Вы находитесь на странице: 1из 9

Mat Res Innovat (2003) 7:1018 DOI 10.

1007/s10019-002-0217-z

O R I G I N A L A RT I C L E

Yiping Duan Anil Saigal Robert Greif Michael A. Zimmerman

Modeling multiaxial impact behavior of a glassy polymer

Received: 20 August 2002 / Revised: 4 November 2002 / Accepted: 5 November 2002 / Published online: 31 January 2003 Springer-Verlag 2003

Abstract In many applications of polymers, impact performance is a primary concern. Impact tests experimentally performed on molding prototypes yield useful data for a particular structural and impact loading case. But, it is generally not practical in terms of time and cost to experimentally characterize the effects of a wide range of design variables. A successful numerical model for impact deformation and failure of polymers can provide convenient and useful guidelines on product design and therefore decrease the disadvantages that arise from purely experimental trial and error. Since the specimen geometry and loading mode for multiaxial impact test provides a close correlation with practical impact conditions and can conveniently provide experimental data, the first step of validating a numerical model is to simulate this type of test. In this paper, we create a finite element analysis model using ABAQUS/Explicit to simulate the deformation and failure of a glassy ABS (acrylonitrile-butadiene-styrene) polymer in the standard ASTM D3763 multiaxial impact test. Since polymers often exhibit different behavior in uniaxial tensile and compression tests, the uniaxial compression or tensile tests are generally not representative of the three-dimensional deformation behavior under impact loading. A hydrostatic pressure effect (controlled by the parameter ) is used to generalize a previously developed constitutive model (DSGZ model) so that it can describe the entire range of deformation behavior of polymers under any monotonic loading modes. The generalized DSGZ model and a failure criterion are incorporated in the FEA model as a user material subroutine. The phenomenon of thermomechanical coupling during plastic deformation is considered in the analysis. Impact load vs. displacement
Y. Duan A. Saigal () R. Greif Department of Mechanical Engineering, Tufts University, Medford, MA 02155, USA, e-mail: anil.saigal@tufts.edu Tel.: +1-617- 6272549, Fax: +1-617- 6273058 M. A. Zimmerman Lucent Technologies, North Andover, MA 01845, USA,

and impact energy vs. displacement curves from FEA simulation are compared with experimental data. The results show good agreement. Finally, equivalent stress, strain, strain rate and temperature distributions in the polymer disk are presented. Keywords Polymer Impact Failure Constitutive model Stressstrain curves Thermomechanical coupling Finite element analysis

Introduction
Due to good thermal and electrical insulation properties, low density, high resistance to chemicals and ease of manufacturing, polymers have increasingly been applied in applications where impact performance is a primary concern [1, 2, 3, 4]. There are several types of standard tests to evaluate the impact strength of polymers, in which the most commonly used are Charpy and Izod [5]. Although the two types of tests provide some information on the relative impact resistance of materials, the particular specimen geometry requirements of both tests make the test results difficult to relate to practical polymer design models. However, the specimen geometry and loading mode for a multiaxial impact test provide a close correlation with practical impact conditions [6]. Impact tests experimentally performed on molding prototypes yield useful data for a particular structural and impact loading case. But, it is generally not practical in terms of time and cost to experimentally characterize the effects of a wide range of design variables. A successful numerical model for impact deformation and failure of polymers can provide convenient and useful guidelines on product design and therefore decrease the disadvantages that arise from purely experimental trial and error. Since the multiaxial impact test has a close correlation with practical impact conditions and can conveniently provide experimental data, the first step of validating a numerical model is to simulate this type of test.

11

Some work has been done with respect to modeling multiaxial impact tests and developing numerical models for impact deformation of polymers during the last two decades [7, 8, 9]. Nimmer [7] created a FEA (finite element analysis) model to simulate a fixed velocity puncture test of Bisphenol A-polycarbonate disk, where a constant bilinear stressstrain curve was used to approximate the material behavior. Good agreement was achieved between model prediction and experimental load vs. deflection data for deflections up to four times the thickness of the test disk. By using the GSell-Jonas constitutive model, Billon and Haudin [8] numerically explored the effects of specimen thickness, friction between striker and specimen disk, and quality of the clamping device during a multiaxial impact test on polypropylene, a semicrystalline polymer. They indicated that the material response during impact is difficult to analyze since the deformation is not homogeneous and not isothermal. After validating the application of the GSell-Jonas constitutive model to a semicrystalline polymer polyamide 12 for a large range of strain rates during uniaxial tensile test, Schang et al. [9] developed a FEA model to simulate the multiaxial impact test. Model predictions were compared with experimental data for the history of impact load. Even though good agreement was achieved at the beginning of the load vs. time curves, the maximum impact load was largely overestimated. They pointed out that this result occurred because either the strain rate dependence in the constitutive model was not correctly taken into account, or the tensile characterizations were not representative of the impact situation. All of the aforementioned simulations are for impact deformation of polymers. At least two material models: a stressstrain constitutive model and a failure model are required in numerical simulation of destructive impact events. Both the material models play a significant role on the reliability and accuracy of simulation results. Using a failure criterion and a new stressstrain constitutive model (generalized DSGZ model) which describe the stressstrain constitutive relationship of polymers under any monotonic loading modes, a FEA model is formulated in this paper using ABAQUS/Explicit to simulate the deformation and failure of a glassy ABS (acrylonitrile-butadiene-styrene) polymer, which has a glass transition temperature of around 378 K, in the standard ASTM D3763 multiaxial impact test.

Relation to previous work


Two major science and engineering databases were searched for previous work that is relevant to the topic of current paper. The first database is ISI (Institute for Scientific Information) SCI-EXPANDED (Science Citation Index Expanded1945-present), and the second database is Ei Compendex (1970-present). The following keyword sequences were used in search of the two databases: multiaxial impact test or multiaxial impact tests,

impact and polymer and FEA, impact and polymers and FEA, constitutive model, stressstrain curves, deformation mode and polymers, and thermomechanical coupling and polymers. Also, books and Ph.D. thesis were searched that talked about the mechanical behavior of polymers, constitutive modeling, and failure of plastics or polymers. Based on the literature search, we found that some work has been done with respect to modeling multiaxial impact tests and developing numerical models for impact deformation of polymers during the last two decades, and the research work that was most relevant were cited in this paper. However, the topic of developing numerical models for impact deformation of polymers and modeling multiaxial impact test of polymeric materials is far from sufficiently explored. It is well known that at least a stressstrain constitutive model and a failure criterion are required in numerical modeling of destructive impact events. Both these material models play a significant role on the reliability and accuracy of simulation results. In the open literature that we searched, some authors used a constant bilinear stressstrain curve to approximate polymer behavior, and some authors used the GSellJonas constitutive model, which can not describe strainsoftening behavior of polymers, characterized by tensile test to approximate polymer behavior. In general, these simulation results are accurate at small deflections, but deviate substantially at relatively large deflections. Using concepts from the Johnson-Cook model, GSell-Jonas model, Brooks model and Matsuoka model, we developed a phenomenological constitutive model (called DSGZ model) to uniformly describe the entire range of deformation behavior of both glassy and semicrystalline polymers under monotonic compressive loading. However, unlike metals, polymers often exhibit different behavior in uniaxial tensile and compression tests. The uniaxial compression or tensile tests are generally not representative of the three-dimensional (multiaxial) deformation behavior under impact loading. The deformation mode should be considered. In this paper, we use a hydrostatic pressure effect (controlled by parameter ) to generalize the DSGZ model so that it can describe the entire range of deformation behavior of polymers under any monotonic loading modes. We consider this work an important innovation in polymer modeling. The generalized DSGZ model was applied in modeling the multiaxial impact behavior of a glassy ABS polymer. Impact load vs. displacement and the impact energy vs. displacement curves from simulation fit well with experimental data up to failure.

Constitutive modeling
Over the past four decades, much effort has been devoted to modeling the stressstrain constitutive relationships for polymers [10, 11, 12, 13]. Using concepts from the Johnson-Cook model, GSell-Jonas model, Brooks model and Matsuoka model, the authors [14] developed

12

a phenomenological constitutive model (called DSGZ model) to uniformly describe the entire range of deformation behavior of both glassy and semicrystalline polymers under monotonic compressive loading. The DSGZ model explicitly gives the compressive true stress c, dependence on true strain , true strain rate and temperature T and is given by:

and

is defined to be the dimensionless form of . The equivalent stress is defined by Eq. 4, the hydrostatic stress p is defined as and the equivalent strain is defined as (8) where [e] is the deviatoric part of a true strain tensor []. The equivalent strain rate is defined as the derivative of the equivalent strain with respect to time t,

(1) where, (2) (3) and is defined to be the dimensionless form of . The eight material coefficients in this model are Kc (Pa sm), C1, C2, C3 (sm), C4, a (K), m, and . It has been well known that the deformation behavior of polymers is not only sensitive to strain, strain rate and temperature but also relate to the deformation mode [15, 16, 17, 18]. Unlike metals, polymers often exhibit different behavior in uniaxial tensile and compression tests. Suppose [s] is the deviatoric part of a true stress tensor []. For materials that exhibit the same behavior in uniaxial tension and uniaxial compression, the hydrostatic pressure part of the stress tensor is generally assumed to have no effect on the mechanical behavior of the material. The equivalent stress for the stress tensor [] is given as, (4) However for materials that exhibit different behavior in uniaxial tension and uniaxial compression, the effect of hydrostatic pressure has to be included. Using this hydrostatic pressure effect, a generalized DSGZ model is proposed to describe the stressstrain constitutive relationship of polymers under any monotonic loading modes, and is given by

(9) and , called the hydrostatic sensitivity parameter, is a material coefficient accounting for the effect of loading mode. The other eight material coefficients in the generalized DSGZ model are K (Pa sm), C1, C2, C3 (sm), C4, a (K), m and . In the generalized DSGZ constitutive model for a uniaxial compression test, the compressive stress c can be written in the form

(10) and for a uniaxial tensile test, the tensile stress t can be written as

(11) Note that Eq. 1 for a uniaxial compression test can be obtained from Eq. 10 by substituting the term by Kc. The value of can be calculated by combining Eqs. 10 and 11. For a given strain , strain rate and temperature T, has the form (12) From Eq. 12, it can be seen that is a function of strain , strain rate and temperature T over the entire range of deformation. Because of the failure of polymers during uniaxial tensile tests at low strains in comparison to the large strains obtained in compression tests, cannot be calculated over a large strain range. Therefore is calculated at the yield stress and assumed to be constant in the form (13) (6) (7) where cy is the yield stress in uniaxial compression test and ty is the yield stress in uniaxial tensile test. The val-

(5) where,

13

ues of the two yielding stresses can be obtained experimentally. The other eight material coefficients in the generalized DSGZ model can be deduced from uniaxial compression stressstrain curves following the procedures given in reference [14].

rate during plastic deformation of polymers. Since there is a lack of available experimental data for vs. strain and strain rate for the glassy polymer, a constant value of = 0.5 is used for all the simulations except otherwise stated. Macdougall and Harding [21] used a similar approximation in their numerical modeling of the high strain rate torsion tests on Ti6Al4V bars.

Thermomechanical coupling during plastic deformation Failure criterion


Since the mechanical properties of polymers are sensitive to temperature, an accurate estimate of the temperature rise during plastic deformation is important. Arruda et al. [4] did a series of uniaxial compression tests in which the specimen surface temperatures were monitored using an infrared detector to investigate the relationship between strain rate and temperature rise for the glassy polymer polymethyl-methacrylate. It was found that the specimen was nearly isothermal up to a true strain of 0.8 at a strain rate of 0.001/s, but significant temperature rise (around 30 C) were observed up to the same true strain at strain rates of 0.01/s and 0.1/s. The rise of temperature had a dramatic effect on stressstrain curves. Rittel [19] embedded a small thermocouple in polycarbonate specimen disks to record the transient temperature during impact tests with strain rates ranging from 5000/s to 8000/s. Within a time order of 104 s, a true strain of 0.45 was obtained and the recorded temperature increased by nearly 25 C. The temperature rose significantly in the softening region of the corresponding stressstrain curve. Using a fast response infrared radiometer to monitor the surface temperature of epoxy specimens in a Split Hopkinson Pressure Bar impact test, Trojanowski et al. [20] observed that there was a temperature increase of approximately 50 C. These experimental results indicate that for polymers, the temperature rise is significant during high strain rate large plastic deformation. For strain rates of 102 ~103/s, it is reasonable to assume that the deformation process is essentially adiabatic. The governing equation for the increase of temperature T at each increment of plastic strain is (14) where is material mass density, c is heat capacity, is the fraction of dissipated plastic energy which converts into thermal energy, is the equivalent stress at the beginning of an increment, is the equivalent stress at the end of the increment and is the increment of equivalent plastic strain. At each increment of plastic deformation, the local temperature of the plastic deformation zone will increase by an amount governed by Eq. 14. The increase in temperature decreases the equivalent stress through the generalized DSGZ model given by Eqs. 5, 6, and 7. This gives a framework to account for the thermomechanical coupling during high strain rate plastic deformation. The issue is how to decide the value of in Eq. 14 for each increment of strain. Rittel [19] found that is dependent on strain and strain Polymer failure in the multiaxial impact test is exhibited by the sudden significant reduction of its load carrying capability. Some significant work has been done with respect to understanding the impact strength and the dynamic crack propagation process of polymers [22, 23]. Through considering failure resulting from crack propagation in Charpy impact test, Brostow [22] proves that the impact transition temperature, which is highly related to impact strength, of a polymer is not only related to the polymeric material property but also related to the dimensions of artificial notches and other imperfections that lead to stress concentration. In the same reference, he gives formulas that relate the impact transition temperature to stress concentration factor. Very recently, Brostow et al. [24] numerically simulated the crack initiation and propagation in polymer liquid crystals during tensile deformations through a molecular dynamics simulation procedure. A material failure criterion has to be combined together with the stressstrain constitutive model in order to simulate the impact failure. To the authors knowledge, there are few reports found in the open literature for numerical simulation of polymer failure in multiaxial impact tests. Strictly speaking, molding polymer specimens always have certain defects, such as scratches. However, unlike the Charpy impact test specimens that include artificial notches, the multiaxial impact test specimens do not have any artificial notches and are treated to include as few defects as possible. Therefore, the effect of stress concentration is not a significant factor in the multiaxial impact tests. The current simulations involve a macroscopic approach to the failure of polymers. There are a variety of proposed macroscopic material failure criteria such as maximum tensile stress, maximum shear strain, and maximum strain energy density. Since the polymer investigated in this paper has significant ductility, the maximum plastic strain failure criterion is used. A failure indicator is created and defined as (15) where is a prescribed maximum equivalent plastic strain, and is the increment of equivalent plastic strain. When the sum of the equivalent plastic strain increment at a material point is equal to or greater than the prescribed value of , i.e. when 1, the material point fails and is permanently removed from future calculations.

14

Finite element analysis model


In the ASTM D3763 multiaxial impact test, a cylindrical striker with hemispherical end is dropped from a given height onto the center of a clamped polymer disk. Figure 1 shows the geometrical parameters used. The clamp assembly consists of two circular parallel plates with a 76 mm diameter hole in the center. Sufficient pressure is applied to prevent slippage of the polymer disk from the clamp assembly during impact. The striker, consisting of a 12.7 mm diameter steel rod with a hemispherical end of the same dimension, is positioned perpendicular to and centered on the clamped disk. During impact, the striker moves down and the polymer disk is totally penetrated. History of impact load, impact energy and displacement of the striker are recorded. Figure 2 shows a FEA model for the ASTM D3763 multiaxial impact test. Nine hundred eight-noded linear brick, reduced integration elements with a total of 1519 nodes are used to mesh the polymer disk. An analytical rigid surface is used to model the geometry of the striker. The rigid surface is associated with a rigid body reference node that defines the mass and the motion of the striker. The boundary conditions are set as fixed support around the outer edge of the disk and the six degrees of freedom of those nodes located on the circular edge are set to be zero. The striker can only move along the vertical axis, all the other five degrees of freedom of the rigid body reference node are set to be zero. The friction coefficient between the striker and the polymer disk is set to be a constant value of 0.3. An algorithm of elastic prediction plastic correction is applied to update the stress tensor of each material point. At the end of each increment of strain, the stress tensor []new is calculated. The corresponding equivalent stress is obtained from Eq. 4. Equation 14 is then applied to calculate the increase of the local temperature of polymer. The failure indicator is updated through Eq. 15 and when 1 the material point fails and is permanently removed from future calculations. Based on this framework of calculation, a user material subroutine is developed and applied in the FEA model to implement the generalized DSGZ constitutive model, the thermomechanical coupling model and the maximum plastic strain failure criterion.

Fig. 1 Geometrical configuration and parameters used in ASTM D3763 multiaxial impact test

Fig. 2 FEA model for ASTM D3763 multiaxial impact test

Results and discussion


Using available stressstrain curves of uniaxial tension and compression tests, we calibrated the nine material coefficients in the generalized DSGZ constitutive model for the glassy ABS polymer [18]. Table 1 shows the calculated material coefficients. The compressive stress c in a uniaxial compression test can be predicted by Eq. 10 for various strain, strain rate and temperature. Figure 3 shows comparison of the DSGZ constitutive model prediction with available uniaxial compression test data for
Fig. 3 Comparison of DSGZ model prediction with uniaxial compression test data (T=296 K, )

the polymer. It can be seen that the DSGZ constitutive model accurately predicts the stressstrain behavior of the polymer over a wide range of strains. As previously stated, a constant value of 0.5 is used for the in Eq. 14 in the simulations of the multiaxial

Fig. 4 Comparison of the FEA model predictions with experimental impact load vs. displacement data (impact velocity: 3 m/s)

Fig. 5 Comparison of the FEA model predictions with experimental impact energy vs. displacement data (impact velocity: 3 m/s)

Fig. 6 Predicted equivalent stress (Pa) distributions on the bottom surface of the polymer disk

16

Fig. 7 Equivalent plastic strain contour maps on the bottom surface of the polymer disk Table 1 Material coefficients for the glassy ABS polymer C1 0.8 C2 0.83 m 0.07 a (K) 900 K (MPa sm) 0.95 C3 (sm) 0.0028 C4 7 100 0.25

impact test. In order to estimate the importance of thermomechanical coupling, simulations were performed in which the effect of the thermomechanical coupling is ignored by setting = 0. To evaluate the effect of hydrostatic pressure, simulations were done in which the constitutive model was calibrated using uniaxial compression data and the parameter set to zero. Figure 4 shows comparison of the FEA model predictions with experimental impact load vs. displacement data, while Fig. 5 shows comparison of the FEA model predictions with experimental impact energy vs. displacement data. From Figs. 4 and 5, we can see that the FEA model prediction, with = 0.5 and appropriately taken from Table 1, agree well with the impact load vs. displacement and impact energy vs. displacement experimental data of the glassy polymer up to the maximum impact loading (failure). It is also observed that not accounting for and/or in the simulation model tends to overestimate the im-

pact load and impact energy especially at large displacements. From Figs. 4 and 5, it is found that the effect of thermomechanical coupling () is small. Even though a maximum temperature rise of 18 C was obtained in the simulation, the effect on the predicted impact load and impact energy is less than 5%. The simulation results show that for materials that exhibit different behavior in uniaxial tension and uniaxial compression, the effect of hydrostatic stress () has to be included. Neglecting the hydrostatic pressure effect will lead to a much greater error for the simulations of the multiaxial impact test. Uniaxial compression or tensile tests taken individually are not representative of the three-dimensional deformation behavior of polymers under impact loading. In the application of the generalized DSGZ constitutive model, it is important to experimentally calibrate the parameter .

17

Fig. 8 Predicted temperature (K) distributions on the bottom surface of the polymer disk

Figures 6, 7, 8 and 9 show the simulated equivalent stress distributions, the simulated equivalent plastic strain contour maps, the simulated temperature distributions, and the contour maps of the strain rates, respectively, on the bottom surface of the polymer disk. The simulation results show that the highest strain rates are of the order of 102~103/s during most of the deformation. Even though the material coefficients in Table 1 are calculated from low strain rates (104/s) test data, the calibrated constitutive model can be used to extrapolate to predict the deformation behavior of the glassy polymer at high strain rates. This has been shown to be true for other polymers such as polymethyl-methacrylate, polycarbonate and polyamide 12 [14]. From the simulation results, it is found that most regions of the disk undergo elastic deformation. The plastic deformation and temperature rise is localized in the impact region with the maximum values in the center of the disk. However, the maximum equivalent stress is located in a circular zone around the striker.

Concluding remarks
A FEA model is created using ABAQUS/Explicit to simulate the standard ASTM D3763 multiaxial impact test. The DSGZ constitutive model is generalized to describe the stressstrain constitutive relationship of polymers under any monotonic loading modes. The thermomechanical coupling during high strain rate plastic deformation and the failure criteria for polymers are discussed. An ABAQUS/Explicit user material subroutine is developed and applied to implement the generalized DSGZ constitutive model, the thermomechanical coupling model and the maximum plastic strain failure criterion in the FEA simulation. Multiaxial impact test on a glassy ABS polymer was simulated. The predicted impact load vs. displacement curve and impact energy vs. displacement curve were compared with experimental data. They agree well. The results indicate that the generalized DSGZ constitutive model accurately predicts the stressstrain behavior of the polymer over a wide range of strains and it correctly extrapolates over a large range of strain rates. For polymers, the uniaxial compression or tensile tests are generally not representative of the three-dimensional deformation behavior under impact loading. The

18

Fig. 9 Contour maps of the strain rates (/s) on the bottom surface of the polymer disk

deformation mode should be considered. In the application of the generalized DSGZ constitutive model, it is important to experimentally calibrate the parameter so that hydrostatic stress effects can be included.
Acknowledgements The support of Lucent Technologies, located in North Andover, Massachusetts, USA, during this research is gratefully acknowledged.

References
1. Delatycki O (1971) Mechanical performance and design in polymers. Interscience Publishers, New York 2. Walley SM, Field JE, Pope PH, Safford NA (1991) J Phys III 1: 1889 3. Kalpakjian S (1995) Manufacturing engineering and technology, 3rd edn. Addison-Wesley, Reading, Massachusetts 4. Arruda EM, Boyce MC, Jayachandran R (1995) Mech Mater 19: 193 5. Perkins WG (1999) Polymer Eng & Sci 39: 2445 6. Tai CM, Li RKY, Ng CN (2000) Polymer Testing 19: 143 7. Nimmer RP (1983) Polymer Eng & Sci 23: 155

8. Billon N, Haudin JM (1992) NUMIFORM 92 9. Schang O, Billon N, Muracciole JM, Fernagut F (1996) Polymer Eng & Sci 36: 541 10. GSell C, Jonas JJ (1979) J Mater Sci 14: 583 11. Boyce MC (1986) PhD thesis, MIT 12. Amoedo J, Lee D (1992) Polymer Eng & Sci 32: 1055 13. Hasan OA (1994) PhD thesis, MIT 14. Duan Y, Saigal A, Greif R, Zimmerman MA (2001) Polymer Eng & Sci 41: 1322 15. Bowden PB, Jukes JA (1972) J Mater Sci 7: 52 16. Boyce MC, Arruda EM, Jayachandran R (1994) Polymer Eng & Sci 34: 716 17. Bauwens JC (1967) J Polymer Sci A2 5: 1145 18. http://www.geplastics.com/resins/designsolution/tools/index.html 19. Rittel D (1999) Mech Mater 31: 131 20. Trojanowski A, Ruiz C, Harding J (1997) J Phys IV 7: 447 21. Macdougall DAS, Harding J (1999) J Mech Phys Solids 47: 1157 22. Brostow W (1986) Impact strength: determination and prediction. In: Brostow W, Corneliussen RD (eds) Failure of plastics. Hanser, Munich Vienna New York, p 196 23. Rosakis AJ, Samudrala O, Coker D (2000) Mat Res Innovat 3: 236 24. Brostow W, Donahue III M, Karashin CE, Simoes R (2001) Mat Res Innovat 4: 75

Вам также может понравиться