This is a repository copy of Closed-form solutions for large strain analysis of cavity

contraction in a bounded Mohr-Coulomb medium.

White Rose Research Online URL for this paper:

https://eprints.whiterose.ac.uk/173543/

Version: Accepted Version

Article:
Song, X-G, Yang, H, Yue, H-Y et al. (3 more authors) (2022) Closed-form solutions for
large strain analysis of cavity contraction in a bounded Mohr-Coulomb medium. European
Journal of Environmental and Civil Engineering, 26 (10). pp. 4548-4575. ISSN 1964-8189

https://doi.org/10.1080/19648189.2020.1857309

© 2020 Informa UK Limited, trading as Taylor & Francis Group. This is an author produced
version of a journal article published in European Journal of Environmental and Civil
Engineering. Uploaded in accordance with the publisher's self-archiving policy.

Reuse
Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless
indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by
national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of
the full text version. This is indicated by the licence information on the White Rose Research Online record
for the item.

Takedown
If you consider content in White Rose Research Online to be in breach of UK law, please notify us by
emailing [email protected] including the URL of the record and the reason for the withdrawal request.

[email protected]
https://eprints.whiterose.ac.uk/
 Closed-form solutions for large strain analysis of cavity
contraction in a bounded Mohr-Coulomb medium
Xiu-Guang Song 1 a, He Yang1 b, Hong-Ya Yue1,3 c,

Xu Guo2 d, Hai-Sui Yu3 e, Pei-Zhi Zhuang*1,3

1
School of Qilu Transportation, Shandong University, Jinan, 250002, China
2
Geotechnical and Structural Engineering Research Center, Shandong University, Jinan, 250061,
China
3
School of Civil Engineering, University of Leeds, LS2 9JT Leeds, UK

Corresponding author, Research fellow, E-mail: [email protected]
a
Professor, E-mail: [email protected]
b
Master student, E-mail: [email protected]
c
Ph.D. candidate, E-mail: [email protected]
d
Professor, E-mail: [email protected]
e
Professor, E-mail: [email protected]

Abstract: This paper presents rigorous analytical solutions for cavity contraction analysis of a
thick-walled cylinder/sphere after an arbitrary magnitude of expansion. Closed-form solutions are
given for the distribution of stress and displacement within the cylinder/sphere of soil that is
subjected to constant external pressure and monotonically decreasing internal pressure. The soil is
modelled as an elastic-perfectly plastic material obeying the Mohr-Coulomb yield criterion and a
non-associated flow rule. Large strain effects are taken into account by adopting the logarithmic
strain definition in the plastic deformation analysis. The new solutions are validated with published
results at first, then parametric studies are carried out. It is shown that the reference stress state (e.g.
in-situ, elastic, partially plastic and fully plastic) and the cavity geometry ratio may greatly affect
the unloading behaviour, in particular, when the cavity geometry ratio is smaller than a limit value.
Finally, three typical applications of the solutions are demonstrated, including (i) design of the
thickness of frozen cylinder walls accounting for large deformation effects, (ii) interpretation of
laboratory pressuremeter tests with consideration of effect of the constant stress boundary, and (iii)
shakedown analysis of a soil cylinder/sphere considering its geometry changes upon cyclic loading
and unloading.

Keywords: Cavity contraction; Boundary effect; Shakedown; Pressuremeter; Frozen earth wall

1
 1 Introduction

2 Cavity contraction theory is concerned with the theoretical study of changes in stresses and
3 displacements of cylindrical and spherical cavities upon unloading. It provides a versatile and
4 accurate tool for study of a variety of geotechnical problems. Among them are the interpretation
5 of pressuremeter tests (Ferreira & Robertson, 1992; Houlsby et al., 1985; Houlsby & Withers,
6 1988; Jefferies, 1988; Schnaid et al., 2000; Shuttle, 2007; Withers et al., 1989; Yu, 1996) and
7 analysis of stability and deformation associated with underground excavation, tunneling and
8 drilling (Brown et al., 1983; Carter, 1988; Chen et al., 2012; Jirari et al., 2020; Mair & Taylor,
9 1993; Mo & Yu, 2017; Ogawa & Lo, 1987; Vrakas & Anagnostou, 2014; Yu & Rowe, 1999).
10 In the modelling of pressuremeter tests during unloading, the cavity contraction analysis
11 normally starts from a residual (elastic-plastic) stress state that was induced by previous loading
12 (e.g. installation and expansion of pressuremeters). However, it is usually assumed that the soil
13 is unloaded from an in-situ elastic stress state in the stability and deformation analysis of tunnels
14 and wellbores. Hence, corresponding cavity contraction analyses involve different complexity.
15 Both cases will be considered in this study.

16 Cavity contraction approaches for modelling pressuremeter tests were advocated mainly
17 because the unloading response of pressuremeters is less sensitive to the initial soil disturbance
18 (Hughes & Robertson, 1985; Schnaid & Houlsby, 1992). Over the years, a number of solutions
19 has been developed to derive soil properties from the unloading portion of pressuremeter curves
20 for both sand (Houlsby et al., 1985; Schnaid et al., 2000; Withers et al., 1989; Yu, 1996) and
21 clay (Ferreira & Robertson, 1992; Houlsby & Withers, 1988; Jefferies, 1988). For example,
22 assuming the cavity unloading from the limit expansion state, Houlsby and Withers (1988)
23 derived an analytical solution for both cylindrical and spherical cavities in clays obeying the
24 Tresca failure criterion. Using a non-associated Mohr-Coulomb plasticity model, Houlsby et al.
25 (1985) first developed an approximate small strain solution for interpreting the unloading
26 portion of pressuremeter curves in sands, which was extended later by Withers et al. (1989) to
27 include the case of a spherical cavity. Later on, Yu and Houlsby (1995) presented a more
28 rigorous large-strain solution for the analysis of unloading from any elastic-plastic stress state
29 adopting the same soil model. These unloading solutions are also of great importance for the
30 interpretation methods that consider both the loading and unloading portions of pressuremeter
31 tests (Jefferies, 1988; Schnaid et al., 2000). Nevertheless, almost without exception, the
32 previous analytical solutions for elastic-plastic contraction analysis have been developed with
33 the idealization that the surrounding soil is infinitely large. This assumption may approximately
34 represent the field conditions of site pressuremeter tests but is not suitable for tests performed
35 in small-sized containers due to the possible lateral boundary effects (Alsiny et al., 1992; Fahey

1
36 & Carter, 1993; Geng et al., 2012; Jewell et al., 1980; Schnaid & Houlsby, 1991). To capture
37 the boundary effects, many analytical/semi-analytical expansion solutions for cavities within a
38 bounded soil mass have been proposed (Cheng & Yang, 2019; Fahey, 1986; Juran & BenSaid,
39 1987; Pournaghiazar et al., 2013; Salgado et al., 1998; Yu, 1992, 1993). However, the progress
40 in developing counterpart contraction solutions in bounded soils lags much behind due to the
41 presence of residual stresses that makes the mathematics of the unloading analysis more
42 complex than that of loading. Existing elastic-plastic loading-unloading studies into this
43 problem mainly focused on the shakedown behaviour of a thick-walled cylinder or sphere,
44 which usually involves elastic unloading only and is lack of consideration for the deformation
45 (Gao et al., 2015; Hill, 1950; Wen et al., 2017; Xu & Yu, 2005; Zhao & Wang, 2010).

46 Cavity contraction analysis from an in situ stress state can be regarded as a reverse process
47 of traditional cavity expansion analysis (Chadwick, 1959; Collins & Yu, 1996; Yu & Rowe,
48 1999). As such, the solution methods between them are transferable. Meanwhile, the elastic
49 initial stress state is relatively simple. Hence, many relevant analytical/semi-analytical
50 contraction solutions have been developed in this case over the years (Brown et al., 1983; Chen
51 & Abousleiman, 2016; Mo & Yu, 2017; Park, 2014; Sharan, 2008; Vrakas & Anagnostou, 2014;
52 Yu & Rowe, 1999; Yu et al., 2019). Likewise, most of them concentrated on the case of a cavity
53 embedded in an infinite soil mass, which represents a reasonable simplification for the problem
54 of deep tunnels and un-reinforced boreholes. However, this is not suitable for the unloading
55 analysis of thick-walled soil cylinders or shallow tunnels (Abdulhadi et al., 2011; Franza et al.,
56 2019; Grant, 1998; Mair, 1979), for example, in the stability and deformation analysis of
57 controlled ground freezing involved tunnels and boreholes, in which the finite thickness of the
58 frozen earth wall must be well accounted for (Andersland & Ladanyi, 2004; Sanger & Sayles,
59 1979; Zhang et al., 2018).

60 In this paper, we present analytical large strain solutions for contraction analysis of a thick-
61 wall cylinder/sphere of dilatant elastic-plastic soils using the Mohr-Coulomb yield criterion and
62 a non-associated flow rule. Without loss of generality, an arbitrary residual stress state
63 (including in-situ, elastic, partially plastic and fully plastic) induced by loading prior to the
64 unloading is considered. The solutions at first are compared with other solutions in the special
65 case of an infinite soil mass for validation. This is followed by parametric studies with a focus
66 on the effects of soil thickness and loading history on cavity contraction behaviour. Finally,
67 three typical applications of the new solutions are presented to show their usefulness, including
68 (i) preliminary design of the thickness of frozen cylinder walls, (ii) prediction of pressuremeter
69 curves measured in calibration chambers in sand, and (iii) determination of the optimal
70 thickness of a hollow cylinder/sphere based on the shakedown concept considering the large

2
 71 deformation effects.

72 Problem definition and reference stress state

73 Problem definition

74 Initially, the inner and outer radii of the soil cylinder/sphere are a0 and b0 , respectively, and
75 a hydrostatic pressure p0 acts throughout the soil which is assumed to be isotropic and
76 homogeneous. An additional radial pressure pin  p0 (  0 ) is then applied at the inner wall of
77 the cavity and increased gradually (i.e. loading). At the end of the loading process (i.e.
78 pin  p20 ), the inner and outer radii of the cylinder/sphere are a20 and b20 , respectively.
79 Subsequently, the radial pressure acting on the inner cavity wall reduces monotonically (i.e.
80 unloading). During the loading and unloading processes, the internal cavity pressure is applied
81 or removed sufficiently slowly, thus the dynamic effects are negligible, and the radial confining
82 pressure at the outer wall of the cavity remains unchanged as p0 . The major concern of this
83 paper is the distribution of stress and displacement in the cylinder/sphere of soil during the
84 unloading process.

85 The unloading analyses of cylindrical and spherical cavities are conducted simultaneously
86 by the introduction of a parameter k which takes 1 for a cylindrical cavity and 2 for a spherical
87 cavity. For convenience, the behaviour of the cylindrical cavity is described in terms of
88 cylindrical polar coordinates (r, θ, z) and the behaviour of the spherical cavity is described in
89 terms of spherical polar coordinates (r, θ, ϕ). As a long cylindrical cavity is considered, its
90 expansion and contraction occur under plane strain conditions with respect to the z-direction of
91 the cylindrical coordinates.

92 Under axisymmetric/spherically-symmetric conditions, the equilibrium equation in the radial

93 direction can be expressed as:

d r k
94   r      0 (1)
dr r

95 where σr and σθ represent the radial and circumferential stresses, respectively.

96 The configuration of the system at the completion of loading is used as a reference state from
97 which the stress and displacement of the unloading process are measured. For clarity, subscripts
98 ‘0’, ‘1’ and ‘2’ are used in this paper to distinguish the in-situ (or initial) state, the loading
99 process and the unloading process (e.g. Figure 1). The residual stresses and radial displacement
100 during unloading are expressed as:

3
101  r 2   r 20   r (2)

102   2    20    (3)

103 u2  u20  u =r2  r0 (4)

104 where  r 2 ,   2 and u2 represent radial stress, circumferential stress, and radial displacement

105 during unloading, and  r 20 ,   20 and u 20 represent their values at the end moment of loading

106 (or beginning of unloading), respectively.  r ,   and u are changes in the radial stress,

107 the circumferential stress and the radial displacement due to unloading, respectively; r2 is the

108 radial coordinate of a soil particle during unloading, and r0 indicates its original location.

109 Taking tensile as positive, the stress boundary conditions during unloading are defined as:

110  r 2 r a   p2 ,  r 2 r b   p0 (5 a,b)
2 2 2 2

111 The surrounding soil is modelled as a linearly elastic-perfectly plastic material obeying the
112 Mohr-Coulomb criterion and a non-associated flow rule. The soil behaves elastically and obeys
113 Hooke's law until the onset of yielding. Depending on the loading history, possible reference
114 stress states within the finite soil medium at the beginning of unloading can be generally divided
115 into three cases as shown in Figure 1, namely (I) purely elastic state (including the case of
116 p20  p0 ); (II) partially plastic state; and (III) fully plastic state. In Figure 1, c1 denotes the
117 radius of the elastic-plastic boundary during the loading phase, c20 represents its value at the
118 end of loading (or the beginning of unloading), and c2 denotes the outer radius of the loading-
119 induced plastic zone during unloading. Note that the axial stress σz in the cylindrical case is
120 assumed to be the intermediate principal stress for the sake of analytical solutions, and it has
121 been shown that the errors that may be caused in the associated stress and displacement fields
122 by this simplification are negligible for practical purpose (Reed, 1986; Vrakas & Anagnostou,
123 2014).

4
 p0 p0 p0
b0 b2 b2
p2 d2
p20 p2
I
a0 a2 a2

Partially plastic
(a1,b1,c1; p1)
expansion
p0 p0
p0 b20 b2 b2
p20 p2 d2 p2
II c20 c2
a20 a2 a2

Fully plastic
(a1,b1; p1)
expansion
p0 p0
p0 b20 b2 b2
p20 p2 p2
d2
III
a20 a2 a2

Initial state of Partially plastic Fully plastic

Elastic unloading
unloading unloading unloading

Figure1. Definition of the loading and unloading processes

124 As the cavity pressure p2 reduces from p20, the surrounding soil contracts purely elastically
125 at first (i.e. the elastic unloading phase in Figure 1). With further removal of the internal
126 pressure, plastic yielding occurs in the reverse direction (referred to as ‘reverse yielding’) once
127 the residual stresses satisfy the Mohr-Coulomb yielding criterion under unloading, which
128 initiates from the inner wall of the cavity (i.e. the partially plastic unloading phase in Figure 1).
129 The outer radius of the reverse plastic zone is denoted as d2, and its corresponding position at
130 the fully loaded state (i.e. the reference state) is denoted as d20. Upon further unloading, the
131 entire soil annulus or spherical shell may enter the plastic state (i.e. fully plastic unloading phase
132 in Figure 1). In the reverse plastic zone, the circumferential stress becomes the major principle
133 stress and the yielding function can be expressed as:

134  r 2    2  Y (6)

135 where  = 1  sin   1  sin   ; Y  2 c cos  1  sin   .  is the angle of friction and c is

136 the cohesion of soil. It has been proven that Equation (6) is satisfied throughout the whole
137 unloading process for any soil (Vrakas & Anagnostou, 2014; Yu & Houlsby, 1995).

5
138 Reference stress state for unloading

139 The soil model adopted in this study is the same as that was used by Yu (1992, 1993) in the
140 loading analysis of a cylinder/sphere. At first, following Yu (1992, 1993), the distribution of
141 stress in the soil during loading is presented to provide a reference for the analysis of the
142 subsequent unloading process.

143 Upon uniform and monotonic loading, the surrounding soil behaves elastically before the
144 cavity pressure reaches the elastic limit p1elim , and the distribution of the elastic stresses is
145 known as:

 b1 r1   1
k 1

146  r1   p0  ( p1  p0 ) (7)
 b1 a1   1
k 1

 b1 r1 k 1  k  1
147   1   p0  ( p1  p0 )  
(8)
 b1 a1   1
k 1

Y    1 p0   b1 a1   1
k 1

148 p1elim  p0    (9)

 / k  1 b1 a1 
k 1
  1

149 When the cavity pressure increases to be larger than p1elim , the distribution of stresses is
150 defined as:

k  1
151  r1  Y   1  A1  c1 r1   (10)

k  1
152   1  Y   1   A1   c1 r1   (11)

153 for a1≤r1≤c1 (i.e. the loading plastic zone), and

 r1   p0  B1  b1 r1   1
k 1
154 (12)
 

  1   p0  B1  b1 r1  k  1
k 1
155 (13)
 

156 for c1≤r1≤b1 (i.e. the loading elastic zone). In which,

  k  1 Y    1 p0 
157 A1  (14)
k   1  k  1    1 c1 b1  
k 1
 

6
 Y    1 p0
158 B1  (15)
 k  1 b1 c1 
k 1
  1

159 The radius of the elastic-plastic boundary upon loading (i.e. c1 ) can be determined by:

k  1  k   1 c1 b1 k 1    k  Y    1 p1 
  
160  c1 a1    (16)
  k  1 Y    1 p0 

161 With sufficiently large loading, the entire soil mass of a finite radial extent may become
162 fully plastic (i.e. c1  b1 ). The distribution of stress in this phase can also be determined by
163 Equations (5 a,b), (10) and (11) by replacing c1 with b1 therein.

164 At the beginning of unloading, a1  a20 , b1  b20 and p1  p20 . According to the difference
165 in the residual stress state and the corresponding geotechnical applications, solutions for the
166 unloading analysis from an in-situ (or elastic) stress state and a plastic reference stress state will
167 be derived separately in the following two sections for clarity.

168 Solutions for unloading from an in-situ/elastic stress state

169 This section focuses on the analysis of a cavity unloading from an elastic stress state (namely,
170 case I in Figure 1, p0  p20  p1elim ). In this case, the unloading from p20 back to p0 is a

171 reverse process of the previous elastic loading (fully recoverable). Thus, the distribution of
172 stress and strain during this process can be readily obtained by the corresponding loading
173 solution (e.g. (Yu, 1992, 1993)). Hence, the following analysis is carried out with reference to
174 an in-situ stress state for brevity.

175 Elastic unloading analysis

176 In the elastic unloading analysis, the small strain definition (e.g. Equations (17) and (18)) is
177 adopted as the elastic deformation is rather small (Houlsby & Withers, 1988; Yu & Houlsby,
178 1995). This is commonly used in quasi-static cavity expansion and contraction analyses and
179 consistent with the counterpart expansion analysis of Yu (1992, 1993). Hence, the elastic stress-
180 strain relationships in rate forms are:

du2 1  kv 
181 r     r 2    2  (17)
dr2 M  1 v2  k  

u2 1  v 
182  =    r  1  v  kv    (18)
r2 M 1  v  2  k  

7
183 where εr and εθ denote radial and circumferential strains, respectively. v is the Poisson’s ratio.
184 M  E 1  v 2  2  k  ; E is the elastic modulus of soil, and E=2G(1+v); G is the shear

185 modulus of soil.

186 Changes in the stresses and radial displacement due to purely elastic unloading from p0 to

187 p2 can be determined by solving the stress equilibrium equation (i.e. Equation (1)) and the
188 strain compatibility equations (i.e. Equations (17) and (18)) with the stress boundary conditions
189 defined in Equation (5 a,b) as:

 b2 r2  k 1 1
(19)
190  r 2   p0  ( p2  p0 )
 b2 a2  k 1 1

 b2 r2  k 1 k  1 (20)
191   2   p0  ( p2  p0 )
 b2 a2  k 1 1

 p2  p0    r k 1 1  v   r k 1  
192 u2  r2   k 1 
1  v  kv  
 2
   1  2   (21)
Mr2 1 a2  1 b2  
k k 1
 2 
b k  1  v  2  k    b2   
  


193 Equations (19) and (20) show that σr2 increases and σθ2 decreases upon unloading. Reverse
194 plasticity occurs once the Mohr-Coulomb yield criterion Equation (6) is satisfied. This
195 condition is firstly satisfied at the inner wall of the cavity when the internal pressure reaches its
196 reverse elastic limit p2e lim 1 :

Y    1 p0   b2 a2   1
k 1

197 p2elim 1  p0    (22)

  1 / k  b2 a2 
k 1
1

198 Elastic-plastic unloading analysis

199 Stress analysis

200 As the cavity pressure p2 further reduces, a plastic unloading zone a2≤r2≤d2 forms and spreads
201 outwards from the inner wall of the cavity. The distribution of stress in the soil can be obtained
202 by considering the plastic zone and elastic zone (i.e. d2≤r2≤b2) separately and matching at the
203 elastic-plastic interface r2=d2.

204 In the elastic zone (i.e. d2≤r2≤b2), the unloading-induced changes in the principal stresses
205 can be obtained by solving the stress equilibrium equation (i.e. Equation (1)) and the strain
206 compatibility equations (i.e. Equations (17) and (18)) as:

8
  r 2   p0  B21  b2 r2   1
k 1
207 (23)
 

  2   p0  B21  b2 r2  k  1
k 1
208 (24)
 

209 where the integration constant B2-1 can be determined as the yield criterion of Equation (6) is
210 satisfied at r2=d2, namely

Y    1 p0
211 B21  (25)
  1 k  b2 d2 
k 1
1

212 In the reverse plastic zone (a2≤r2≤d2), the principal stresses are determined by jointly solving
213 the stress equilibrium equation and the Mohr-Coulomb yield criterion as:

k 1 
214  r 2  Y   1 +A21  d2 r2  (26)

k 1 
215  2  Y   1 + A21  d2 r2  (27)

216 where the integration constant A2-1 can be determined from the continuity condition of stress
217 components at the outer radius of the elastic-plastic interface (i.e. r2=d2) as:

  k  1 Y    1 p0 
218 A21  (28)
k   1   1 k  1    d 2 b2  
k 1
 

219 Then the cavity pressure p 2 and the radius of the elastic-plastic interface d2 during
220 unloading can be related based on Equations (5 a), (26) and (28) as:

k 1 
k Y    1 p2    d2  
k 1
 d2  1
221       1       (29)
 a2   k  1 Y    1 p0   k  b2  

222 While p 2 reduces to the fully plastic limit value p2fp lim , the entire soil medium will

223 become plastic. Equation (29) gives:

Y    1 p0 
 b a     1
k 1
224 p2fp lim  p0  (30)
 1  2 2

225 Displacement analysis

226 In the elastic unloading zone (d2≤r2≤b2), the radial displacement at r2 can be obtained by

227 integrating Equations (17) and (18) with the inputs of Equations (23) and (24) as:

9
 B r 
 v  v
k 1
1  v  kv   b2   
228 u2  r2   21 2 1  v  kv       (31)
M  1  v  2  k  1  v  2  k  k   r2  
 

229 At the elastic-plastic interface and the outer wall of the cavity, we have:

230 d20  d2  u2  d2  , b20  b2  u2  b2  (32a, b)

231 In the displacement analysis within the plastic unloading zone (a2<r2<d2), a non-associated
232 flow rule (i.e. ψ is not necessarily equal to  ; ψ is the dilation angle of soil) is adapted, and ψ
233 is assumed to be constant to limit the complexity of the model so that closed-form solutions
234 can be obtained. The non-associated flow rule is expressed as:

 rp  r   re
235  =  k (33)
p   e

236 where  = 1  sin  1  sin  . Note that the dilatancy of soils is in fact not constant (e.g.

237 tends to zero at critical state). Hence, the above assumption on the dilation angle may lead to
238 overprediction on the volumetric deformation at large deformation. The superscripts ‘e’ and ‘p’
239 represent the elastic and plastic components of strain, respectively. The distribution of stress
240 and strain in the soil at the initiation of reverse plastic yielding is known with Equations (17)-
241 (20) by putting p2  p2elim 1 . Subject to this initial condition, the total stress-strain relation in
242 the reverse plastic zone is obtained by integrating Equations (17), (18) and (33) as:

1 
 kv   kv  

243  r 2  k  2  1    r 2  p0    k  1  v  kv      2  p0 
M   1  v  2  k  
  1  v  2  k   

244 (34)

245 The definition of logarithmic strain is adopted to account for the effects of large strain in the
246 axisymmetric plastic deformation analysis (Chadwick, 1959; Yu & Houlsby, 1995), namely:

247  r 2  ln  dr2 dr0  ,  2  ln  r2 r0  (35 a,b)

248 Substituting Equations (10), (11), (26), (27) and (35 a,b) into Eq. (34) leads to:

r2k  dr2 k 1 

249 ln  ln    d 2 r2  (36)
r0k  dr0

250 where

Y    1 p0   kv(   1) 
251 ln   1   k  1  v  kv   (37)
  1 M  1  v  2  k  

10
  k  1 Y    1 p0   kv(    ) 
252 = 1  k 1  v  kv    (38)
kM   1   1 k  1    d 2 b2 
k 1
  1  v  2  k  
 

253 With the aid of the transformation variable  in Equation (39),

k 1 
254  (r2 )=  d2 r2  (39)

255 the integration of Equation (36) over [r2, d2] leads to:

k  1 k  1
  d2    d  r  
  (40)
   1
256  e d   20   0  
r2    d 2   d2  


257 where  =  k   1 k   1 .

258 Then putting r0=a0 and r2=a2, Equation (40) can be solved with the aid of series expansion

of e (i.e. e   n 1 n / n!, n ! represents the factorial of n ) as:


259

k  1 k  1

n k 1  n     d20   a0  
260 
n  0 n! n   


1   d2 a2    
   d 2 
   (41)
  d 2  

261 It is worth noting that the above displacement analysis can be significantly simplified by
262 ignoring the contribution of elastic strain and/or using the small strain definition within the
263 plastic zone. For example, the right-hand side of Equation (34) will become zero under the
264 former assumption (Vrakas & Anagnostou, 2014; Yu & Houlsby, 1995; Yu & Rowe, 1999).
265 However, to avoid possible errors that are accompanied by these simplifications (Vrakas &
266 Anagnostou, 2014), they are not attempted in this study.

267 Fully plastic unloading analysis

268 When the surrounding soil enters the fully plastic unlading phase (i.e. d2=b2), it is found that
269 the stress distribution can be obtained directly by replacing d2 with b2 from Equations (26), (27)
270 and (28) as:

k 1 
Y Y    1 p0  b2 
271  r2     (42)
 1  1  r2 

k 1 
 Y    1 p0   b2 
 
Y
272  2    (43)
 1  1  r2 

273 A large-strain displacement solution for a fully plastic soil cylinder/sphere can be obtained

11
274 by integrating Equation (36) over [a2, b2] as:

 
n
   b  k 1  n      b  k  1  a  k  1 

d 2  b2
275 1   2     0   0   (44)
n0 n ! n      2 
a    2 
b  b2  

276 For a cavity unloading from an in-situ stress state (also applicable for unloading analysis
277 from a loading-induced elastic stress state), all the necessary information for determining the
278 complete pressure-contraction curve and stress distributions has been given. Results can be
279 readily obtained following a similar procedure given by Yu (1992, 1993) or in a simplified way
280 of that will be detailed in the next section.

281 Solutions for unloading from a partially/fully plastic state

282 This section presents solutions for the analysis of a cavity unloading from a partially or fully
283 plastic stress state (namely cases II and III in Figure 1, p20  pl e lim ).

284 Elastic unloading analysis

285 Initially, the unloading is purely elastic. Upon elastic unloading, changes in the stress and
286 displacement can be determined by solving Equations (1), (17) and (18), which gives:

 b2 r2  k 1 1
(45)
287  r  ( p2  p20 )
 b2 a2  k 1 1

 b2 r2  k 1 k  1 (46)
288    ( p2  p20 )
 b2 a2  k 1 1

 p2  p20    r k 1 1  v   r  k 1  
289 u  r2   k 1 
1  v  kv      
2
1   2    (47)
Mr2 1 a2  1 b2  
k k 1
 2 
b k  1  v  2  k    b2   


290 Reverse plasticity occurs once the Mohr-Coulomb yield criterion Equation (6) is satisfied.
291 This condition is firstly satisfied at the inner wall when the cavity pressure reaches its elastic
292 limit p2elim  2 . Combining Equation (6), (10), (11), (45) and (46), it gives:

k  1
A1   1   c20 a 20    b2 a2 k 1  1
293 p2e lim  2  p20    (48)
  1 / k  b2 a2   1  
k 1

294 Elastic-plastic unloading analysis

295 As the cavity pressure p2 further reduces, a reverse plastic zone forms and spreads outwards

12
296 from the inner wall of the cavity (i.e. a2≤r2≤d2). Equations (2) and (3) define that the current
297 stress state in the soil depends on both the residual stresses due to previous loading and the
298 stress changes due to unloading. Hence, for a cavity unloading from a partially plastic state (i.e.
299 case II in Figure 1), the solution needs to be discussed according to the relative size of the
300 loading-induced plastic zone and the reverse plastic zone as illustrated in Figure 2. The elastic-
301 plastic unloading behaviour of a cavity unloading from a fully plastic state (i.e. case III in Figure
302 1) is studied simultaneously as follows.

(a) (b)
r2 r2
o o
a2 d2 c2 b2 a2 c2 d2 b2

(a) Phase with d 2  c2  b2 (b) Phase with c2  d 2  b2

Figure2. Distribution of stress states

303 Unloading Phase with d 2  c2  b2

304 (1) Stress analysis

305 The unloading-induced changes of the principal stresses in the elastic unloading zone (i.e.
306 d2≤r2≤b2, Figure 2a) can be determined by solving Equations (1), (17) and (18) as:

 r  B2 2  b2 r2   1
k 1
307 (49)
 

    B22  b2 r2  k 1
k 1
308 (50)
 

309 In the reverse plastic zone (a2≤r2≤d2), the principal stresses are determined by jointly solving
310 Equations (1) and (6) as:

k 1 
311  r 2  Y   1  A22  d2 r2  (51)

k 1 
312  2  Y   1  A22  d2 r2  (52)

313 The constants B2-2 and A2-2 of integration can be determined based on the continuity
314 condition of stress components at the outer radius of the reverse plastic interface (i.e. at r2=d2)
315 as:

13
 k  1
A1   1   c20 d 20  
316 B2  2  (while d 2  c2  b2 ) (53)
  1 k  b2 d2 
k 1
1

k  1
A2  2  B2  2  b2 d 2   1  A1  c20 d 20 
k 1
317  (while d 2  c2  b2 ) (54)
 

318 (2) Displacement analysis

319 In the elastic unloading zone (d2≤r2≤b2), the radial displacement can be obtained by integrating
320 Equations (17) and (18) with inputs of Equations (49) and (50) as:

B r 
 v  v 1  v  kv   b2  
k 1

321 u  r2   22 2 1  v  kv        (55)
M  1  v  2  k  1  v  2  k  k   r2   


322 The below relationship between strain and stress is established with reference to the state at
323 the completion of unloading, taking a procedure akin to that of obtaining Equation (34).

1 
 kv   kv  

324  r  k   1    r   k  1  v  kv       (56)
M  1  v  2  k  
  1 v2  k   


325 According to the definition of logarithmic strains,  r and  are expressed as:

326  r  ln  dr2 dr20  ,   ln  r2 r20  (57 a,b)

327 Substituting Equations (57 a,b) into Equation (56) gives:

k  1
ln  r2k  dr2 r20 k  dr20   1  d 2 r2 
k 1 
328  1  c20 r20   (58)

329 in which

A2 2 
 kv  kv 
330 1 =  1     k  1  v  kv    (59)
M  1 v2  k 
  1  v  2  k   


A1 
 kv 1 kv 
331 1 = 1    k  1  v  kv    (60)
M  1  v  2  k   
 1  v  2  k   


332 In this case, two transformation variables are introduced, namely:

k  1
k 1 
333   r2  =1  d2 r2  ฀  (r20 )=1  c 20 r20   (61 a,b)

334 Then integrating Equation (58) over the interval [r2, d2] leads to:

14
  ( d2 )   ( d20 ) 
335 d2k  1  e   11 d   c1k  1  e 1   1d =0 (62)
 ( r2 )  ( r20 )

336 Placing r20=a20 and r2=a2, Equation (62) can be solved with the aid of infinite series as:


d 2k  11n  k 1  n   

337 
n0  n    n ! 
1   d 2 a2  

 
n0
1n  0 (63)

338 where

  c20 1 k   1 a20
k  1 n

 ln , if n = 
 n !  d
339 1n  
20
(64)
k  1 n k  1 n   k  1 n  
  c20 1  c d      

  n    n!  20 20  , otherwise
c20 a20
 

340 While taking b0   , the above solution reduces to the large-strain solution of Yu and
341 Houlsby (1995) for the cavity contraction analysis in an infinitely large soil mass.

342 Note that the above solution can also be applied in the analysis of a cavity unloading from
343 a fully plastic stress state before it reaches the reverse fully plastic phase because the condition
344 of d 2  c2 is always fulfilled in this case. However, care should be exercised in the calculation
345 of the residual stress field in the soil as it is different between the partially plastic expansion
346 state and the fully plastic expansion state.

347 Unloading Phase with c2  d 2  b2

348 (1) Stress analysis

349 The unloading phase of c2  d 2  b2 (Figure 2b) is likely to occur in a lightly pre-loaded soil
350 mass. In this phase, the stress solutions of Equations (49) and (50) for the elastic unloading
351 zone (i.e. d2≤r2≤b2) and Equation (51) and (52) for the reverse plastic zone (a2≤r2≤d2) are still
352 valid. However, the constants of integration B2-2 and A2-2 need to be replaced by B2-3 and A2-3
353 below respectively as the stress conditions at r2=d2 changed.

Y    1 p0  B1   1 k  b20 d 20  1 

k 1
 
354 B23 = (while c2  d 2  b2 ) (65)
  1 k  b2 d2   1  
k 1

Y    1 p0  b k 1   b k 1 
355 A23    B1  20
  1  B23    1
2
(while c2  d 2  b2 ) (66)
  1  d 20    d 2  

356 Then according to Equation (2) and (3) (i.e.  r   r 2   r 20 and      2    20 ), the

15
357 changes in stress due to unloading can be readily determined using Equation (10), (11), (51),
358 (52), (65) and (66) for the zone a2≤r2≤c2 (named as loading plastic and reverse plastic zone)
359 and Equations (12), (13), (51), (52), (65) and (66) for the zone c2≤r2≤d2 (named as loading
360 elastic and reverse plastic zone), respectively.

361 (2) Displacement analysis

362 In this unloading phase, the radial displacement in the elastic zone (i.e. d2≤r2≤b2) can be
363 calculated by Equation (55) with the value of B2-3 of Equation (65).

364 In the reverse plastic zone a2≤r2≤d2 (Figure 2b), a large-strain displacement analysis can be
365 carried out following the same procedure of deriving Equations (31)-(64). However, due to the
366 difference in the residual stress field between the zone of a2≤r2≤c2 and the zone of c2≤r2≤d2, the
367 distribution of displacement in the soil now needs to be derived by considering these two zones
368 separately, which is continuous at the interface r2=c2.

369 In the loading-elastic and reverse plastic zone of c2≤r2≤d2, substituting Equations (12), (13),
370 (51), (52), (57 a,b), (65) and (66) into Equation (56) leads to:

ln  r2k  dr2 r20k  dr20   ln 1  2  d 2 r2 

k 1 
 2  b20 r20 
k 1
371 (67)

372 where

1  kv kv  Y 
373 ln1  1   k  1  v  kv     p0   B1  (68)
M  1  v  2  k  1  v  2  k      1 

B1  kv v 
374 2 = 1    1  v  kv    (69)
M  1  v  2  k  1  v  2  k  

A23 
 kv  kv 
375 2 =  1     k  1  v  kv    (70)
M   1 v2  k   1  v  2  k   


376 Then the integration of Equation (67) in the interval of [c2, d2] gives:

1  k  d 2k  12 n  k 1  n   

377 
k   1 n  0  n    n! 
1   d 2 c2   
 
n0
 2n  0 (71)

378 where

16
  1b20k  12 n  k  1 c20 k 1
 ln , if n =
 n! d 20 k 1
379 n    b 
2
k  1 n
  k  1   k  1 
k 1 
(72)
 b20 d 20  k 1     b20 c20  k 1    , otherwise
n k 1 n

1 20 2
  
  n  k 1  
 n!

  k 1 

380 In the loading-plastic and reverse plastic zone of a2≤r2≤c2, the displacement at r2=a2 can be
381 obtained by integrating Equation (67) in the interval of [a2, c2], which is:


d 2k  12 n k 1  n    k 1  n   

382   n    n!  d
n 0
2 c2    d 2 a2     3n  0
 n 0
(73)

383 where

  c20 1 k   1 a20
k  1 n

 ln , if n = 
 n!  c20
384 n   k  1 n
3
(74)
k  1 n  
  c20 1 1   c a  

  n    n!  
20 20 , otherwise
 

385 Reverse fully plastic unloading analysis

386 With sufficiently large contraction, the cylinder/sphere of soil will enter a fully plastic
387 unloading phase. This may be reached from either a partially plastic state (i.e. case II in Figure
388 1) or a fully plastic state (i.e. case III in Figure 1) as studied separately below.

389 Reverse fully plastic unloading of Case II

390 This state follows the unloading phase of c2  d 2  b2 that was studied previously. The stress

391 solution of Equations (51) and (52) still holds in this phase, but A2  2 and d2 therein need to be

392 replaced by A2  4 and b2, respectively. The new integration constant A2  4 is determined
393 according to the given stress boundary conditions in Equation (5 a,b) as:

394 A24    p0  Y   1 (while c2  d 2  b2 ) (75)

395 The distribution of displacement in the soil can be obtained with the same procedure of
396 deriving Equations (67)-(74), considering the zone a2≤r2≤c2 and the zone of c2≤r2≤b2 separately.
397 Similarly, it is found that the displacement solution can be obtained by replacing A2  4 and d2

398 with A2  4 and b2, respectively, in Equation (67)-(74).

399 Reverse fully plastic unloading of Case III

400 This is for the fully plastic unloading analysis of the case that the soil cylinder/sphere was
17
401 previously loaded to be fully plastic ( c2  b2 ). In this case, the stress distribution in the soil can

402 be obtained by Equation (51) and (52), replacing d2 and A2  2 with b2 and the new constant

403 A2 5 (i.e. Equation (76)), respectively.

404 A25  A24   p0  Y   1 (while c2  d 2  b2 ) (76)

405 The large-strain displacement of the surrounding soil during this fully plastic unloading
406 phase can be described by Equations (55), (56), (59) and (60) as well while replacing d2, c 1 ,

407 and A2  2 therein with b2, b2 and A2 5 , respectively.

408 Solution procedure

409 All the necessary information for determining the pressure-contraction curve and stress
410 distributions has been given. As the pressure-expansion relationship is not expressed in terms
411 of a single equation, it is instructive to summarize the procedure that can be used to construct
412 the complete pressure-contraction curve as below:

413 (1) Choose input soil parameters: E, v, c, φ, ψ; in-situ stress: p0; cavity geometry information:
414 k, b0 and a0; and the reference state parameter: a20;

415 (2) Calculate the derived parameters G, Y, α, β, γ, M and the reverse limit p2elim  2 (Equation
416 (48));

417 (3) Calculate the pressure-expansion curve in the loading process until the inner cavity radii
418 reaches a20, and then record c20, b20, and p20 for elastic-plastic loading process, or b20 and p20
419 for fully plastic loading process. The solution procedure is available in Yu (1992, 1993);

420 (4) Calculate A1 and B1 from Equations (14) and (15), respectively, with the known value of
421 c20 and b20. The stress field at the end of loading process (σr20 and ∆σθ20) can be obtained from
422 Equations (10) and (11) for the plastic zone and Equations (12) and (13) for the elastic zone;

423 (5) For elastic unloading analysis, choose a pressure p2 (p20<p2< p2elim 2 ) and calculate ∆σr

424 and ∆σθ from Equations (45) and (46), respectively. Then the distribution of stress can be
425 obtained from Equations (2), (3) and (10)-(13). And the relative displacement ∆u could be
426 calculated from Equation (47);

427 (6) If p2< p2elim  2 , elastic-plastic (a20<c20<b20) or fully plastic (c20=b20) unloading analysis is
428 needed;

429 (7) In the case of a20<c20<b20 (Case II in Figure 1):

18
430 (a) while d 2  c2  b2 , choose a value of d20 (a20<d20<c20), calculate d2 and b2 from
431 Equation (55) with the known value of d20 and b20; then calculate a2 from Equations (63) and
432 (64) with the known values of d2, d20, b2 and c20, and p2 from Equations (51) and (54).

433 When d20, d2, b2 are known, B2-2 and A2-2 can be calculated from Equations (53) and (54),
434 and then the distribution of stress can also be derived from Equations (49)-(52).

435 (b) while c2  d 2  b2 , choose a value of d20 (c20<d20<b20), calculate d2 and b2 from

436 Equation (55) with the known value of d20 and b20; calculate c2 from Equations (71) and (72)
437 with the known values of d2, d20, b2 and c20; calculate a2 from Equations (73) and (74); then
438 calculate p2 from Equation (51) by setting r2=a2. Note that in all calculations of this phase,
439 A2-2 and B2-2 need to be replaced by A2-3 and B2-3 (Equations (65) and (66)), respectively;

440 With the values of A2-3, B2-3, b2 and d2, the distribution of stress components can be
441 calculated from Equations (49)-(52) by replacing A2-2 and B2-2 with A2-3 and B2-3, respectively.

442 (c) while d2=b2, choose a value of b2/c2 (greater than b20/c20), calculate c2 and b2 from
443 Equations (71) and (72); calculate a2 from Equations (73) and (74) with the known values of
444 b2 and c2; then calculate p2 from Equation (51) with r2=a2. In this phase, A2-2 in the relevant
445 equations needs to be replaced by A2-4 (Equation (75)), and the stress components can be
446 obtained from Equations (51) and (52).

447 (d) For the case of c20<b20, repetition of steps of (a) and (b) for varying d20, and (c) for
448 varying b2/c2 provides the data for a complete pressure-contraction curve.

449 (8) In the case of c20=b20 (Case III in Figure 1):

450 (e) while d2< c2=b2, step (a) still holds;

451 (f) while c2= d2= b2, choose a value of b2/c2 (greater than the value of b2/c2 in the previous
452 step), calculate b2 and a2 from Equations (63) and (64); then calculate p2 from Equation (51)
453 with r2=a2. In this phase, A2-2 in the relevant equations needs to be replaced by A2-5 (Equation
454 (76)). Similarly, the distribution of stress components can be obtained from Equations (51)
455 and (52) by replacing A2-2 by A2-5.

456 Repetition of steps of (e) for varying d20 and (f) for varying of b2/a2 provides the data for
457 complete pressure-contraction curve. Note that, for the cases in a frictionless soil, results can
458 be calculated with the above solutions using very small φ values.

19
459 Solution validation and result analysis

460 Unloading curves from an in-situ stress state

461 At first, the unloading solution from an in-situ stress state is validated by comparing with the
462 solution of Vrakas and Anagnostou (2014) in the special case of a cavity in an infinite soil mass.
463 Taking the material parameters: p0=22.5MPa, E=2000MPa, c=0.25MPa, φ=23o, ψ=3o and
464 v=0.25 that were used by Vrakas and Anagnostou (2014), both the pressure-contraction curve
465 and the distribution of stresses and the radial displacement at the moment of p2=0MPa were
466 calculated, and they are compared with those obtained by Vrakas and Anagnostou (2014) in
467 Figure 3. The comparisons in Figure 3 indicate that their solution can be recovered by the
468 present solution assuming b0 / a0   , neglecting the out-of-plane plastic flow effect which
469 proved to be insignificant (Reed, 1986; Vrakas & Anagnostou, 2014).

70 70 35
Vrakas and Anagnostou (2014)
60 60 This study 30
Vrakas and Anagnostou (2014)
This study
50 50 25
[u2(a2)/a0] /%

[u2(a2)/a0] /%

plastic zone elastic zone

 2
40 40 20

 / MPa
(r0-r2)/a0
30 30 15

20 20 10
r2
10 10 5

0 0 0
0.0 0.5 1.0 1.5 2.0 2.5 0 1 2 3 4 5 6 7 8 9
(a) p2 / MPa (b) r2 / a0

(a) pressure-contraction curve. (b) distribution of stresses and radial displacement.

Figure3. A cylindrical cavity unloading from an in-situ stress state (Sedrun section of the
Gotthard Base Tunnel (Vrakas & Anagnostou, 2014))

1.0 b0/a0=2 1.0 b0/a0=2

b0/a0=3 b0/a0=3
b0/a0=5 b0/a0=5
0.8 0.8
b0/a0=10 b0/a0=10
b0/a0=10000 b0/a0=10000
p2 / p0
p2 / p0

0.6 0.6

fully plastic point

0.4 0.4

fully plastic point

0.2 0.2

0.0 0.0
0 20 40 60 80 100 0 10 20 30 40 50 60
(a) [u2(a2) / a0] /% (b) [u2(a2) / a0] /%

(a) cylinders (b) spheres

Figure4. Pressure-contraction curves with varying b0/a0

20
470 To indicate the effects of the cavity geometry ratio b0/a0, example cavity pressure-
471 contraction curves with different values of b0/a0 are computed using the same material
472 parameters as above. The results in Figure 4 show that the value of b0/a0 greatly influences the
473 cavity unloading behaviour when it is smaller than a limit value for both hollow cylinders and
474 spheres. The value of this limit ratio varies with soil properties and is generally smaller for a
475 hollow sphere than a cylinder. With the same level of contraction (e.g. (a20-a0)/a0), the
476 magnitude of unloading (e.g. (p20-p0)/p0) is smaller for a thinner cylinder/sphere of soil while
477 b0/a0 is smaller than the limit ratio. In other words, with the same magnitude of unloading,
478 greater radial contraction may occur for a thinner hollow cylinder/sphere. Tunneling involves
479 the removal of soil/rock masses from their initial locations, and this is analogy to the problem
480 of cavity unloading from an in-situ stress state (Mair & Taylor, 1993; Mo & Yu, 2017; Ogawa
481 & Lo, 1987; Vrakas & Anagnostou, 2014; Yu & Rowe, 1999). Experimental results (e.g. (Franza
482 et al., 2019)) have shown that the ground response curves (GRCs) of shallow tunnels in sands
483 vary with the tunnel depth ratio (e.g. soil cover depth/tunnel radius) in a very similar fashion as
484 that is shown in Figure 4.

485 Apart from the geometry ratio b0/a0, soil strength and stiffness parameters also affect the
486 cavity unloading behaviour. Their influences are akin to those observed in unloading analyses
487 of a cavity in an infinite soil mass, which can refer to the results of Yu and Rowe (1999) and
488 Vrakas and Anagnostou (2014).

489 Unloading curves from a partially/fully plastic state

490 Using the same soil model, Yu and Houlsby (1995) developed an analytical large-strain solution
491 for unloading analysis of an infinite soil mass under a loading-induced partially plastic state.
492 The paper extended their solution to the more general case of a hollow cylinder/sphere of soil.
493 For validation, results obtained by these two solutions are compared in Figure 5. It is shown
494 that the present solution can exactly recover to their solution, taking b0 / a0   . Note that the
495 solutions of Yu (1992, 1993) were used to calculate the expansion curves and the reference
496 stress state in the validation and following analyses.

21
 70 300
Symbol: Yu and Houlsby (1995) Symbol: Yu and Houlsby (1995)
60 Line : This study Line : This study
250
50 =50, =25 =50, =25
200
40
=40, =12.5 150
p/p0

p/p0
30
=40, =12.5
20 100
=30, =0

10 50 =30, =0
0 0
1 2 3 4 5
1 2 3 4 5
(a) a/a0 (b) a/a0

(a) cylindrical cavities (b) spherical cavities

Figure5. Comparison with results of Yu and Houlsby (1995) (E/p0=1300, v=0.3 and c=0).

497 To show the effects of the reference stress state on the unloading behaviour of a finite soil
498 mass, a selection of results of loading and unloading curves with different magnitudes of
499 preloading and values of b0/a0 are presented in Figures 6-9. The soil parameters of φ=40o, ψ=20o,
500 v=0.3, c=0 and E/p0=1300 were used. In Figures 6 and 7, two typical ratios of the initial outer
501 to inner radii of a finite soil mass are considered (e.g. b0/a0=10000 or b0/a0=5), and the
502 unloading is assumed to start after different magnitudes of expansion (i.e. a20/a0). In Figures 8
503 and 9, loading and unloading curves with different values of b0/a0 are plotted, in which the
504 applied cavity pressure is removed when the expansion ratio a20/a0 is equal to 3 and 1.15,
505 respectively. In the figures, the triangle represents the point when the elastic-plastic interface
506 reaches the outer boundary of the hollow cylinder/sphere during loading (i.e. c1=b1); the circle
507 for each curve represents the point when plastic unloading occurs (i.e. d2=a2).

10 3.5
fully plastic
expansion
3.0
8
2.5
p /p0

6 2.0
p/p0

1.5
4
1.0
2
0.5

0 0.0
0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05 0.98 0.99 1.00 1.01 1.02 1.03 1.04 1.05
(a) a/a0 (b) a/a0

(a) b0/a0=10000 (b) b0/a0=5

Figure6. Unloading curves of soil cylinders from varying reference stress states

22
 30 12
fully plastic
expansion
25 10

20 8

p/p0
p/p0

15 6

10 4

5 2

0 0
1.00 1.01 1.02 1.03 1.04 1.05 1.00 1.01 1.02 1.03 1.04 1.05
(a) a/a0 (b) a/a0

(a) b0/a0=10000 (b) b0/a0=5

Figure7. Unloading curves of soil spheres from varying reference stress states
40 160
b0/a0= b0/a0=
140
30 120 70
200
100
50
100
p/p0
p/p0

20 80
60 30
50
10 30 40
20
10 20
0 0
1.0 1.5 2.0 2.5 3.0 1.0 1.5 2.0 2.5 3.0
(a) (b)
a/a0 a/a0
(a) cylinders (b) spheres
Figure8. Variation of loading and unloading curves with b0/a0 (a20/a0=3)
20 60
b0/a0= b0/a0=
16 50
15
50
40
12 12
30
p/p0

p/p0

30
10
8 20
20
10 5
4
10

0 0
1.00 1.03 1.06 1.09 1.12 1.15 1.00 1.03 1.06 1.09 1.12 1.15
(a) a/a0 (b) a/a0

(a) cylinders (b) spheres

Figure9. Variation of loading and unloading curves with b0/a0 (a20/a0=1.15)

508 Figures 6-9 show that initially the internal cavity pressure reduces rapidly with cavity

23
509 contractions in the elastic unloading phase (i.e. Equations (45)-(47)). The unloading curves in
510 the non-dimensional plot of p2/p0 against a2/a0 are almost linear and parallel with each other in
511 this phase. Although the slope of the unloading curve is insensitive to the initial geometry ratio
512 b0/a0 and the previous loading history (e.g. a20/a0), these factors affect the maximum amount of
513 the stress reduction during elastic unloading. Due to the residual stresses generated during
514 previous loading (i.e. the so-called ‘overstrain’ effect (Hill, 1950; Zhao & Wang, 2010)), the
515 elastic unloading process is much longer than the initial elastic loading process.

516 The above intrinsic characteristics of the elastic unloading process have been used in various
517 applications such as the control of unloading-reloading loops of pressuremeter tests and the
518 shakedown analysis. As pointed out by Wroth (1982), the soil stiffness can be obtained from
519 the unloading-reloading loop of self-boring pressuremeter tests. While conducting unloading-
520 reloading loops, it is important to ensure that the loop deformation remains in an elastic state.
521 For a linear elastic-perfectly plastic Mohr-Coulomb material, Equations (10) and (48) can be
522 used to determine the maximum reduction of the effective pressure allowed for elastic
523 unloading in pressuremeter tests. In an infinite soil mass, the maximum cavity pressure
524 reduction is a function of soil strength parameters (e.g. friction angle and cohesion) and the
525 loading-induced stress state (e.g. p20) (Wroth, 1982; Zhao & Wang, 2010). However, it also
526 varies with the value of b0/a0 for the unloading of a finite soil mass (e.g. Figures 6(b), 7(b), 8
527 and 9). An example application to the shakedown analysis will be given in the next section.

528 Once the cavity pressure reduces to be smaller than the elastic limit (i.e. Equation (48)), the
529 unloading curve becomes highly non-linear as reverse yielding occurs in the soil. The radial
530 convergence accelerates as the plastic region spreads out with smaller internal confining
531 pressure. When the cavity pressure reaches a sufficiently low value, the radial convergence
532 increases sharply until the inner cavity is filled. In general, the speed of transition from a purely
533 elastic state to the steady or limit state during unloading is much faster than that occurred in the
534 initial loading process, and it varies with the cavity shape (normally, it is faster for a spherical
535 cavity than a cylindrical cavity). The minimum internal pressure that the soil can sustain mainly
536 depends on the soil strength parameters as defined in Equation (51), for example, it is close to
537 zero for cohesionless soils but could be negative for cohesive soils or rocks. Figures 6-9 show
538 that this limit value of unloading pressure does not significantly vary with the value of b0/a0
539 and the loading history.

540 As mentioned previously, experimental studies on both self-boring and full-displacement

541 pressuremeters (Hughes & Robertson, 1985; Schnaid & Houlsby, 1992) have shown the
542 unloading portion of pressuremeter curves is less sensitive to initial soil disturbance than the
543 loading portion. This is consistently observed in the results of Figures 6-9. Besides, the results

24
544 indicate that the shape of the unloading curve is also less sensitive to the geometry ratio b0/a0
545 than the loading curve in a finite soil mass. This suggests that the use of the unloading curve of
546 pressuremeter tests to measure the soil properties may also help to remove the boundary effects
547 that are commonly encountered in small-sized calibration chambers.

548 Example geotechnical applications

549 Thickness of frozen cylinder earth walls

550 Artificial ground freezing has been widely used to stabilize temporarily the ground in order to
551 provide ground support and/or exclude groundwater from an excavation until the final retaining
552 and lining structures are constructed (Andersland & Ladanyi, 2004; Sanger & Sayles, 1979;
553 Viggiani & Casini, 2015; Zhang et al., 2018). From a structural point of view, determination of
554 the geometry and the thickness of a frozen wall is one of the main concerns for practitioners.
555 Because of the relatively high compressive and low tensile strengths of frozen soil, curved arch
556 walls, particularly circular walls, are often selected with priority. The unloading model of a
557 cylinder unloading from an in-situ stress state that was studied previously has been commonly
558 used to determine the thickness of a circular frozen wall (Andersland & Ladanyi, 2004; Klein
559 & Gerthold, 1979; Sanger & Sayles, 1979). For example, assuming  r 2 r b   p0 and
2 2

560  r 2 r a  0 (i.e. no internal support), Sanger and Sayles (1979) proposed Equation (77) to
2 2

561 estimate the minimum thickness of a cylinder wall. Klein and Gerthold (1979) extended this
562 solution to the case where the internal pressure acting on the wall equals p 2 (i.e.

563  r 2 r a   p2 ), thereby Equation (78) was given. These solutions were obtained by solving
2 2

564 the equilibrium equation (1) and the Mohr-Coulomb yield function (6). Therefore, they can be
565 recovered by Equation (26) or (30) considering the boundary conditions they adopted.

p0  c cot 
 b0 a0 
 1
566  (77)
c cot 

p0  c cot 
 b0 a0 
 1
567  (78)
p2  c cot 

568 In both Equations (77) and (78), a hidden assumption is that the cylinder wall of frozen soil
569 becomes unstable once it becomes fully plastic. This is a typical large deformation problem,
570 and displacements of the cavity during elastic-plastic contractions can be calculated by using
571 Equations (30), (32a, b) and (44). Adopting the criterion of Sanger and Sayles (1979) (i.e.
572 p2fp lim  0 in Equation (30)), new results are calculated considering the large deformation

25
573 effects with typical soil properties of v=0.3, E/p0=100 and ψ=min(0, φ-20). The new results are
574 compared with the published results of Sanger and Sayles (1979) in Figure 10. It is shown that
575 Equation (77) tends to give conservative predictions of the minimum thickness of a cylinder
576 wall due to the lack of accounting for its radial convergence, whose effects become more
577 significant at larger values of the in-situ earth pressure and the friction angle of soil.

5 Sanger and Sayles (1979)


This Study =40
4

3
p0/c


=30
2

=20

1 =10

0
1.0 1.1 1.2 1.3 1.4 1.5
b0/a0

Figure10. Comparison of design charts for circular frozen walls

578 In addition to the requirement of stability, frozen earth walls may also be designed under
579 displacement control on the basis of required excavation limits and the available space on site
580 (Andersland & Ladanyi, 2004). For example, taking 2% cavity strain (i.e. a2/a0=0.98) as the
581 maximum allowable radial convergence of a cylinder wall, the minimum internal pressure
582 required can be obtained by the cylindrical unloading solution from an in-situ state. Example
583 results are presented in Figure 11 with varying strength and stiffness parameters of frozen soils.
584 It is shown that the self-stability of a cylinder wall, in general, increases with the thickness ratio,
585 the frictional strength and stiffness of the soil, thus less internal support is required accordingly.

586 Adopting different control standards, the application of the large strain in-situ unloading
587 solution to the preliminary structural design of frozen cylinder walls is illustrated in Figures 10
588 and 11. It needs to be pointed out that the boundary conditions at the outer wall of the cylinder
589 were simplified as a constant radial pressure whose value equals the in-situ stress. In fact,
590 however, the outside confining pressure may reduce with contractions of the cylinder. Thus,
591 this method still tends to be conservative as well. Additionally, it should bear in mind that the
592 above analyses focused on the short-term unloading behaviour. For the long-term stability and
593 deformation analysis, the time-dependent behaviour of frozen soils (e.g. creep strength) needs
594 to be taken into account (Andersland & Ladanyi, 2004; Sanger & Sayles, 1979).

26
 1.0 1.0
=10 E/p0=100
=20 E/p0=150
0.8  0.8
=30 E/p0=200
=40 E/p0=300
p2 /p0

p2 /p0
0.6 =50 0.6
E/p0=

0.4 0.4

0.2 0.2

0.0 0.0
1.0 1.2 1.4 1.6 1.8 2.0 1.0 1.2 1.4 1.6 1.8 2.0
(a) b0/a0 (b) b0/a0

(a) variation with angle of friction (E/p0=100) (b) variation with stiffness (φ=30°)
Figure11. Internal cavity pressure at a radial displacement of 2% (v=0.3 and c/p0=0.3)

595 Prediction of pressuremeter curves

596 Based on the analogy between pressuremeter tests and a long cylindrical cavity upon loading
597 and unloading, cavity expansion and contraction solutions have been used in the interpretation
598 of pressuremeter tests with considerable success (Clarke, 1995; Mair & Wood, 1987; Wroth,
599 1984; Yu, 2000). As summarized by Schnaid et al. (2000), the methods of interpreting
600 pressuremeter tests can be broadly categorised into two groups: in the first group each
601 parameter of soil is assessed independently from one portion of the pressuremeter curve; in the
602 second the whole pressuremeter curve (both loading and unloading portions) is taken into
603 account. Using the closed-form expansion and contraction solutions of Yu and Houlsby
604 (1991,1995), Schnaid et al. (2000) analysed many site pressuremeter tests, from which a set of
605 fundamental parameters of soil can be derived. In modelling site pressuremeter tests, it is
606 reasonable to assume the surrounding soil to be horizontally infinite. However, this might be
607 questionable for the modelling of laboratory pressuremeter tests in small-sized calibration
608 chambers as highlighted previously (Alsiny et al., 1992; Fahey, 1986; Jewell et al., 1980; Juran
609 & BenSaid, 1987; Schnaid & Houlsby, 1991).

610 To account for the possible boundary effects during cavity expansion, Yu (1992, 1993)
611 extended the solution of Yu and Houlsby (1991) to the case in a finite soil mass. Likewise, an
612 extension of the solution of Yu and Houlsby (1995) was obtained in this paper for the analysis
613 of a cavity in a finite soil mass. Now, using the loading solution of Yu (1992) and the present
614 unloading solution in combination, the method proposed by Schnaid et al. (2000) can be
615 extended for the interpretation of pressuremeter tests performed in calibration chambers of a
616 constant lateral stress boundary (i.e. the BC1-type boundary (Ghionna & Jamiolkowski, 1991)).
617 This is evaluated by comparing with the experimental results of pressuremeter tests obtained
618 by Ajalloeian (1996).

27
619 A number of pressuremeter tests were performed by Ajalloeian (1996) in a calibration
620 chamber using dry Stockton Beach sand. The ratio of the chamber diameter (1000mm) to the
621 pressuremeter diameter D (30mm) was 33.3. As the pressuremeter is assumed to be infinitely
622 long in the plane strain cylindrical cavity model, test data obtained with pressuremeters of the
623 longest membrane length L available were selected in the analysis to minimize the possible end
624 effects (Ajalloeian & Yu, 1998; Houlsby & Carter, 1993). In specific, L/D was 15 for the test in
625 the loose sand (test ID: 15LK1P100, relative density Dr=27.5%); L/D was 20 for the tests in the
626 medium dense (test ID: 20MK1P100, Dr=63.3%) and dense sand samples (test ID: 20DK1P100,
627 Dr=86.8%) (see Figure 12).

628 Based on the loading solution of Yu (1992) and the unloading solution of this study
629 (Equations (45)-(76)), the pressuremeter tests are interpreted following the curve fitting method
630 proposed by Schnaid et al. (2000) as follows:

631 (a) Initial stress state. The initial mean effective stress was 100kPa and the initial stress ratio
632 of the effective vertical stress to the effective horizontal stress was 1 in the tests (Ajalloeian,
633 1996). The same initial stress state was used in the modelling.

634 (b) Shear modulus. For the loading portion, the curve fitting was initiated using the tangent
635 stiffness value of the initial portion of the loading curve (Ajalloeian, 1996); for the unloading
636 portion, the value of the shear modulus was estimated by drawing a single line between the
637 point that defines the end of the loading and the representative point of the theoretical plastic
638 reverse of the experimental unloading curve (Schnaid et al., 2000). The Poisson’s ratio was
639 assumed to be 0.3.

640 (c) Strength parameters. The dilation angle  was calculated using the correlation of Rowe

641 (1962) (i.e. Equation (79)). The critical state friction angle cv of the Stockton Beach sand
642 required in Equation (79) is 31o (Ajalloeian, 1996). The cohesion was set as zero for the dry
643 sand. Then based on the cavity expansion and contraction solutions and the measured
644 pressuremeter curves (both loading and unloading portions), the plane strain friction angle ps

645 was back-calculated using a curve fitting technique.

sin ps  sin cv

646 sin  (79)
1  sin ps sin cv

647 Figure 12 presents the predicted and the measured loading and unloading curves. The
648 comparison shows that a good fit (the correlation coefficient R2>0.99 in all three cases
649 compared) was achieved between theory and data over the whole curve of loading and
650 unloading. In Figure 13, the back-calculated friction angles are compared with the data from

28
651 triaxial tests measured by Ajalloeian (1996). The relationship 8ps  9tri (Wroth, 1984) was

652 used to convert the triaxial friction angle  tri and the plane strain friction angle ps for

653 comparison. Results obtained by Ajalloeian (1996) using other common theoretical methods
654 (e.g. (Hughes et al., 1977; Manassero, 1989; Yu, 1994)) are also plotted in Figure 13. It is shown
655 that the combined use of the loading solution of Yu (1992) and the present unloading solution
656 from an elastic-plastic state as well as the curve-fitting technique of Schnaid et al. (2000) gave
657 the best estimate of the soil strength parameters in the compared cases. The close agreement
658 between theoretical and experimental results in Figures 12 and 13 indicates that this method is
659 able to construct a theoretical curve that reproduces a pressuremeter test from which
660 fundamental soil parameters can be derived reasonably well.

661 Note that the maximum cavity strain was less than 11% and a large diameter ratio of the
662 chamber to the pressuremeter was intentionally used in the tests of Ajalloeian (1996) to
663 minimize the boundary effect. Therefore, the difference of the friction angle back-calculated
664 with or without considering the size of the sand sample is not significant in these tests (Figure
665 13). Much stronger boundary effects may appear in tests performed in smaller sand samples as
666 shown in Figure 9 and observed by Jewell et al. (1980), Fahey (1986), Schnaid and Houlsby
667 (1991) and Alsiny et al. (1992), among others. Under this circumstance, the advantages of the
668 present method will be more obvious.

669 It is necessary to bear in mind that as an elastic-perfectly plastic soil model was used, the
670 non-linear elastic (e.g. stress and strain-dependent shear modulus) and strain
671 hardening/softening behaviour of sand (e.g. non-constant dilatancy) cannot be realistically
672 modelled by the present solutions (Fahey & Carter, 1993; Manassero, 1989). The constraints
673 imposed by these simplifications may introduce some limits to the quality of the comparisons
674 but enforce the consistency among all parameters (Schnaid et al., 2000).

700
1000 Experimental data (20MK1P100)
Experimental data (15LK1P100)
600 Fitting curve Fitting curve
G(loading)=10MPa 800 G(loading)=17MPa
500 G(unloading)=25MPa G(unloading)=35MPa
R2=0.9915 R2=0.9900
p / kPa

600
p / kPa

400

300
400
200
200
100

0 0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
(a) (a-a0)/a0 /% (b) (a-a0)/a0 /%

(a) loose sand (b) medium dense sand

29
 1500
Experimental data (20DK1P100)
Fitting curve
1200 G(loading)=26MPa
G(unloading)=40MPa
R2=0.9919

p / kPa
900

600

300

0
0 2 4 6 8 10 12
(c) (a-a0)/a0 /%

(c) dense sand

Figure12. Comparisons between theoretical and experimental pressuremeter curves (test data
from Ajalloeian (1996)).
70
This study (b0/a0=33.3)
This study (b0/a0=
60


Hughes et al. (1977)

Friction angle /

Manassero (1989)
Yu (1994)
50

40

Triaxial tests
30
from Ajalloeian (1998)

20 30 40 50 60 70 80 90
Relative density /%
Figure13. Back-calculated and measured plane strain friction angles

675 Optimal thickness of a hollow cylinder/sphere for overstrain

676 Several stress limits are of great concern in the stress analysis and optimal design of a hollow
677 cylinder/sphere, for example, the elastic limit, the plastic limit and the shakedown limit (Hill,
678 1950; Xu & Yu, 2005; Zhao & Wang, 2010). In the process of loading, the elastic limit p1elim
679 (i.e. while c1=a1) and plastic limit p1plim (i.e. while c1=b1) were given in Equations (9) and

680 (16), respectively. In the subsequent unloading process, reverse yielding occurs at the inner wall
681 of the cavity (i.e. d2=a2) while the cavity pressure reduces to be equal to p2elim  2 (i.e. the
682 unloading elastic limit defined in Equation (48)); the unloading plastic limit (i.e. while d2=b2)
683 can be obtained from Equation (51).

684 As defined previously, due to the additional pressure p20  p0 , residual stresses are
685 generated within the cylinder/sphere (i.e. overstrain effect). Subsequently, as the internal
686 pressured is removed, the soil undergoes elastic unloading until the reverse yielding limit
687 p2elim  2 is reached. Based on the shakedown concept, no new plastic deformation will occur in

30
688 the hollow cylinder/sphere of soil during the subsequent cyclic reloading-unloading under
689 uniform internal pressures varying within the range of ( p2elim  2 , p20 ) (neglecting the possible
690 Bauschinger effects) (Hill, 1950; Zhao & Wang, 2010). Providing that the cavity pressure varies
691 in the range of [ p0 , p20 ], the shakedown limit, within which neither fully plastic state during
692 the initial loading nor reverse plastic state during subsequent unloading will occur in the soil,
693 can be determined by two conditions: (i) p20  p1plim , and (ii) p0  p2e lim  2 , which gives,

694 respectively,

Y    1 p0 
 b a     1
k  1 
695 p20  p0  (80)
 1  20 20

k  1
A1   1   b20 a20   b2 a2 k 1  1
696 p20  p0    (81)
  1 k  b2 a2   1  
k 1

697 The shakedown limit equals the minimum value of Equations (80) and (81) as it requires
698 that the inequalities hold simultaneously (Xu & Yu, 2005). It needs to be pointed out that
699 previous shakedown analyses of a hollow cylinder/sphere were mostly carried out in the
700 framework of small strain theory (Hill, 1950; Zhao & Wang, 2010). However, large deformation
701 may occur during the initial elastic-plastic loading process, which may affect the shakedown
702 limit, particularly in soft materials like soils. For example, Yu (1992) observed that the cavity
703 pressure reaches a peak value before the whole cylinder of soil becomes plastic due to the large
704 strain effects, which can also be seen in Figures 8 and 9. Hence, the large strain effects on the
705 shakedown limit are examined by calculating the optimal thickness of hollow cylinders/spheres
706 as follows.

707 It is known that a hollow cylinder or sphere cannot be too thin to be strengthed (Hill, 1950;
708 Zhao & Wang, 2010). The optimal thickness of a hollow cylinder/sphere for overstrain can be
709 determined by taking the equalities of Equation (80) and (81) simultaneously. Within the
710 shakedown limit, the unloading process is purely elastic. Thus, the radial displacement can be
711 determined by Equation (47), from which another relationship between the cavity pressure and
712 the geometry ratio can be obtained as:

1   p0  p20  M  1  v  kv  k  1 k   b2 a2   1

k 1

713
b20 b2
  
a20 a2 p0  p20 
 v   a2 
k 1
v 
1  v  kv 
1 k 1  
1  v  kv      
M 1   a2 / b2    1  v  2  k    b2  1 v2  k  k
   


714 (82)

31
715 Equations (80)-(82) give the value of b20/a20 after the overstrain, based on which the optimal
716 geometry ratio (i.e. b0/a0) can be readily obtained from the displacement solutions of loading
717 (see the Appendix).

718 At first, the optimal thickness for frictionless soil is investigated. In this special case, the
719 condition of equality of Equations (80)-(81) can be simplified as:

ln  b20 a20    2  k  1 1   a2 b2  

k 1
720 (83)
 

721 Equation (83) will reduce to the expression given by Zhao and Wang (2010) for Tresca
722 materials if ideally regarding the surrounding soil as rigid or adopting the small strain
723 definitions (i.e. a1=a2=a0 and b1=b2=b0). However, when the soil deformation during the elastic-
724 plastic loading and elastic unloading process is considered, the optimal thickness will vary with
725 the soil stiffness index G/su (su represents the shear strength of clay, corresponding to the
726 cohesion strength in the Mohr-Coulomb criterion). The variation of the optimal thickness for
727 both hollow cylinders and spheres with the soil stiffness is presented in Figure 14, taking a
728 broad range of values of G/su for clays. It is shown that the optimal thickness decreases with
729 increases of G/su and gradually converges to the limit value calculated by Zhao and Wang (2010)
730 for rigid clays. In other words, the geometry changes of the cylinder/sphere during loading and
731 unloading apply insignificant influence on the optimal thickness ratio in stiff clays, whereas
732 this effect cannot be ignored when the soil is relatively soft (e.g. G/su<200).

3.5
Zhao and Wang (2010)
This study
3.0
b0/a0

2.5
Cylinder (k=1)

2.0
Sphere (k=2)

10 100 1000
G/su

Figure14. Optimal thickness ratio b0/a0 of cohesive soil (p0/su=1 and v=0.5)

733 For frictional soils, Zhao and Wang (2010) observed that the optimal thickness ratio is a pure
734 function of the friction angle based on the small strain theory. However, Equations (80) and (81)
735 show that it is also dependent on the stiffness and compressibility of materials while taking the
736 geometry changes of the cylinder/sphere into consideration. A selection of results was
737 computed taking the soil cohesion as 0, the Poisson’ ratio as 0.3 and the dilation angle as 0 and

32
738 is plotted in Figure 15. The example results indicate that due consideration should be given to
739 the soil deformation in the shakedown analysis. Its influences on the optimal thickness become
740 greater for a larger value of the friction angle φ and a smaller value of the shear modulus. Not
741 surprisingly, results predicted by the present solution converge to those reported by Zhao and
742 Wang (2010) when the soil is sufficiently stiff (e.g. G/p0=500 in Figure 15).

16 5.6
Zhao and Wang (2010)
14 Zhao and Wang (2010)
G/p0=50 4.8 G/p0=50
12 G/p0=100
G/p0=100
G/p0=200
4.0 G/p0=200
b0 /a0

10

b0 /a0
G/p0=500 G/p0=50
G/p0=50 G/p0=500
8 3.2
500
6 500
2.4
4

2 1.6
0 5 10 15 20 25 30 35 40 45 0 5 10 15 20 25 30 35 40 45
(a) / (b) /

(a) hollow cylinders (b) hollow spheres

Figure15. Optimal thickness ratio b0/a0 at various shear moduli and friction angles

743 Conclusion

744 This paper presents analytical solutions for quasi-static contraction analysis of a thick-walled
745 cylinder/sphere of dilatant soils. The unloading is assumed to start after an arbitrary magnitude
746 of loading. The logarithmic strain definition is adopted in the plastic zone so that large strain
747 effects are taken into account. A linear elastic perfectly-plastic model is used. The plasticity of
748 the soil is described by adopting the Mohr-Coulomb yield criterion with a non-associated plastic
749 flow rule. The solutions are able to calculate the stress and displacement distribution in the soil
750 at any stage of the unloading process. They are validated by comparing with corresponding
751 analytical solutions for the case of an infinite soil mass. Parametric studies showed that both
752 the reference stress state and the cavity geometry parameters may greatly influence the cavity
753 contraction behaviour, in particular, for thin cylinders and spheres.

754 The new solutions are useful in modelling many geotechnical problems. Among them, three
755 typical applications are demonstrated, including: preliminary design of the thickness of the
756 frozen cylinder walls under either stress or displacement control, interpretation of laboratory
757 pressuremeter tests with consideration of the finite boundary effect, and determination of the
758 optimal thickness of cylinders/spheres based on the shakedown concept considering large
759 deformation effects. Additionally, the closed-form solutions can also be used to verify elastic-
760 plastic numerical methods in analysing the Mohr-Coulomb materials, in particular, in a finite
761 soil mass.
33
762 Disclosure statement

763 No potential conflict of interest was reported by the authors.

764 Funding

765 The work presented in this paper has been supported by the ‘Taishan’ Youth Scholar Program
766 of Shandong Province, China (No. tsqn201909016, No. tsqn201909044) and ‘Qilu’ Scholar
767 Program of Shandong University.

768 References

769 Abdulhadi, N. O., Germaine, J. T., & Whittle, A. J. (2011). Thick-walled cylinder
770 testing of clays for the study of wellbore instability. Geotechnical Testing Journal,
771 34(6), 746-754
772 Ajalloeian, R. (1996). An experimental study of finite pressuremeter length effects in
773 sand. (PhD), University of Newcastle, Australia.
774 Ajalloeian, R., & Yu, H. S. (1998). Chamber studies of the effects of pressuremeter
775 geometry on test results in sand. Géotechnique, 48(5), 621-636
776 Alsiny, A., Vardoulakis, I., & Drescher, A. (1992). Deformation localization in cavity
777 inflation experiments on dry sand. Géotechnique, 42(3), 395-410
778 Andersland, O. B., & Ladanyi, B. (2004). Frozen ground engineering (Second Edition).
779 Hoboken, New Jersey: John Wiley & Sons.
780 Brown, E. T., Bray, J. W., Ladanyi, B., & Hoek, E. (1983). Ground response curves for
781 rock tunnels. Journal of Geotechnical Engineering, 109(1), 15-39
782 Carter, J. (1988). A semi‐analytical solution for swelling around a borehole.
783 International Journal for Numerical and Analytical Methods in Geomechanics, 12(2),
784 197-212
785 Chadwick, P. (1959). The quasi-static expansion of a spherical cavity in metals and
786 ideal soils. The Quarterly Journal of Mechanics and Applied Mathematics, 12(1),
787 52-71
788 Chen, S. L., & Abousleiman, Y. N. (2016). Drained and undrained analyses of
789 cylindrical cavity contractions by bounding surface plasticity. Canadian
790 Geotechnical Journal, 53(9), 1398-1411
791 Chen, S. L., Abousleiman, Y. N., & Muraleetharan, K. K. (2012). Closed-form
792 elastoplastic solution for the wellbore problem in strain hardening/softening rock
34
793 formations. International Journal of Geomechanics, 12(4), 494-507
794 Cheng, Y., & Yang, H.-W. (2019). Exact solution for drained spherical cavity expansion
795 in saturated soils of finite radial extent. International Journal for Numerical and
796 Analytical Methods in Geomechanics, 43, 1594–1611. doi: doi:
797 org/10.1002/nag.2924
798 Clarke, B. G. (1995). Pressuremeters in geotechnical design. London: Blackie
799 Academic & Professional.
800 Collins, I. F., & Yu, H. S. (1996). Undrained cavity expansions in critical state soils.
801 International Journal for Numerical and Analytical Methods in Geomechanics, 20(7),
802 489-516
803 Fahey, M. (1986). Expansion of a thick cylinder of sand: a laboratory simulation of the
804 pressuremeter test. Geotechnique, 36(3), 397-424
805 Fahey, M., & Carter, J. P. (1993). A finite element study of the pressuremeter test in
806 sand using a nonlinear elastic plastic model. Canadian Geotechnical Journal, 30(2),
807 348-362
808 Ferreira, R., & Robertson, P. (1992). Interpretation of undrained self-boring
809 pressuremeter test results incorporating unloading. Canadian Geotechnical Journal,
810 29(6), 918-928
811 Franza, A., Marshall, A., & Zhou, B. (2019). Greenfield tunnelling in sands: the effects
812 of soil density and relative depth. Géotechnique, 69(4), 297-307. doi: doi:
813 org/10.1680/jgeot.17.P.091
814 Gao, X. L., Wen, J. F., Xuan, F.-Z., & Tu, S. T. (2015). Autofrettage and Shakedown
815 Analyses of an Internally Pressurized Thick-Walled Cylinder Based on Strain
816 Gradient Plasticity Solutions. Journal of Applied Mechanics, 82(4), 041010
817 Geng, Y., Yu, H. S., & McDowell, G. R. (2012). Discrete element modelling of cavity
818 expansion and pressuremeter test. Geomechanics and Geoengineering, 8(3), 179-190.
819 doi: doi: org/10.1080/17486025.2012.735375
820 Ghionna, V. N., & Jamiolkowski, M. (1991). A critical appraisal of calibration chamber
821 testing of sands the First International Symposium on Calibration Chamber Testing
822 (pp. 13-40). Potsdam, New York: Elsevier.
823 Grant, R. J. (1998). Movements around a tunnel in two-layer ground. (PhD thesis), City
824 University London, London.
825 Hill, R. (1950). The mathematical theory of plasticity. London: Oxford University Press.

35
826 Houlsby, G. T., & Carter, J. P. (1993). The effects of pressuremeter geometry on the
827 results of tests in clay. Géotechnique, 43(4), 567-576
828 Houlsby, G. T., Clarke, B. G., & Wroth, C. P. (1985). Analysis of the unloading of a
829 pressuremeter in sand the 2nd International Symposium on The Pressuremeter and
830 its Marine Applications (pp. 245-262). West Conshohocken, PA: ASTM International.
831 Houlsby, G. T., & Withers, N. J. (1988). Analysis of the cone pressuremeter test in clay.
832 Géotechnique, 38(4), 575-587
833 Hughes, J. M. O., & Robertson, P. K. (1985). Full-displacement pressuremeter testing
834 in sand. Canadian Geotechnical Journal, 22(3), 298-307
835 Hughes, J. M. O., Wroth, C. P., & Windle, D. (1977). Pressuremeter tests in sands.
836 Géotechnique, 27(4), 455-477
837 Jefferies, M. G. (1988). Determination of horizontal geostatic stress in clay with self-
838 bored pressuremeter. Canadian Geotechnical Journal, 25(3), 559-573
839 Jewell, R., Fahey, M., & Wroth, C. (1980). Laboratory studies of the pressuremeter test
840 in sand. Geotechnique, 30(4), 507-531
841 Jirari, S. E., Wong, H., Deleruyelle, F., Branque, D., Berthoz, N., & Leo, C. (2020).
842 Analytical modelling of a tunnel accounting for elastoplastic unloading and reloading
843 with reverse yielding and plastic flow. Computers and Geotechnics, 121, 103441.
844 doi: doi.org/10.1016/j.compgeo.2020.103441
845 Juran, I., & BenSaid, M. (1987). Cavity expansion tests in a hollow cylinder cell.
846 Geotechnical Testing Journal, 10(4), 203-212
847 Klein, J., & Gerthold, A. (1979). Die Fliessbedingung von Drucker/Prager im Vergleich
848 zu anderen Bruchkriterien bei der Bemessung von Gefrierschächten. Bautechnik,
849 11(1979), 368-375
850 Mair, R. J. (1979). Centrifugal modelling of tunnel construction in soft clay. (PhD
851 thesis), Cambridge University, Cambridge, UK.
852 Mair, R. J., & Taylor, R. N. (1993). Prediction of clay behaviour around tunnels using
853 plasticity solutions Predictive Soil Mechanics: Proceedings of the Wroth Memorial
854 Symposium (pp. 449-463). Oxford, UK: Thomas Telford.
855 Mair, R. J., & Wood, D. M. (1987). Pressuremeter testing: methods and interpretation.
856 London, UK: CIRIA-Butterworths.
857 Manassero, M. (1989). Stress-strain relationships from drained self-boring
858 pressuremeter tests in sand. Geotechnique, 39(2), 293-307

36
859 Mo, P. Q., & Yu, H. S. (2017). Undrained Cavity-Contraction Analysis for Prediction
860 of Soil Behavior around Tunnels. International Journal of Geomechanics, 17(5),
861 04016121-04016121-04016110. doi: DOI: 10.1061/(ASCE)GM.1943-
862 5622.0000816.
863 Ogawa, T., & Lo, K. Y. (1987). Effects of dilatancy and yield criteria on displacements
864 around tunnels. Canadian Geotechnical Journal, 24(1), 100-113
865 Park, K. H. (2014). Similarity solution for a spherical or circular opening in elastic-
866 strain softening rock mass. International Journal of Rock Mechanics and Mining
867 Sciences, 71, 151-159
868 Pournaghiazar, M., Russell, A. R., & Khalili, N. (2013). Drained cavity expansions in
869 soils of finite radial extent subjected to two boundary conditions. International
870 Journal for Numerical and Analytical Methods in Geomechanics, 37(4), 331-352
871 Reed, M. B. (1986). Stresses and displacements around a cylindrical cavity in soft rock.
872 IMA Journal of Applied Mathematics, 36(3), 223-245
873 Rowe, P. W. (1962). The stress-dilatancy relation for static equilibrium of an assembly
874 of particles in contact the royal society of London a: mathematical, physical and
875 engineering sciences (Vol. 269, pp. 500-527): The Royal Society.
876 Salgado, R., Mitchell, J. K., & Jamiolkowski, M. (1998). Calibration chamber size
877 effects on penetration resistance in sand. Journal of Geotechnical and
878 Geoenvironmental Engineering, 124(9), 878-888
879 Sanger, F., & Sayles, F. (1979). Thermal and rheological computations for artificially
880 frozen ground construction. Engineering geology, 13(1-4), 311-337
881 Schnaid, F., & Houlsby, G. (1991). An assessment of chamber size effects in the
882 calibration of in situ tests in sand. Geotechnique, 41(3), 437-445
883 Schnaid, F., & Houlsby, G. (1992). Measurement of the properties of sand by the cone
884 pressuremeter test. Géotechnique, 42(4), 587-601
885 Schnaid, F., Ortigao, J. A., Mantaras, F., Cunha, R., & MacGregor, I. (2000). Analysis
886 of self-boring pressuremeter (SBPM) and Marchetti dilatometer (DMT) tests in
887 granite saprolites. Canadian geotechnical journal, 37(4), 796-810
888 Sharan, S. K. (2008). Analytical solutions for stresses and displacements around a
889 circular opening in a generalized Hoek–Brown rock. International Journal of Rock
890 Mechanics and Mining Sciences, 1(45), 78-85
891 Shuttle, D. (2007). Cylindrical cavity expansion and contraction in Tresca soil.

37
892 Géotechnique, 57(3), 305-308
893 Viggiani, G., & Casini, F. (2015). Artificial ground freezing: from applications and case
894 studies to fundamental research the XVI ECSMGE: Geotechnical Engineering for
895 Infrastructure and Development (pp. 65-92). Edinburgh.
896 Vrakas, A., & Anagnostou, G. (2014). A finite strain closed‐form solution for the
897 elastoplastic ground response curve in tunnelling. International Journal for
898 Numerical and Analytical Methods in Geomechanics, 38(11), 1131-1148
899 Wen, J.-F., Gao, X.-L., Xuan, F.-Z., & Tu, S.-T. (2017). Autofrettage and shakedown
900 analyses of an internally pressurized thick-walled spherical shell based on two strain
901 gradient plasticity solutions. Acta Mechanica, 228(1), 89-105
902 Withers, N., Howie, J., Hughes, J., & Robinson, P. (1989). Performance and analysis of
903 cone pressuremeter tests in sands. Geotechnique, 39(3), 433-454
904 Wroth, C. (1982). British experience with the self-boring pressuremeter the 1st
905 International Symposium on The Pressuremeter and its Marine Applications (pp.
906 143-164). Paris.
907 Wroth, C. P. (1984). The interpretation of in situ soil tests. Géotechnique, 34(4), 449-
908 489
909 Xu, S. Q., & Yu, M. H. (2005). Shakedown analysis of thick-walled cylinders subjected
910 to internal pressure with the unified strength criterion. International journal of
911 pressure vessels and piping, 82(9), 706-712
912 Yu, H. S. (1992). Expansion of a thick cylinder of soils. Computers and Geotechnics,
913 14(1), 21-41
914 Yu, H. S. (1993). Finite elastoplastic deformation of an internally pressurized hollow
915 sphere. Acta Mechaniea Solida Sinica (English Edition), 6(1), 81–97
916 Yu, H. S. (1994). State parameter from self-boring pressuremeter tests in sand. Journal
917 of geotechnical engineering, 120(12), 2118-2135
918 Yu, H. S. (1996). Interpretation of pressuremeter unloading tests in sands. Géotechnique,
919 46(1), 17-31
920 Yu, H. S. (2000). Cavity expansion methods in geomechanics. Dordrecht, the
921 Netherlands: Kluwer Academic Publishers.
922 Yu, H. S., & Houlsby, G. T. (1991). Finite cavity expansion in dilatant soils: loading
923 analysis. Géotechnique, 41(2), 173-183
924 Yu, H. S., & Houlsby, G. T. (1995). A large strain analytical solution for cavity

38
925 contraction in dilatant soils. International Journal for Numerical and Analytical
926 Methods in Geomechanics, 19(11), 793-811
927 Yu, H. S., & Rowe, R. K. (1999). Plasticity solutions for soil behaviour around
928 contracting cavities and tunnels. International Journal for Numerical and Analytical
929 Methods in Geomechanics, 23(12), 1245-1279
930 Yu, H. S., Zhuang, P. Z., & Mo, P. Q. (2019). A unified critical state model for
931 geomaterials with an application to tunnelling. Journal of Rock Mechanics and
932 Geotechnical Engineering, 11(3), 464-480. doi: org/10.1016/j.jrmge.2018.09.004
933 Zhang, B., Yang, W., & Wang, B. (2018). Plastic Design Theory of Frozen Wall
934 Thickness in an Ultradeep Soil Layer Considering Large Deformation Characteristics.
935 Mathematical Problems in Engineering, 2018, 8513413. doi: doi:
936 org/10.1155/2018/8513413
937 Zhao, J., & Wang, G. (2010). Unloading and reverse yielding of a finite cavity in a
938 bounded cohesive–frictional medium. Computers and Geotechnics, 37(1), 239-245

939 Appendix

940 At the moment that the elastic-plastic boundary reaches the outer radius of the cavity upon
941 loading, the displacement at r=b20 can be calculated from the elastic displacement solution of
942 Yu (1992, 1993) as:

 1  v  kv  Y    1 p0  
 
943 b0  b20 1   (84)

 M 


944 The displacement of the inner wall of the cavity (i.e. r=a20) can be obtained from the large
945 strain plastic displacement solution as below.

 k
1
k
1 

 0 b    a  

946  n 
4

    
0
  (85)
 20   20  
n 0 b b


947 in which  
 k    , and
k    1

 k   1 3n b20
 ln , if n =
  n! a20
948 n  
4
(86)
k  1 n  
 3  
n


  n    n!   b20 a20    1 , otherwise
 

39
 Y    1 p0   kv v 
949 3     k 1  v  kv   k   (87)
M    1  1 v2  k  1  v  2  k  

Y    1 p0  kv kv 
950 ln     k 1  v  kv    (88)
M    1  1 v2  k  1  v  2  k  

40
951 Notation

952 a0, b0 initial inner and outer radii of a cylinder/sphere

953 p0, p20 in-situ stress and cavity pressure at the beginning of unloading

954 pin internal cavity pressure

955 a20, b20 initial inner and outer radii of a cylinder/sphere at the beginning of
956 unloading

957 k k=1 for a cylinder and k=2 for a sphere

958 r, θ, z cylindrical polar coordinates

959 r, θ, ϕ spherical polar coordinates

960 σr, σθ, σz radial, circumferential and axial stresses

961 Δσr, Δσθ incremental radial and circumferential stresses due to unloading

962 u20, u2 radial displacement at the beginning of unloading and during unloading

963 Δu incremental radial displacement due to unloading

964 σr2, σθ2 radial and circumferential stresses during unloading

965 σr20, σθ20 radial and circumferential stresses at the beginning of unloading

966 r0 initial radius of a given soil particle

967 a2, b2 inner and outer radii of the cavity during unloading

968 p1, p2 internal cavity pressures during loading and unloading

969 c1, c20, c2 radii of the elastic-plastic interface caused by loading in the loading
970 process, at the end of loading and during unloading

971 d2, d 20 radius of the elastic-plastic interface caused by unloading and its initial

972 value

973 φ, c,  soil friction angle, cohesion and dilation angle

974 α, Y functions of soil cohesion and friction angle

975  function of dilation angle

976 p1elim elastic limit pressure in the loading process

41
 977 a1, b1 inner and outer radii of a cavity in the loading process

978 r1, r2 radial radii during loading and unloading

979 σr1, σθ1 radial and circumferential stresses during loading

980 A1, B1 constants of integration in the loading process

981 εr, εθ radial and circumferential strains

982 v, E, G Poisson’s ratio, elastic modulus and shear modulus

983 M function of Poisson’s ratio and elastic modulus

984 a2, b2 inner and outer radii of a cavity in the unloading process

985 p2elim-1 elastic unloading limit of case I

986 A2-1, B2-1 constants of integration in the unloading process of case I

987 p2fp-lim fully plastic unloading limit of case I

988  rp ,  p plastic radial and circumferential strains

989  re ,  e elastic radial and circumferential strains

990 η, ω, θ, γ non-dimensional coefficients of case I

991 n! factorial of n

992 A2-1, B2-1, A2-2, B2-2,

993 A2-3, B2-3, A2-4, A2-5 constants of integration in the unloading process

994 p2elim-2 elastic unloading limit of cases II and III

995 Δεr, Δεθ incremental radial and circumferential strains due to unloading

996 ω1, λ1,  ,ρ non-dimensional coefficients of case II

997 1n , 2n , 3n , 4n infinite power series

998 η1, ω2, λ2 non-dimensional coefficients of case III

999 L, D length and diameter of pressuremeters

1000 Dr relative density of sand

1001 φcv, φps, φtri critical state friction angle, plane strain friction angle and

42
1002 triaxial friction angle of sand

1003 R2 the correlation coefficient

1004 p1plim plastic limit during loading

1005 su shear strength of clay

1006  , ω3 ,  non-dimensional coefficients in Appendix

43