Dissipative propagation of pressure waves along the ...

0 downloads 0 Views 2MB Size Report
Jun 8, 2016 - For a Von Mises material, this leads to a telegraphy equation in the. 58 pressure. In many ways our approach follows the derivation for an ...
1

2

Dissipative propagation of pressure waves along the slip-lines of yielding material Jack Moorea,b,c,1,, Ali Karrechd , Michael Smalla,b , Emmanouil Veveakise,b , Klaus Regenauer-Liebe

3 4

a

School of Mathematics and Statistics, The University of Western Australia b Mineral Resources, CSIRO c School of Earth and Environment, The University of Western Australia d School of Civil, Environmental and Mining Engineering, The University of Western Australia e School of Petroleum Engineering, The University of New South Wales

5 6 7 8 9

10

Abstract

11

Traditional slip-line field theory considers steady-state pressure distributions. This work

12

examines the evolution of pressure at yield. We show that, for a two phase material in plane

13

strain at the point of plastic yield of the main constituent, pressure develops according to

14

dissipative wave mechanics. The pressure wave equations are examined in more detail for the

15

special case of radially symmetric plane strain, and for Von Mises and Drucker-Prager yield

16

envelopes. The positions of pressure troughs are compared to the separation of slip-lines

17

formed during the expansion by internal pressurisation of a cylinder of mild steel.

18

Keywords: slip-line field theory, wave mechanics, two phase material

19

Acknowledgements

20

JMM thanks Martin Paesold, Neville Fowkes and Thomas Poulet. JMM was supported

21

by the Prescott Postgraduate Scholarship of the University of Western Australia and the

22

OCE PhD Scholarship of CSIRO.

1

[email protected]

Preprint submitted to International Journal of Engineering Science

June 8, 2016

1

2

Dissipative propagation of pressure waves along the slip-lines of yielding material

3

Abstract

4

Traditional slip-line field theory considers steady-state pressure distributions. This work

5

examines the evolution of pressure at yield. We show that, for a two phase material in plane

6

strain at the point of plastic yield of the main constituent, pressure develops according to

7

dissipative wave mechanics. The pressure wave equations are examined in more detail for the

8

special case of radially symmetric plane strain, and for Von Mises and Drucker-Prager yield

9

envelopes. The positions of pressure troughs are compared to the separation of slip-lines

10

formed during the expansion by internal pressurisation of a cylinder of mild steel.

11

Keywords: slip-line field theory, wave mechanics, two phase material

12

List of symbols

13

Latin. Symbol

Description

a/b

Constant coefficients used in separation of variables.

A/B

The matrices describing the part of the rate of change of momentum which is proportional to the derivative of (p, ✓)T with respect to x1/2 .

B

The Biot coefficient.

c

The cohesion of a Drucker-Prager material.

C

The drag coefficient.

cn /dn /fn

Fourier coefficients.

cs

The speed of second sound.

cu

The wave speed under undrained conditions.

Preprint submitted to International Journal of Engineering Science

June 8, 2016

d

Half the distance separating wide parallel plates. (This symbol appears only in the appendices.)

d

The negative of the constant of proportionality between the -volume fraction

and specific interaction force f . (This symbol appears only

in a footnote.) D(↵/ ) Dt

The ↵/ material derivative.

f

The interaction force (exerted by

on ↵).

g(p)

The shear stress at which yield occurs at hydrostatic pressure p.

G↵

The shear modulus of ↵. (This symbol appears only in the appendices.)

H

A signature time scale.

k

The shear strength of a Von Mises material.

K

The bulk modulus relating pressure and volumetric strain.

Ks

The bulk modulus of the skeleton.

K↵/

The bulk modulus of ↵/ . (This symbol appears only in the appendices.)

l

The distance of steady-state laminar flow of an incompressible Newtonian fluid. (This symbol appears only in the appendices.)

L

A geometric length scale; the total length of a slip-line.

n

A positive integer.

n⇤

An upper/lower bound of the range of n corresponding to hyperbolic/trigonometric pressure eigenmodes pn .

nmax

The index of the highest pressure eigenmode used to produce plots.

p

The hydrostatic component of the e↵ective stress.

P

The pressure on an incompressible Newtonian fluid. (This symbol appears only in the appendices.)

pD

The discontinuous part of the pressure.

pf

The pore fluid pressure. (This symbol appears only in a footnote.)

2

po

The constant total pressure at the outer wall of the cylinder.

pQ

The steady state solution to the pressure for vanishing specific discharge, V.



The overpressure, p



The dimensionless overpressure.



The rescaled dimensionless overpressure.

\

The continuous part of the overpressure.

p

\

|ˆ p|max / p

max

pQ .

Attenuation factors used to plot overpressures/the continuous parts of overpressures.

q

The shear component of the e↵ective stress.

r

The radial polar coordinate.

r0

The value of r where s1 = s10 .

ri

The value of the radial coordinate at the inner wall of the cylinder.

ro

The value of the radial coordinate at the outer wall of the cylinder.

S

A function of s˜1 used in separation of variables.

s10

The value of s1 where r = r0 .

si

Coordinates the level sets of which are characteristics of the momentum balance.

s˜i

The dimensionless coordinates.

T

A function of t˜ used in separation of variables.



The dimensionless time.

u

The displacement.



The mean speed of the incompressible Newtonian fluid. (This symbol appears only in the appendices.)

v (1/2)

The velocity of ↵/ .

V

The specific discharge of .

w

The width of wide parallel plates. (This symbol appears only in the appendices.)

3

w(J)

A left eigenvector corresponding to the left eigenvalue µ(J) .

xi

Cartesian coordinates.

y

A coordinate describing location between wide parallel plates. (This symbol appears only in the appendices.)

y

The column vector of coordinates of the stress state, (p, ✓)T .

z/z

A variable used to express identities or introduce notation.

zs

The expression in coordinates (s1 , s2 ) of a vector written z in coordinates (x1 , x2 ).

14

Greek. Symbol

Description

↵/

The phases. The di↵usivity.

ij

s1

The Kronecker delta. A signature distance between characteristics.

✏ij

The Eulerian infinitesimal strain.

"

The dilatation.



The proportionality between the initial dimensionless rate of change of dimensionless overpressure and the initial dimensionless pressure.



The dynamic viscosity of the incompressible Newtonian fluid. (This symbol appears only in the appendices.)



An angle such that ✓ + ⇡/4 is the angle counter-clockwise from the positive x1 -axis to the direction of the maximum tensile stress.



The Heaviside theta function.



The dimensionless parameter defined as the ratio of the total length of a slip-line L to the material length scale 2 /cs . n

The inverse of the time constant corresponding to the hyperbolic pressure eigenmode pn .

µ(J)

A left eigenvalue corresponding to the left eigenvector w(J) . 4

⌫↵

The Poisson ratio of ↵. (This symbol appears only in the appendices.)



The sinuosity. (This symbol appears only in the appendices.)

$

A constant used in separation of variables.

⇢1/2

The relative densities of ↵/ .

⇢↵/

The densities of ↵/ .

1

The maximum tensile stress.

(1/2)

The partial stress of ↵/ .

00

The e↵ective stress (experienced by ↵). The stress in .

&

Half the angle by which the s1 and s2 axes di↵er from normality. The volume fraction of . The azimuthal polar coordinate. A di↵erence between angles, ✓

!n

.

The angular frequency corresponding to the trigonometric pressure eigenmode pn .

⌦t

15

Other. Symbol

16

A material volume.

Description



The di↵erential operator

rs ·

The di↵erential operator

⇣ ⇣

@ , @ @x1 @x1 @ , @ @s1 @s1





·. ·.

1. Introduction

17

This work examines the evolution of pressure of plastically yielding material. It will

18

show that, for a two phase material in plane strain, pressure propagates along slip-lines in

19

dissipative waves.

20

Slip-lines are the characteristics of the quasistatic momentum balance and are used in

21

the design or description of, among other things, welds (Khan et al., 2014), hardness tests 5

22

(Jackson et al., 2015), retaining walls (Vo and Russell, 2014) and pipeline foundations (Gao

23

et al., 2015). A particularly elegant application of slip-line field theory (SFT) for Von Mises

24

and Drucker-Prager materials, and one relevant to the plastic design of pressure vessels

25

(Sowerby and Johnson, 1970), arises for radial stress of a cylinder in plane-strain. As Hill

26

(1950) and Shield (1953) discuss, slip-lines are both the only curves across which pressure

27

and velocity fields can have discontinuities, and are directed at a constant angle to the

28

direction of maximal tensile stress. Thus SFT predicts that any strain localisations will

29

occur along logarithmic spirals. Sowerby and Johnson (1970) and Papamichos et al. (2001)

30

verify these predictions via pressurisation experiments from which come beautiful networks

31

of intersecting failure lines following logarithmic spirals. Figure 9.3a of Sowerby and Johnson

32

(1970) appears here as figure 4a.

33

Thus the quasistatic momentum balance r ·

= 0 determines the geometry of lines of

34

failure for a steel cylinder internally pressurised to yield. However, the analysis does not

35

identify the spacing of failure lines which, at least in figure 4a, is intriguingly regular. There

36

are a few di↵erent approaches with the potential to add information about the spacing of

37

slip-lines to the existing, correct prediction of their logarithmic spiral shape.

38

An example is the dynamic strain ageing model of McCormick (1988). He considers the

39

interplay of local solute concentration, which limits the dislocation motion which enables

40

plastic deformation, and strain rate, an increase of which allows more rapid di↵usion of

41

solute particles. Feedback between the two local properties can lead to sinusoidal strain

42

oscillations. (McCormick, 1988) From a dynamic strain ageing model, Estrin et al. (1991)

43

deduce that in tensile extension these spatially homogenous strain rate oscillations can lead

44

to strain inhomogeneous in space.

45

Any material deformation description which leads to predictions of spatial undulations

46

of strain rate has the potential, if incorporated into SFT, to resolve the spacing between

47

logarithmic spirals formed during radially symmetric plane strain. However, lines of strain

48

localisation in mild steel are an upper yield point phenomenon (Hertzberg et al., 2012, pp.

49

99-100), and the upper yield point is sensitive to pressure (Capp et al., 1973). It follows

6

50

that the locations of lines of strain localisation are sensitive to pressure inhomogeneities. For

51

this reason we seek to understand regular formation of the logarithmic spirals of figure 4a

52

by describing pressure.

53

The focus on pressure emulates Veveakis and Regenauer-Lieb (2015), who identified spa-

54

tially periodic pressure singularities in a two phase material in which the primary component

55

underwent plastic volumetric changes at a rate proportional to a power of the pressure. In-

56

stead of following Veveakis and Regenauer-Lieb (2015) in considering only the plastic com-

57

ponent of the volumetric response, we consider only the e↵ect of the elastic part of the

58

volumetric response. For a Von Mises material, this leads to a telegraphy equation in the

59

pressure. In many ways our approach follows the derivation for an elastic saturated material

60

of a telegraphy equation in the dilatation of Vardoulakis and Sulem (2005, Chapter 5).

61

The structure of this paper follows. Section 2.1 derives an approximate momentum

62

balance for a two phase material. Assuming a state of plane strain, section 2.2 identifies the

63

characteristics of this equation at the point of plastic yield. Along these curves the evolution

64

of pressure can be reduced from a two to a one-dimensional problem. Section 3 applies the

65

results of section 2, derived (with some simplifying assumptions) for a two phase material in

66

a state of plane strain, to a convenient material type and geometry. Section 4 uses separation

67

of variables to describe the evolution of the pressure distribution for a Von Mises cylinder

68

expanded via internal pressurisation. Section 4.2 employs parameters and initial conditions

69

pertaining to expansion of a cylinder of mild steel. As Appendix A discusses, the viscosity is

70

difficult to estimate precisely. Hence the section considers three distinct values of viscosity

71

spanning two orders of magnitude. It compares the output of each realisation of the model

72

to experimental observations. The final section, section 5, discusses results and suggests

73

ways this work could be continued.

74

2. Describing pressure evolution using conservation equations

75

76

Section 2.1 ends with a relationship between e↵ective stress and the specific discharge of the less prominent phase,

, which is a form of the momentum balance. Section 2.2 uses

7

77

the method of characteristics to reduce the spatial dimension of the relationship from two

78

to one.

79

2.1. The mass and momentum balance 1

Consider a composite of two phases ↵ and

. The volume fraction of the second phase

will be denoted . The relative densities ⇢1 and ⇢2 are related to the densities ⇢↵ and ⇢ of constituents ↵ and

by ⇢1 = (1

Balances of the mass of phases ⇢↵ and

)⇢↵ ,

⇢2 = ⇢ .

lead to

@⇢1 + r · ⇢1 v (1) = 0, @t @⇢2 + r · ⇢2 v (2) = 0, @t

(2) (3)

where v (1) and v (2) denote the velocities of phases ↵ and . Using the definition (1), 1 ⇢

(1)

1 (2) + ⇢↵

(3) can be written r·V =

r · v (1)

⇢↵

D(1) ⇢↵ Dt

v (2)

v (1)

1

D(2) ⇢ , ⇢ Dt

(4)

where V=

D(J) Dt

=

@ @t

+ v (J) · r is a material derivative.

80

is the specific discharge of

81

The material derivative indexed J is the rate of change which would be measured by an

82

observer travelling with the phase indexed J. The next section (section 2.2) uses the mass

83

balance (4). 1

and for J = 1, 2,

(5)

The phases do not need to be chemically distinct. For example, Vardoulakis (1989) partitions a granular

material into a “frail” phase and a “competent” phase, which contribute di↵erently to the total stress, according to number of grain contacts.

8

In the absence of body forces, balances of the linear momentum of ↵ and r·

(1)



(2)

lead to

D(1) v (1) Dt

(6)

D(2) v (2) f = ⇢2 , Dt

(7)

+ f = ⇢1

and

where f is the specific interaction force exerted on the primary constituent ↵ by the secondary constituent , and

(1)

and

(2)

are the partial stresses of ↵ and . The interaction force

f is assumed to oppose relative motion between the phases. Specifically, it is assumed that the interaction force is linear in the relative velocity; f=

C v (1)

v (2) ,

(8)

where C ⌘ C( ) is a function of the volume fraction of .2 Appendix A considers estimation of the drag coefficient C. Decompose the total stress (1)

where

00

+

(2)

=

00

+B

,

(9)

is the e↵ective stress which ↵ experiences, B is the Biot coe↵cient and

stress in constituent , related to the partial stress of (2)

=

is the

by

.

(10)

84

Coussy (2004, pp. 78–79) shows that the Biot coefficient of a linear isotropic material can

85

be estimated from the total bulk modulus K and the skeleton bulk modulus Ks . Appendix

86

B, upon which relies the application comprising section 4.2, uses this result. 2

Vardoulakis and Sulem (2005) consider a situation in which the secondary component is a fluid. In this

case shear stresses are negligible, so that the partial stress tensor is

(2)

=

pf 1. They show that if spatial

variations of -volume fraction and convective and inertial terms can be neglected then the choice C = d where d is constant recovers, via (7), Darcy’s law V = pressure.

9

1 d rpf .

In this expression, pf is the fluid pore

The fields

(1)

(2)

,

, f and v (2) will be eliminated from (5), (6), (7), (8), (9) and

,

(10). Substitution of (10) into (7) gives r · the -volume fraction

D(2) (2) v . Dt

f = ⇢

Hence, assuming that

varies slowly in space and neglecting convective terms, r·

Together, (9) and (10) show that

= (1)

1

=

f +⇢

00

@v (2) . @t

+ (B

(11)

)

. Using (11), treating the Biot

coefficient B as constant, and again assuming that the -volume fraction space, the divergence of this equation can be written r· Substituting this expression for r·

(1)

into (6), r·

By (8), assuming that the -volume fraction

00

(1)

= r·

+ B f +(B

00

+B

)⇢

varies slowly in f +(B

@v (2) @t

= (1

)⇢

@v (2) . @t (1)

)⇢↵ @v@t .

varies slowly in time and using the definition

(5) of the specific discharge, this can be written r·

00

=

C

B

V 2

B



@ V + ((1 @t

)⇢↵

(B

)⇢ )

@v (1) . @t

(12)

87

Via the method of characteristics and a choice of constitutive relationship, the next

88

section (section 2.2) reduces the momentum balance (12) from two to one spatial variables.

89

2.2. Pressure at plastic yield

90

This section (section 2.2) considers a material the yield condition of which is an alge-

91

braic relationship between the hydrostatic and the shear components of the Biot e↵ective

92

pressure. Following Paesold et al. (2016), it derives a description of the characteristics of

93

(12), the momentum balance, at the point of plastic yield. As the section will show, along

94

the characteristics3 of the momentum balance, termed slip-lines, pressure evolves in a single

95

dimension. Assume plane strain in the plane parameterised by Cartesian coordinates (x1 , x2 ) (see figure 1a or figure 1 of Shield (1953)). Assume that ↵ is in a state of yield described4 by q = g(p), 3

Note that if the right hand side of the momentum balance (12) is considered a function not only of s1 ,

s2 , t, p (s1 , s2 , t) and

@p @t

(s1 , s2 , t) but also of the spatial derivatives

@p @s1

(s1 , s2 , t) and

@p @s2

(s1 , s2 , t) then the

coming considerations would not necessarily identify characteristics. 4 Here arises a major assumption of this work: that the isotropic and deviatoric components of the Biot

10

1 3

where p =

P3

i=1

00

ii

is the hydrostatic and q =

qP 3

i,j=1

(

00

ij

+p

00

ij ) (

ij

+p

ij )

is the

shear component of the e↵ective stress. Then (because conservation of angular momentum implies symmetry of the stress tensor), as Johnson et al. (1982, p. 31) show, 00 11

=

p

00 12

g(p) sin(2✓),

= g(p) cos(2✓) =

00 21 ,

00 22

=

p + g(p) sin(2✓),

where ✓ + ⇡/4 is the angle counter-clockwise from the positive x1 -axis to the direction of the maximum principal stress

1

(see figure 1a). Following Paesold et al. (2016)5 , the momentum

balance (12) can be written A

@y @y +B = @x1 @x2

C

B

V 2

B



@ V + ((1 @t

)⇢↵

(B

)⇢ )

@ (1) v , @t

(13)

where 0

A,@

g 0 (p) sin(2✓)

1 0

g (p) cos(2✓)

2g(p) cos(2✓) 2g(p) sin(2✓)

1

0

A, B , @

g 0 (p) cos(2✓) 0

2g(p) sin(2✓)

1 + g (p) sin(2✓)

2g(p) cos(2✓)

1

A,

(14)

y1 , p and y2 , ✓. Let w(1) , w(2) denote two linearly independent left eigenvectors with corresponding eigenvalues µ(1) , µ(2) which satisfy, for J = 1, 2, µ(J) w(J) T · A = w(J) T · B. Assuming that |g 0 (p)| < 1, solutions to µ(J) A µ(1) = tan(✓

&),

B = 0 are

µ(2) = tan(✓ + ⇡/2 + &),

where the angle & 2 ( ⇡/4, ⇡/4) is defined by sin(2&) = g 0 (p). The corresponding eigenvectors are 0

w(1) = sec(2&) @

sin (✓ + ⇡/2 + &) cos (✓ + ⇡/2 + &)

1

A,

0

w(2) = sec(2&) @

sin (✓ cos (✓

&) &)

1

A.

e↵ective stress given by (9) describe (via a yield envelope) yield of the material ↵. Of course, there are more general descriptions of yield in a two phase material. For example, stress could be partitioned according to a version of (9) with the scalar B replaced by a tensor. 5 Paesold et al. (2016) actually considered a steady state momentum balance of a single phase material, for which the right hand sides of (12) and (13) would be zero.

11

Of course, any multiple of an eigenvector is also an eigenvector. However, these choices 0 1w(J) 0 1 have the convenient property @ defined such that directed at ✓

@ @s1

and

@ @s2

w(1) T T

A=@

@s1 @x1

@s1 @x2

@s2 @x1

@s2 @x2

A, where (s1 , s2 ) are coordinates

w(2) have the same Euclidean magnitude as

@ @x1

and

@ , @x2

but are

& and ✓ + ⇡/2 + & to the positive x1 -direction (see figure 1a); 0 1 0 1 @x1 @x1 cos(✓ &) cos(✓ + ⇡/2 + &) @ @s1 @s2 A = @ A. @x2 @x2 sin(✓ &) sin(✓ + ⇡/2 + &) @s1 @s2

(15)

96

If & = 0 then this transformation is a rotation. The si axes intersect at an angle ⇡/2 + 2&

97

and are oriented symmetrically about the direction of maximum stress (see figure 1a).

Figure 1: Coordinate systems. (a). The angle ✓ is defined such that the maximum tensile stress

1

occurs

at ✓ + ⇡/4 counter-clockwise from the positive x1 axis, where (x1 , x2 ) is a Cartesian coordinate system. The si axes intersect at an angle ⇡/2 + 2& and are oriented symmetrically about the direction of maximum tensile stress. (b). Coordinate directions for the Von Mises cylinder of section 4. Since & = 0, the s1 - and s2 -directions are everywhere orthogonal. The direction of maximum tensile stress is everywhere orthogonal to the radial direction. At the point considered in the diagram s1 = 0 and r = ro .

For any vector written z in coordinates (x1 , x2 ), let z s be its expression in coordinates

12

0

w(1) T

(s1 , s2 ). With this notation, the momentum balance (13) left-multiplied by @ 0 @

@p @s1 @p @s2

1

@✓ 2g(p) sec(2&) @s 1

A=

@✓ + 2g(p) sec(2&) @s 2

C

B 2

B

Vs

w(2)

T

1

A is

@V s @t



(1)

+ ((1

)⇢↵

(B

)⇢ )

@v s . @t

(16)

98

Thus the transformation (15) decouples the momentum balance. Lines parameterised by

99

s1 and s2 according to (15) are the characteristics of the momentum balance (12), which

100

are called slip-lines. As shown in (15) or figure 1a, the orientations of the characteristics

101

parameterised by s1 and s2 are ✓

102

g 0 (p) = 0, characteristics intersect at right angles.

103

& and ✓ + ⇡/2 + & respectively. In particular, where

Where the fields comprising the right hand side of (16) ( , V s ,

@V s @t

and

(1)

@v s @t

) are known

104

as functions of space and time the decoupled momentum balance (16) would represent an

105

informative constraint between geometry and pressure. As will be seen, spatial di↵erentiation

106

allows more general use of the decoupled momentum balance (16). The chain rule shows that 0

r · z = rs · z s + sec(2&)z s T · @ where rs · ,



@ , @ @s1 @s2







sin(2&) @s@ 1

@ @s1

P





(✓

&)

+ sin(2&) @s@ 2 (✓ + ⇡/2 + &)

1

A,

(17)

· is a di↵erential operator. Hence, neglecting convective terms, rs · v (1) s =

where ✏ ,

+

@ @s2

i "ii is the dilatation, "ij ,

1 2



@ui @xj

@✏ , @t

+

@uj @xj

(18) ⌘

is the Eulerian infinitesimal strain,

and u is displacement. So, by (4) and (17), neglecting convective terms, rs · V s =

@✏ @t

Assuming that the -volume fraction

1 ⇢↵

@⇢↵ @t

@⇢ . ⇢ @t

(19)

and densities ⇢a and ⇢ vary slowly in space, rs · 13

(16) can be written 2

rs p

2rs ·

g(p) sec(2&)



@✓ , @s1

@✓ @s2

◆T !

=

C

B 2

B

rs · V s

⇢ rs ·

@V s @t (1)

+ ((1

)⇢↵

)⇢ ) rs ·

(B

Hence, assuming commutativity of spatial derivatives rs and the time derivative

@v s . @t (20)

@ , @t

by the

approximations (18) and (19), ◆T ! @✓ @✓ rs p =2rs · g(p) sec(2&) , @s1 @s2 ✓ ✓ ◆ ◆ 2 B @✏ B @ ✏ +C 2 + (1 ) ⇢↵ + ⇢ @t @t2 ✓ ◆ ✓ B 1 @⇢↵ @⇢ B +C 2 + + (1 ⇢↵ @t ⇢ @t ✓

2

107

Some geometries imply relationships between @ @si

@ @s1

and

⇢ @ 2 ⇢↵ @ 2⇢ ) + ⇢↵ @t2 @t2

@ @s2



.

(21)

which allow elimination of the

108

dependence on one of the

in the above. The next section (section 3) shows how assuming

109

radial symmetry and particular forms for the yield envelope q = g(p) assists further analysis.

110

3. Radially symmetric plane strain For convenience, introduce polar coordinates (x1 , x2 ) = (r cos , r sin ). With this definition and the matrix of derivatives (15), the chain rule shows that 0 1 0 1 @r @r cos( &) cos( + ⇡/2 + &) @ @s1 @s2 A = @ A, @ @ 1 1 sin( &) r sin ( + ⇡/2 + &) @s1 @s2 r ,✓

(22)

111

where

. As Johnson et al. (1982, p. 68) observe, for a homogeneous and isotropic

112

material in plane strain under radially symmetric loading, the directions of principal stress

113

are radial and azimuthal. Hence the angle

114

with the instance depending on whether the azimuthal or radial tensile stress is greater.

115

Note that in either case, tan2

is either

= 1. 14

=

3⇡/4 or

=

⇡/4 (a constant),

If g 0 (p) is constant then & is constant. Hence the orientations of characteristics relative to the radial direction, which by the coordinate transformation (15) are + ⇡/2 + & = ✓ + ⇡/2 + &

&=✓

&

and

, are constant. In this case, by (22), along the characteristic

parameterised by s1 the radial coordinate is r = cos( 116

&) (s1

s10 ) + r0 ,

(23)

where s10 is the value of s1 at r = r0 . A computation using (15) and (22) shows that the derivative in the azimuthal direction is

@ @

= r sec 2&(sin( + &) @s@ 1 + cos(

&) @s@ 2 ). Since the radial symmetry of the problem

means that the pressure does not change in the azimuthal direction, ✓ ◆ @p sin( + &) @p sin2 ( + &) @ 2 p 2 = , rs p = 1 + . @s2 cos( &) @s1 cos2 ( &) @s1 2 117

(24)

3.1. Quasistatic pressure of Von Mises and Drucker-Prager materials Consider the steady-state form with vanishing specific discharge (V s = 0) of the momentum balance (16). Letting pQ denote this quasistatic pressure solution, 0 1 0 1 @pQ @✓ 2g(pQ ) sec(2&) @s1 0 @ @s1 A = @ A. @pQ @✓ + 2g(pQ ) sec(2&) @s 0 @s2 2

(25)

118

This section (section 3.1) identifies the pressure profile pQ along the characteristic parame-

119

terised by s1 of this momentum balance for the two classes of materials for which the slope

120

of the yield envelope g 0 (p) is constant: Von Mises and Drucker-Prager. The yield envelope of a non-hardening Von Mises material satisfies g(p) = k, where k is constant.6 Observing that since g 0 (p) = 0, & = 0, equations (22), (23) and (25) show that along the slip-line parameterised by s1 , (2k tan ) Hence, by (23) with & = 0, the pressure is ✓ ✓ s1 pQ = p0 2k tan ln cos 6

and

s10 r0



+1



1

dp =

= p0

(s1

2k tan

s10 + sec r0 )

ln



In this case, (25) reduces to Hencky’s equations (see, for example, Hill (1950, p. 135)) @ @s2

(pQ

2k) = 0.

15

r r0



@ @s1

,

1

ds1 .

(26)

(pQ + 2k) = 0

121

122

123

where p0 denotes the pressure at s1 = s10 (and thus, at r = r0 ). Recall that since either =

3⇡/4 or

=

= ±1. Sowerby and Johnson (1970) and Johnson et al.

⇡/4, tan

(1982) reached results equivalent to (26), also via SFT.

124

The yield envelope of a Drucker-Prager material satisfies g(p) = c+sin(2&)p, where c and &

125

are constant and sin(2&) 6= 0. In this case, (22) and (25) show that along the slip-line param-

126

eterised by s1 , (c csc(2&) + p)

129

2 tan(2&) tan( &) (s1 s10 + sec( &)r0 ) 1 ds1 . ⇣ ⇣ ⌘ ⌘ 2 tan(2&) tan( &) s1 s10 Hence the pressure is pQ = (c csc(2&) + p0 ) cos + 1 c csc(2&) = r0 ⇣ ⌘ 2 tan(2&) tan( &) (c csc(2&) + p0 ) rr0 c csc(2&), where p0 denotes the pressure at s1 = s10

130

results from which, in the case of radial symmetry, this expression for pQ would easily follow.

131

3.2. Pressure evolution of Von Mises and Drucker-Prager materials

127

128

1

dp =

(and thus, at r = r0 ), and the second equality follows from (23). Gao et al. (2015) reach

132

This section (section 3.2) and those which succeed it rely upon (21). This equation is

133

a derivative of the reformulation along its characteristics of the momentum balance (12).

134

The momentum balance (12) was derived by neglecting convective terms (that is, making

135

the approximation

136

The derivation also assumes that

137

remainder of the document it will be assumed that the drag coefficient C is constant in

138

space.

D(J) Dt

=

@ ) @t

and assuming that the Biot coefficient B is constant in space. varies slowly in space and time. Additionally, for the

Let pˆ denote the overpressure pˆ (s1 , s2 , t) , p (s1 , s2 , t)

pQ (s1 , s2 ), where pQ satisfies the

quasistatic momentum balance (25). The derivative (21) of the momentum balance can be written 0

rs 2 pˆ =rs · @

@pQ @s1 @pQ @s2

+

@✓ 2g(pQ ) sec 2& @s 1 @✓ 2g(pQ ) sec 2& @s 2

1

0

A + tan 2&rs · pˆ @

✓ ✓ ◆ ◆ 2 B @✏ B @ ✏ +C 2 + (1 ) ⇢↵ + ⇢ @t @t2 ✓ ◆ ✓ B 1 @⇢↵ @⇢ B +C 2 + + (1 ⇢↵ @t ⇢ @t

@✓ @s2

1 A

⇢ @ 2 ⇢↵ @ 2⇢ ) + ⇢↵ @t2 @t2

Using (24), twice using (22), and noting that because either 16

@✓ @s1

=

3⇡/4 or



.

(27)

=

⇡/4,

cos 2 = 0, it can be shown that 0 0 11 rs · @pˆ @

@✓ @s1

@✓ @s2

AA = 2 sin( r

&)

@ pˆ @s1

sin 2 cos 2&

pˆ . r2

(28)

139

Classically, increments of plastic deformation can be resolved as normal to the yield en-

140

velope when it is written as a function of the elastic component of strain (Hosford, 2010, p.

141

73). For a Von Mises material this corresponds to vanishing plastic volumetric strain (Rud-

142

nicki and Rice, 1975). Spitzig et al. (1975) investigated for martensic steel the relationships

143

between volumes and stresses at yield. They found volumetric changes orders of magnitude

144

less than those predicted by normality. Also, as Rice (1976) notes, to describe localisation

145

conditions for a range of geological materials, Rudnicki and Rice (1975) consider dilatancy

146

factors which are always less than or equal to, and usually less than, the corresponding co-

147

efficients of internal friction g 0 (p). This choice corresponds to plastic volumetric change less

148

than that associated with a normal flow rule (Rudnicki and Rice, 1975). Cases such that plastic volumetric changes can be neglected admit the volumetrically linear and isotropic volumetric relationship @ pˆ = @t

K

@✏ , @t

(29)

where K is the bulk modulus. In such cases and in radial symmetry, by (24) and (28), (27) can be written ✓ ◆ 1 sin2 ( + &) @ 2 pˆ 1 @ pˆ 1 pˆ 1+ tan 2& sin( &) + sin 2 sin 2& 2 2 2 2 cos ( &) @s1 r @s1 2 r ✓ ✓ ◆ ◆ 2 C B @ pˆ 1 B @ pˆ = (1 ) ⇢↵ + ⇢ 2 2K @t 2K @t2 ✓ ◆ ✓ ◆ B 1 @⇢↵ @⇢ B 1 @ 2 ⇢↵ @ 2⇢ +C 2 + + ⇢ + . 2 ⇢↵ @t ⇢ @t 2 ⇢↵ @t2 ⇢ @t2 149

(30)

If desired then (23) could be used to write the above in terms of s1 instead of r.

150

Von Mises materials correspond to & = 0 and so to a form of (29) the left hand side of

151

which has a single term instead of three. The next two sections (3.2.1 and 3.2.2) consider

152

ways to eliminate the densities ⇢↵ , ⇢ from the above, and write it in terms of the overpressure 17

153

pˆ alone. For simplicity only Von Mises materials are considered. The same assumptions and

154

very similar methods could be used to eliminate densities from (30) for a Drucker-Prager

155

material.

156

3.2.1. Neglecting changes in density Following the derivation of a telegraphy equation in the dilatation ✏ of Vardoulakis and Sulem (2005), consider the case of negligible changes in time of the densities ⇢↵ and ⇢ . For a Von Mises material, the spatial derivative (30) of the momentum balance becomes @ 2 pˆ 1 @ pˆ 1 @ 2 pˆ = + , @t cs 2 @t2 @s1 2 where

,

2K 2 C B

is defined by

1 cs 2

(31)

is a di↵usivity and the positive quantity cs , called the speed of second sound, ⇣ ⇣ ⌘ ⌘ 1 = 2K (1 ) ⇢↵ + B ⇢ . The right hand side of the equation defining

cs is assumed to be positive. Equation (31) is a telegraphy equation in the overpressure pˆ. To express (31) dimensionlessly, define the dimensionless variables p˜, t˜ and s˜i by pˆ = k p˜,

t = H t˜ and si = L˜ si , where H , 2 /cs 2 , and L is a geometric length scale which will be chosen in section 4.2. Assuming

> 0, the telegraphy equation (31) becomes 1 @ 2 p˜ @ p˜ @ 2 p˜ = 2 + , 2 @˜ @ t˜ @ t˜2 s1 2

(32)

157

where  ,

158

Table 4 lists values of the scales H, 2 /cs and  for the physical parameters considered in

159

section 4.2.

L 2 /cs

is the ratio of the geometric length scale L to the material length scale.

The dimensionless overpressure p˜ has greater physical significance, but since (32) is to solved by separation of variables it will be easier to consider pˇ defined by p˜ s˜1 , s˜2 , t˜ = exp

t˜ pˇ s˜1 , s˜2 , t˜ .

(33)

In terms of pˇ the telegraphy equation (32) can be written as the Klein-Gordon equation 1 @ 2 pˇ = 2 @˜ s1 2

pˇ +

@ 2 pˇ 2. @ t˜

(34)

160

Section 4.2 considers the solution of (34) under boundary conditions which correspond to

161

sudden application of internal pressure. As section 4.2 explains, this is pertinent to the 18

162

expansion of a cylinder of mild steel. Appendix C supplements this solution with the outcome

163

of a very di↵erent sort of initial condition; one corresponding to a particular choice of spatially

164

smooth pressure perturbation.

165

3.2.2. Neglecting changes in -content

166

This section (section 3.2.2) describes pressure evolution for the case of small changes

167

in the mass composition. The assumption is analogous to that of undrained conditions

168

in poromechanics, which occur when in any reference volume of the skeleton ↵ there is a

169

constant mass of . The rate of change apparent to phase ↵ of the amount of phase in a material volume R (1) R (1) ⌦t is DDt ⌦t ⇢2 d⌦t = ⌦t DDt ⇢2 d⌦t . This integral is zero for every material volume ⌦t if and only if the integrand is everywhere zero. Coussy (2004) shows, and expresses in his equation (1.42), that the material derivative of a material volume is

D(1) d⌦t Dt

the condition for -content to remain fixed can be written

@⇢2 @t

= r · v (1) . It follows that

+ r · ⇢2 v (1) = 0. So, by the

fluid mass balance (3), r · (⇢ V) = 0,

(35)

170

where one of the definitions of the relative densities (1) and the definition of the specific

171

discharge (5) have been used. Hence, by (17) and (19), neglecting convective terms,

172

@⇢ @t



=

1 @⇢↵ ⇢↵ @t

+

@✏ . @t

By this result and the constitutive equation (29), assuming that the -volume fraction and densities ⇢↵ and ⇢ change slowly in time, the derivative (30) of the momentum balance ⇣ ⌘ 2 (1 )⇢↵ (B )⇢ @ 2 pˆ @ pˆ can be written (for a Von Mises material) @s . Assuming that 2 = 2K @t2 1 (1

)⇢↵

(B

)⇢ > 0, this leads to a wave equation in the overpressure pˆ , p cu 2

q

173

where cu ,

174

to the wave equation (36) is pˆ = F (s1

(1

2K )⇢↵ (B

)⇢

@ 2 pˆ @ 2 pˆ = . @s1 2 @t2

pQ ;

(36)

is a wave speed for undrained conditions. The general solution cu t) + G(s1 + cu t) for arbitrary F and G.

175

In contrast to (31), (36) predicts non-dissipative propagation of pressure waves. When

176

-content is constant the divergence of the specific discharge is small. Hence the dissipative 19

177

interactive force (8) does not enter the spatial derivative of the balance of momentum. Of

178

course, conservative pressure propagation is only possible because the volumetric pressure-

179

strain relationship (29) neglects any plastic contribution to volumetric strain.

180

4. A Von Mises cylinder internally pressurised to yield

181

Consider a cylinder of a Von Mises material held in plane strain and internally pressurised

182

to the point of yield. Note that because the azimuthal is greater than the radial tensile stress,

183

=

3⇡/4, and because the material is Von Mises, & = 0.

184

Section 4.1 uses separation of variables to find a form of the pressure distribution con-

185

sistent with overpressure pˆ vanishing at the inner and outer wall of the cylinder. Section

186

4.2 considers instances of the pressure distribution corresponding to parameter values and

187

initial conditions relevant to the experiment with mild steel of Sowerby and Johnson (1970),

188

a product of which is shown in figures 4a and C.7a.

189

4.1. The solution modulo an initial condition To identify boundary conditions, assume that, from time t = 0, a constant pressure is applied at the inner wall (r = ri ) of the cylinder. Choosing s1 = 0 at the outer wall (r = ro ) and denoting the pressure p there po , it follows from (23) that 1 p s1 2

r = ro

and hence from (26) that the quasistatic pressure pQ = po

(37) ⇣ 2k ln 1

p1 s1 2 ro



. Following

Sowerby and Johnson (1970), consider the minimum pressure at the inner wall to establish quasistatic plastic yield. Equivalently, assume zero overpressure at the inner and outer walls (r = ri and r = ro ); pˇ|s˜1 =1 = 0 = pˇ|s˜1 =0 .

(38) p

190

Note that, as figure 1b indicates, the geometric length scale L ,

191

the arclength of a slip-line. For this reason s˜1 = 1 at the inner wall of the cylinder.

20

2 (ro

ri ) is set equal to

Following the method of separation of variables, consider solutions of the form pˇ s˜1 , t˜ = 00 s ) T 00 (t˜) 1 S (˜ s1 ) T t˜ . With this substitution, (ˇ p) 1 ⇥ (34) is 12 SS(˜(˜ = 1 + = $2 , where $ s1 ) T (t˜) is a real constant. Hence S (˜ s1 ) = a cos ($˜ s1 ) + b sin ($˜ s1 ), for constant a and b. The boundary conditions (38) are satisfied by a = 0, $ = n⇡/, n 2 Z. So the form of the solution of T 00 t˜ = (1

where

$2 )T t˜ depends on n as 8 < c cosh ˜ ˜ n n t + dn sinh nt T t˜ = : c cos ! t˜ + f sin ! t˜ n n n n n

q , 1

!n ,

(n/n⇤ )2 ,

q

, n < n⇤ , n > n⇤

(n/n⇤ )2

,

1

(39)

and n⇤ , /⇡. The full solution for the dimensionless overpressure p˜ = exp

t˜ pˇ is



p˜ = exp



bn c X

sin (n⇡˜ s1 ) cn cosh

˜ + dn sinh

nt

˜

nt

n=1

+

1 X

sin (n⇡˜ s1 ) cn cos !n t˜ + fn sin !n t˜

n=bn⇤ c+1

!

,

(40)

where for z 2 R, bzc denotes the greatest integer less than or equal to z and for z 2 / R,

bzc , 0. By (40) and orthonormality of the odd functions sin (n⇡x), the coefficients cn are given by cn = 2

Z

1 0

sin (n⇡˜ s1 ) ( pˇ|t˜=0 ) d˜ s1 .

(41)

Assume that the initial dimensionless rate of change of dimensionless overpressure is proportional to the initial dimensionless pressure; @ p˜ @ t˜

t˜=0

= ⇣ p˜|t˜=0 .

(42)

192

The choice ⇣ = 0, which is perhaps most natural, describes the case for which the initial

193

rate of change of pressure is zero. A choice ⇣ < 0 could reflect the fact that the application

194

of pressure can not be instantaneous, and that the material will begin to react against its

195

perturbation from equilibrium. A choice ⇣ > 0 could describe the continuation of the external 21

196

process which brought the material to its t˜ = 0 disequilibrium state. By the definition (33)

197

of pˇ, (42) can be written

@ pˇ @ t˜ t˜=0

= (⇣

Assume n⇤ 2 / Z, so that for n 2 Z,

1) pˇ|t˜=0 . n

and !n are non-zero. In this case it follows, by the

definition (33) of pˇ, (40), and L2 ([ 1, 1]) orthonormality of the odd functions sin (n⇡x), that the coefficients dn and fn are dn = ⇣ n1 cn , fn = ⇣!n1 cn , where cn is given by (41). Hence (40) P becomes p˜ = 1 ˜n , where p˜n is the component of overpressure with an s˜1 -wavelength of n=1 p 2/n,

p˜n = cn exp

With the expressions (39) for

198

199

200

201

8 ⇣ ⌘ < cosh n t˜ + ⇣ 1 sinh n t˜ , n < bn⇤ c n ⌘ t˜ sin (n⇡˜ s1 ) ⇣ . : cos ! t˜ + ⇣ 1 sin ! t˜ ⇤ , n > bn c n n !n n

(43)

and !n , the eigenmode solution (43) shows that if ⇣ 6=

1

then there can be a local peak in the amplitude of the components of pressure around n = n⇤ . ⇣ ⌘ n⇤ n⇤ 2 ⇤ n ˜ q The coefficient of sin (n⇡˜ s1 ) has a stationary point at t , arccos 2 n 1 7 and ⇤ 2 1 (n ) n so limn!n⇤ t˜⇤ = 2. Hence, if the components of overpressure around n = n⇤ are of interest

202

then it will be worthwhile considering the pressure profile at dimensionless time t˜ = 2. Figure

203

2a includes the profile at this time, as do several plots of figures 3 and C.6.

204

4.2. A mild steel cylinder suddenly loaded to yield

205

This section (section 4.2) identifies the pressure evolution during expansion of a mild steel

206

cylinder. The following ideas and observations motivate the modelling approach. Firstly,

207

note that the main constituent is assumed to be a Von Mises material, a useful description

208

for steel8 (Hill, 1950). Next, note that the experiment of Sowerby and Johnson (1970)

209

sought strain localisation in the form of the L¨ uders bands pictured in figure 4a. At the

210

stage of deformation in which localisation is observed, the preponderance of total plastic

211

deformation can be taking place within the L¨ uders bands. (Hertzberg et al., 2012) This

212

suggests importance of the bands to the overall deformation process. At least, a complete

213

description of the experiment would e↵ectively identify L¨ uders lines. Following are reasons Since cos(ix) = cosh(x), this formula holds both for n < n⇤ and n > n⇤ . 8 Admittedly, the major component of steel considered as a two phase material is alpha-ferrite. 7

22

214

supporting the working hypothesis that the chosen model is suitable for describing L¨ uders

215

lines.

216

The experiments and observations of Price and Kelly (1964), producing L¨ uders bands in

217

single crystals of aluminium alloys, helped to delineate the simplest framework which could

218

encapsulate the physics integral to L¨ uders bands. In their experiments deformation localised

219

despite small volumes and slow deformation, showing that shear band formation need not

220

hinge upon temperature e↵ects. Also, repeated deformation to the point of strain localisation

221

of the same crystal did not result in shear bands localising in the same places, showing that

222

localisation was not the result of permanent work softening. These results admit neglect of

223

work hardening and softening, and of any dependence of constitutive laws on temperature.

224

The results of Cox and Low (1974) suggested that the critical mechanism leading to

225

strain localisation during their loading of steel was di↵erential deformation of metallic and

226

non-metallic material elements. As Rice (1976) observes from the figures of Cox and Low

227

(1974), the cavities which nucleate around inclusions are prominent in localisation bands

228

but not visible elsewhere. The influential role of material disparity suggests a description

229

which accounts separately for the two phases of steel. Admittedly, the saturated mixtures

230

description of this work does not explicitly account for cavities.

231

Pressure impedes shear strain (Frost and Ashby, 1982), and so the maxima of shear

232

strain, L¨ uders bands, may be sought via the minima of pressure. An additional motivation

233

for seeking to describe pressure is the sensitivity to small stress inhomogeneities of the

234

positions of L¨ uders bands (Hertzberg et al., 2012).

235

This section (section 4.2) considers the pressure solution p˜ of the last section (section 4.1)

236

for the expansion of a mild steel cylinder. Figure 4a shows the cylinder after expansion and

237

etching by Sowerby and Johnson (1970). Appendix B discusses estimation of the physical

238

parameters corresponding to their experiment. Because the drag coefficient C is difficult to

239

estimate, values of C around an order of magnitude greater than and less than that roughly

240

estimated in Appendix B are also considered. Table 4 shows signature quantities which

241

correspond to these choices of parameters.

23

C



kg m3 s



220 · 106 2.2 · 109 22 · 109

H [s] 2.0 · 10 200 · 10 20 · 10

6 9 9

2 /cs [m]



n⇤

3

1

0.318

3

10

3.18

6

100

31.8

22 · 10 2.2 · 10 220 · 10

Table 4: Signature quantities of the computations of section 4.2. The first, second and third rows come from the parameter values estimated for mild steel in Appendix B, with the drag coefficient chosen so that the dimensionless parameter  is 1, 10 and 100 respectively. Note that, by definition, n⇤ = /⇡. 242

243

244

To compute pressure profiles at particular times this section (section 4.2) employs the P max truncation p˜ = nn=1 p˜n , where nmax = 105 . The calculations of this section assume that expansion began from static equilibrium. That is, the parameter ⇣ of (42) is chosen ⇣ = 0.

245

The initial condition to be considered corresponds to sudden loading. As will be seen,

246

the sudden application of pressure leads to a propagating discontinuity. Appendix C instead

247

considers an initial condition for which both the overpressure and the spatial derivative of

248

overpressure at the inner and outer wall are zero. It is included to show that results do not

249

depend essentially upon the discontinuous application of pressure. Consider the case in which the cylinder is suddenly subjected to internal pressure. To identify initial conditions, assume that prior to time t = 0 the pressure is zero. By (44), the definitions of overpressure pˆ and the dimensionless fields p˜ and pˇ and the dimensionless variable t˜,



pˇ|t˜=0 = 2 ln 1



ro

ri ro





s˜1 .

(44)

This initial rescaled overpressure pˇ is zero at the outer wall of the cylinder (˜ s1 = 0) and negative between the inner and outer wall. By (41) it follows from (44) that ✓ ✓ ◆ ◆ Z 1 ro r i cn = 4 sin (n⇡˜ s1 ) ln 1 s˜1 d˜ s1 . ro 0

(45)

250

Computation of the coefficients dn and fn requires a choice of the dimensionless parameter

251

.

252

Figures 2a and 2b show, for sudden loading and the choice  = 10, the dimensionless

253

total pressure and overpressure profiles at a range of dimensionless times in 0  t˜  16. In 24

1

p

 = 10 (n⇤ = 3.18)

k ( p

max )

-1

p

1.0

1

1.5 0.5

0.5

1

0.5 -0.5

0.5

0

1/4

1

1

s1

p

max

1 0.50

1.0

t˜ =

p

k

s1

0.10 0.05 0.01

-1.0

2

0.5

2

p 2 2

4

8

1

5

10

˜ t

16

Figure 2: Time evolution of dimensionless (a) total pressures and (b) overpressures of a hollow cylinder suddenly loaded to yield. Parameters are those estimated for mild steel in Appendix B and listed in table B.5, and which correspond to the signature quantities listed in the second row of table 4. The factors (|ˆ p|max )

1

are chosen so that the maximum magnitude of the scaled overpressure profiles of (b) are about unity. The coincidence of the discontinuity in and the local maximum of the overpressure at t˜ = 2 distinguishes this time. (c) Values of the inverse of the scaling factor |ˆ p|max at di↵erent dimensionless times t˜. 254

particular, they include the profile at the dimensionless time t˜ = 2. As section 4.1 showed,

255

this time was a critical point of the amplitude of the component of pressure p˜n as n ! n⇤ .

256

However, as this section (section 4.2) will discuss, figure 3c, a plot of the continuous parts

257

of the overpressures, indicates more clearly the potential significance of the time t˜ = 2.

258

The plots of figures 2, 3 and C.6 correspond to the parameters for mild steel estimated in

259

Appendix B and listed in table B.5.

260

The profiles of figures 2a and 2b, corresponding to  = 10, show that the initial pressure

261

discontinuity at the inner wall of the cylinder (where s˜1 = 1) propagates and attenuates.

262

The same occurs for  = 1 and  = 100. The considerations of Zauderer (2006, pp. 700

263

– 716) allow identification of the discontinuity. It travels at the speed of second sound cs ,

264

attenuates as exp

265

(where s˜1 = 0) wall of the cylinder. Let pD denote the discontinuous part of the pressure.

t˜ and is reflected when it reaches the inner (where s˜1 = 1) or outer

25

266

Specifically, let pD 9 denote the spatially piecewise constant function satisfying pD |s˜1 =0 = 0

267

such that p , pˆ

268

overpressure.

\

\

pD is continuous. The function p will be called the continuous part of the

269

The left column of figure 3 shows scaled dimensionless overpressure profiles at a range of

270

dimensionless times for sudden loading and choices  = 1, 10 and 100. The scaling factors

271

\

p

1 max

are chosen so that the maximum magnitude of the scaled pressure profiles are \

272

about unity. The right column of figure 3 shows values of p

273

times t˜. The factor exp

274

that, as the right column suggests, as time passes the magnitude of the pressure profile

275

approaches zero.

max

at di↵erent dimensionless

t˜ in the expression (43) of the components of pressure means

278

The expression (43) of overpressure components shows that for 1  n < n⇤ , p˜n has a ⇣ ⌘ factor cosh n t˜ + ⇣ n1 sinh n t˜ . Also, the expressions (39) of time constants shows that

279

shape of the pressure profile approaches that of the spatial factor of p˜1 , sin (n⇡˜ s1 ). This is

280

seen in figures 3c and 3e.

276

277

for 1 < n < n⇤ ,

1

>

n.

Hence, as time passes the mode p˜1 comes to dominate and so the

The profiles of figure 3a, which also pertains to sudden loading but to the choice of the

281



282

dimensionless parameter  = 1, show di↵erent behaviour. Of course, the factor exp

283

in the expression (43) of the components of pressure means that, as figure 3b suggests, the

284

magnitude of the pressure profile approaches 0 as time approaches infinity. However, because

285

n⇤ = 0.318 < 1, no components of overpressure have additional hyperbolic factors. Hence,

286

no single term comes to dominate the shape of the profile. The experiments inspiring this work produced L¨ uders bands. How and where does the

287

9

The discontinuous, piecewise constant function pD can be written (regrettably, fairly opaquely) ⇣ ⌘ pD = pQ |s˜1 =1 exp

t˜ ⇥ s˜1

1+

t˜/ + 1

8 < 0 where ⇥ is the Heaviside theta function satisfying ⇥(z) = : 1

,

mod 2

z0

1 ,

and for real a and positive n, a , z>0 mod n (which can be read ‘a modulo n’) is the unique real number r satisfying 0  r < n for which there exists an integer q such that a = nq + r.

26

288

preceding analysis suggest they will form? Towards answering this question, note that ob-

289

served sensitivity of the position of L¨ uders bands to small stress inhomogeneities (Hertzberg

290

et al., 2012) justifies consideration of critical points of the pressure distribution. By (29),

291

consideration of minima of pressure is equivalent to examining the maxima of volumetric

292

strain. The peaks of volumetric expansion may provide a more intuitive way to think about

293

shear strain localisation than do the minima of pressure.

294

For a constant shear the rate of shear strain increases with decreasing pressure. (Frost

295

and Ashby, 1982) However, material which has reached a quasistatic pressure distribution

 = 1 (n⇤ = 0.318)

( p⋂

max )

p

1.0

1

p⋂

max

k

0.5

2 1 0.5

1

s1

0.5

-0.5

0.2 0.1

-1.0

t˜ =

0.05

0 ( p⋂

 = 10 (n⇤ = 3.18)

-1 ⋂

max )

1/2

1

p

2

0.5

p 2 2

2

1

2

64

256

˜ t

-1 ⋂

p

1.0

1

p⋂

max

k 0.5

1 0.50 0.5

1

s1 0.10 0.05

-0.5

0.01 -1.0

t˜ =

0.5

0

1/4

1

p

2

2

p 2 2

1

5

4

8

10

˜ t

16

Figure 3: Continues on a later page which contains a descriptive caption.

27

( p⋂

max )

-1 ⋂

p

 = 100 (n⇤ = 31.8)

1.0

1

0.5

max

1.

0.5

1

0.1

s1

0.01

-0.5

0.001 0.0001

-1.0

t˜ =

p⋂

k

0

1

2

16

256

1

10

100

1000

˜ t

1024

Figure 3: Continued from an earlier page. Left column. Time evolution of the scaled dimensionless continuous \

parts of the overpressures p of a hollow cylinder suddenly loaded to yield, using the parameters for mild steel estimated in Appendix B and listed in table B.5. For  = 10 and 100, the coincidence of the discontinuity in and the local maximum of the overpressure pˆ at about t˜ = 2 distinguishes this time. Right column. Values \

of the inverse of the scaling factor p

max

at di↵erent dimensionless times t˜. Both abscissa and ordinate use

a logarithmic scale. The initial profile is scaled by the inverse of

1 k

\

p

max

t˜=0

= 1.91. Any times t˜ > 0 \

which are represented in a pressure profile on the left but absent on the right have values of p

max

which are

cumbersomely many orders of magnitude less than those which are shown. Each row of the figure pertains to the signature quantities listed in the corresponding row of table 4.

296

p = pQ is not expected to deform further. A simple way to affirm the latter result is by

297

assuming that the tendency towards shear strain depends upon the overpressure pˆ. Assume

298

that a point favourable for strain localisation is one for which the left and right hand limit

299

of the overpressure, pˆ, is less than at points in a neighbourhood just to its left and right

300

respectively. Then the discontinuous part of the overpressure which was identified does not

301

a↵ect the location of points favourable for strain localisation. Hence, for clarity, consider the

302

continuous part of the overpressure p rather than the total overpressure pˆ.

\

The following paragraph applies for n⇤ > 1. For positive times t˜ < 2, there is no local \

spatial minimum of the continuous part of the overpressure p of pressure in the interior of the cylinder. For an open interval of times with infimum t˜ = 2, the continuous part of the

28

\

overpressure p exhibits a local minimum at the site of the discontinuity in pˇ, s˜1 = 1

t˜/.

Figure 3c suggests this but, unfortunately, scale means that it is not possible to discern for  = 100 from figure 3e. The scaled arc-length s˜1 of this relationship for t˜ = 2, around which the interior pressure minimum is forecast, is a reasonable location to anticipate initialisation of strain localisation. So is s˜ = 1 (which corresponds to the inner wall of the cylinder), the location of the global minimum of the continuous part of the overpressure. The distance s1 between characteristics intersecting at these locations can be written in terms of the physical parameters of the problem p

s1 2 (ro

ri )

=1

p

2p

4 K⇢↵ C B

1

s

1+

B

⇢ . ⇢↵

(46)

303

Figure 4b shows characteristics which intersect at the values s˜1 = 1 and s˜1 = (1

304

0.8. It corresponds reasonably closely to figure 4a, which shows the product of the operation

305

of Sowerby and Johnson (1970). However, it is coincidence that for the relevant parameters

306

the choice  = 10 corresponds so closely to the experiment. Still, the process suggests a way

307

to use (46) to estimate the drag coefficient C relevant to the experiment of Sowerby and

308

Johnson (1970) and to similar applications.

309

For  = 100 the location of the discontinuity at t˜ = 2 is s˜1 = (1

2/)|=10 =

2/)|=100 = 0.980,

310

which would correspond to curves of strain localisation far more closely spaced than those

311

indicated by figure 4b or seen in figure 4a.

312

Several links in the chain of reasoning which locates lines of strain concentration are

313

insecure. The initial condition of section 4.2 (that is, (44)) may not conform precisely with

314

the experiment of Sowerby and Johnson (1970), some details of which we do not know.

315

Having obtained from this initial condition a pressure profile, and aware that that spatially

316

regular strain localisation indeed occurs during the expansion, the reasoning of section 4.2

317

represents one way to identify the spacing of lines of concentrated strain. It may be that

318

the identified pressure profiles are not, after all, accurate, or that in the expansion pressure

319

is not the most appropriate property whereby to locate the lines of strain concentration.

320

Even if pressure is, as the observations of Capp et al. (1973) and Hertzberg et al. (2012)

321

might suggest, a fitting way to pinpoint strain localisation, it may not be the case that the 29

Sudden loading,  = 10

x2 [in.] 1.0 0.5

-1.0 -0.5

0.5

1.0

x1 [in.]

-0.5 -1.0 Figure 4: Characteristics and slip-lines. (a) Figure 9.3a of Sowerby and Johnson (1970). Slip-lines in a mild steel cylinder of outer diameter ro = 2.2500 and inner diameter ri = 100 . The inner diameter was inferred from the image and knowledge of the outer diameter. (b) Characteristics which intersect at s˜1 = 1 and s˜1 = 0.8, positions which section 4.2 argues are favourable for strain localisation if  = 10. \

322

continuous part of the overpressure, p, should be used. For example, either the continuous

323

part of the total pressure, p

324

However, their local minima (where strain would tend to concentrate) might need to be

325

found numerically.

pD , or the total pressure, p, may play a more important role.

326

Pressure minima also develop for the smoother initial condition of Appendix C. Lines of

327

strain localisation to which these could correspond are shown in figure C.7b. Their locations

328

had to be found numerically, but the section shows that results do not depend essentially

329

upon the discontinuous application of pressure.

330

5. Conclusion

331

This work focused on expanded cylinders of mild steel, but the relationship (21) between

332

overpressure and its spatial and time derivatives could be applied to describe pressure evo-

333

lution in other conditions of plane strain. The relationship is not restricted to Von Mises 30

334

and Drucker-Prager materials.

335

It would be interesting to assess the role of dissipative volumetric deformation by mod-

336

ifying the volumetric pressure-strain relationship (29) to include a plastic contribution to

337

volumetric strain. Although, as discussed in section 4.2, strain localisation need not de-

338

pend upon temperature e↵ects, more precise realism could also be sought by including heat

339

di↵usion and the dependence of material properties on temperature. These dissipative phe-

340

nomena could be useful in describing the width of deformation bands. Another possible

341

extension could be to materials containing more than two phases. Since, as discussed in 4.2,

342

in deforming steel cavities can nucleate around inclusions in localisation bands more than

343

elsewhere, it would be particularly interesting to consider a component comprising void.

344

Applying SFT to yielding two phase materials identified dissipative pressure waves. The

345

case of radially symmetric plane strain and a Von Mises or Drucker-Prager yield envelopes

346

allowed particularly simple expression of the wave equations along a slip-line. When the main

347

phase is Von Mises this was a telegraphy equation. The pressure solution associated with

348

radially symmetric, sudden loading of the alpha-ferrite-cementite mixture which comprises

349

steel uncovered pressure troughs which may be linked with strain concentrations. For one of

350

the surveyed values of an imprecisely known physical parameter called the drag coefficient,

351

the pressure troughs would lead to strain concentrations similar to those of figure 4a. The

352

process could be used to estimate drag coefficients for this and other systems.

353

It would be interesting to load rapidly to yield cylinders of di↵erent sizes and di↵erent

354

compositions, and see how closely the observed variation of the spacing of lines of strain

355

concentration matches that predicted by (46). Likewise, it would be interesting to record

356

pressure during loading and use it as the boundary condition at the inner wall of the cylinder

357

for numerical integration of the pressure. The resulting pressure profiles could be compared

358

with the observed spacings of lines of strain concentration. Pressurisation of confined rather

359

than free materials would also provide extra sets of boundary conditions to test.

31

360

References

361

Aboudi, J., Arnold, S. M., Bednarcyk, B. A., 2013. Micromechanics of Composite Materials:

362

363

364

A Generalized Multiscale Analysis Approach. Butterworth-Heinemann, Oxford. Bhadeshia, H., Honeycombe, R., 2006. Steels: Microstructure and Properties. ButterworthHeinemann, Oxford.

365

Biot, M. A., 1956a. Theory of propagation of elastic waves in a fluid-saturated porous solid.

366

I. Low-frequency range. The Journal of the Acoustical Society of America 28 (2), 168–178.

367

Biot, M. A., 1956b. Theory of propagation of elastic waves in a fluid-saturated porous solid.

368

II. Higher frequency range. The Journal of the Acoustical Society of America 28 (2),

369

179–191.

370

371

372

373

Callister, W. D., Rethwisch, D. G., 2010. Materials Science and Engineering: An Introduction. Wiley, New York. Capp, D. J., McCormick, P., Muir, H., 1973. The e↵ect of pressurization and quenching on the yield behaviour in low carbon steel. Acta Metallurgica 21 (1), 43–47.

374

Coussy, O., 2004. Poromechanics. John Wiley & Sons.

375

Cox, T., Low, J. R., 1974. An investigation of the plastic fracture of AISI 4340 and 18

376

377

378

379

380

381

382

Nickel-200 grade maraging steels. Metallurgical Transactions 5 (6), 1457–1470. Estrin, Y., Ling, C., McCormick, P., 1991. Localization of plastic flow: Spatial vs temporal instabilities. Acta Metallurgica et Materialia 39 (11), 2943–2949. Frost, H. J., Ashby, M. F., 1982. Deformation Mechanism Maps: The Plasticity and Creep of Metals and Ceramics. Pergamon Press, Exeter. Gao, F.-P., Wang, N., Zhao, B., 2015. A general slip-line field solution for the ultimate bearing capacity of a pipeline on drained soils. Ocean Engineering 104, 405–413.

32

383

Godunov, S. K., Deribas, A. A., Zakharenko, I. D., Mali, V. I., 1971. Investigation of the

384

viscosity of metals in high-velocity collisions. Combustion, Explosion and Shock Waves

385

7 (1), 114–118.

386

387

Hertzberg, R. W., Vinci, R. P., Hertzberg, J. L., 2012. Deformation and Fracture Mechanics of Engineering Materials. Wiley, Hoboken.

388

Hill, R., 1950. The Mathematical Theory of Plasticity. Clarendon Press, Oxford.

389

Hosford, W. F., 2010. Solid Mechanics. Cambridge University Press.

390

Jackson, R., Ghaednia, H., Pope, S., 2015. A solution of rigid-perfectly plastic deep spherical

391

392

393

394

395

indentation based on slip-line theory. Tribology Letters 58 (3), 1–7. Johnson, J. N., 1981. Dynamic fracture and spallation in ductile solids. Journal of Applied Physics 52 (4), 2812–2852. Johnson, W., Sowerby, R., Venter, R., 1982. Plane-Strain Slip-Line Fields for MetalDeformation Processes: A Source Book and Bibliography. Pergamon Press, Exeter.

396

Khan, I. A., Bhasin, V., Chattopadhyay, J., Singh, R. K., Vaze, K. K., Ghosh, A. K.,

397

2014. An insight of the structure of stress fields for stationary crack in strength mismatch

398

weld under plane strain mode–I loading - Part ii: Compact tension and middle tension

399

specimens. International Journal of Mechanical Sciences 87, 281–296.

400

401

402

403

404

405

Koot, L., Dumberry, M., 2011. Viscosity of the earth’s inner core: Constraints from nutation observations. Earth and Planetary Science Letters 308 (3), 343–349. Laszlo, F., Nolle, H., 1959. On some physical properties of cementite. Journal of the Mechanics and Physics of Solids 7 (3), 193–208. McCormick, P., 1988. Theory of flow localisation due to dynamic strain ageing. Acta Metallurgica 36 (12), 3061–3067.

33

406

407

Miodownik, A. P., 1994. Young’s modulus for carbides of 3d elements (with particular reference to Fe3 C). Materials Science and Technology 10 (3), 190–192.

408

Paesold, M. K., Bassom, A. P., Veveakis, E., Regenauer-Lieb, K., 2016. Conditions for the

409

localization of plastic deformation in temperature sensitive visco-plastic materials. Journal

410

of Mechanics of Materials and Structures 11 (2), 113–136.

411

Papamichos, E., Vardoulakis, I., Tronvoll, J., Skjaerstein, A., 2001. Volumetric sand produc-

412

tion model and experiment. International Journal for Numerical and Analytical Methods

413

in Geomechanics 25 (8), 789–808.

414

415

Price, R., Kelly, A., 1964. Deformation of age-hardened aluminium alloy crystals-ii. Fracture. Acta Metallurgica 12 (9), 979–992.

416

Rice, J. R., 1976. The localization of plastic deformation. In: Theoretical and Applied Me-

417

chanics (Proceedings of the 14th International Congress on Theoretical and Applied Me-

418

chanics). pp. 207–220.

419

Rudnicki, J. W., Rice, J. R., 1975. Conditions for the localization of deformation in pressure-

420

sensitive dilatant materials. Journal of the Mechanics and Physics of Solids 23 (6), 371–394.

421

Savenkov, G. G., 2010. Dynamic viscosity and material relaxation time during shock loading.

422

423

424

425

426

Journal of Applied Mechanics and Technical Physics 51 (2), 148–154. Savenkov, G. G., Meshcheryakov, Y. I., 2002. Structural viscosity of solids. Combustion, Explosion and Shock Waves 38 (3), 352–357. Shield, R., 1953. Mixed boundary value problems in soil mechanics. Quarterly of Applied Mathematics 11 (1), 61–75.

427

Sowerby, R., Johnson, W., 1970. Use of slip line field theory for the plastic design of pressure

428

vessels. In: Experimental stress analysis and its influence on design : Proceedings of the

429

Fourth International Conference on Experimental Stress Analysis. pp. 303–317.

34

430

Spitzig, W. A., Sober, R. J., Richmond, O., 1975. Pressure dependence of yielding and

431

associated volume expansion in tempered martensite. Acta Metallurgica 23 (7), 885–893.

432

Vardoulakis, I., 1989. Shear-banding and liquefaction in granular materials on the basis of a

433

434

435

436

437

438

439

440

441

442

Cosserat continuum theory. Ingenieur-Archiv 59 (2), 106–113. Vardoulakis, I., Sulem, J., 2005. Bifurcation Analysis in Geomechanics. Blackie Academic & Professional, Glasgow. Veveakis, E., Regenauer-Lieb, K., 2015. Cnoidal waves in solids. Journal of the Mechanics and Physics of Solids 78, 231–248. Vo, T., Russell, A. R., 2014. Slip line theory applied to a retaining wall-unsaturated soil interaction problem. Computers and Geotechnics 55, 416–428. Zauderer, E., 2006. Partial Di↵erential Equations of Applied Mathematics. John Wiley & Sons, Hoboken.

Appendix A. Estimating the interaction force

443

Section 2.1 considers an interaction force between cementite and alpha-ferrite which op-

444

poses relative motion of the phases; the force, given by (8), is proportional to the relative

445

velocity. This appendix considers a method of estimation of the coefficient of proportionality

446

C, and shows that estimates of C could range across several orders of magnitude.

447

As figure A.5a suggests, mild steel consists of a mixture of grains of alpha-ferrite (nearly

448

pure iron) and grains of the two phase material pearlite. Pearlite comprises alternating layers

449

of alpha-ferrite and cementite (a ceramic). The lamellar structure of mild steel motivates

450

investigation of flow between layers.

451

Consider steady-state laminar flow over a distance l of an incompressible Newtonian

452

fluid between stationary wide, long parallel plates separated by a distance 2d. Figure

453

A.5b illustrates the flow. Denote the width of the plates by w. Since the cross-section

35

Figure A.5: Lamellae and laminar flow in mild steel. (a). Following figure 9.29 of Callister and Rethwisch (2010). Mild steel is composed of grains of metallic alpha-ferrite and grains of the mixture pearlite. Pearlite is composed of alternating layers of alpha-ferrite and the ceramic cementite. Alpha-ferrite (↵) in grey and cementite (Fe3 C) in black. (b). Following Biot (1956a). Laminar flow between parallel plates.

454

is constant and the phase is assumed to be incompressible, the material does not accel-

455

erate. It follows that the force balance on the layer of material between y and y + dy is

456

0 = P |x=0 wdy



dv dy

y

wl

P |x=l wdy + ⌘

dv dy

y+dy

wl, where P is pressure ⌘ is the vis-

457

cosity of the phase. In the limit of small dy this is

458

which can be written

( P |x=0

2

( P |x=0

2

d v P |x=l ) wdy = ⌘wl dy 2 dy,

d v P |x=l ) /l = ⌘ dy 2 . Assuming that the material is stationary

462

where it contacts the plates, v|y= d = 0 = v|y=d , and so the solution to this di↵eren1 ( P |x=0 P |x=l ) tial equation is v(y) = 2⌘ (d2 y 2 ). Hence the average speed of the phase is l R y=d 1 ( P |x=0 P |x=l ) 2 1 1 ( P |x=0 P |x=l ) 2 v¯ = 2d (d y 2 ) dy = 3⌘ d. l l y= d 2⌘

463

cementite v (1)

464

force per unit volume of alpha-ferrite

465

Hence, the proportionality factor between specific force and discharge, C =

466

estimated from the viscosity ⌘ and the plate separation distance d.

459

460

461

Associating the average speed v¯ with the relative motion between the alpha-ferrite and v (2) and the spatial rate of pressure drop ( P |x=0 f /(1

) gives f =

P |x=l ) /l with the drag

(3⌘)/d2 (1

) v (1)

3⌘(1 d2

)

v (2) .

, can be

467

Of course, not all lamellae will be parallel to the flow. Biot (1956b) accounted for the

468

anisotropy of flow channels by multiplying his result for the drag coefficient by a factor ⇠ > 1

36

469

which he called the sinuosity. Incorporating this, the result of this appendix for the drag

470

coefficient would be C = ⇠ 3⌘(1d2

471

drag coefficient which depended on the frequency of the relative motion between the phases.

472

For simplicity this work considers a constant drag coefficient and does not explicitly consider

473

the sinuosity ⇠. A greater viscosity ⌘ or lesser plate separation d achieves the e↵ect of a

474

sinuosity ⇠ greater than unity.

)

. Biot (1956a,b) further increased accuracy by considering a

475

Estimated viscosities of solid metals vary, and depend not just upon temperature pressure

476

and loading conditions but upon the models used. (Savenkov, 2010) Godunov et al. (1971)

477

cite the use and estimation of viscosities of steel in the range ⌘ = 10

478

Koot and Dumberry (2011) mention that calculations of the viscosity of the iron core of

479

the Earth range between ⌘ = 1011

480

that viscosity decreased with the length scale of the flow. At the smallest length scale they

481

consider, 10

482

figure 5), but they do not indicate a positive lower bound. To reproduce observed spalling

483

under plate impact of copper Johnson (1981) identifies a viscosity of 1.0 Pa s. He notes that

484

this is orders of magnitude lower than the value observed for macroscopic metallic flow, and

485

speculates that the discrepancy may be due to localisation e↵ects.

7

3±1

106 Pa s, while

1021 Pa s. Savenkov and Meshcheryakov (2002) found

10 6 m, the pertinent viscosity is bounded above by 100 Pa s (see their

486

The e↵ective interplate distance d will lie somewhere between the separation of neigh-

487

bouring cementite plates within a pearlite grain and the separation of cementite plates of

488

neighbouring pearlite grains (separated by ferrite). Figure 3.18 of Bhadeshia and Honey-

489

combe (2006) shows interlamellar spacings of carbon steels spanning the range d = 4. ⇥

490

10

491

d = 1. ⇥ 10

492

8

5. ⇥ 10 7 m, while their figure 11.6 shows grain sizes of mild steel spanning the range 5

1. ⇥ 10 3 m.

With viscosity in the range ⌘ = 10 8

4

1021 Pa s and plate spacing in the range d =

1. ⇥ 10 3 m, it would be easier to estimate C =

493

4. ⇥ 10

494

constituent factors.

37

3⌘(1 d2

)

directly than from its

Property

ri [m] 2.9 · 10

Value

ro [m] 2

1.3 · 10

⇢↵ 2

1.3 · 10

2

⇥ kg ⇤ m3

7.69 · 103



⇥ kg ⇤ m3

7.85 · 103

K [Pa] 1.70 · 1011

B 2.2 · 10

3

Table B.5: Estimates of the parameters of the cylinder of mild steel, used for the computations of section 4.2. Appendix B discusses the choice of parameters.

495

Appendix B. Estimating parameters for a cylinder of mild steel

496

Inner radius, ri and outer radius, ro . The radii considered are those pertaining to

497

figure 9.3a of Sowerby and Johnson (1970) (figure 4a of this document); ro = 1.12500 and

498

inner diameter ri = 0.5000 . The inner radius was inferred from their figure 9.3a and knowledge

499

of the outer radius.

500

Volume fraction of

,

. Densities, ⇢↵ and ⇢ . The

-volume fraction

was es-

501

timated assuming the carbon content of the mild steel of Laszlo and Nolle (1959), 0.088%

502

(by weight), along with the densities of alpha-ferrite ⇢↵ and cementite ⇢ they used. These

503

densities are those listed in table B.5. The calculation treats the alpha-ferrite fraction as

504

pure Fe and the cementite fraction as pure Fe3 C. The molecular weight of Fe and C used

505

were 55.85g/mol and 12.01g/mol.

506

Bulk modulus, K. Biot coefficient, B. The -volume fraction

of table B.5 is orders

507

of magnitude less than unity, motivating use of the formula for dilute bulk modulus of Aboudi

508

et al. (2013, pp. 109–110). This is K = K↵

509

is the shear modulus and ⌫↵ is the Poisson ratio of phase ↵.

510

(K↵

↵ +4G↵ K ) 3K , where G↵ = 3K +4G↵

3 1 2⌫↵ K↵ 2 1+⌫↵

The Biot coeffient was calculated using (4.35) of Coussy (2004) for a linear isotropic mateK . Ks

511

rial; B = 1

512

the skeleton, Ks , with the bulk modulus of alpha-ferrite, K↵ . With the formula above for the

513

dilute bulk modulus K , the formula of Coussy (2004) becomes B =

514

The

Any occluded

-volume fraction

was neglected, allowing identification of the bulk modulus of

(1

↵ +4G↵ K /K↵ ) 3K . 3K +4G↵

is that discussed earlier in this appendix and given in table

515

B.5. The other values in the formulae are those of Miodownik (1994); bulk moduli K↵ =

516

1.71 ⇥ 109 Pa and K = 1.45 ⇥ 109 Pa, and Poisson ratio ⌫↵ = 0.29.

517

Coefficient of the interaction force, C. Appendix A shows that the drag coefficient

38

518

C can be estimated from the viscosity ⌘, characteristic distance d and -volume fraction .

519

However, it also demonstrates the difficulty of choosing the parameters precisely, and in this

520

way precisely estimating the drag coefficient C.

521

As figure 4b shows, the choice

= 10 produces curves of strain concentration with spacing

522

similar to those observed in the experiment of Sowerby and Johnson (1970). As table 4

523

indicates,

524

there is strong resistance to relative motion between alpha-ferrite and cementite. As table 4

525

suggests, this work also considers values of C an order of magnitude lesser and greater.

526

= 10 corresponds to C = 2.2⇥109 kg m 3 s 1 . This suggests that, unsurprisingly,

The process of choosing

and C suggests a method by which (46), an expression for

527

the predicted distance along slip-lines between the first and second intersection of slip-lines,

528

could be used to estimate the drag coefficient C from experimental or real world observations.

529

Appendix C. A C 1 pressure perturbation in a yielding cylinder of mild steel

530

As seen in section 4.2, the sudden application of pressure leads to a propagating dis-

531

continuity. This appendix supplements section 4.2 by considering an initial condition very

532

di↵erent from sudden loading: one for which both the overpressure and the spatial derivative

533

of overpressure at both the inner and the outer wall vanish. It shows that the local pressure

534

minima which could locate lines of strain concentration do not depend essentially upon the

535

discontinuous application of pressure. Consider an initial pressure perturbation which will be referred to as a C 1 perturbation; ◆ 2✓ ◆2 ✓ 1 1 s˜1 1 s˜1 . (C.1) pˇ|t˜=0 =   By (41) it follows that cn = 2

Z

1

sin (n⇡˜ s1 ) 0

s˜12 (1

s˜1 )2 d˜ s1 .

(C.2)

536

The left column of figure C.6 shows scaled dimensionless overpressure profiles at a range of

537

dimensionless times for a C 1 perturbation and choices  = 1, 10 and 100. The scaling factors

538

(|ˆ p|max )

539

unity. The right column of figure C.6 shows values of |ˆ p|max at di↵erent dimensionless times

1

are chosen so that the maximum magnitude of the scaled pressure profiles are about

39

540

t˜. The factor exp

t˜ in the expression (43) of the components of pressure means that,

541

as this column suggests, as time passes the magnitude of the pressure profile approaches

542

0. For  = 10 and 100, as time passes the mode p˜1 begins to dominate and so the shape

543

of the pressure profile approaches that of the spatial factor of p˜1 , sin (n⇡˜ s1 ). For  = 1,

544

n⇤ , /⇡ = 0.318 < 1, and so no single term comes to dominate the shape of the profile.

545

Figures C.6c and C.6e show a di↵erence from the case of sudden loading; the dimensionless

546

time t˜ = 2 is not distinguished by being the time at which the discontinuity in and the

547

local maximum of the overpressure coincide. It follows that for a C 1 perturbation the

 = 1 (n⇤ = 0.318)

( p

max )

-1

p

1.0

1

p

k

max

0.5 10-2 0.5

1

s1

10-5

-0.5

10-8 0.5

-1.0

t˜ =

0

 = 10 (n⇤ = 3.18)

( p

max )

1/2 -1

1

p

2

8

16

1

2

5

10

20

1024

p

1.0

1

p

k

0.5

0.5

1

max

s1

-0.5 0.001 1

-1.0

t˜ =

˜ t

0

1

p

2

2

p 2 2

4

2

5

10

˜ t

8

Figure C.6: Continues on a later page which contains a descriptive caption.

40

( p

max )

-1

p

 = 100 (n⇤ = 31.8)

1.0 1

p

k

0.5

max

0.001 0.5

1

s1 0.0001

-0.5 0.00001 1

-1.0

t˜ =

0

2

p 2 2

4

64

256

10

100

˜ t

1024

Figure C.6: Continued from an earlier page. Left column. Time evolution of the scaled dimensionless overpressures of a yielding hollow cylinder the pressure profile of which is perturbed from equilibrium according to the initial condition (C.1). The plots apply to the parameters for mild steel estimated in Appendix B and listed in table B.5. For  = 10, the troughs are particularly clear at t˜ = 2. Right column. Values of the inverse of the scaling factor |ˆ p|max at di↵erent dimensionless times t˜. Both abscissa and ordinate use a logarithmic scale. The initial profile is scaled by the inverse of

1 k

(|ˆ p|max )|t˜=0 = 6.25 ⇥ 10

2

, the maximum

magnitude of the initial pressure profile (C.1) (attained at s˜1 = 0.5). Any times t˜ > 0 which are represented in a pressure profile on the left but are absent on the right have values of |ˆ p|max which are cumbersomely many orders of magnitude less than those which are shown. Each row of the figure pertains to the signature quantities listed in the corresponding row of table 4.

548

dimensionless time t˜ = 2 is not the earliest positive dimensionless time for which there is an

549

interior point most favourable for strain in its local neighbourhood. However, figure C.6c

550

shows that when  = 10 the troughs of pressure, which correspond to regions more favourable

551

for strain, are particularly prominent when t˜ = 2. Figure C.7b shows characteristics which

552

first intersect at the inner wall of the cylinder (where s˜1 = 1), and next at the pressure

553

trough for t˜ = 2 which is closer to the inner wall of the cylinder; s˜1 = 0.860.

554

Interestingly, although figure C.6e, corresponding to sudden loading and the choice  =

555

100, exhibits sharp troughs in overpressure, the dimensionless time when they are most

556

prominent is not t˜ = 2. Note that the pressure extrema evident after a C 1 perturbation are

557

only troughs because the initial perturbation involves a initial reduction in pressure relative 41

C 1 perturbation,  = 10

x2 [in.] 1.0 0.5

-1.0 -0.5

0.5

1.0

x1 [in.]

-0.5 -1.0 Figure C.7: Characteristics and slip-lines. (a) Figure 9.3a of Sowerby and Johnson (1970). Slip-lines in a mild steel cylinder of outer diameter ro = 2.2500 and inner diameter ri = 100 . The inner diameter was inferred from the image and knowledge of the outer diameter. (b) Characteristics which intersect at s˜1 = 1 and s˜1 = 0.860. The latter is a prominent trough of the overpressure after the C 1 perturbation which the initial condition (C.1) describes if  = 10.

558

to quasistatic pressure. If the negative of this perturbation were considered then peaks in

559

overpressure would be identified in place of troughs.

42