You are on page 1of 7

CEE 471, Fall 2019: HW6 Solutions

Bhavesh Shrimali, Aditya Kumar


Department of Civil and Environmental Engineering, University of Illinois, Urbana–Champaign, IL 61801, USA

Problem 1:

We consider the combined extension and inflation of a circular cylindrical tube. As depicted on Figure
1(a), the circular cylindrical tube has a length L and inner and outer radii A and B in its undeformed
configuration Ω0 and length l and inner and outer radii a and b in its deformed configuration Ω as indicated
on Figure 1(b). This tube is subjected to inner and outer Cauchy pressures Pi and Po on ∂Ωi and ∂Ωo
with respective unit normal outward vectors ni and no as well as an extension force N normal to its

section on ∂Ω+z and ∂Ωz with respective unit normal outward vectors nz = ±e3 . Recall the term Cauchy
pressure means the pressure remains constant in deformed configuration. The tube is made of an isotropic
incompressible nonlinear elastic material with stored-energy function W = W (λ1 , λ2 , λ3 ). We therefore

Figure 1: Combined extension and inflation of a circular tube in its (a) undeformed and (b) deformed configurations.

want to solve the following boundary value problem

div T = 0

 in Ω
 T


 T =T in Ω
∂W T


 T= F − qI in Ω


∂F (1)
 J = det F =1 in Ω
Tn = −P i ni on ∂Ωi



 i
Tn = −P o no on ∂Ωo



 o R
N = ∂Ω± Tn z ds on ∂Ω±

z z

where the two first equations correspond to the balance of linear and angular momemta (BLM and BAM),
the following two to the stress-deformation relation and the incompressibility constraint and the final
three to the boundary conditions. Note that the Lagrange multiplier q is introduced in the constitutive
relation to account for additional equation coming from incompressibility constraint. Also recall here that
balance of angular momentum is automatically satisfied when W is objective. As we have already said that
W = W (λ1 , λ2 , λ3 ), it is indeed objective and we don’t need to consider the BAM equation. Further the

Email addresses: bshrima2@illinois.edu (Bhavesh Shrimali), akumar51@illinois.edu (Aditya Kumar)


choice is made here to carry out the calculations on the deformed configuration Ω with the Cauchy stress
tensor T for simplicity but the problem could equivalently be solved on the undeformed configuration Ω0
using the first Piola-Kirchoff stress tensor S.
We start with the kinematic description associated with the geometry of the problem. The undeformed
configuration Ω0 is described as Ω0 = {X| A ≤ R ≤ B and 0 ≤ X3 ≤ L} with R the radial distance to
the axis of the cylinder defined through R2 = X12 + X22 . Similarly, Ω = {x | a ≤ r ≤ b and 0 ≤ x3 ≤ l}
with r the radial distance to the axis of the cylinder defined through r2 = x21 + x22 . Based on this
representation of deformed configuration which is obtained through inspection, we assume a mapping
x = χ (X) corresponding to a radial extension in the plane (0, e1 , e2 ) transverse to the cylinder and a
longitudinal extension along e3 . It reads as

χ (X) = f (R)(X1 e1 + X2 e2 ) + λz X3 e3
= f (R)Ru + λz X3 e3 (2)

X1 X2 X X3
with the unit radial vector (in this cylindrical setting) u = e1 + e2 = − e3 or in component
R R R R
Xi X3
form ui = − δi3 and a constant λz yet to be determined. Note that the first term of Equation (2)
R R
corresponds to the radial extension while the second one is the longitudinal extension1 .
The next step is to compute the deformation gradient F. From Equation (2) we have

∂R ∂R ∂ui
Fij = f 0 (R) Rui + f (R) ui + f (R)R + λz δi3 δj3 . (3)
∂Xj ∂Xj ∂Xj

∂R
We compute with
∂Xj

R2 = X12 + X22 = Xj Xj − X32


∂R
⇒2R = 2δij Xj − 2X3 δi3
∂Xi
∂R Xi X3
⇒ = − δi3 = ui , (4)
∂Xi R R
∂ui
and R with
∂Xj

Rui = Xi − X3 δi3
∂R ∂ui
⇒ ui + R = δij − δi3 δj3
∂Xj ∂Xj
∂ui
⇒R = δij − δi3 δj3 − ui uj . (5)
∂Xj

Plugging Equations (4-5) in Equation (3) yields

Fij = [Rf 0 (R) + R]ui uj + f (R)[δij − δi3 δj3 − ui uj ] + λz δi3 δj3 . (6)

It is important to recognize that F is already expressed in its spectral form. This is a generalization
of the form discussed in HW5-Pbm3 to the case of three different eigenvalues. Hence the eigenvalues
of F are Rf 0 (R) + f (R), λz and f (R) with the associated unit eigenvectors ui and δi3 and eigentensor
δij − δi3 δj3 − ui uj .

1 One could start with a more general mapping of the form χ (X) = f (R)Ru+g(X )e . In such a case, the incompressibility
3 3
constraint (16 ) to be used later would lead to g(X3 ) = λz X3 with λz = l/L via a separable partial differential equation
similarly to the bending-of-a-block problem encountered in HW3-Pbm3. This is “guessed” right away here to simplify the
derivation.
2
Finally, since F is symmetric, we have from the polar decomposition of F, R = I and F = U = V.
Therefore, the eigenvalues of F are nothing but the principal stretches
λ1 = Rf 0 (R) + f (R), λ2 = f (R) and λ3 = λz , (7)
and its eigenvectors and eigentensor are the Eulerian (or Lagrangian since they are the same in this
problem) principal directions.
To make further progress, we next focus on the behavior of the isotropic incompressible nonlinear
elastic material with stored-energy function W = W (λ1 , λ2 , λ3 ).
The incompressibility constraint (14 ) imposes det F = (Rf 0 (R) + f (R))f (R)λz = 1. Hence
1 − λz f (R)2
Rf 0 (R) =
λz f (R)
Z f (R) Z R
λz s ds
⇒ ds =
f (A) 1 − λ z s2
A s
1 1
⇒ − ln |1 − λz f (R)2 | + ln |1 − λz f (A)2 | = ln R − ln A
2 2
2 2
λ z f (A) A − A2
⇒λz f (R)2 = 1 + . (8)
R2
Using the mapping χ(X), we have

x1 = f (R)X1
⇒ r = f (R)R. (9)
x2 = f (R)X2
a
Thus f (A) = and
A s
1 λz a2 − A2
f (R) = + , ... (20 points) (10)
λz λz R 2
which actually completes the kinematic description for the problem. Once a and λz are determined, the
deformation field would be known for all points in the tube.
Next, the Cauchy stress-deformation relation for an incompressible material given in Equation (17 )
reads in indicial notation as
∂W
Tij = Fjk − qδij . (11)
∂Fik
We proceed with
∂W ∂W ∂λ1 ∂W ∂λ2 ∂W ∂λ3
= + + . (12)
∂Fik ∂λ1 ∂Fik ∂λ2 ∂Fik ∂λ3 ∂Fik
But from Equations (6) and (7), we have
∂λ1
λ1 = Fpq uq up ⇒ = ui uk ,
∂Fik
∂λ2
λ2 = Fpq (δpq − δp3 δq3 − up uq ) ⇒ = δik − δi3 δk3 − ui uk ,
∂Fik
∂λ3
and λ3 = Fpq δp3 δq3 ⇒ = δi3 δk3 .
∂Fik
Therefore with the deformation gradient (6) and Equation (12), Equation (11) reads as
Tij = t1 ui uj + t2 (δij − δi3 δj3 − ui uj ) + t3 δi3 δj3 (13)
with t1 , t2 and t3 the principal Cauchy stresses given by
∂W ∂W ∂W
t1 = λ1 − q, t2 = λ2 −q and t3 = λ3 − q. (14)
∂λ1 ∂λ2 ∂λ3
3
where recall that q is a Lagrange multiplier. Note that T is coaxial with V = F as expected for an
isotropic material.
We finally introduce for later convenience

Ŵ (λ, λz ) = W (λ−1 λ−1


z , λ, λz ) (15)

where λ is defined by transforming Equation (7) into


1
λ1 = Rf 0 (R) + f (R) = , λ2 = f (R) = λ and λ3 = λz . (16)
λλz
To relate the kinematic description of the deformation of the structure to the applied loads, we next
consider the equilibrium of the structure given by Equations (11 -12 ). Making use of Equation (13), balance
of linear momentum reads as
∂Tij
Tij,j =
∂xj
dt1 ∂r ∂ui uj dt2 ∂r ∂
= ui uj + t1 + (δij − δi3 δj3 − ui uj ) + t2 (δij − δi3 δj3 − ui uj )
dr ∂xj ∂xj dr ∂xj ∂xj
dt3 ∂r ∂δi3 δj3
+ δi3 δj3 + t3 . (17)
dr ∂xj ∂xj

Because the Lagrangian and Eulerian principal directions are the same (R = I) we can rewrite
xi x3
ui = − δi3
r r
∂r ∂ui
making use of Equation (9). Hence, similarly to Equations (4) and (5), = ui and r = δij −
∂xj ∂xj
δi3 δj3 − ui uj . It follows that

∂ui uj ∂ui ∂uj


r =r uj + rui
∂xj ∂xj ∂xj
= (δij − δi3 δj3 − ui uj )uj + ui (δjj − δj3 δj3 − uj uj ) = ui ,

and
∂ ∂ui uj
r (δij − δi3 δj3 − ui uj ) = −r = −ui .
∂xj ∂xj

Plugging in Equation (17) yields


dt1 t1 dt2 t2 dt3
Tij,j = ui uj uj + ui + (δij − δi3 δj3 − ui uj )uj − ui + δi3 δj3 uj
dr r dr r dr
t1 − t2
 
dt1
= + ui = 0
dr r
dt1 t1 − t2
⇒ + = 0. . . . (20 points) (18)
dr r
Making use of Equations (15-16) we compute

∂ Ŵ ∂W ∂λ1 ∂W 1 ∂W ∂W t2 − t1
= + =− 2 + =
∂λ ∂λ1 ∂λ ∂λ2 λ λz ∂λ1 ∂λ2 λ
with help of Equation (14). Hence from Equation (18) the ordinary linear differential equation

dt1 λ ∂ Ŵ
= , (19)
dr r ∂λ
4
with the boundary conditions from Equations (15 ) and (16 ). A similar simplification of the equilibrium
condition can not be performed in 1st P-K stress tensors. That is the reason we have chose to carry out
calculations in deformed configuration. Now since no = u and ni = −u, the boundary conditions read as
t1 (r = a) = −Pi and t1 (r = b) = −Po . Integrating Equation (18) gives for the radial Cauchy stress
Z r
λ ∂ Ŵ
t1 (r) = dy − Pi . (20)
a y ∂λ
Evaluating at r = b renders
Z b
λ ∂ Ŵ
∆P = Pi − Po = dy,
a y ∂λ
y
which after the change of variable s = (going from r to R as a and b are yet to be determined)
f (s)
becomes,
Z B
1 ∂ Ŵ
∆P = Pi − Po = ds. ... (10 points) (21)
A sλλz ∂λ

The traction boundary condition (17 ) on ∂Ω+


z for example leads to
Z
N = N e3 = t3 ds e3
∂Ω+
z
Z b
⇒ N = 2π rt3 (r) dr.
a

Making use of Equations (15-16) we compute

∂ Ŵ ∂W ∂λ1 ∂W 1 ∂W ∂W t3 − t1
= + =− 2 + =
∂λz ∂λ1 ∂λz ∂λ3 λλz ∂λ1 ∂λ3 λz

with help of Equation (14). It follows with Equation (20)


!
Z b Z r
∂ Ŵ λ ∂ Ŵ
N = 2π r λz + dy − Pi dr,
a ∂λz a y ∂λ

r y
or with the changes of variables R = and s = ,
f (R) f (s)
!
Z B Z R
R ∂ Ŵ 1 ∂ Ŵ
N = 2π λz + ds − Pi dR. (22)
A λz ∂λz A sλλz ∂λ

Note that Equations (21) and (22) together with λ = f (R) and Equation (10) provide an implicit way
to determine the unknowns constants a and λz . b follows directly from the incompressibility constraint
(B 2 − A2 )L = (b2 − a2 )l and l = λz L from the mapping. The hoop Cauchy stress t2 is solely determined
after the calculation of the Lagrange multiplier q with Equations (20) and (14), which closes the problem.

Finally, we define a parameter δ via B = A(1 + δ), and in the limiting case of a thin-walled shell as
δ → 0, any function g of B can be approximated through its Taylor series

∂g
+ O (B − A)2

g(B) = g(A) + (B − A)
∂B B=A


∂g
+ O(δ 2 ).

= g(A) + Aδ
∂B B=A
5
Specializing to ∆P (21) renders
! !
∂∆P 1 ∂ Ŵ 1 1 ∂ Ŵ
= =
∂B B=A Bλλz ∂λ
B=A Aλz λ ∂λ
R=A

!
δ 1 ∂ Ŵ
+ O(δ 2 ).

⇒ ∆P =
λz λ ∂λ
R=A

Similarly, specializing to N (22) renders


" Z B !#
∂N B ∂ Ŵ 1 ∂ Ŵ
= 2π λz + ds − Pi
∂B B=A λz ∂λz A sλλ z ∂λ
B=A
" !#
A ∂ Ŵ
= 2π λz − Po with Equation (21)
λz ∂λz R=A

!
∂ Ŵ Po
⇒ N = 2πA2 δ − + O(δ 2 ). ♣
∂λz R=A λz

Also note here that as ∆P = O(δ), an alternative expression for N is


!
2 ∂ Ŵ Pi
N = 2πA δ − + O(δ 2 ). . . . (25 points)
∂λz R=A λz

6
Problem 2:

Fig. (2a) gives a plot of the numerical solution computed via the “Shooting” method. It compares
fairly good with the asymptotic result, even though the latter corresponds to a completely incompressible
hyperelastic solid (κ = +∞), especially for lower values of radial stretch f (R) (Fig. (2b)).
This is due to the fact that the shell is fairly incompressible (κ/µ = 20). It can be readily verified that
increasing the value of the bulk modulus (κ), such that κ/µ > 100, will render a solution very close to the
asymptotic result, more conspicuous for smaller values of the applied internal pressure. For comparison
purposes, the solution corresponding to κ/µ = 200 is plotted in Fig. (2b) (bottom-right corner) and the
reference solution for a completely incompressible material, i.e. when Rf 0 + f = 1/f 2 , is plotted in red.
. . . (25 points)

Numerical Asymptotic 1.00 κ/µ = 20 κ/µ = 20 κ/µ = +∞


0.08 κ/µ = 200
0.95

0.90
0.06
The solution on the top-left corresponds to
0.85 p = 0.01 MPa, whereas the one on the
Rf (R) + R

The numerical solution


p (MPa)

bottom-right correspond to p = 0.075MPa


is valid only for 0.976

0.04 p < 0.08 0.80 0.975


κ/µ = 20 κ/µ = +∞

0.974

0.75 0.973

0.972
0.02
Near p = 0.08 MPa, the thickness of the shell rapidly 0.70 0.971
approaches zero. Any more pressure results in the complete
rupture/failure of the shell. Therefore the numerical solution 0.970

beyond p = 0.08 MPa is non physical. 0.65 0.969

0.00 1.0140 1.0145 1.0150 1.0155 1.0160

1.0 1.1 1.2 1.3 1.4 0.60


b = B . f (B) 1.00 1.05 1.10 1.15 1.20 1.25 1.30
f (R)
(a) Deformed radius (b) vs the Cauchy pressure on
the internal surface of the shell (p) (b) Relative incompressibility

You might also like