Beruflich Dokumente
Kultur Dokumente
www.elsevier.nl/locate/polytest
Property Modelling
Modelling the behaviour of plastics for design under impact
G. Dean, B. Read
Centre for Materials Measurement and Technology, National Physical Laboratory, Teddington, Middlesex TW11 0LW, UK
Abstract
Computer methods based on finite element analysis are able to predict the performance of plastics under impact
loading. Although the accuracy of results depends on the model used to describe the deformation behaviour of the
polymer, whichever model is used, the analysis requires stress/strain data over a wide range of strain rate. These data
are most conveniently measured in tension, but procedures are currently not available for determining results at high
strain rates. ISO standards for tensile property measurement are applicable for strain rates up to around 0.1 s⫺1. To
simulate behaviour under impact, data are required at rates that are 3 or 4 orders of magnitude higher than this. For
accurate data acquisition at these higher rates, attention needs to be paid to apparatus design in order to minimise
contributions from transient forces arising from resonances and the propagation of shock waves in the apparatus. In
addition, procedures and extensometers are not routinely available for determining strains at the higher rates of defor-
mation. This paper illustrates the acquisition of data over a wide range of strain rates through a combination of measure-
ments at low and moderate strain rates with extrapolation of these data to higher rates. In order to maximise accuracy
at moderate strain rates, suitable designs need to be selected for the transducers, the test specimen geometry and the
test assembly. Extrapolation is achieved by the use of mathematical functions to model the stress/strain curves and
their rate dependence. Reference is also made to the development of a new materials model for describing the defor-
mation behaviour of toughened plastics at large strains. 2001 NPL. Published by Elsevier Science Ltd. All rights
reserved.
1. Computer aided design competent design rather than the moulding and testing
of prototypes. Computer methods are available based on
Most modern grades of plastics are very tough finite element analysis that enable predictions to be made
materials that are capable of sustaining large strain levels of performance under realistic loading situations.
prior to failure, making them an attractive option for There are two factors that limit the wider use of these
many applications. Such applications include housings computer methods. Firstly, a lack of recognised and rou-
for domestic and electronic appliances, which may have tine test methods for measuring some of the data required
to sustain accidental impact without showing signs of by the analysis, and secondly, limited knowledge of the
damage, and interior components of motor vehicles, validity for plastics of the materials models employed by
which may be impacted by occupants in a crash and must the computer analysis, and thus uncertainty in the accu-
minimise damage to the human body part involved. In racy of derived results.
these applications and others, it is important that the per- This paper discusses the experimental issues associa-
formance and the safe operating limits of the component ted with data measurement, and a procedure for
under impact are known. Substantial savings can be determining properties by modelling and extrapolation
made in both time and cost if this can be done through is proposed. The limitations of elastic–plastic materials
0142-9418/01/$ - see front matter 2001 NPL. Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 1 4 2 - 9 4 1 8 ( 0 1 ) 0 0 0 0 3 - 4
678 G. Dean, B. Read / Polymer Testing 20 (2001) 677–683
models for describing non-linear deformation in plastics A simpler approach to determining strain in such tests is
are also discussed and the basis for a new model is to derive values from cross-head movement or the
explained. change in grip separation. An apparent or average strain
is obtained from the ratio of the change in grip separation
to the length of the specimen between the grips prior to
2. Determination of data required for impact load application. [Note, if the measurement of cross-head
analysis movement is used then a correction will, in general, be
necessary arising from the compliance of the test
2.1. Materials models and data assembly.] Since the test specimen is usually waisted in
the gauge region, this average strain will differ from
Non-linear stress/strain behaviour of plastics is gener- actual strain in the gauge region of the specimen, where
ally associated with an enhanced stress–relaxation, or the corresponding stress is calculated. It is possible to
creep arising from an increase in molecular mobility estimate the actual strain values from the average strain
induced by elevated stresses. Much of the strain is recov- using a correlation of these quantities determined from
erable at levels below the yield strain. Satisfactory mod- lower speed tests, where actual strain can be measured
els of non-linear viscoelasticity have yet to be developed, with extensometers. The correlation obtained will be spe-
and most FE packages consider material non-linearity in cific to a particular test machine, the specimen size and
terms of elastic–plastic models that were developed for geometry and the strain level in the specimen when this
metals and adapted for use with plastics and other is above the limit for linear behaviour. There will also
materials. be a small dependence upon material properties and
With elastic–plastic models, calculations of stress and hence test speed.
strain distributions at low strains (ⱕ0.01) are based on There remains some concern over the use of this
the theory of linear elasticity using, for isotropic approach for determining strain in materials where the
materials, values of the Young’s modulus E and elastic strain is not uniform in the gauge section of the specimen
Poisson’s ratio ne. The onset of non-linearity in a stress– beyond the peak in the stress/strain curve. Materials that
strain curve is then ascribed to plastic deformation and show a visible neck region are an extreme example of
occurs at a stress level regarded as the first yield stress. this.
The subsequent increase in yield stress with strain is In this way, it is generally possible to obtain valid
associated with the effects of strain hardening. The hard- tensile data up to test speeds of around 1 m/s which will
ening behaviour is required by the computer analysis and correspond to strain rates of typically 10 s⫺1. At test
is most commonly defined by a plot of tensile yield stress speeds above this, additional problems arise. Owing to
against the tensile plastic strain. This is derived from a the finite time for the hydraulic ram to accelerate to the
measured tensile stress/strain curve by subtracting the set speed, the test will not be performed under constant
elastic component of strain at each stress level. Since strain rate. In general, this is not a major problem since
plastics are viscoelastic materials, the hardening curves in the initial part of the test where the behaviour is linear,
are dependent upon the strain rate associated with the the strain rate is less than in the latter part where plastic
applied load. deformation is dominant. Hence, elastic and plastic
behaviour should be analysed using different values of
2.2. Experimental methods strain rate. A more serious problem arises with the rapid
accelerations associated with load application that give
Much of the material property data required by a finite rise to transient forces. Such forces are recorded by the
element analysis of the performance of a component force transducer and superimposed onto the measured
under impact loading can be derived from a series of force on the specimen. These transient forces arise from
tensile stress/strain curves measured over a range of excitation of a resonance of the force transducer and
strain rates. For many applications, the strain rate should from multiple reflections of the acoustic pulse generated
reach or exceed 100 s⫺1. Standard test methods for meas- at one end of the specimen by the sudden load appli-
uring tensile behaviour are confined to strain rates below cation. The influence of these forces on the accuracy of
0.1 s⫺1. Using servo-hydraulic test machines, smaller measurements can be minimised by using components
specimens and light weight extensometers, reliable of low mass in the test assembly and a high stiffness
results can generally be obtained to strain rates around force transducer and by filtering or smoothing the
1 s⫺1. Above this value experimental difficulties, force signal.
especially with strain measurement, can significantly
reduce the accuracy of results. 2.3. Data acquisition by materials modelling
The development of high-speed extensometers that
operate at higher strain rates has been reported, although The measurement of data characterising rate-depen-
these devices are not used routinely for data acquisition. dent behaviour suitable for impact analyses is laborious
G. Dean, B. Read / Polymer Testing 20 (2001) 677–683 679
Table 1
Values for the parameters in Eq. (2) giving the fits shown to the data in Fig. 2
Table 2
Values for the parameters in Eqs. (7) and (5) used to model strain and strain-rate dependent yield in the propylene copolymer
4. Conclusions