv, then for exponential
decay, the pressure p at any time t after initiation can be found approximately
from the expression
—0 . 1 3 8 f
P = Pv e (6.3)
where t is in milliseconds. The expression gives p approaching zero when
t = 20. The impulse I c to cause collapse is given by
(6.4)
and if this is applied to both steel and brass specimens it implies that the
collapse time at any depth of cover for both types is 20 ms. This suggests that
the time taken for the pressure wave to travel through the soil is not signi
ficant, and that the collapse time is governed mainly by soil properties and
the elastic response of the specimen. These conclusions might not apply at
full scale because the distances of travel of the pressure wave would be an
order of magnitude greater than those used in the experiments. In addition,
this was a very limited series of tests, and more experimental data are desir
able before firm conclusions can be drawn.
A second facility of the AWRE used for dynamic tests on tubular structures
was the large horizontal shock tube. It was operated by two 9.2 inch guns
fired into a 6 ft diameter steel tunnel 119 ft long linked by an asymmetric
cone into an 8 ft diam eter section 137 ft long. This section was linked by a
further expansion cone into a 16 ft diam eter section, 217 ft long. A short
length of the 8 ft diam eter section ran over a tank of soil, and at this point
the lower wall of the tunnel was removed so that the shock wave ran trans
versely over the soil surface. Figs 6.14 and 6.15 show general views of the
shock tube, a cross section and the internal arrangement inside the test
zone.
Three gauges were used to record the overpressure in the vicinity of the
Driver umTs -. 2x9*2 inch guns
Figure 6.14 G eneral view of large horizontal shock tube.
@seismicisolation
@seismicisolation
Figure 6.15 C ross section of shock tube at the test zone.
test rig, which consisted of the 5 ft square static tank, described in Section
4.3.4, buried to the level of its surface in the soil under the open area of the
tube. The positive phase duration was about 80 ms.
Similar specimens to those used in the U K static tests were buried in
com pacted sand or soft clay, and subjected to increasing dynamic loads
from the transverse blast wave until collapse took place. As in the tests
under the vertical shock tube, a zone containing the peak pressure levels
Cover dephh (inch)

Figure 6.16 C ollapse zone for brass tubes in well com pacted sand u n d er transverse
blast pressure.
@seismicisolation
@seismicisolation
at which collapse first occurred was established, and the results for brass
specimens of 4 inch diameter, buried in sand and clay, are shown in Figs 6.16
and 6.17, and com pared with the tests under vertical dynamic overpressure
in Figs 6.18 and 6.19. The results may be summarized as follows:
1. Vertical blast, 20 ms positive phase: peak dynamic collapse pressures in
sand and clay are about equal at all cover depths investigated.
2. Transverse blast, 80 ms positive phase: peak dynamic collapse pressures
in clay are about half their value under vertical blast.
Cover depth (inch)

Figure 6.17 C ollapse zone for brass tubes in soft clay under transverse blast pressure.
Cover depth (inch)

Figure 6.18 C om parison of vertical and transverse dynam ic tests on brass tubes in
sand.
@seismicisolation
@seismicisolation
Collapse zone, vertical
dynamic surface pressure
Figure 6.19 C om parison of vertical and transverse dynam ic tests on brass tubes in
soft clay.
3. Transverse blast, 80 ms positive phase: peak dynamic collapse in sand

are about one-third of their value under vertical blast.
If the peak instantaneous transverse pressure is pt, then, as before,
P = pte _0 0345t, (6.5)
where t is in milliseconds.
In this expression p approaches zero when t = 80. Assuming that a similar
impulsive load level is required to collapse the specimen in either vertical
or transverse blast conditions, we may write
’20 *20
K =P, e - ° 138fd t = Pt e - u-ui45fdt, (6.6)

o
which gives
p = 0.47 p . (6.7)
This elementary analysis gives a good indication of the behaviour in clay,
but overestimates by some 15% the results in sand. A possible explanation
is that an appreciable quantity of the upper layers of sand was removed
during each transverse firing by the forward airblast, effectively reducing
the cover to about one-quarter of its original depth. This removal of sand
layers might be less im portant at full scale, but it underlines the need to
consolidate or to protect the sand cover from dispersion.
The periodic time of one complete vibration of a thin circular ring, having
a mode given by n circumferential full waves, has been discussed by Lord
@seismicisolation
@seismicisolation
Rayleigh [6.11]. He showed that, to a close approxim ation, the angular
velocity of the vibration is given by
2 0n2- 1f E t 2
w = 7KT, 2T-^4> (6-8>
where a> is expressed in rad/s and density p in lb sec2/in4. The periodic time
T{ = In/co) is thus expressed (in seconds) as
T = — ^ Y ^ n a - v 2) ] 1'1. (6.9)
(n2 - 1) t \ E
W hen n grows large,
n* ( 6 . 10)
co2 "—
— 2p
Rtp
The response time of the ring, i.e. the time that elapses from the application
of the disturbance to the maximum elastic deflection of the circumference
in a radial direction, is equal to T/4. A typical value of n for the thin-walled
tubes is 7, so that the response of steel specimens, t = 0.0055 inch, R = 2 inch,
pg = 0.28 lb/in3, is about 0.4 ms.
Since the time to collapse of the specimen was 20 ms, it follows that the
elastic response time was negligible and that the resistance of the specimens
to dynamic loading was due mainly to the rotation of the buckle hinges
under fully plastic conditions.
6.5 Dynamic analysis

The response of a cylindrical shell to transverse shock waves, taking into
account extensional and inextensional modes of vibration, was examined
theoretically in the 1950s by Baron [6.12], and later he examined the diffrac
tion of a pressure wave by a cylindrical shell buried in an elastic medium
(Baron and Parnes [6.13]). In 1960 W iedermann [6.14] discussed the
forces acting on a buried structure due to idealized waveforms, and soon
after Yoshihara et al. [6.15] considered the response of buried structures
subjected to dilatational and shear types of plane wave. A useful contribution
by C onstantino et al. [6.16] presented an approxim ate soil-structure interac
tion model applicable to buried flexible circular tunnels and silos subjected
to air-blast-induced ground shock. To examine the response of a vertical
silo, they assumed that the loading would consist of the following radial
com ponents:
1. the free field stress;
2. a stress dependent on the radial displacement of the cylinder; and
3. a stress dependent on the radial velocity of the cylinder.
@seismicisolation
@seismicisolation
This work has been reviewed by Duns and Butterfield [6.17], who themselves
presented an analysis using a dynamic version of the finite element method
in conjunction with a mode superposition technique. They used about 200
finite elements in the region of a cylinder buried with its axis horizontal,
and the m otion under transient loading was represented by the first eight
modes of vibration. The cylinder, having a diameter/thickness ratio of 50,
was represented by a 12-sided polygon.
Their results suggested that the dynamic load factor (by which static thrusts
and bending m oments are magnified) was about 1.2 for depth of cover
exceeding 0.5 of the radius. This ratio compares reasonably well with the
results of the tests by M arino and Riley, Albritton and Allgood, presented
earlier in this chapter. The authors repeated their analysis for a cylinder
having a diameter/thickness ratio of 600, with similar results.
The dynamic analysis of many configurations of buried structure is often
linked to survival and protection against nuclear or large high-explosive
bomb. The worst conditions usually stem from the explosion of a weapon
on the surface of the ground close to the structure, from which the structure
is subjected to an overpressure that varies with time as the air-blast wave
front passes over the earth’s surface above it. If parts of the structure or its
cover project above ground level at too steep an angle, then it will be subjected
in addition to a dynamic pressure. This also varies with time. The minimum
earth cover required to define the structure as ‘underground’ has been given
with respect to arch or dome structures, and to rectangular structures, in
ref. 6.18. This cover is sufficient to eliminate the dynamic pressure loading
and is shown in Fig. 6.20.
The structure can also be loaded by direct ground shock, which starts
L / e (min.)
Arch or dome
Ground level
Rectangular structure
Figure 6.20 M inim um earth cover to define a structure as ‘underground’(Newmark).
@seismicisolation
@seismicisolation
Pressure wave
Air-blast" wave propagation

or shock velocity
Z
Seismic , Outrunning
velocity ^ ground shock
Logging ground shock
Figure 6.21 ‘Lagging’ and ‘o u trunning’ ground shock (Newmark).
where the air-blast wave front meets the surface, as shown in Fig. 6.21.
G round shock may ‘lag’ behind the wave front or precede it, depending on
the relationship between the air-blast wave velocity and the seismic velocity.
This is indicated in the figure. At depths of less than 100 ft, and at distances
from the explosion at which the air-blast overpressure is less than 200 psi,
the ground shock has low acceleration, and is ignored in the calculations.
The vertical stresses in the soil induced by air blast from megaton weapons
is considered in ref. 6.18 to be not attenuated at depths up to 100 ft.
The position of the water table is also im portant. It is not advisable to
bury the structure under the water table, as the saturated medium will
result in much higher loads reaching it. It is better to set it above the water
table and use an earth mound to obtain the required depth of cover, taking
care to avoid dynamic pressures, as indicated in Fig. 6.22.
The impulsive loads due to blast can be linked to the strain energy absorbed
by the structure in deforming and folding during the collapse process,
particularly when the structure takes the shape of a thin-walled steel tube
with good ductile properties. This was illustrated by the author in a field
test in Canada, in which square-sectioned tubes were buried in varying covers
of com pacted sand and subjected to blast from the explosion of a large
Blast wave
Earth mound
Water table J
Figure 6.22 Setting an ‘u n derground’ structure above the w ater table.
@seismicisolation
@seismicisolation
0 0
(a ) (b )
i------------- 1
I I
(c )
Figure 6.23 D ynam ic collapse m odes for 8 inch square steel tubes in sand (Bulson).
(a) 3 inch cover, less pressure; (b) 3 inch cover, high pressure; (c) 6 inch cover, high
pressure.
quantity of TNT. The tubes were unstiffened, made from thin mild steel
sheet, 0.0165 inch thick, and 8 inches square, and the tests were conducted
at two depths of burial, 3 and 6 inches (see Bulson [6.19]). The duration of
the positive phase of the blast wave varied from 193 ms at a peak pressure
of 10.5 psi, to 100 ms at a peak pressure of 51 psi.
The dynamic collapse modes are shown in Fig. 6.23, and, with the aid
of the static load-deflection curve at zero cover and the results of subsidiary
tests on the collapse of tubes under two-point loading, it was possible to
estimate the strain energy (in lb ft) absorbed by each specimen in deforming
to its final shape. Fig. 6.24 compares this strain energy with H 2, where H
Figure 6.24 R elationship between strain energy and (impulse)2 for 8 inch square tubes
(Bulson).
@seismicisolation
@seismicisolation
is the total impulse of the external pressure over the plan area of the tube.
H is the area under the pressure-duration curve of each explosion multiplied
by plan area. The linear experimental relationship is shown in Fig. 6.24.
Because the peak value of the external pressure is so large compared
with the plastic resistance of the system, the load must be nearly a pure
impulse, and the response of the system can be neglected. Then, for an
equivalent system having one degree of freedom,
H = W ev/g (6.11)
where W e is the equivalent weight and v the velocity resulting from the
impulse. If it is also assumed that the kinetic energy is all absorbed by the
strain energy U, then U = W cv2/2g, so that, from equation (6.11)
U = H 2g/2W e. (6.12)
This indicates the linear relationship.
6.6 Underground concrete structures under localized explosive loads

This problem has given rise to a num ber of design and test studies over the
past 30 years, with much of the basic inform ation stemming from experience
during the Second W orld War. This was summarized by Christopherson
[6.20] in 1946, who dealt with all types of explosions of the non-nuclear
variety, with the theory of structural behaviour under suddenly applied
loads and with the design of ‘bom bproof’ shelters and buildings. He pointed
out that the confinement afforded by the earth, when a bomb explodes in
contact with a concrete structure below ground level, will increase the
dynamic forces on the structure. It was concluded that underground walls
must be made heavier than walls above ground to resist the same explosive
force. Three full-scale experiments were made to determine the critical
thickness of concrete required to resist side-on contact explosion in earth,
and these indicated that a concrete wall, reinforced by 0.75% by volume of
steel, with two-thirds of the reinforcement at the inner face, will bulge a
maximum distance equal to about half the span if its thickness is given by
t = 13w1/3, where w is the charge weight in TN T of the bomb (t in inches,
w in pounds). If it is desired to limit the bulge to one-seventh of the span,
then t = 17 w1/3. F or example, a 250 kg bomb, with a charge weight of 275 lb,
was found to cause a bulge of 0.43 x span when exploding against an
underground vertical slab of 84 inch thickness.
The effects of bombs exploding in the earth near subsurface walls are much
greater than those due to bombs exploding in the air. The thickness of concrete
@seismicisolation
@seismicisolation
15
Bomb weighf
2000 lb
O 5 10 15 20
Distance of explosion from wall ( ft)
Figure 6.25 Thickness of reinforced concrete to resist spalling from tam ped explosions.
to resist spalling from tamped explosions was given in terms of the weight of
the charge in pounds and the distance of the explosion from the wall in feet, and
is shown in Fig. 6.25. In a tam ped explosion, the bomb is sufficiently far below
the surface for no crater to be formed. Tamped explosions occur at depths
greater than 5, 8 and 10 ft for 500, 1000 and 2000 lb bombs respectively.
In 1965 a m anual on the Fundamentals o f Protective Design (Non-nuclear)
was issued by the US Departm ent of the Army (Tech. M anual TM5-855-1),
and this gave advice on the structural disign of an underground shelter.
The shelter, made of reinforced concrete, was 25 ft long, 10 ft wide, 8.5 ft
high (internal dimensions), in the shape of a rectangular box, and was buried
with a minimum of 2 ft cover. This depth of cover was taken as sufficient
to support vegetation, yet avoid the effects of a tam ped explosion. The
protective requirem ent was ‘slight damage (fine cracks)’ in the face of a 550
lb general purpose (GP) bom b detonating at distances greater than 35 ft
from the side walls. The charge weight of the bomb is 250 lb. The peak
pressure from an underground explosion of w pounds, occurring at a distance
r feet from a structure, is given as
(6.13)
where k is the soil constant, and is defined as follows: compacted sand,

30 000; sandy clay, 20 000; clay loam, 5000; loose soil (moist), 2000; loose
soil (dry), 1000.
@seismicisolation
@seismicisolation
o
T>
o
o
0 1 2 3 4 5
t/T
Figure 6.26 M axim um positive dynam ic load factors for underground explosions.
The duration of the pressure wave (in seconds) is given by

t = 0.002wOAr l/3k l/6. (6.14)
F or a charge weight of 250 lb, r = 35 ft and k = 8000 (gravelly soil), equations
(6.13) and (6.14) give 62 psi and 0.051 s respectively.
In order to design the side wall it is necessary to assume a thickness and
calculate the period of a uniformly loaded simple slab from the well known
equation
(6.15)
where W is the slab weight per foot, L is the slab span, and E l the flexural
rigidity of the slab. The relationship between this period and the duration
of the pressure wave is used to find the m aximum positive dynamic load
factor. This is found from Fig. 6.26 in terms of t/T. The equivalent static
pressure on the slab is then p x dynamic factor.
F urther work in the USA by Fuehrer and Keeser [6.21, 6.22] examined
the minimum breaching range for buried concrete slabs, and had shown that
this range r could be expressed as
(6.16)
where t is the thickness of the concrete slab, L is its span and w is the charge
weight.
@seismicisolation
@seismicisolation
6.7 Underground concrete structures under surface pressures

from nuclear explosions
A m ethod of analysing the behaviour of reinforced concrete bunkers using
the pressure-im pulse characterization has been proposed by Florence
[6.23]. This analytical method, due to Abraham son and Lindberg [6.24],
uses the relationship between ground surface peak pressure and impulse
with range from a specified nuclear weapon. The impulse from a nuclear
explosion is the area under the pressure-duration curve, and if this is com
pared with the peak instantaneous pressure for a series of ranges from the
centre of the explosion, there will be a decrease of both pressure and impulse
as the range from the centre increases. If the range is eliminated from the
plotting coordinates the ‘free field’ relationship of Fig. 6.27 is obtained,
with range an intrinsic coordinate, measured along the curve and increasing
towards the origin. The impulse for a given overpressure is assumed to be
less at the interface between the roof of the bunker and the backfill, due to
dynamic arching. Taking this into account would produce a relationship
of the type shown as ‘interface’ in Fig. 6.27; once again, range is an intrinsic
coordinate increasing towards the origin.
The curve labelled ‘structure’ is an ‘isodamage’ curve, containing all the
com binations of pressure and impulse that are known to produce the same
damage level. From the analysis of simple structures it has been shown that
the curve is of the reciprocal variety, i.e. pressure x impulse = constant,
and this is indicated in Fig. 6.27. The point of intersection of this curve with
the ‘interface’ curve gives the range within which the specified nuclear
detonation will destroy the structure. It has been suggested from the analysis
of tests that an approxim ate generalization of the structural pressure-
impulse curve is
(P/Po- ! ) ( / / / „ - 1 ) = 1 . (617)
Figure 6.27 F ailure prediction by p ressure-im pulse relationship (Florence).
@seismicisolation
@seismicisolation
Figure 6.28 P ressu re-im p u lse m ethod applied to 4 ft x 4 ft x 16 ft bunker models

(Florence): p0 = 228 psi, I 0 = 2329 lb m s/in 2.
where p0 is the static collapse pressure and I 0 is the impulse that produces
the same dam age level as that used in the ‘structure’ curve. N ote that, for
buried structures, the pressure-im pulse curve must take account of load
transfer from the surface to the roof of the structure. The generalized diagram
deduced by Florence [6.23] for 4 ft x 4 ft x 16 ft bunker models referred
to in ref. 6.22 is shown in Fig. 6.28, and it indicates catastrophic damage at
a range of 4 ft for a localized explosion.
The effect of blast loading on one-tenth scale models of reinforced concrete
slabs was examined in a field test in 1961 by Purdie [6.25]. Panels with 0.75%
reinforcement, measuring 15 inches square, 0.6 inch thick, were buried with
the following range of cover depths of well compacted sand: 0.5, 1.0, 1.5
and 2.0 inch. They were exposed to the effects of a surface blast wave from
the detonation of 100 tons of TN T at Suffield, Canada, and the peak pressures
varied from 16 to 74 psi depending on the position of the slab relative to the
centre of the explosion. Slabs were also set in the surface as a comparison.
The surface results w ere:
P eak pressure (psi) 12 16 19 22 25

State of slab residual slight residual destroyed destroyed
deflection cracking deflection
N one of the slabs under soil cover were destroyed or severely cracked.
Residual deflections in the range 0.06-0.09 inch were measured in the
following cases:
@seismicisolation
@seismicisolation
Peak pressure (psi) 16 45 45 35 to 55

C over depth 0.5 x span 1 x span 1.5 x span 2 x span
Thus, burial at 1 x span (15 inches) was sufficient to double the peak pressure
at which slight residual damage occurred. N ote that a 100 ton hemispherical
TN T charge produces a blast wave similar to that obtained from a kiloton
nuclear surface burst.
Reference 6.18, which has become a standard work over the years on the
design of hardened structures to resist nuclear weapon effects, points out
that the loading on underground structures comes from direct ground shock
and air-induced ground shock. At the pressures of m ajor interest, i.e. 2000
psi peak overpressures and less, the direct ground shock is of relatively
little im portance and is not normally considered. A conservative design
philosophy is to design the underground structure to have static strength
equal to the peak load applied to it by the blast.
6.8 Structural design of domestic nuclear shelters

The U K Home Office issued a technical guide to the design and installation
for nuclear shelters in 1981, prepared by a working group under the chair
m anship of J.C. Cotterill, and published by Her Majesty’s Stationery Office.
The structural design chapter concentrated mainly on the behaviour of
reinforced concrete shelters of a rectangular shape, but with some reference
to the use of heavy steel construction. Thin-walled metal and plastic designs
were left for a subsequent publication which is in draft at the time of writing.
The design blast pressures for shallow buried construction were given in
terms of the peak overpressure, ps0, as: dry g ro u n d -ro o f and floors, ps0 ;
walls, 0.5psO ; high water le v e l-ro o f and floors, ps0 ; walls, ps0. The figures
given here should be added to the dead loads, soil and water loads on the
structure. The dynamic analysis of the structure was based on the assumption
that for ductile thick-walled structures exposed to overpressures of up to 3
atm from nuclear explosions, the load duration is relatively long in relation
to the period of inelastic response of the elements. The following relationship
was recom m ended:
(6.18)
where ru is the ultim ate unit resistance, F is the blast load, and p is the ductility
ratio
permanent deflection
^ deflection at the elastic limit
@seismicisolation
@seismicisolation
F or m oderate damage (i.e. slight deformation and rupture of the structural
elements), is taken as 3, and equation (6.18) becomes
ru = 1.2 F. (6.19)
The ultim ate resistance of an element varies with load distribution, geometry,
reinforcement and support, and for ‘one-way’ elements such as cantilevers
and beams can be specified in terms of the moment capacity at the supports
and midspan. Thus, for example, for a simple cantilever
ru = 2 M J L \
where M y is the ultim ate unit negative moment capacity at the support, and
L is the span length. For a beam with fixed supports,
r u = J } ( M V + M p)>
where M p is the ultim ate unit positive m oment capacity at midspan.

F or a two-way element such as a simply supported slab :
short span
8(Mu + M p) ( 3 L - x )
ru - '■' » (6.20)
B 2( 3 L - 4 x)
long span
5(Mu + M p)
(6.21)
where B is the short span side length of the slab, and L - 2x is the length of
the central yield line. N ote that when the slab is square x = L/2, B = L, and
Ground level
Earth cover 1-5 m
Stairs, etc.
? ////////////////////& &
2-5 m
6 '4 m
Figure 6.29 Vertical section thro u g h U K H om e Office proposed layout of a shelter for
12 persons.
@seismicisolation
@seismicisolation
equations (6.20) and (6.21) both give
_ 20(MU + M p)
r“ “ L2
This is the well known ‘yield-line theory’ for calculating the ultimate strength
of thin concrete slabs. For further discussion the reader should see J. Am.
Concrete Inst., 74, February 1977, Appendix C - N uclear concrete structures.
The recommended general arrangem ent for a shelter for 12 persons,
under 1.5 m of earth cover, is shown in Fig. 6.29. As well as protection from
the overpressure due to the blast wave, this design also gives a high level of
protection against fall-out and initial nuclear radiation.
6.9 Backpacking
Interface pressures under dynamic loading on buried structures can be
considerably reduced by the proper use of ‘backpacking’. Energy-absorbing
materials can be introduced as slabs immediately above the structure, or
as complete external casings around it, and this will help in the ‘shock isola
tion’ of the installation. Typical materials are granular in form, or honey
combs, flexible and rigid foamed plastics and rubbers, and low-density
concretes. The ideal elastoplastic load-com pression relationship is shown
in Fig. 6.30. Tests were carried out on a wide range of materials by Hoff
[6.26], who also reviewed other published reports, and he concluded that
foamed or cellular m aterials exhibited the desirable elastoplastic behaviour.
A year or two later tests were conducted by Foster [6.27] to study the
response of a stiff, horizontally oriented steel cylinder that had been back-
packed around its entire circumference with low-strength cellular concrete.
It was buried in well compacted sand and subjected to static and dynamic
overpressures, using the small blast load generator at the US Waterways
Figure 6.30 Ideal stre ss-stra in characteristic for elastoplastic backpacking (Hoff).
@seismicisolation
@seismicisolation
2-875 inch! \ J
'V . ^ ,/ L O w -s fr e n g fh
cellular concrete
Figure 6.31 C ross section of test cylinder w ith cellular concrete backpacking (Foster).
Experiment Station (described earlier in this chapter). The cylinder had an

outside diam eter of 6 inches, a 0.12 inch wall thickness, and was surrounded
by a layer of cellular concrete either 1.375 inch or 2.875 inch thick, as shown
in Fig. 6.31. There was a noticeable difference in the compressive yield
strengths of the backpacking, depending on whether it was bonded to the
cylinder or not. The yield strength (bonded) was 26 psi, and unbonded 40
psi. The cylinder was buried to a depth of 12 inches (surface to crown).
The static overpressures ranged from 0 to 250 psi, and the dynamic over
pressures from 100 to 250 psi (peak).
The conclusions were that the backpacking reduced the peak strains
in the cylinder by 20% for the 1.375 inch thick layer, and 40% for the 2.875 inch
thick layer. The moments and thrusts in the latter case were 50% of those of a
cylinder without backpacking. The affect on acceleration response was
significant, and accelerations were only 10-50% of those in a cylinder with
no backpacking.
Allgood [6.28] introduced backpacking calculations into his recommenda
tions for the design of underground structures, by assuming that the depth
of cover/diam eter ratio must be large enough to permit the formation of a
soil arch above the slab of cellular material and between the slab and the
roof of the structure. Providing the integrity of this soil arch is maintained,
he deduced that the peak pressure on the structure would equal the yield
stress of the backpacking. Referring to Fig. 6.32, he showed that the required
thickness of backpacking was given by
(6.22)
where Qm = A gH'{sJss - 1),
@seismicisolation
@seismicisolation
M M I
;W /////////A r Backpacking
— Sfruct-ure
Figure 6.32 Backpacking notatio n (Allgood).
so that
= BA l . (6.23)
where ec is the average strain over the height of the inclusion, and ehL is the
hardening strain of the backpacking.
Allgood also investigated theoretically the effect of backpacking around
the entire circumference of a cylinder, with the thickness of the packing
equal to half the diameter. The effect is to reduce stresses in the cylinder by a
factor of 3. There is a danger, however, with a complete annular ring of
packing, that under blast conditions the cylinder might vibrate within the
ring.
6.10 Military shelters

There have been many tests in the USA and elsewhere on the behaviour
of military field shelters under blast loading. Some results are classified
and cannot be discussed here, but there are two series of tests that give an
indication of the behaviour of lightweight underground structures under
dynamic loading.
The requirem ent for speed and simplicity of construction in modern
field defence systems is of overriding importance, and to achieve this the
field shelter shown in Fig. 6.33 is used by the U K Army. It is known as the
MEXE Shelter, and consists of a num ber of single steel frame members
that support a flexible revetting material. Corrugated iron sheeting as a
revetment was unsuitable because edge distortion made it vulnerable to drag
pressures. These in turn caused serious displacements. The construction
shown in the figure consists of a main two-bay chamber 5 ft 4 inch square,
assembled from three frames separated by spacer bars. The covered entrance
trench is 5 ft 4 inch long and 2 ft 8 inch wide, with an entrance shaft 2 ft
@seismicisolation
@seismicisolation
Figure 6.33 C ross section of U K m ilitary shelter.
8 inch square at the end remote from the chamber. The frameworks consist
of steel pickets, spacer bars and curved roof arches.
The structure is built in an excavation 4 ft 6 inch deep and the sides and
roof are clad in a flexible revetting material made from a combination of
jute, wire and polyvinylchloride (PVC). The sides of the excavation are
backfilled with sand, and the roof covered with sand to a depth of 18 inches
over the main frames. Over the 2 ft 8 inch square entrance shaft an aluminium
cover is fitted with a central aperture 1 ft 2f inch diameter. This sufficiently
restricts blast entry to cause a pressure differential between the inside and
the outside.
W hen a shelter of this type was subjected to a pure pressure loading
of 43 psi from the detonation of 500 tons of TNT, the heavy metal com
ponents and the revetting material withstood the strain satisfactorily. The
am ount of debris falling into the shelter would be uncomfortable for an
occupant, but not lethal.
A one-third scale model was tested in the nuclear blast sim ulator at
AWRE, Foulness (described in Section 6.4), to study the variation of pressure
with time inside the shelter, to assess the degree of protection from over
pressure and translation afforded by the shelter to its occupants, and to
investigate the structural response. One-third scale dummy men, which
presented the correct area, degree of movement and weight distribution,
were suspended by a bom b release from one of the arch members of the
shelter framework. F our separate positions of the dummies were chosen,
and at each of these the dummies were the object of tests at 17 psi nominal
overpressure. The movements were small and limb velocities were low enough
not to cause damage on impact. The dummies were subjected to the blast
forces within the shelter from a step shock wave of overpressure 4.5 psi,
travelling down the length of the structure from the entrance to the far wall.
The observed response to blast loading of a selection of the model shelters
was as shown in Table 6.1 (see also Fig. 6.34). In general, there was little
@seismicisolation
@seismicisolation
Table 6.1 Response to blast loading of models
P eak pressure 10 psi Little disturbance of shelter or cover
Peak pressure 16 and 19 psi Large percentage of overhead sand cover was
removed, the roof revetting m aterial lifted and
vented along the ridge. The entrance shaft and
debris trap partly collapsed
Peak pressure 24.5 and 26.5 psi There was evidence from the films th at the
sand cover was removed. W ith the revetting
m aterial wired at the centre to the frame there
was a reduction in venting, but this increased
the internal lift forces on the roof framework
Figure 6.34 Pressure characteristics for m ilitary shelter tests (Emery et al).
damage to the model shelters, and with the compacted cover rebuilt to the
required depth, the shelters would have been serviceable. The internal lift
forces were due to the ‘overshooting’ of the blast air flow into the shelter,
so that the maximum internal pressure was higher than the external free
field pressure. The total upthrust on the roof, calculated for full scale, would
be about 22 tons, which would be sufficient to lift some of the pickets,
throw off parts of the overburden, and move the roof revetting material.
The conclusions of the study, by Emery et al. [6.29], were as follows:
1. F or charges of 100 tons TN T giving peak overpressures of 10 to 30 psi
with a duration of 100 ms, the pressure rise inside the field shelter varies
between 6 and 12 psi with durations similar to that of the incident over
pressure, but with a rise-time to maximum pressure of 20 to 30 ms.
2. The personnel inside would suffer negligible injury through being thrown
to the floor, but would suffer eardrum damage at pressures over 5 psi
and be on the threshould of lung damage at 12 psi.
@seismicisolation
@seismicisolation
12 ft 8 inch
Door
Figure 6.35 P refabricated end walls of US personnel shelter (C arre and Ballard).
3. The tests suggest that the M EXE Shelter M k 2 would be able to withstand
up to 20 psi peak overpressure from a weapon yield in the region of
5 kiloton, provided the revetting material is wired to the framework,
and an extra piece of revetting material is added to the roof.
A study in the USA to evaluate end-wall designs for underground pro
tective military structures was reported by Carre and Ballard [6.30] in 1968.
An underground personnel shelter made of galvanized steel structural
plate pipe-arch sections has been adopted for field use, and the sections
were of the type used on large drainage or conduit projects where no end-wall
bulkhead is required. Prefabricated end walls of the type shown in Fig. 6.35
were tested by detonating 16 lb spherical TN T charges at distances of 5 ft
and 6 ft from the wall, as shown in Fig. 6.36.
The pipe arches were made of galvanized steel structural plate with corruga
tions of 6 inch pitch and 2 inch depth. The thickness of the plate was 0.1644
inch (8 gauge). The assembled pipe arch was 12 ft 8 inch span, with a rise of 8 ft
1 inch, and was 12 ft long. Each shelter had a structural plate underpass
entrance having a span of 5 ft 8 inch, and a rise of 5 ft 9 inch. The structure
was placed in a 4 ft deep excavation, and covered with silty clay soil to an
elevation of 5 ft above the crown of the arches.
At a detonation 6 ft away, at the m idspan of the north end wall (see Fig.
6.36), there was m oderate damage. Perm anent deformation of about 5 inches
was m easured at the centre. There was some crushing of arch section cor-
h 12 f t H
6 ft
Figure 6.36 Test geom etry of US personnel shelter (C arre and Ballard).
@seismicisolation
@seismicisolation
rugations and fracture of welds. At a detonation 5 ft away, at the midspan
of the south end wall, there was severe damage. There was total failure of the
end wall, with a perm anent deformation of 30 inches at the wall centre, and
18 inches at the top end. O ther configurations of end wall were also tested,
but as the final recom m endation proposed that they should not be used,
they have not been discussed here.
6.11 Burster slabs

A technique often used to avoid the effects of localized explosions on under
ground structures is to provide a ‘burster slab’ at ground level immediately
over the structure. If the slab is made sufficiently strong it will either prevent
or reduce the penetration of the bom b into the soil. The penetration path is
rarely completely straight, but often takes the form of a J, as shown in Fig.
6.37. The straight portion is usually about two-thirds of the total path
length, and the radius of the curved portion is between one-fifth and one-third
of the total path length.
It has been shown theoretically and subsequently checked with experi
mental results that
(6.24)
where L is the path length in soil, L « P/0.8 cos 0, P is the vertical penetration,
R is the angular linear penetration, R « 0.8L, k is an empirical constant
Total pa
lengl-h (
=L
Straight length (AB)

= 21/3
Figure 6.37 ‘J ’ p a th p en etratio n of high explosive (HE) bom bs.
@seismicisolation
@seismicisolation
E 40 Bomb weight*, IV (lb)

o
A
2000
O
c 1000
2 500
8 20
o
250
3
0) 100
0 10 20 30*103
Altif*ude (ft*)
Figure 6.38 P en etratio n of US G P bom bs in loam.
depending on type of soil, k « a —0V, V is the impact velocity, A is the

cross-sectional area of the projectile, W h is the weight of the bomb, a = 154
for sand, 190 for sandy loam, 227 for loam and 341 for clay, and = 0.07
for sand, 0.09 for sandy loam, 0.11 for loam and 0.19 for clay.
Bomb penetration therefore depends on the size of the bomb, the type
of soil into which it is dropped, and the impact velocity. The latter depends
on the altitude from which it is dropped; the relationship between altitude
and vertical penetration for a range of general-purpose (GP) bombs is shown
in Figs 6.38, 6.39 and 6.40.
The effect on this penetration of a burster slab set at ground level has been
30 Romh\A/oinhh (lb)
2000
1000
500
250
100
0 10 20 30x103
Alh'fude (ft)
Figure 6.39 P en etratio n o f US G P bom bs in sand.
@seismicisolation
@seismicisolation
Figure 6.40 P enetration of US G P bom bs in clay.
Slab thickness (in)

Figure 6.41 P en etratio n of a 500 lb G P bom b (dropped from 20 000 ft) below a concrete
slab (McNeil).
@seismicisolation
@seismicisolation
calculated by McNeil [6.31] for a 500 lb general-purpose bomb. There is a
relationship between the thickness of the slab and the exit velocity, given as:
Thickness of concrete slab (inch) 8 10 12 14 16 18

Exit velocity (ft/s) 708 620 525 422 307 173
Using this exit velocity in conjunction with equation (6.24) gives Fig. 6.41,
which shows the relationship between penetration depth in inches and
concrete thickness in inches, for a series of subgrades having a range of
clay content between 0 and 40%. It is interesting to note that the value of
k in equation (6.24) is related linearly to the clay content of the soil. Fig. 6.41
shows that for a 500 lb general-purpose bom b dropped from 20 000 ft, the
penetration varies from 20 inches in a clay-free soil under an 18 inch thick
slab, to 270 inches in a 40% clay soil under an 8 inch thick slab. By choosing
the appropriate thickness of slab it is possible to ensure that the projectile
will come to rest in the soil under the slab before it penetrates to the depth of
the underground structure. The relationship between k and ‘percentage clay’
is given in Fig. 6.42 for two thicknesses of burster slab. Note that, for a 500 lb
general-purpose bomb, W J A in equation (6.24) is approximately 2.5 psi.
Equation (6.24) is a form of the theory of penetration, originally due to
Poncelet (6.32) in the early nineteenth century. He assumed that the resisting
Clay content" (•/•)

Figure 6.42 R elationship between the soil constant k and percentage clay content
(McNeil).
@seismicisolation
@seismicisolation
force on the projectile was given by
F = - A ( a + b V 2\ (6.25)
so that the equation of m otion became
dV
m — = —A(a + b V 2), (6.26)
where m is the mass of the projectile. Integration leads eventually to the

relationship
(6.27)
The version of this expression due to Petry is equation (6.24), where a/b —
2.15 x 105 in FPS units, and b = (loge10)/2%.
The Poncelet-Petry equation can be used to calculate the penetration
of concrete, but it is generally accepted that an empirical formula derived by
Whiffin from an extensive analysis of experimental results is more accurate.
Thus,
(6.28)
in inches, where ac is the compressive strength of concrete in psi, D is the

diam eter or ‘calibre’ of projectile in inches, c is the maximum size of coarse
aggregate used in the concrete, in inches, V is the striking velocity in ft/s,
n = 10.7/<rc1/4, and </> is the angle between the path of the projectile immedi
ately before impact and the norm al to the target face.
P is the vertical penetration depth, but of course if the projectile explodes
after reaching this depth, fragments will penetrate further into the concrete
or soil. A rough estimate is that the further penetration in concrete will be
D/2. The formula, for total penetration, which is thought to be less accurate
than equation (6.28), is
(6.29)
for </> = 0°. This is often known as the ‘US Corps of Engineers’ equation.
The additional penetration (in inches) due to the explosion of the bomb
charge is often written as
(6.30)
where w is the weight of the explosive charge in the bomb.
@seismicisolation
@seismicisolation
References
6.1 Egger, W . (1957) 60 Kip C apacity, Slow or Rapid Loading Apparatus, D ept o f
Civil Engineering, U niv. o f Illinois, Rep. SRS 158.
6.2 Sinnam on, G .K . and N ew m ark, N .M . (1961) Facilities fo r Dynamic Testing o f
Soils, D ep t o f Civil Engineering, U niv. o f Illinois, Rep. SRS 244.
6.3 D orris, A .F . (1965) Response o f Horizontally Orientated Buried Cylinders to
Static and Dynamic Loading, US A rm y W aterw ays Experim ent Station, Tech.
R ep. 1-682.
6.4 H anley, J.T . (1964) Interaction betw een a sand and cylindrical shells under static
an d dynam ic loading. Proc. Symp. on Soil-Structure Interaction, Univ. o f A rizona,
T ucson.
6.5 B oynton A ssociated (1960) Operation Manual fo r 250 p si 4 Foot Diameter Dyna
mic Load Generator, B oyntor A ssoc., L a C anada, CA.
6.6 A lb ritto n , G .E ., et al. (1966) The Elastic Response o f Buried Cylinders; Critical
Literature Review and Pilot Study, US A rm y W aterw ays E xperim ent Station,
Tech. Rep. 1-720.
6.7 D orris, A .F . an d A lb ritto n , G .E. (1966) Response o f a Buried Prototype Communi
cations Conduit to Static and Dynamic Loading, U S A rm y W aterw ays E xperim ent
S tation, Tech. R ep. 1-750.
6.8 W alker, R .E ., A lb ritto n , G .E . and K ennedy, T.E. (1966) The Elastic Response o f
Buried Cylinders in Sand, US A rm y W aterw ays Experim ent Station, Misc. Paper
1-810.
6.9 A b b o tt, P. A. (1967) Calibration o f a Vertical Shock Tube and its Associated Soil
Bin, US A ir F orce W eapons Lab., K irtlan d A ir Force Base, Tech. Rep. A FW L-
TR-66-134.
6.10 C lare, R. (1966) A Face-on Vertical Blast Loading Simulator fo r Structural Response
Studies, A tom ic W eapons R esearch E stablishm ent, Rep. E l/6 6 .
6.11 Rayleigh, L ord (1945) The Theory o f Sound, vol. I, D over, N ew Y ork, p.407.
6.12 B aron, M .L. (1955) The response o f a cylindrical shell to a transverse shock wave.
Proc. 2nd Nat. Congr. o f Applied Mechanics, A SM E.
6.13 B aron, M .L. an d Parnes, R. (1962) D iffraction o f a pressure wave by a cylindrical
shell in an elastic m edium . Proc. 4th Natl. Congr. o f Applied Mechanics, A SM E.
6.14 W iederm ann, A .H . (1960) The Interaction o f Buried Structures with Ground
Shock, U S A F Special W eapons C enter, K irtlan d A ir Force Base, Tech. Rep.
A FSW C -TR -60-3.
6.15 Y oshihara, T., R obinson, A .R . and M erritt, J.L. (1963) Interaction o f Plane
Elastic Waves with an Elastic Cylindrical Shell, U niv. o f Illinois, Structural Res.
Rep. 261.
6.16 C on stan tin o , C .J., R obinson, A .R . and Salm on, M .A . (1964) A simplified so il-
structure in teraction m odel to investigate the response o f buried silos and cylinders.
6.17 D uns, C.S. and B utterfield, R. (1971) Flexible buried cylinders, P art 2 - D ynam ic
response. Int. J. Rock Mech. Min. Sci ., 8 , 601 .
6.18 N ew m ark, N .M ., et al. (1961) Design o f structures to resist nuclear weapons
effects, A S C E Manual o f Engineering Practice, no. 42.
6.19 Bulson, P.S. (1965) Blast loading o f buried square tubes. Proc. Symp. on Vibration
@seismicisolation
@seismicisolation
in Civil Engineering, Im perial College, L ondon, B utterw orths.
6.20 C hristopherson, D .G . (1946) Structural Defence, 1945, M inistry o f H om e Security,
Rep. R C 450.
6.21 F uehrer, H .R . and K eeser, J.W . (1977) Investigation o f Oblique Shocks and Edge
Effects fo r Underground Targets, US A ir Force A rm am ent Lab., Rep. A FA T L -
TR-77-1, vol. 2.
6.22 F uehrer, H .R . and K eeser, J.W . (1977) Response o f Buried Concrete Slabs to
Underground Explosions, US A ir Force A rm am ent L ab., Rep. A F A T L -T R -
77-115.
6.23 Florence, A .L. (1977) Critical Loads fo r reinforced Concrete Bunkers, SRI In ter
n ational, Rep. PYO-6203.
6.24 A braham son, G .R . and L indberg, H .E. (1976) Peak lo ad -im p u lse characterisa
tion o f critical pulse loads in structural dynam ics. Nucl. Eng. Design, 37(1).
6.25 Purdie, A .C. (1964) U K A tom ic W eapons R esearch E stablishm ent, Rep. E l/64.
6.26 H off, G .C . (1967) Shock-Absorbing M aterials, US A rm y W aterw ays Experim ent
S tation, Tech. Rep. 6-763.
6.27 F oster, D .C . (1969) Elastic Response o f Shock-Isolated Cylinders in a Dense, Dry
Sand, US W aterw ays Experim ent Station, Tech. Rep. N -69-6.
6.28 A llgood, J.R . (1972) Summary o f Soil-Structure Interaction, US N aval Civil
E ngineering Lab., Tech. Rep. R771.
6.29 Em ery, R ., Leys, I. and Rowe, R .D . (1970) U K A tom ic W eapons Research
E stablishm ent, Rep. E l/70.
6.30 C arre, G .L . and Ballard, J.T. (1968) Evaluation o f End Wall Designs fo r Under
ground Protective M ilitary Structures, US A rm y Engineer W aterw ays Experim ent
S tation, Misc. P aper N.68.7.
6.31 M cN eil, G. (1982) Effect o f a burster slab on bom b penetration (Private com m u
nication).
6.32 Poncelet, J.V. (1829) M em. Acad. Sci., 15, 55.
Further reading
H ashm i, S.J. and A l-H assani, S.T.S. (1975) Large deflexion response o f square frames
to d istributed im pulsive loads. Int. J. Mech. Sci., 17, 513.
H w ang, R .N . and Lysmer, J. (1981) R esponse o f buried structures to travelling waves.
J. Geotech. Eng. Div., ASCE, 107, G T2, Feb., 183.
K uesel, T .R . (1969) E arth q u ak e criteria for subways. J. Struct. Div., ASCE, 95, ST6,
June, 1213.
L onginow , A ., C hu, K .H . and T hom opoulos, N .T. (1982) Probability o f survival in
blast environm ent. J. Eng. Mech. Div., ASCE, 108, EM 2, A pr.
M anolis, G .D . and Beskos, D .E. (1982) D ynam ic response o f fram ed underground
structures. Comput. & Struct., 15 (5).
W illiam son, R .A . and H uff, P.H . (1960) Tests o f Buried Structural-Plate Pipes Subjected
to Blast Loading, R ep o rt to Test D irector, H olm es and N arver Inc., Los Angeles.
@seismicisolation
@seismicisolation
CHAPTER SEVEN
Loads, strength and safety
We have reviewed the fundamentals of stress distribution in soil and rock,

and the behaviour of simple buried structures under static and dynamic
surface loads. This survey, coupled with inform ation on the m ajor experi
m ental studies of the past 30 years, has been written to help the designer use a
knowledge of the science of the subject to attack new problems. But under
ground construction has, of course, been going on for countless centuries,
and over this period a fund of experience has been accumulated. Much of
this still forms the backbone of our design practice, particularly in the field
of heavy tunnel linings, and although tunnel design has not been included in
the scope of this book, an examination of safety in terms of load and strength
factors should naturally start with this form of construction.
7.1 Rail and road tunnels

During the construction of most of the British rail network in the years
between 1850 and 1900, the linings were designed so that their resistance to
crushing was about four times as great as the forces in them due to the
earth pressures on the walls. For certain buried arches exposed to shock
loading, the factor was increased to 8 or 10.
As cast iron began to be used for linings, in conjunction with the shield
m ethod of driving in Britain and the USA, the safety margins were still
m aintained in the range 4 to 10, as discussed above. Because cast or ‘grey’
iron is a material of low tensile strength, it was necessary to take special
precautions against excessive bending stresses during the construction
phase.
In the days before the Second W orld War, precast concrete began to
replace cast iron, and in the early days of the transition the concrete segments
were m anufactured to a similar profile, but with an increased thickness to
accom m odate steel reinforcing bars. Analysis and design of concrete linings,
which has recently considered the changes of loading in the ground due to the
presence of the tunnel, has led to the calculation of stresses that still have a
factor of safety of 4 applied to them.
In terms of safety against the limit state of inward collapse of the lining,
@seismicisolation
@seismicisolation
we note a variation in practice in the m ajor tunnel-building countries. The
approach varies from factoring the load and varying this factor with the
m aterial of the overburden, to modifications of the conventional ‘allowable
stress’ in the extreme fibres of the lining. The overall safety margins against
collapse, taking account of characteristic loads and strengths from statistical
distributions of these factors, are rarely pursued. A m ajor problem, of course,
is the range of scatter associated with the properties of disturbed soil and
rock, and with the properties of the backfill when cut-and-cover methods
of construction are employed. There are also the time-dependent problems
of settlement, subsurface movement and pore-water pressure change, and a
good deal of research is aimed at m onitoring stress changes in the linings due
to these actions. The great variation in m ethods of lining construction -
bolted, expanded, grouted and sp ra y ed -m ak e consistent design rules
difficult to formulate.
This is highlighted by inform ation given in ref. 7.1, which lists the annual
length of tunnels constructed in the UK, divided into forms of lining, for the
calendar years 1970 to 1976. Between 70 and 80% of all tunnels in terms of
length (not volume) were constructed for sewers, and the remainder were
mainly used for the conveyance of water. Only a small percentage of the
total was reserved for road and rail tunnels. The smaller-diameter sewers
and water tunnels were mainly constructed from concrete, bolted and smooth
bore, or expanded. The road tunnels were mainly precast or ‘cast in situ’
concrete. The percentage use of cast iron was very low, and the use of brick
as a prim ary lining had virtually disappeared during the period, except in
conditions of aggressive effluent associated with high scour or fluid velocity.
The extensive use of precast elements has meant that a minimum level
of robustness is needed to prevent damage and breakages during the handling
process. Experience shows that if the segments are strong enough to be
handled, then the complete lining has ample reserves of structural strength
against the pressures applied by the overburden, ground water and surface
loads. In fact, structural design plays a relatively m inor role compared
with the problem s of the damage caused by movement of the ground around
the tunnel as it is driven forward. In these circumstances it is unlikely that
a strict code of structural safety based on limit state principles will ever
find much support from the industry.
7.2 Large water pipelines

The economic development of the water resources of the UK, using large-
diam eter thin-walled steel or glass-reinforced plastic, was examined by the
W ater Resources Board in the early 1970s, and their projections suggested
a total of over 2500 km of pipeline in the range 0.6 to 1.9 m diameter over a
period of 25 years. M ost of this pipeline would be installed by cut-and-cover
@seismicisolation
@seismicisolation
Loads, strength and safety 215
m ethods at a relatively shallow depth of burial, and since two-thirds of the
overall cost of straight pipeline is attributed to the pipe shell itself, there is
clearly a case for reducing the thickness as much as possible. This require
ment led to the review of design and construction of buried thin-walled
pipes discussed earlier in C hapter 4 (ref. 4.7). Four main areas were surveyed:
a critical review of design m ethods; a description of pipe materials and
jointing techniques; procedures to obtain adequate backfill strength;
and relative costs of different m aterials and construction methods.
With regard to loading, the recommended design procedure considered
the following conditions:
1. Total weight of a prism of backfill equal in width to the horizontal dia
meter of the pipe extending up to the surface, plus the self-weight of half
the pipe above the horizontal diameter, less the effect of water outside the
pipe above the level of the centreline which can introduce buoyancy.
2. The pressure difference between that due to the atmosphere at ground
level, and the minimum pressure in the pipe during partial vacuum
conditions.
3. The net difference between the pressure of external ground water and
internal fluid contents.
4. Distributed pressure due to wheel loads on the surface of the ground,
with any impact factors to be independent of cover depth.
The loads were not factored, but were converted to a total external pressure
applied radially to the wall of the pipe. Under this load, the allowable
compressive stress in the pipe wall set at one-third of the stress that would
cause buckling. The proposals thus used a factor of safety against buckling
of 3, on the grounds that the design procedure uses relatively high values of
soil properties, the consistent achievement of which has not been properly
established. An alternative approach is to use a more conservative value for
the elastic m odulus of the backfill soil, with a safety factor of 2 (as proposed
by M eyerhof and Baikie in ref. 4.3).
It is im portant, in choosing the appropriate safety factor, that long-term
changes will not decisively alter any of the loading conditions, the properties
of the soil, or the strength of the thin-walled pipe. Care must also be taken
to ensure that the pipe is not damaged or distorted during backfilling.
The US Federal Highway Administration, following a review of corrugated
pipe culverts, proposed that a soil modulus of 10 x 106 N /m 2 should be
used in design, and this value has been recommended for U K thin-walled
water pipes.
In an elemental cost breakdown of thin-walled pipe systems designed
to the standards described above, it was found that for a 1.82 m diameter
steel pipe, with a cover depth of 0.3 m above the crown, the supply and
protection of the pipe accounted for about 80% of the cost, the excavation
@seismicisolation
@seismicisolation
and backfilling about 16%, and the remaining cost covered jointing and
handling charges. In poor, soft ground, however, where a compact granular
material has to be provided, the backfilling costs can be noticeably increased.
If a granular backfill is not available, and concrete has to be employed,
the cost of the project can double.
7.3 Corrugated steel pipes

The American Iron and Steel Institute provide information on design in their
Handbook o f Steel Drainage and Highway Construction Products [7.2].
In design procedures illustrated by worked examples, they do not factor
loads, but give an allowable compressive stress in the wall equal to half
the ultim ate stress. The ultimate stress takes account of the possibility of
wall buckling, as indicated in Fig. 7.1. W orking in pounds per square inch,
the figure gives three distinct zones: wall crushing, the interaction zone,
and the ring buckling zone. Thus
<7b = cry = 33 000 when D'/R < 294,
crb = 40 0 0 0 - 0 M \ ( D ' / R ) 2 when 294 < D'/R < 500, (7.1)
4.93 x 109
when D'/R > 500.
ffb “ (D’/R)2
In these equations D' is the flexural rigidity of the wall of the pipe. Finally,
</)
a
Figure 7.1 U ltim ate buckling stress for corrugated steel pipe (American Iro n and Steel
Institute).
@seismicisolation
@seismicisolation
the design stress f c is given for pipes by
(7.2a)
(7.2b)
The design rules specify a minimum of 12 inches of cover over most cross
sections, although for pipes installed under a rigid pavement, such as an
airport runway, the minimum is increased to 1.5 ft.
W ith regard to backfill, the density of soil is taken as 120 lb/ft3 at a specified
com paction of 80%. F or other compactions, the density is assumed to vary
according to Fig. 7.2. The design guide also m entions life in corrosive and
abrasive atm ospheres of galvanized corrugated steel pipe culverts, in terms of
‘projected years to perforation’. If the installations are silted, and in standing
water in an acid soil, a typical life is 30-55 years. If the installations are dry,
in alkaline soil, life can be expected to exceed 100 years.
Very large-span corrugated structures have been built in recent years
using ARM CO corrugated sheeting. In order to prevent excessive deflection
during installation and prem ature wall buckling, the minimum cover depths
have had to be increased and linked to the span. Thus, a typical arch-shaped
tunnel with a top radius of 15 ft needs a minimum wall thickness of 0.109 inch
and a minimum cover of 2.5 ft. An arch with a top radius of 23-25 ft needs a
wall thickness of 0.249 inch and a minimum cover of 4 ft. These figures
18
At 85°/o standard
density, k '- 0-86
0-6
65 70 75 80 85 90 95
Soil compaction (%> Standard AASHO T-9 9)
Figure 7.2 R elationship between com paction and density (American Iron and Steel
Institute).
@seismicisolation
@seismicisolation
assume a granular backfill over the crown, compacted to 90% AASHTO
T-180 density.
7.4 Underground blast shelters

Guidelines for the structural design of shallow buried shelters constructed
of reinforced concrete or steel plates were drawn up in the late 1970s by a
working group set up by the Emergency Services Division of the U K Home
Office [7.3]. The group drew on work from the USA on structural dynamics
and on nuclear concrete structures. The dynamic analysis of shelter elements
was approached by assuming that load duration is long in relation to the
period of inelastic response of the elements, and by employing the ‘ductility
ratio’ concept. The ductility ratio /i is given by
permitted deflection
u —--------------------------------- ,
deflection at elastic limit
and for m oderate damage ji = 3.
Further, the total load on the structure can be given in terms of the blast
load F and //, in the form
total load = F [ !— (7.3)

V1 - 1/2^
so that, when ji — 3,
total load = 1.2F. (7.4)
The factor 1.2 can be thought of as a partial load factor yF. N ote that if
m oderate damage were not permissible, and fi — 1, then yF would increase
to 2.0. A dynamic increase factor is applied to the characteristic strength
of the material, so that the yield stress under dynamic conditions is equal to
the static yield multiplied by this factor. It is normally set at 1.10 for steel
in bending, and 1.25 for concrete in compression.
The author attem pted to bring the principles of limit state design into the
design of underground protective structures (Bulson [7.4]). The following
proposals were m ade for partial safety factors.
1. Deviations of loading, y x :
y l = 1.2 for perm anent loads (i.e. pressure due to backfill),
= 1.5 for dynamic surface pressure (i.e. pressure due to explosions
on the surface, or similar dynamic loads).
2. Deviations of structural strength, y2 :
y2 = 1.6 for strength variations due to uneven soil compaction or the
presence of ground water.
@seismicisolation
@seismicisolation
3. Seriousness of collapse, y3 :
y3 = 1.05 for pipelines (no hum an life lost at collapse),
y3 = 1.6 for a larger structures with hum an habitation (possibility of
loss of life at collapse).
F or glass-reinforced plastic construction, it was suggested that y2 should
be increased to 2.0 to take account of possible deterioration in properties
with age. The collapse load factor y is then y xy2y3, which yield the following
overall safety margins for thin-walled underground construction.
1. Large thin-walled cylinders housing civilians:
y = 3.0 (metal),
y = 3.75 .(glass-reinforced plastic).
2. Smaller thin-walled cylinders for pipelines and conduits:
y = 2.0 (metal),
y = 2.5 (glass-reinforced plastic).
The variations in strength due to bad soil com paction and substandard
bedding can be reduced if close attention is paid to emplacement. It is
im portant that thin-walled tubes are properly supported by internal struts
during burial and backfill, to ensure that the shape of the cross section is
maintained. In culvert design and practice in the USA and Britain, it is
norm al to allow some variation in dimensions, and based on experience it is
proposed that, for cylindrical tubes, the total increase in horizontal diameter
after com paction should not exceed 2%. F or square tubes the total increase
in width at the transverse axis of symmetry should not exceed 1.3%.
Particular attention during bedding should be paid to the angular zone
± 45° from the invert. Note, too, that however adequate the precautions to
ensure proper support for square tubular structures, there will always be a
Figure 7.3 O ptim u m soil cover for thin-w alled pipes under surface pressure.
@seismicisolation
@seismicisolation
noticeable downward deflection of the roof due to the local weight of soil.
Thin-walled tubes and containers, carefully bedded and backfilled,
are capable of surviving under high surface loads. The geometry of the
structure has an influence on the optimum depth of burial. There is not much
to be gained by burying circular blast shelters more than one diameter deep
(surface to crown), but square section thin-walled shelters should be buried
at least two side depths in order to m arshall their full load-carrying capacity.
The cost of excavation then becomes an im portant factor when structural
efficiency is assessed (Fig. 7.3).
7.5 Conclusions
Engineers coming new to the subject will have found by reading this book
that the theory and practice of underground structures falls into two divisions.
The lessons of history, stretching back thousands of years, have gradually
been assimilated into the art of thick-walled construction. Although there
have been many worthy attem pts to inject analytical thought into the
design process, civil engineers tend to build on the knowledge gained by
experience. The gradual evolution of the design of road and railway tunnels,
sewers and conduits, illustrate the way that proven, safe methods are handed
down from generation to generation. The margins of safety are not known,
only that there have been no catastrophic failures within memory. The
m ain steps forward have resulted from the development of new materials,
cast iron and concrete instead of brick and masonry, and from innovative
m ethods of construction. These have usually resulted from the insight and
dedication of individuals. We read, for example, of the development of the
‘shield’ m ethod of tunnelling by the elder Brunei (Marc Isambard) in 1818.
Its use in the first tunnel under the River Thames was fraught with difficulties,
and it is said that Brunei and his son virtually lived underground while the
interm ittent work was in progress between 1825 and 1843. Structural
testing is difficult, and the measurements of deflection subsidence and thrust
in the lining are costly to make.
The second division of the subject has quite a different background.
The advent of the thin-walled culvert at the beginning of this century was
accom panied by testing to destruction. F or the first time it was possible to
study the ultim ate loads and collapse behaviour, and university departm ents
in the USA began to take an interest in the science of the subject. Their work
has been given an unforeseen boost by the coming of the nuclear age and the
need for hardened structures to withstand blast and ground shock. Because
this requirem ent was linked to the needs of defence, finance was available
to study the behaviour of lightweight underground installations from an
analytical and experimental standpoint. It was possible to cause catastrophic
collapse by subjecting buried structures to very large field explosions,
@seismicisolation
@seismicisolation
set off by the detonation of nuclear bombs in the late 1950s and early 1960s,
and later by the controlled detonation of very large assemblies of TNT
charges. There was money available, too, for instrum entation and for
detailed analysis of the results. The field began to attract applied m athe
maticians, physicists and com puter scientists with limited experience of the
civil engineering aspects of buried structures. Large blast simulators were
constructed, a num ber of which have been described in Chapter 6. Interest
in the blast protection field has been variable during the middle years of this
century. In the 1940s one solution to the problem of the civilian air-raid
shelter was a thin-walled structure in the form of the Anderson Shelter.
There was a resurgence of interest in the late 1950s and early 1960s, and now,
20 years on, the need for design information on the building of domestic
nuclear shelters has resulted in the initiation of new methods of analysis
and testing.
The divisions are beginning to draw together, too. F or the first time,
new underground railway systems, employing conventional thick-walled
concrete and cast iron elements, have also to serve as shelters, rather as the
London U nderground did in the Second W orld War. The problems of
blast penetration and filtration in a nuclear environment may begin to
enter the world of civil engineering more frequently as time goes by, and
research expertise in this field is limited. No doubt there will be increasing
pressure to specify safety margins more closely than hitherto. Limit state
analysis, now a natural way forward for thin-walled construction, may have
to be applied to the more conventional fields of road and rail tunnels, and
large-diameter water pipes. Collapse of tunnels is rare, and most of the
disasters associated with roof falls and wall ‘caving’ have, of course, occurred
in the unlined galleries of mines, an area of construction outside the scope of
this book. In the very long term the hum an race may turn rather more to
living below ground. Underground shopping precincts are becoming more
common, particularly in countries with a severe climate. We read of schemes
to build global tunnels and the long-awaited Channel Tunnel may yet be
constructed. Meanwhile the pursuit of economy, efficiency and safety in the
design of underground works will continue to attract scientific and engineer
ing interest.
References
7.1 Craig, R .N . and M uir W ood, A .M . (1978) A review o f Tunnel Lining Practice in the
United Kingdom, T ran sp o rt and R oad R esearch L ab., Suppl. Rep. 335.
7.2 A m erican Iro n and Steel Institute (1971) Handbook o f Steel Drainage and Highway
Construction Products. A m erican and Steel Institute.
7.3 Em ergency Services D ivision, H om e Office (1981) Domestic Nuclear Shelters,
Technical Guidance, A H om e Office G uide, H M SO , L ondon.
7.4 Bulson, P.S. (1976) Lightw eight u n derground structures. Struct. Eng., 54 (9), Sept.
@seismicisolation
@seismicisolation
@seismicisolation
@seismicisolation
Author index
Abbott, P.A., 164, 180 Bush, R.Y., 81

Abel, J.F., 131, 144 Butler, B.E., 131
Abrahamson, G.R., 196 Butterfield, R., 190
Agrawal, J.S., 88
Ahmed, S., 87 Carre, G.L., 205
Albritton, G.E., 178, 179 Casagrande, A., 45
Al-Hassani, S.T.S., 212 Cates, W.H., 132
Allen, H.G., 167 Chak, J.S., 81
Allgood, J.R., 12, 36,49, 74, 82, 108, 123, Chaplin, T.K., 166
124, 126, 132, 151, 154, 164, 201 Chelepati, C.V., 32, 108, 132
Allison, C. J., 119 Cheney, J.A., 106, 108, 116, 128, 132
Anand, S.C., 74 Christensen, N.H., 57
Anderson, A.O., 52 Christopherson, D.G., 193
Anzo, Z., 16 Chu, K.H., 214
Atkinson, J.H., 87 Chu, S.L., 132
Auld, H.E., 48 Ciani, J.B., 108
Azar, J.J., 87 Clare, R., 181
Clarke, N.W.B., 4, 55, 61, 63, 66
Bacher, A.E., 87 Coker, E.G., 25
Baikie, L.D., 107, 215 Compston, D.G., 4, 108
Bakht, B., 171 Constantino, C.J., 132, 163, 189
Ballard, J.T., 205 Cooling, L.F., 39
Batjansky, A., 51 Corotis, R.B., 88
Barker, H.A., 131 Craig, R.N., 221
Barnard, R.E., 95
Baron, M.C., 189 Da Deppo, D.A., 132
Bart, R., 50 Dar, S.M., 51
Bartlett, J.V., 4 Davis, R.E., 80, 87
Bates, R.C., 51 Davisson, M.T., 36, 42
Bathe, K., 103 Demmin, J., 172
Beskos, D.E., 212 Doderer, E.S., 146
Bomba, J.G., 87 Donnellan, B.A., 123
Boussinesq, J., 30, 69 Dorris, A.F., 116, 123, 175, 177, 179
Brassow, C.L., 87 Doyle, J.M., 132
Braune, G.M., 77 Duncan, J.M., 147, 151
Britto, A.M., 120 Duns, C.S., 108, 128, 190
Brock, J.S., 22 Duvall, W.E. 18, 20, 23, 32, 41
Brockenbrough, R.L., 131
Brody, O., 131 Egger, W., 173
Brown, C.B., 87, 131 Eggwertz, S., 72
Brown, E.T., 4 Einstein, H.H., 51
Bulson, P.S., 108, 112, 116, 128, 142, 160, Eitan, R., 51
167, 168, 183, 192, 218, 221 Emery, R., 204
Bums, J.Q., 97 English, R.J., 120
@seismicisolation
@seismicisolation
224 A uthor index
Fernandez, P., 172 Kane, H., 49
Filon, L.N.G., 25 Kastner, H., 4
Finn, W.D., 32 Katona, M.G., 100
Fisher, C.L., 132 Keeney, C., 87
Flathau, W.J., 156 Keeser, J.W., 195
Fleck, H., 172 Kennedy, T.E., 171, 211
Florence, A.L., 196 Kloppel, K., 165
Forrestal, M.J., 107, 116 Kovach, T.M., 87
Foster, D.C., 200 Krizek, R.J., 74
Fuehrer, H.R., 195 Kuesel, T.R., 212
Gates, W.E., 159, 162
Gaunt, S., 74 Lambe, G.W., 4
Gellert, M., 51 Larsen, H., 57, 87
Gere, J.M., 105 Layer, J.P., 110
Getzler, S., 51, 167 Lefebvre, G., 154
Ghavami, M., 133 Leonards, G.A., 87
Gilboy, G , 69 Leonhardt, G., 73
Gill, H.L., 11, 12, 128, 154 Leys, I., 212
Gough, H.J., 30 Lindberg, H.E., 51, 196
Green, D.R., 131 Ling, C.B., 26
Green, W.E., 132 Linger, D.A., 172
Greenspan, M , 20 Longhurst, G.R., 133
Gumbel, J.E., 172 Longinow, A., 163, 212
Gunn, M.J., 132 Lupu, L., 167
Luscher, U., 36, 107, 111, 122, 129
Habib, P., 108 Lysmer, J., 211
Hainsworth, I.H., 78
Hall, W.J., 46
Haitiwanger, J.P., 46 McCavour, T.C., 122, 124
Hanley, J.T., 175 McMickle, R.W., 87
Hanna, M.M., 87 McNeil, G., 209
Hardin, B.O., 103 McNulty, J.W , 10
Hashmi, S.J., 212 McQuade, P.V, 79
Ha veil, R.F., 87 Manolis, G .D , 212
Heller, S.R., 23 Marino, R.L, 158
Hendron, A.J., 36, 42, 48 Mark, R , 131
Hennessey, R.L., 70 Marston, A , 52, 53, 68
Hermann, H.G., 82 Mason, H .G , 12
Herrmann, G., 107, 116 Matsuzaki, K , 132
Herrmann, L.R, 101 Megaw, T.M , 4
Hess, J.D., 81 Merritt, J.L, 211
Hoeg, K., 87, 103, 107, 123, 125 Meyer, G .D , 156
Hoek, E , 4 Meyerhof, G.G., 94, 107, 123, 124, 129, 132,
Hoff, G.C., 200 152, 172, 215
Howard, A.K., 78, 96, 120, 132 Mindlin, R.D, 16, 18
Howland, R.C.J, 25 Moser, A.P, 133
Huff, P.H., 212 Muir Wood, A.M., 221
Hwang, R.N., 212 Munn, J.F , 155
Murtha, R.N., 51
James, R.G., 87
Janson, J.E., 79 Nasir, G.A, 131
Janssen, H.A, 8, 53, 58 Nath, P , 87
Jeffery, G.B., 18 Nayak, G .C, 74
Jester, G.E., 14 Newmark, N.M., 46, 47, 69, 85, 191
Johnson, A.M., 81 Nielson, F.D., 132, 148
Jungbluth, O., 165 Norman, D.C, 82
@seismicisolation
@seismicisolation
Author index 225
Obermuller, J.C., 87 Shenkman, S., 132
Obert, L., 18, 20, 23, 32, 41 Shrock, B.J., 132
Odelio, R.S., 171 Sinnamon, G.K., 173
Olson, R.E., 43, 45 Skempton, A.W., 39
Sonntag, G., 108, 132, 151, 172
Page, J., 81, 87 Southwell, R.V, 30
Parmalee, R.A., 88 Spang, J , 172
Parnes, R., 189 Spangler, M .G, 52, 61, 63, 68, 70, 71, 74, 75,
Parola, J.F., 43 91, 96, 124
Pawsey, S., 131 Steinbrenner, W , 70
Pearson, F.H., 88 Sternberg, E , 30
Peterson, F.E., 129 Szechy, K , 4
Pettibone, H.C., 78, 120
Phong, L.M., 108 Takahashi, S.K, 151, 159, 162
Poncelet, V., 209 Terzaghi, K„ 4, 6, 7, 9, 12, 14, 30, 37
Potts, D.M., 87 Thomopoulos, N .T, 212
Prendergast, J.P., 82 Timoshenko, S.P., 105
Pruska, M.L., 72 Trott, S.S., 74
Purdie, A.C., 197 True, D .G , 12
Rajasekar, H.L., 132 Valentine, H.E, 132

Randolph, V.G., 87 Valliappan, S, 132
Rankine, 14, 53 Valsangkar, A.J, 120, 132
Rayleigh, Lord, 189 Van Horn, A.D, 10, 12
Richard, R.M., 97 Vey, E„ 132
Richards, R., 88, 170 Voellmy, A , 71
Richardson, A.M., 45
Richart, F.E., 30 Walker, R.E., 179
Riley, W.F., 158 Wang, C.T., 28
Robinson, A.R., 211 Wastlund, G., 72
Robinson, R.R., 103 Watkins, R.K., 96, 110, 132
Rowe, R.D., 212 Welch, R.E., 132
Rude, L.C., 88 Wetzorke, M., 57
Whiffen, A.C., 210
Sadowsky, M.A., 30 White, H.L., 110, 123
Salmon, M.A., 211 Whitman, R.V., 4, 36, 45, 164
Savin, G.N., 23 Wiedermann, A.H., 189
Schlick, W.J., 55, 65 Williamson, R.A., 212
Schmidt, H., 15 Wilson, E.L., 129
Schofield, A.N., 120 Wilson, J., 172
Schwartz, C.W., 51 Wong, K.S., 151
Schwinn, K.H., 132
Selig, E.T., 143, 148 Yamaguti, S., 16
Sevin, E., 132 Yoshihara, T., 189
Shannon, W.L., 45 Young, O.C., 74
@seismicisolation
@seismicisolation
Subject index
Arches Air Force Weapons Laboratory,

thick-walled 156-158 180-181
thin-walled 152-156 Illinois University, 173-176
Arching 6-15 Waterways Experiment Station, 176-180
Backpacking, 64, 200-202 Elliptic sections, 143-147

Bedding Experimental studies, thin-walled cylinders,
class D, 75 114-121
concrete cradle, 74
first class, 75 Finite element analysis, thick-walled
Burster slab, 206-210 conduit, 74
Frictionless interface, 107
Capsules, 162-164 Full scale tests, thick-walled conduits, 75-81
Concentrated surface load, 29-30, 68-71
Conduit bedding, 74-75 Incremental elastic soil model, 102
Conduits
ditch condition, 53-57 Large water pipes, 214-216
full scale tests, 75-81 Linear elastic soil model, 47, 101
incomplete ditch condition, 59-61 Localized explosive loads, 193-195
model tests, 82-85
positive projecting, 57-61 Military shelters, 202-206
Corrugated steel pipe Moments in tunnel linings, 85
allowable stress, 216-218 Multiple pipes, 65-67
design rules, 121-123
Cover depth, 110-113 Negative projecting conduits, 61-65
Cylinders
closed, thick-walled, 158-160 Partial safety factors, 218-219
closed, thin-walled, 160-162 Penetration formulae, 206
elastic buckling, 104-109 Pipe arches, 147-151
ultimate strength, 109-114 Properties of soils (static)
cohesive soils, 39-40
Deflection analysis granular soils, 36-39
recent developments, 97-103 rock, 41
traditional, 90-97
Design rules
CIRIA, 126-127 Rail tunnels, 213-214
military code (UK), 127-129 Road tunnels, 213-214
US Naval Civil Engineering Laboratory,
124-126 Spherical shells, 168-170
Domes, 164-168 Square sections (thin-walled), 134-143
Domestic nuclear shelters, 198-200 Stress distribution
Dynamic analysis, 189-193 circular tunnel, 15-19
Dynamic soil properties, 41-45 non-circular tunnel, 20-25
Dynamic test rigs spherical cavity, 30-32
@seismicisolation
@seismicisolation
Subject index
Submerged pipes, 67-68 Underground blast shelters, 218-220
Variable modulus soil model, 103
Thick cylinder theory, 27-28
Thin-walled pipes Wave propagation in soils, 45-49
buckling analysis 104-109 Wide trenches, 65-67
static loads, 89
Two circular tunnels, 25-26 Yielding trapdoor, 32-35
@seismicisolation
@seismicisolation

Buried Structures Static and Dynamic Strength Bulson 1984

Uploaded by

Document Information

Copyright

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Buried Structures Static and Dynamic Strength Bulson 1984

Uploaded by

Copyright:

@seismicisolation

Taylor & Francis

All rights reserved. No part of this book may be reprinted,

British Library Cataloguing in Publication Data

Library of Congress Cataloging in Publication Data

Bulson, P.S., 1925-

3 Thick-walled pipes under static loads 52

4 Thin-walled pipes under static loads 89

6 Structures under dynamic loads 173

7 Loads, strength and safety 213

As a schoolboy I frequently journeyed to the Dorset coast through the road

P ro p erty SI U nits f.p.s. U nits

L ength 1.0 m m 0.0394 in.

1.1 Early history

1.2 Contemporary structures

1.3 The future

1.4 Purpose of the book

Before specific types of structures are considered, it will be useful to devote

Arch ing, A = 1- p. /p^

Stiffness -geom etry factor, R

* = 2 [ 1 - exp ( - 4> f ) ] + q e x p ( - kt tan * f ) . (2.13)

2.2 Elastic stress distribution

2.2.1 Stress distribution around a circular cavity

Figure 2.8 Axes of reference for M indlin’s analysis.

Table 2.1 Three states of stress th at can exist in a semi-infinite solid

State A Each elem ent is in a state of isotropic pressure due to the

State C C om plete absence of lateral stress, so th a t a x = 0

Figure 2.10 Values of oJay and oJ<Jy (O bert and Duvall).

where r 0 is the hole radius, and r is distance from hole centre.

Figure 2.12 V ariation of a0 with depth of burial.

2.2.2. S tr e ss d istrib u tio n a ro u n d n o n -circu la r c a v itie s

A solution for the stress distribution in a uniformly loaded elastic plate

= K + a )(p2 sin2 ft + q2 cos2 fi - 9r2) - r (p + q)2 sin 20

2.2.3 Two circular cavities in close proximity

2.2.4 Thick-cylinder theory

When r = rQ, or = 0, and

Figure 2.23 C oncentrated line load on the b o u n d ary of a semi-infinite plate.

2.2.5 Concentrated free surface loads

so that or is inversely proportional to radius.

a = — —(ft — sin /?),

2.2.6 Stress distribution around a spherical cavity

Figure 2.25 Spherical cavity notation.

so that, when 6 = 0, r = r0 (the top of the cavity),

and when 6 = tc/2, r = r0 (the equatorial plane),

By superimposing the results, the concentrations for biaxial and triaxial

2.2.7 Stress distribution above a yielding trapdoor

considering a purely elastic medium, there will be no ‘plane of equal settle­

2.3 Properties of soils

2.3.1 Static properties o f granular soils

Figure 2.31 Typical curve from confined com pression test.

where v is Poisson’s ratio. N ote that when v = f , E = § M s .

s = c + a tan </>, (2.38)

s = c + (o —w)tan </>, (2.39)

Decrease possible for soft soils

the excessive pore pressure reduced by consolidation during loading.

Table 2.2 Typical characteristics of sand

Unified Soil Classification G ro u p Symbol SP

Particle size distribution

2.3.2 Static properties o f cohesive soils

Table 2.3 Typical characteristics of clay

Unified soil C lassification G ro u p Symbol C H

C om position characteristics (BS 1377: 1967, test 12)

Unified Soil Classification G ro u p Symbol CL

C om position characteristics (BS 1377: 1967, test 12)

Table 2.5 Typical characteristics of rocks

Static elastic m odulus max 11.0 x 106 psi

Static elastic m odulus max. 7.3 x 106 psi

considering a purely elastic medium, there will be no ‘plane of equal settle

F or a further discussion of wave propagation in soils and a review of elemen