84 Mechanical Behaviour of Plastics

This would suggest that D / h = 4 offers the best balance of transverse and
axial properties. In fact the ‘optimum’ ratio is essentially independent of the
thickness, h, but it depends on B and b. Fig. 2.33 shows this relationship and
reflects the combinations of B, b, D and h which will give the best design of
corrugations. Generally it is desirable to have B and b approximately equal and
it is also good design practice to keep D / h well below 10 (to avoid buckling).
Hence values of B, h and D in the shaded area tend to be best. A combination of
D / h = 4,b / h = 5 and B / h = 4 is known to give good results for rotationally
moulded products where shape must be used very effectively to ensure good
performance.

Fig. 2.33 Variation of optimum D/h with b/h

2.11 Mathematical Models of Viscoelastic Behaviour

Over the years there have been many attempts to simulate the behaviour of
viscoelastic materials. This has been aimed at (i) facilitating analysis of the
behaviour of plastic products, (ii) assisting with extrapolation and interpolation
of experimental data and (iii) reducing the need for extensive, time-consuming
creep tests. The most successful of the mathematical models have been based on
spring and dashpot elements to represent, respectively, the elastic and viscous
responses of plastic materials. Although there are no discrete molecular struc-
tures which behave like the individual elements of the models, nevertheless
Mechanical Behaviour of Plastics 85

they do aid in the understanding and analysis of the behaviour of viscoelastic

materials. Some of the more important models will now be considered.

(a) Maxwell Model

The Maxwell Model consists of a spring and dashpot in series at shown in
Fig. 2.34. This model may be analysed as follows.

Fig. 2.34 The Maxwell model

Stress-Strain Relations
The spring is the elastic component of the response and obeys the relation
(TI = 6 E1
f
(2.27)
where LTI and ~1 are the stress and strain respectively and 6 is a constant.
The dashpot is the viscous component of the response and in this case the
stress a2 is proportional to the rate of strain .42, ie
a2 = r] * &2 (2.28)
where r] is a material constant.

Equilibrium Equation
For equilibrium of forces, assuming constant area
Applied Stress, (T = (TI = 0 2 (2.29)

Geometry of Deformation Equation

The total strain, E is equal to the sum of the strains in the two elements.
so
E=&1+&2 (2.30)
86 Mechanical Behaviour of Plastics

From equations (2.27), (2.28) and (2.30)

1. 1
6 = - g 1 + -a2
t 9
1 1
&=-.&+-.a (2.3 1)
6 9
This is the governing equation of the Maxwell Model. It is interesting
to consider the response that this model predicts under three common time-
dependent modes of deformation.

0) creep
If a constant stress, a,, is applied then equation (2.31) becomes
. 1
E=-.*, (2.32)
9
which indicates a constant rate of increase of strain with time.
From Fig. 2.35 it may be seen that for the Maxwell model, the strain at any
time, 1, after the application of a constant stress, a,, is given by

E(t) =-
t
*o
+ -9t
go

Fig. 2.35 Response of Maxwell model

Mechanical Behaviour of Plastics 87

Hence, the creep modulus, E @ ) , is given by

a0
E(t)= - =- 6rl (2.33)
€0) r l + &
(ii) Relaxation
If the strain is held constant then equation (2.31) becomes

o = -1 . b + -1 - 0
6 rl
Solving this differential equation (see Appendix B) with the initial condition
a = a, at t = to then,
-1,
a(t)= a,e 1) (2.34)
a(t) = a0e-f/TR (2.35)
where T R = q / 6 is referred to as the relaxation time.
This indicates that the stress decays exponentially with a time constant of
q/c (see Fig. 2.35).

(iii) Recovery
When the stress is removed there is an instantaneous recovery of the elastic
strain, E ' , and then, as shown by equation (2.31), the strain rate is zero so that
there is no further recovery (see Fig. 2.35).
It can be seen therefore that although the relaxation behaviour of this model
is acceptable as a first approximation to the actual materials response, it is
inadequate in its prediction for creep and recovery behaviour.

(b) Kelvin or Voigt Model

In this model the spring and dashpot elements are connected in parallel as
shown in Fig. 2.36.

Fig. 2.36 The Kelvin or Voigt Model

88 Mechanical Behaviour of Plastics
Stress-Strain Relations
These are the same as the Maxwell Model and are given by equations (2.27)
and (2.28).

Equilibrium Equation
For equilibrium of forces it can be seen that the applied load is supported
jointly by the spring and the dashpot, so

0=0
1 +02 (2.36)

Geometry of Deformation Equation

In this case the total strain is equal to the strain in each of the elements, i.e.

&=E1 =E2 (2.37)

From equations (2.27), (2.28) and (2.36)

or using equation (2.37)

0.=6.E+Q.i (2.38)

This is the governing equation for the Kelvin (or Voigt) Model and it is
interesting to consider its predictions for the common time dependent defor-
mations.

(i) Creep
If a constant stress, a,, is applied then equation (2.38) becomes

a, = c E + vi
*

and this differential equation may be solved for the total strain, E , to give

where the ratio q / 6 is referred to as the retardation time, T R .

This indicates an exponential increase in strain from zero up to the value,
uo/6,that the spring would have reached if the dashpot had not been present.
This is shown in Fig. 2.37. As for the Maxwell Model, the creep modulus may
be determined as
(2.39)
Mechanical Behaviour of Plastics 89

Fig. 2.37 Response of KelvinNoigt model

(ii) Relaxation
If the strain is held constant then equation (2.38) becomes
Q=<.E

That is, the stress is constant and supported by the spring element so that the
predicted response is that of an elastic material, i.e. no relaxation (see Fig. 2.37)

(iii) Recovery
If the stress is removed, then equation (2.38) becomes
0 =6 . E + r)&
Solving this differential equation with the initial condition E = E’ at the time
of stress removal, then
$.!
E ( t ) = de- i) (2.40)
This represents an exponential recovery of strain which is a reversal of the
predicted creep.

(c) More Complex Models

It may be seen that the simple Kelvin model gives an acceptable first approx-
imation to creep and recovery behaviour but does not account for relaxation.
The Maxwell model can account for relaxation but was poor in relation to creep
90 Mechanical Behaviour of Plastics
and recovery. It is clear therefore that some compromise may be achieved by
combining the two models. Such a set-up is shown in Fig. 2.38. In this case
the stress-strain relations are again given by equations (2.27) and (2.28). The
geometry of deformation yields.
Total strain, E = ~1 + ~2 + &k (2.41)

Fig. 2.38 Maxwell and Kelvin models in series

where &k is the strain response of the Kelvin Model. From equations (2.27),
(2.28) and (2.41).

E(f)=-+-+-
t1
ao ar,
l lt

From this the strain rate may be obtained as

uo
t2 [l-e
-“I
12 (2.42)

a, a, -h,
&=-+-e t)2 (2.43)
111 r12

The response of this model to creep, relaxation and recovery situations is the
sum of the effects described for the previous two models and is illustrated in
Fig. 2.39. It can be seen that although the exponential responses predicted in
these models are not a true representation of the complex viscoelastic response
of polymeric materials, the overall picture is, for many purposes, an acceptable
approximation to the actual behaviour. As more and more elements are added
to the model then the simulation becomes better but the mathematics become
complex.
Example 2.12 An acrylic moulding material is to have its creep behaviour
simulated by a four element model of the type shown in Fig. 2.38. If the creep
curve for the acrylic at 14 MN/m* is as shown in Fig. 2.40, determine the
values of the four constants in the model.
Mechanical Behaviour of Plastics 91

Fig. 2.40 Creep curve for acrylic at 20°C

Solution The spring element constant, (1, for the Maxwell model may be
obtained from the instantaneous strain, ~ 1 Thus
.

l4 -
a0 = -
6, = - -2800MN/m2
E] 0.005
The dashpot constant, q1, for the Maxwell element is obtained from the slope
of the creep curve in the steady state region (see equation (2.32)).
00 14
7 ) 1 = y = = 1.2 x 10' m.hr/mz
E 1.167 x 1O-6
= 4.32 x 10" MN.s/m2
92 Mechanical Behaviour of Plastics
The spring constant, 62. for the Kelvin-Voigt element is obtained from the
maximum retarded strain, ~ 2in, Fig. 2.40.
a0 14
6 2 z - 1 = 7000 MN/m2
~2 (0.7 - 0.5)10-2
The dashpot constant, q2, for the Kelvin-Voigt element may be determined
by selecting a time and corresponding strain from the creep curve in a
region where the retarded elasticity dominates (Le. the knee of the curve
in Fig. 2.40) and substituting into equation (2.42). If this is done then r)2 =
3.7 x lo8 MN.s/m2.
Having thus determined the constants for the model the strain may be
predicted for any selected time or stress level assuming of course these are
within the region where the model is applicable.

(d) Standard Linear Solid

Another model consisting of elements in series and parallel is that attributed to
Zener. It is known as the Standard Linear Solid and is illustrated in Fig. 2.41.
The governing equation may be derived as follows.

Fig. 2.41 The standard linear solid

Stress-Strain Relations
As shown earlier the stress-strain relations are
0 1 = 61E1 (2.44)
02 = 62.52 (2.45)
03 = v3&3 (2.46)

Equilibrium Equation
In a similar manner to the previous models, equilibrium of forces yields.
01 =a 3
a =0 1 +a 2 (2.47)
Mechanical Behaviour of Plastics 93
Geometry of Deformation Equation
In this case the total deformation, E , is given by
E = E2 = E1 + E3 (2.48)
From equation (2.48)
€=&1+&3

but from equation (2.47)

irl = ir - 6 2

and a 3 = a - a 2
. ir - 6 2 E
E=-
61
+-a -62.5
r/3

Rearranging gives
r/3& + 610 = r/3(61 + 62% + 6261E (2.49)
This is the governing equation for this model.
The behaviour of this model can be examined as before

0) CWP
If a constant stress, a,, is applied then the governing equation becomes
i{r/3(61 + $2)) + 6162s - 6 l a o = 0
The solution of this differential equation may be obtained using the boundary
+
condition E = a,/(61 62) at t = 0. So

(2.50)

It may be seen in Fig. 2.42 that this predicts the initial strain when the stress
is first applied as well as an exponential increase in strain subsequently.

(ii) Relaxation
If the strain is held constant at E', then the governing equation becomes
q3ir + (la - 6162E' =0
This differential equation may be solved with the boundary condition that
+
a = a, = ~'(61 62) when the strain is first kept constant.

(2.51)

This predicts an exponential decay of stress as shown in Fig. 2.42.

94 Mechanical Behaviour of Plastics

Fig. 2.42 Response of standard linear solid

(iii) Recovery
If the stress is at a value of a’and then completely removed, the governing
equation becomes
q3(61 e2)i +
e1626 = 0+
The solution of this differential equation may be obtained using the boundary
condition that when the stress is removed, the strain is given by
Ef = af/O1 +t2)
-hht
strain, E ( t ) =
{ E’ -
+h)
(61 (TI 1 e q 3 (5+b )

This predicts an instantaneous recovery of strain followed by an exponential

(2.52)

decay.
It may be observed that the governing equation of the standard linear solid
has the form
alu +
a,a = b l i bo& +
Mechanical Behaviour of Plastics 95

where a l , a,, bl and bo are all material constants. In the more modern theories
of viscoelasticity this type of equation or the more general form given in
equation (2.53) is favoured.

The models described earlier are special cases of this equation.

2.12 Intermittent Loading

The creep behaviour of plastics considered to date has assumed that the level of
the applied stress is constant. However, in service the material may be subjected
to a complex pattern of loading and unloading cycles. This can cause design
problems in that clearly it would not be feasible to obtain experimental data to
cover all possible loading situations and yet to design on the basis of constant
loading at the maximum stress would not make efficient use of material or be
economical. In these cases it is useful to have methods of predicting the extent
of the recovered strain which occurs during the rest periods of conversely the
accumulated strain after N cycles of load changes.
There are several approaches that can be used to tackle this problem and
two of these will be considered now.

2.12.1 Superposition Principle

The simplest theoretical model proposed to predict the strain response to a
complex stress history is the Boltzmann Superposition Principle. Basically this
principle proposes that for a linear viscoelastic material, the strain response to
a complex loading history is simply the algebraic sum of the strains due to each
step in load. Implied in this principle is the idea that the behaviour of a plastic
is a function of its entire loading history. There are two situations to consider.

(a) Step Changes of Stress

When a linear viscoelastic material is subjected to a constant stress, C T ,~ at time,
at any subsequent time, c, may be expressed as
t l , then the creep strain, &(t),

(2.54)

where E(t - t l ) is the time-dependent modulus for the elapsed time (t - t l ) .

Then suppose that instead of this stress 0 1 , another stress, 0 2 is applied at
some arbitrary time, t 2 , then at any subsequent time, t , the stress will have
been applied for a time (t - t 2 ) so that the strain will be given by
96 Mechanical Behaviour of Plastics
Now consider the situation in which the stress, 01, was applied at time, t l ,
and an additional stress, a2, applied at time, t2, then Boltzmanns’ Superposition
Principle states that the total strain at time, t, is the algebraic sum of the two
independent responses.

This equation can then be generalised, for any series of N step changes of
stress, to the form
i=N
1
(2.55)
i= 1
E(? - ti)
where ai is the step change of stress which occurs at time, ti.
To illustrate the use of this expression, consider the following example.
Example 2.13 A plastic which can have its creep behaviour described
by a Maxwell model is to be subjected to the stress history shown in
Fig. 2.43(a). If the spring and dashpot constants for this model are 20 GN/m2
and 1000 GNs/m2 respectively then predict the strains in the material after 150
seconds, 250 seconds, 350 seconds and 450 seconds.
Solution From Section 2.11 for the Maxwell model, the strain up to 100s is
given by
a at
E(t) = - + -
( ‘ I
Also the time dependent modulus E ( t ) is given by

(2.56)

Then using equation (2.54) the strains may be calculated as follows:

(i) at t = 150 seconds; ai = 10 MN/m2 at tl = 0, a 2 = -10 MN/m2 at t2 =
100 s

= 0.002 - 0.001 = 0.1%

(ii) at 250 seconds; 01,a 2 as above, a3 = 5 MN/m2 at tg = 200 s

+ 4 . (250 - 200)
= 0.003 - 0.002
1
+ 0.0005 = 0.15%
Mechanical Behaviour of Plastics 97

Fig. 2.43 Strain predictions using superposition theory

(iii) at 350 seconds; a1,a2, a3 as above, a4 = 10 MN/m2 at t4 = 300 s

so,
~ ( 3 5 0=
) 0.003 = 0.3%
(iv) and in the same way
~ ( 4 5 0=
) 0.004 = 0.4%
The predicted strain variation is shown in Fig. 2.43(b). The constant strain
rates predicted in this diagram are a result of the Maxwell model used in
this example to illustrate the use of the superposition principle. Of course
superposition is not restricted to this simple model. It can be applied to any
type of model or directly to the creep curves. The method also lends itself to a
graphical solution as follows. If a stress 01 is applied at zero time, then the creep
curve will be the time dependent strain response predicted by equation (2.54).
When a second stress, 0 2 is added then the new creep curve will be obtained
by adding the creep due to 0 2 to the anticipated creep if stress 01 had remained
98 Mechanical Behaviour of Plastics

alone. This is illustrated in Fig. 2.44(a). Then if all the stress is removed this
is equivalent to removing the creep strain due to 01 and a2 independently as
shown in Fig. 2.44(b). The procedure is repeated in a similar way for any other
stress changes.

Fig. 2.44(b) Predicted strain response using Boltzmann’s superposition principle

(b) Continuous Changes of Stress

If the change in stress is continuous rather than a step function then equa-
tion (2.55) may be generalis4 further to take into account a continuous loading
cycle. So

& ( t )=
i
--bo
E(t
1
-tl)
.-d d t ) . dt
dt
(2.57)

where a(t) is the expression for the stress variation that begins at time, tl.
The lower limit is taken as minus infinity since it is a consequence of the
Mechanical Behaviour of Plastics 99
Superposition Principle that the entire stress history of the material contributes
to the subsequent response.
It is worth noting that in exactly the same way, a material subjected to a
continuous variation of strain may have its stress at any time predicted by

~ ( t=)
1
--bo
E ( t - t) . -
dWdt
dz
(2.58)

To illustrate the use of equation (2.57) consider the following Example.

Example 2.14 A plastic is subjected to the stress history shown in Fig. 2.45.
The behaviour of the material may be assumed to be described by the Maxwell
model in which the elastic component c = 20 GN/m2 and the viscous compo-
nent r) = lo00 GNs/m2. Determine the strain in the material (a) after u1 seconds
(b) after u:! seconds and (c) after u3 seconds.

Fig. 2.45 Stress history to be analysed

Solution As shown in the previous Example, the modulus for a Maxwell

element may be expressed as
tr)
E(t) = -
rl +B
(a) The stress history can be defined as
--oo < t < 0, a(t)= 0 + do(t)/dt = 0
0 < t < T, a(t)= Kit + da(t)/dt = K1
100 Mechanical Behaviour of Plastics

Substituting into equation (2.57)

-02 0

then €or u1 = 50 seconds and K1 = 0.1 MN/m2 s

~ ( 5 0=) 0.1(50) ( 1
2 0 x 103
+ A)
2 x io6
= 0.037%

€(loo) = O . l ( l O 0 )
(20 lo3
+X
2 x lo6
)= 0.1%
It is interesting to note that if K1 was large (say K1 = 10 in which case
T = 1 second) then the strain predicted after application of the total stress
(10 MN/m2) would be ~ ( 1= ) 0.0505%. This agrees with the result in the
previous Example in which the application of stress was regarded as a step
function. The reader may wish to check that if at time T = 1 second, the stress
was held constant at 10 MN/m2 then after 100 seconds the predicted strain
using the integral expression would be ~(100)= 0.1495% which again agrees
with the previous example.
(b) After the time T, the change in stress is given by
Change in stress, a(t)= -Kl(t - T) + K2(t - T )

Hence,

where tl = 0 and t2 = T = 100 s.

Mechanical Behaviour of Plastics 101

then for K2 = -0.2 MN/m2s, T = 100 seconds and u:! = 125 seconds
~ ( 1 2 5=
) 0.094%
(c) After the time TI, the change in stress is given by
Change in stress, o(t) = K2(t - TI)
d4t)
-= K2
dt
Hence,

and from this, for u3 = 200 seconds and K2 = -0.2 MN/m2s

~ ( 2 0 0=
) 0.075%
This will in fact be constant for all values of u3 because the Maxwell Model
cannot predict changes in strain if there is no stress. The overall variation in
strain is shown in Fig. 2.46.
Example 2.15 In the previous Example, what would be the strain after
125 seconds if (a) the stress remained constant at 10 MN/m2 after 100 seconds
and (b) the stress was reduced to zero after 100 seconds.
Solution
(a) If the stress was kept constant at 10 MN/m2 after 100 seconds as shown
in Fig. 2.47 then the effective change in stress would be given by
change in stress, a(t) = -Kl(t - T)
ddt)
-= -K1
dt
so

K I U ~ T KIT K1T2
E(U2) = ___ +---
tl e 2tl
~ ( 1 2 5=
) 0.125%
102 Mechanical Behaviour of Plastics

Fig. 2.46 Variation of strain with time

Fig.2.41 Stress history: Example 2.15(a)

(b) If the stress was completely removed after 100 seconds as shown in
Fig. 2.48 then the effective change in stress would be given by
change in stress, a(t)= - K l ( t - T ) - Aa
ddt)
- = -K1
dt
Mechanical Behaviour of Plastics 103

Fig. 2.48 Stress history: Example 2.15(b)

so

&3) = J'
u2

e'
+ e t - et1 K i d t + 1' "6'
u2
+ - et2 ( - K l ) d t - A0
E(U2 - T)
fl f2

K,T~
E(U3) =-
2'
~ ( 1 2 5=
) 0.05%
It is apparent therefore that the Superposition Principle is a convenient
method of analysing complex stress systems. However, it should not be
forgotten that the principle is based on the assumption of linear viscoelasticity
which is quite inapplicable at the higher stress levels and the accuracy
of the predictions will reflect the accuracy with which the equation for
modulus (equation (2.33)) fits the experimental creep data for the material.
In Examples (2.13) and (2.14) a simple equation for modulus was selected in
order to illustrate the method of solution. More accurate predictions could have
been made if the modulus equation for the combined MaxwellKelvin model
or the Standard Linear Solid had been used.

2.12.2 Empirical Approach

As mentioned earlier, it is not feasible to generate test data for all possible
combinations of load variations. However, there have been a number of ex-
perimental investigations of the problem and these have resulted in some very