Sei sulla pagina 1di 10

J. Non-Newtonian Fluid Mech.

129 (2005) 7584

Re-examination of the approximate methods for interconversion


between frequency- and time-dependent material functions
I. Emri a, , B.S. von Bernstorff b , R. Cvelbar a , A. Nikonov a
a Center for Experimental Mechanics, University of Ljubljana, Cesta na Brdo 49, SI-1000 Ljubljana, Slovenia
b BASF, Carl-Bosch-Strasse 38, Ludwigshafen, Germany

Received 4 October 2004; received in revised form 28 May 2005; accepted 28 May 2005

Abstract

The paper examines the correctness of the best-known approximate methods of interconversion between the frequency- and the time-
dependent material functions. Approximate interconversions are compared to the close-form viscoelastic interrelations through the relaxation
and/or retardation spectra. The analysis showed that the simple method, G(t)  G ()|=1/t , should be avoided. The most successful was the
method of Schwarzl yielding results with maximum relative error within 2%. The magnitude of the error of interconversion is related to
the magnitude of the response function first derivative. When the first derivative = |d(log G ())/d(log())| < 0.5 all approximate methods
except the simple method furnish interconversions within the acceptable error, i.e., less than 5%.
2005 Elsevier B.V. All rights reserved.

Keywords: Interconversion; Material functions; Material characterization

1. Introduction excitation (displacement or force), are interrelated in a close


form via the corresponding relaxation and retardation spec-
Polymers are becoming increasingly important materi- trum. The two spectra are not measurable directly; they must
als in mechanical-, electrical- and civil-engineering. In such be calculated from the appropriate response function, mea-
applications, materials are expected to carry loads over sured either in time or frequency-domain. These calculations
extended period of time. This requires means of predict- require solution of an inverse problem, which happens to be
ing their long-term reliability, which furthermore demands ill-posed. Until recently, there was no appropriate solution
knowledge of viscoelastic material functions. To determine for these problems. Mainly for that reason several approx-
material functions in time-domain we need, in general, six imate methods for the interconversion between frequency-
independent experiments, three for force (or stress) excita- and time-dependent material functions have been developed
tion and three when the excitation is displacement (or strain). in the past. Many of these algorithms are still in use, usually
In addition, we need the same number of experiments to as part of the software packages supporting different appa-
determine material functions in frequency-domain (dynamic ratuses for the viscoelastic material characterization, e.g.,
material functions). These material functions are listed in rheometers and DTMA apparatuses. Most of these com-
Table 1. mercial apparatuses are dynamic, i.e., they measure mate-
Measurements in frequency-domain are common for char- rial functions in the frequency-domain. The time-domain
acterization of melts, while the time-domain experiments are response is then predicted using one of the approximate algo-
usually used for characterization of solids. Material func- rithms [1].
tions in time and frequency-domain, within a given mode of The interrelations between frequency- and time-domain
material functions are schematically presented in Fig. 1 [1].
Corresponding author. The ultimate goal of this paper is to examine the correct-
E-mail address: ie@siol.net (I. Emri). ness of the approximate methods of interconversion between

0377-0257/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnnfm.2005.05.008
76 I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584

Table 1
Time- and frequency-dependent viscoelastic material functions
Mode of loading Time-dependent Frequency-dependent

Modulus Compliance Modulus Compliance


Shear G(t) J(t) G () G () J () J ()
Volumetric K(t) B(t) K () K () B () B ()
Uniaxial E(t) D(t) E () E () D () D ()
Poissons ratio (t)  ()  ()

kind, which is an ill-posed problem. Until recently [214],


we had no stable solution for this group of problems. Thus,
for the interconversion between the frequency- and the time-
dependent material functions many approximate methods
have been developed in the past. Some of these methods are
still in use, mostly in the industrial laboratories. We examine
here some of the most popular algorithms.

2.1. Conversion from frequency- to time-dependent


functions

The simplest approximation to G(t) and J(t) using the real


part of the frequency-dependent modulus, G (), and com-
pliance, J (), is [16]:
Fig. 1. The interrelations between frequency- and time-dependent material
G(t)  G ()|=1/t , (1)
functions.

the frequency- and the time-dependent material functions. and


We therefore compare the approximate interconversions to
J(t)  J  ()|=1/t . (2)
the close-form viscoelastic interrelations through the discrete
relaxation and/or retardation spectra. The latter can be now For the compliances similar interrelation was proposed by
readily calculated using one of the algorithms for calcula- Riande and Markovitz [17] and by Koppelman [18]:
tion of mechanical spectra from the dynamic and/or static
response functions, e.g., [211]. J(t)  J()|
=1/t , (3)
We will first examine the approximate interrelations
between frequency- and time-dependent material functions where
[1225], using synthetic data generated from the known dis- 
crete spectrum. We carefully examine the effect of the shape J()
= (J  ())2 + (J  ())2 . (4)
of the spectrum on the accuracy of the interconversion. Next,
we analyze the accuracy of approximate methods for four dif- Christensen [12] has proposed slightly modified algorithm
ferent materials, i.e., PVAc, NR, EPDM and PA66 [2628]. for prediction of G(t) and J(t) from the frequency-dependent
modulus, G (), and compliance, J (). He replaced fre-
quency by = 2/t to obtain:
2. Approximate interrelations G(t)  G ()|=2/t , (5)
Mechanical spectrum cannot be measured directly. It has and
to be calculated from one of the response functions, usu-
ally from the harmonic (dynamic) response functions, e.g., J(t)  J  ()|=2/t . (6)
G () and G (). Dynamic measurements are most widely
used in industry, because they are easily performed on one of Ninomiya and Ferry [19,29] developed an algorithm for cal-
the commercially available apparatuses.1 Calculation of the culating G(t) and J(t) that comprises contributions from the
relaxation and retardation spectra requires an inverse solu- real and the imaginary part of the dynamic modulus and
tion of the Fredholm integral equation of the first or second dynamic compliance, respectively,

1 G(t)  [G () 0.4G (0.4) + 0.014G (10)]=1/t , (7)


It has to be noted that due to the heat dissipation, the accuracy of these
measurements may be a weak point, particularly when experiments are per-
formed in the vicinity of glass-transition temperature. J(t)  [J  () + 0.4J  (0.4) 0.014J  (10)]=1/t . (8)
I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584 77

Schwarzl and Struik [20] obtained several approximations of and


similar form. One, which is most often used, is:
G ()  0.470G(4t) 2.144G(2t) + 1.476G(t)
  t t t
G(t)  G () 0, 337G (0, 323)|=1/t . (9)
0.422G + 0.608G 0.160G
In the same paper, they proposed also an approximation to 2 4 8
 t 
G(t) derived solely from G ():
+ 0.172G |t=1/ . (16)
16
G(1.44t)  G () 0, 4[G (1, 59) G (0, 193)]|=1/t ,
Relations for J () and J () may be again obtained through
(10)
the substitutions (14). Similar relations were proposed by
which is however less precise. Schwarzl [22,31] developed Yagii and Maekawa [23,29]:
several approximations for calculation of time-dependent
G ()  G(t) + 1.080[G(1.585t) G(2.512t)]
material functions from the corresponding frequency-
dependent functions. For those cases when damping, tan , + 0.159[G(0.251t) G(0.398t)]|t=1/ , (17)
is small he proposed a simple formula:
and
G(t)  G () 0.566G (0.5) + 0.203G ()|=1/t ,
G ()  2.7[G(0.631t) G(t)]
(11)
and for all other cases, a more complex relation: + 0.794[G(0.1t) G(0.159t)]|t=1/ . (18)
 
G(t)  G () 0.00807G 0.00719G They have also proposed relations for the frequency-
16 8 dependent (dynamic) creep compliance functions:
 
+ 0.00616G 0.467G + 0.0918G () J  ()  J(t) + 1.08[J(1.585t) J(2.512t)] + 0.159
4 2
+ 0.0534G (2) 0.08G (4) + 0.0428G (8)|=1/t , [J(0.251t) J(0.398t)] + 0.025{f t}|t=1/ , (19)
(12) and
or
 J  ()  2.7[J(t) J(0.631t)] + 0.794[J(0.159t)
 
G(t)  G () 0.496G 0.0651
2 J(0.1t)] + 0.043{f t}|t=1/ . (20)
      
G G 0.0731 G G () The curly brackets in Eqs. (19) and (20) denote presence or
4 2 2
absence of the corresponding physical quantity, depending
0.111[G () G (2)] 0.03[G (8) G (16)] on the arrheodictic,2 or rheodictic behavior of the material
0.00683[G (32) G (64)]|=1/t . (13) [29].
Plazek and Raghupathi [25,29,30] proposed relations for
The relations for the creep functions in Eqs. (11)(13) may the frequency-dependent creep compliance, which takes into
be obtained through the substitutions: account the slope of the time-dependent creep compliance
G(t) J(t); G () J  (); function:

G () J  (). (14) J  ()  [1 m(2t)]0.8 Jr (t)|t=1/ , (21)


   0.8

There are of course many approximations in opposite direc-

2 1

tion, i.e., predictions of frequency-dependent material func- J  ()  m t Jr (t) + 0


, (22)
3

tions from the corresponding time-dependent ones. We t=1/


review those below. where
t
2.2. Conversion from time- to frequency-dependent Jr (t) = J(t) , (23)
functions 0
d(log Jr (t))
Let us start with one of the approximate relations proposed m= . (24)
d(log t)
by Schwarzl [22,31]:
It should be noted that when the slope of the creep compliance
G ()  0.142G(8t) 0.86G(4t) + 0.674G(2t) function, J(t), is small then the loss tangent, tan , becomes
t t
+ 0.942G(t) + 0.001G + 0.101G
2 4
t  t  2 The term rheodictic refers to a material showing steady-state flow and

0.00855G + 0.00855G |t=1/ , (15) arrheodictic denotes a material which does not ([29], p. 93).
8 16
78 I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584

independent of frequency, and may be approximated as:


d(log Jr (t))
tan  . (25)
2 d(log t)
Schapery and Park [15] recently developed new relations:

G ()   G(t)|=1/t , and G ()   G(t)|=1/t ,


(26)

where
 
 = (1 n) cos n , and Fig. 2. Schematics of the numerical experiments procedure.
2
 
 = (1 n) sin n , (27) on the arrheodictic,3 or rheodictic behavior of the material
2
[29].
with Next, we use the exact frequency response functions as
d(log G(t)) a starting point to predict the time-dependent response func-
n= , (28) tions, Gap (t), using the approximate relations (1), (5), (7), (9),
d(log t) (12) and (13). We used the subscripts ex and ap to indicate
which is a slowly varying function (1 < n < 1) and is a the so-called exact and approximate response functions,
Gamma function. respectively. The goodness of the approximate interconver-
sions we analyze by calculating the relative error:
Gap (t) Gex (t)
3. Numerical experiments = , (32)
Gex (t)
We first examine the approximate interrelations between and the average relative error:
frequency- and time-dependent material functions using syn- n

thetic data generated from the known discrete spectrum 1

Gap (ti ) Gex (ti )

=



. (33)
Hex (). Next, we analyze the accuracy of the approxi- n Gex (ti )
i=1
mate methods using measured relaxation functions, G(t),
for four different materials. From them, we first calculate This scheme of analysis is schematically presented in Fig. 2.
discrete spectrum Hex () using EmriTschoegl algorithm
[610]. As in previous case, we consider that Hex () is the 3.1. Synthetic spectra
correct known spectrum. Using the spectra, we gener-
ate the exact response functions in the frequency-domain We start the accuracy analysis of the approximate inter-
[29]: conversions with known synthetic spectra with different
distributions: (i) box, HB (), (ii) wedge, HW (), (iii) ramp,
i=n
i=n

2 i2 1 HR () and (iv) Lorentzian, HL (). In all four cases, we have
Gex () = {Ge } + Hi = G g Hi , generated 57 spectrum lines, ranging from log = 14 s to
i=1
1 + i
2 2
i=1
1 + 2 i2
log = 14 s, i.e., two lines per decade, using Gg = 109 Pa and
(29)
Ge = 99900.999 Pa as the glassy and the equilibrium modu-
lus, respectively. Strengths of the spectrum lines were in all
i=n
i cases normalized according to:
Gex () = Hi , (30)
1 + 2 i2 i=57
i=1
H i = Gg G e . (34)
and the time-domain, Gex (t): i=1
i=n
 
t According to this we have:
Gex (t) = {Ge } + Hi exp 
i
i=1 H
i
   H() = 57 (Gg Ge ), i ; i = 1, 2, 3, . . . , 57 ,
j=1 Hj
i=n

t
= Gg Hi 1 exp . (31) (35)
i
i=1

The curly brackets in Eqs. (29)(31) denote presence or 3 The term rheodictic refers to a material showing steady-state flow and

absence of the corresponding physical quantity depending arrheodictic denotes a material which does not ([29], p. 93).
I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584 79

for the Lorentzian distribution. The envelopes of the four


spectra are shown in Fig. 3.
Using the four spectra we can now generate the exact
response functions in the frequency-, Gex () and Gex () and
the time-domain, Gex (t), using the Eqs. (29)(31) respec-
tively. The exact response functions corresponding to the
four spectra were calculated for the logarithmic frequency-
and time-span ranging from 12 to +12, which is two decades
shorter on both sides than the corresponding spectrum to
account for the truncation error. The exact response func-
tions are shown in Fig. 4.
Starting from the exact frequency-dependent response
Fig. 3. Envelopes of the four spectra: HB (), HW (), HR (), HL (). functions, G () and G (), we can now calculate the corre-
sponding time-dependent material functions, Gap (t), using
where six different approximate interrelations, which we named
Gg Ge as: Simple (1), Christensen (5), Ninomiya and Ferry (7),
H
i = HB (i ) = , for the box distribution, (36) Schwarzl and Struik (9), Schwarzl2 (12) and Schwarzl3 (13),
57
respectively. The goodness of the approximate predictions,
i = HW (i ) = 109 i0.5 ,
H for the wedge distribution, expressed as relative-error are presented in Fig. 5.
From each of the four diagrams, one may extract the max-
(37) imum relative error, which is shown as bar diagram in Fig. 6a.
In Fig. 6b, we also show the corresponding average relative
9 1.35
10 i ; for logi < 6

Hi = HR (i ) = 7.94328; for 6 logi 6 ,
for the ramp distribution, and (38)

9 1.35
10 i ; for logi > 6

error calculated from the Eq. (33). It is interesting to


i = HL (i ) =  r 2  r ,
H with 0 = 1 and r = 0.5, note that the shape of the response function affects the
i
0 + 0i maximal relative error and the averaged relative error
(39) differently.

Fig. 4. Stress relaxation functions, G(t), storage, G () and loss moduli, G (), corresponding to different spectra.
80 I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584

Fig. 5. Comparison of different approximate interconversions of the response functions in terms of the relative error .

3.2. Interconversion of experimental data From the calculated spectra, we now again calculate the
exact frequency-dependent material functions Gex () and
We now proceed with the analysis of the approximate Gex (), using Eqs. (29) and (30), which are not shown. The
interconversions by using our own experimental data on frequency-dependent response functions Gex () and Gex (),
shear relaxation modulus, G(t), measured on four different served as a starting point to predict the time-dependent
materials: PVAc, NR, EPDM and PA66. The time-dependent response functions, Gap (t), using the approximate relations
response functions are displayed on the left-hand side in (1), (5), (7), (9), (12) and (13), as schematically presented in
Figs. 710. Details of the experimental procedures and data Fig. 2.
reduction are presented and discussed elsewhere [2628]. The relative errors of the approximate interconversions are
From the time-dependent response functions, we first cal- presented in Fig. 11. In the error analysis, we have skipped
culate the corresponding relaxation spectra, H(), using the one decade on both sides of the response function to avoid the
algorithm proposed by Emri and Tschoegl [6,10]. The cor- effect of the truncation error resulting from the calculation of
rectness of the calculated spectra was checked by calcu- spectra.
lating back the shear relaxation functions, G(t), using Eq. From each of the four diagrams, we can again extract the
(31). The maximum relative error was in all cases less than maximum relative error, which is shown as bar diagram in
0.5%. The four spectra are shown on the right-hand side in Fig. 12a. In Fig. 12b, we show the corresponding average
Figs. 710. relative error calculated from the Eq. (33).

Fig. 6. Comparison of approximate interconversions, in terms of the maximum relative error, max (a) and in terms of the average relative error,
(b).
I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584 81

Fig. 7. The relaxation function, G(t,T), on PVAc (a) and the corresponding relaxation spectrum, H() (b).

Fig. 8. The relaxation function, G(t,T), on NR (a) and the corresponding relaxation spectrum, H() (b).

Fig. 9. The relaxation function, G(t,T), on EPDM (a) and the corresponding relaxation spectrum, H() (b).

Fig. 10. The relaxation function, G(t,T), on PA66 (a) and the corresponding relaxation spectrum, H() (b).
82 I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584

Fig. 11. Comparison of different approximate interconversions of the PVAc, NR, EPDM and PA66 frequency-response functions, in terms of the relative error
.

Fig. 12. Comparison of different approximate interconversions of the material response functions, presented with the maximum, max (a) and the average
relative error,
(b).

4. Discussion and conclusions we have generated two additional response functions using
the wedge and the Lorentzian shaped spectra.
The first set of analysis was performed on synthetic data The analysis showed that the error of interconversion in
generated from the four different synthetic spectra. In order general corresponds to the first derivative of the response
to minimize the truncation error, the time span of spectra function:
was chosen to be four decades broader than the correspond-


d(log G ())

ing response functions (see [6] for the detailed analysis of =




(40)
d(log())

this problem). The spectra have been selected such that the
corresponding response functions mimic the shape of the The derivatives of the response functions are displayed in
most commonly used polymers in solid and molten state, i.e., Fig. 13.
very slowly and very rapidly decreasing response functions Comparing Figs. 5 and 13 shows that the location of the
in respect to the logarithmic time scale. maximum error for each of the four response functions coin-
The slowest and the fastest response functions corre- cides with the location of the maximum of the response
spond to the box and ramp shaped spectra. Between the two function derivative, max . The latter may be correlated with
I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584 83

ative error in interconversions of experimental response


functions, as function of the corresponding first derivative
maximum.
Fig. 15 confirms that the error of interconversion for
all approximate methods except the simple method remains
within the acceptable range when, = |d(log G ())/
d(log())| < 0.5.
From the presented analysis, several useful conclusions
may be drawn:

(i) The so called simple method, G(t)  G ()|=1/t , should


be avoided. For the synthetic (data with no experimen-
Fig. 13. First derivatives of the response functions. tal error), and the experimental data for the selected
materials, the relative error of interconversion was up
to 80%.
(ii) In all cases, the most successful was the method of
Schwarzl2 described with the Eq. (12). The relative error
was in all cases within 2%. The other two methods,
Christansen and NinomiyaFerry, yielded relative errors
in the order of 10 and 5%, respectively.
(iii) The location of the response function first derivative
maxima coincides with the location of the maximum
error of interconversion along the logarithmic time or
frequency scale.
(iv) The magnitude of the error of interconversion is
related to the magnitude of the response function first
derivative. When the first derivative = |d(log G ())/
Fig. 14. Maximum relative error as function of the maximum first derivative
of the synthetic response functions. d(log())| < 0.5 all approximate methods except the sim-
ple method furnish interconversions within the accept-
able error.
the maximum relative error, max (see Fig. 6), as shown in
Fig. 14.
From Fig. 14, we may conclude that all analyzed approx-
References
imate methods, except the Simple method, give equally
acceptable interconversions when the absolute value of the [1] TA Orchestrator Help, TA Instruments (help with the software for
response functions first derivative, , is smaller than 0.5. This the measuring apparatuses TA instruments), Copyright 2003, TA
may be used as a criterion in using approximate methods for Instruments, Waters LLC, 2003.
the interconversion of true experimental data. [2] J. Honerkamp, Ill-posed problems in rheology, Rheol. Acta 28 (1989)
363371.
This observation is confirmed in the analysis of the
[3] T.L. Cost, E.B. Becker, A multidata method of approximate laplace
four experimental response functions for materials: PVAc, transform inversion, Int. J. Numer. Methods Eng. 2 (1970) 207
NR, EPDM and PA66. Fig. 15 shows the maximum rel- 219.
[4] M. Baumgaertl, H.H. Winter, Determination of discrete relaxation
and retardation spectra from dynamic moduli, Rheol. Acta 28 (1989)
511519.
[5] H.H. Winter, Analysis of dynamic mechanical data: inversion into
a relaxation time spectrum and consistency check, J. Non-Newton.
Fluid Mech. 68 (1997) 225239.
[6] I. Emri, N.W. Tschoegl, Generating line spectra from experimental
responses. Part 1. Relaxation modulus and creep compliance, Rheol.
Acta 32 (1993) 311321.
[7] N.W. Tschoegl, I. Emri, Generating line spectra from experimental
responses. Part 2. Storage and loss functions, Rheol. Acta 32 (1993)
322327.
[8] N.W. Tschoegl, I. Emri, Generating line spectra from experimental
responses. Part 3. Interconversion between relaxation and retardation
behavior, Int. J. Polym. Mater. 18 (1992) 117127.
[9] I. Emri, N.W. Tschoegl, Generating line spectra from experimental
Fig. 15. Maximum relative error as function of the maximum first derivative responses. Part 4. Application to experimental-data, Rheol. Acta 33
of the experimental response functions. (1994) 6070.
84 I. Emri et al. / J. Non-Newtonian Fluid Mech. 129 (2005) 7584

[10] I. Emri, N.W. Tschoegl, Generating line spectra from experimental [21] F.R. Schwarzl, The numerical calculation of storage and loss com-
responses. Part 5. Time-dependent viscosity, Rheol. Acta 36 (1997) pliance from creep for linear viscoelastic materials, Rheol. Acta 8
303306. (1969) 6.
[11] A.Y. Malkin, V.V. Kuznetsov, Linearization as a method for deter- [22] F.R. Schwarzl, On the interconversion between viscoelastic material
mining parameters of relaxation spectra, Rheol. Acta 39 (2000) functions, Pure Appl. Chem. 23 (1970) 219234.
379383. [23] K. Yagii, E. Maekawa, Nippon Gomu Kyokaish 40 (1967) 46.
[12] R.M. Christensen, Theory of Elasticity, Academic Press, New York, [24] F.R. Schwarzl, Numerical-calculation of stress relaxation modulus
1982. from dynamic data for linear viscoelastic materials, Rheol. Acta 14
[13] R.A. Schapery, Aproximate methods of transform inversion for vis- (1975) 581.
coelastic stress analysis, Proceedings Fourth US National Congress [25] D.J. Plazek, N. Ragupathi, S.J. Orborn, Determination of dynamic
of Applied Mechanics, vol. 2, 1962, pp. 10751085. storage and loss compliances from creep data, J. Rheol. 23 (1979)
[14] S.W. Park, R.A. Schapery, Methods of interconversion between linear 477.
viscoelastic material functions. Part I. A numerical method based on [26] T. Prodan, The effect of hydrostatic pressure and temperature on the
Prony series, Int. J. Solids Struct. 36 (1999) 16531675. shear modulus of time-dependent materials, Ph.D. Dissertation, Fac-
[15] R.A. Schapery, S.W. Park, Methods of interconversion between linear ulty of Mechanical Engineering, University of Ljubljana, Ljubljana,
viscoelastic material functions. Part II. An approximate analytical Slovenia, 2003.
method, Int. J. Solids Struct. 36 (11) (1999) 16771699. [27] A. Kralj, T. Prodan, I. Emri, An apparatus for measuring the effect
[16] W. Retting, M.H. Laun, Kunststoff-Physik, Hanser, Munchen, 1991. of pressure on the time-dependent properties of polymers, J. Rheol.
[17] E. Riande, H. Markovitz, Approximate relations among compliance 45 (4) (2001) 929943.
functions of linear viscoelasticity for amorphous polymers, J. Polym. [28] I. Emri, T. Prodan, A measuring system for characterization of poly-
Sci. Polym. Phys. Ed. 13 (1975) 947. mers at elevated pressures and different temperatures, Exp. Mech.,
[18] J. Koppelmann von, Uber die Bestimmung des dynamischen Elas- in press.
tizitatmodulus und des dynamischen Shubmodulus im Frequenzbere- [29] N.W. Tschoegl, The Phenomenological Theory of Linear Viscoelastic
ich von 105 bis 101 Hz, Rheol. Acta 1 (1958) 20. Behavior, Springer-Verlag, Berlin, Germany, 1989.
[19] K. Ninomija, J.D. Ferry, Some approximate equations useful in the [30] J.D. Ferry, Viscoelastic Properties of Polymers, third ed., Willey,
phenomenological treatment of linear viscoelastic data, J. Colloid New York, 1981.
Sci. 14 (1959) 36. [31] F.R. Schwarzl, Werkstoffkunde der Kunststoffe II, Scriptum zu einer
[20] F.R. Schwarzl, L.C.E. Struik, Analysis of relaxation measurements, Vorlesung, University of Erlangen-Nuernberg, Technical University,
Adv. Mol. Relax. Process. 1 (1967) 210. Institute for Material Research, Erlangen, 1982.

Potrebbero piacerti anche