Analysis and Simulation of the Free-stretch-blow Process of PET Nixon J., Yan S., Menary G.H.

Transcription

1 Key Engineering Materials Online: ISSN: , Vols , pp doi: / Trans Tech Publications, Switzerland Analysis and Simulation of the Free-stretch-blow Process of PET Nixon J., Yan S., Menary G.H. Queen s University of Belfast School of Mechanical and Aerospace Engineering Ashby Building, Stranmillis Road, Belfast, UK g.menary@qub.ac.uk Keywords: ISBM; PET; Digital image correlation; ABAQUS /Explicit FEA, viscoelastic Abstract This paper is concerned with understanding the behaviour of Polyethylene Terephthalate (PET) in the injection stretch blow moulding (ISBM) process where it is typically bi-axially stretched to form bottles for the packaging industry. Preforms which have been pre sprayed with a pattern and heated in an oil bath have been stretched and blown in free air using a lab scale ISBM machine whilst being monitored via high speed video. The images have subsequently been analysed using a digital image correlation system (VIC 3D). The results have been used to validate appropriate simulations of the free-blow process using ABAQUS /Explicit FEA with a suitable viscoelastic material model, along with experimental process data obtained using an instrumented stretch rod. 1.0 Introduction Injection stretch blow moulding (ISBM) is one of the most common methods of producing rigid, thin-walled bottles, mainly utilized polyethylene terephthalate (PET) as the appropriate material. PET preforms with a suitably designed geometry are manufactured using an injection moulding process, heated in an infrared oven above the polymer s glass transition temperature, before being placed in a mould where the bottle is formed using a combination of linear stretching from a stretch-rod activated inside the preform and pressure rise from airflow. With the formation of the bottle lasting a matter of milliseconds, the process conditions employed are critical and selecting the correct parameters dictates the performance of the final bottle. Key variable that constitute the input parameters are preform shape, preform temperature distribution, pressure and flow rate values, and the timing relationship between the pressure and stretch-rod activation. In turn, these parameters control the desired output characteristics such as material thickness distribution and the amount of orientation and crystallinity within the final bottle. The goal of multiple research groups, both academic and industrial, has been to examine and comprehend the deformation of PET during the ISBM process and then successfully model the procedure using a suitable numerical solver; thus leading to a method of accurately predicting the finished bottle properties and therefore allowing for optimization of the initial process parameters. 1.1 Free-blow Analysis Method With so many parameters contributing to the overall ISBM process, eliminating certain aspects can simplify the analysis procedure. Previous attempts to visually study the blowing process include utilising a transparent mould [1]. A single camera recorded the bubble propagation as the bottle filled the mould, studying the effect of pre-blow pressure timing. Another method is to remove the mould from the process altogether, therefore allowing the bottle to be blown freely. This then allows the observer to examine the bottle shape as it inflates which would not be possible with the mould attached. All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of Trans Tech Publications, (ID: , Pennsylvania State University, University Park, USA-09/05/16,01:28:38)

2 1730 The Current State-of-the-Art on Material Forming Tan et al. [2] used this free-blow method in conjunction with a video to record the inflation and instrumentation to record pressure values within the preform cavity. ABAQUS FEA solver along with a UMAT material model was used to create an appropriate simulation. When comparing the free-blow results of both bottle shape evolution and cavity pressure profile with the simulation results, it was clear that applying the pressure in the form of a direct load input on the preform inner surface was not appropriate. Regarding the cavity pressure as an output determined from an air mass flow rate input resulted in greater accuracy in pressure prediction and bottle shape evolution. Conclusions were made that the values of fluid flow during the blowing phase were required; a feat not easily achieved considering the time frame of the stretch-blow process. Nagarajappa et al. [3] also performed free-blow trials in order to gain a greater perspective into the ISBM process. Utilising an process parameter array of two input pressures, two flow restrictor values and two preform temperature values, the cavity pressure and stretch-rod force were recorded with instrumentation and the free-blow bottle formation was captured using 3D stereoscopic high speed cameras. The experimental data was then used to validate the equivalent simulation. Billion et al.[4] also studied the free-blow process using high speed video analysis, utilising a grid pattern to evaluate the strain and strain rates achieved. The effect of pre-blow pressure timing on the bubble kinematics during blowing was subsequently studied using preforms heated in an IR oven. 1.2 Digital Image Correlation Digital image correlation (DIC) is an analysis technique that tracks the movement of a specific speckle pattern applied to an object as a means of determining the displacements and strain values during loading. Chevalier et al. [5] used this speckle tracking techniques to capture the strain in biaxial stretched specimens of PET. The specimens were sprayed black and a white speckle pattern was then applied. The motion of the pattern during deformation were then be tracked using a CCD camera resulting in a precision in strain values of around With an accurate picture of the strain map, and knowing the stress throughout the deformation, an accurate experimental stress-strain profile at different temperatures could be derived. A similar experimental technique was used by Nagarjappa et al. [3] to track the 3D strain evolution of the free-blow bottle formation. A suitable pattern was applied to the preforms which were heated and then blown in a stand-alone industrial SBM machine. The strain data was then processed by analysing the images gathered from the high speed cameras using Vic3D software. The resulting strain mapping data was then used to provide appropriate stress-strain behaviour to develop a constitutive model for PET. The limited process parameters used for these trials leaves room to develop a more in-depth experimental trial array that can account for more values of input pressure, flow setting and preform temperature. This paper follows on from the VIC3D analysis performed by Yan et al [6] where free-blow analyses were performed at various preform temperatures, pressure/flow settings and pressure timings values. A specifically designed pattern was then applied to the preforms; thereafter the preforms were heated in a Grant GR-150 oil bath and then manually transported to a lab scale SBM machine. The hoop and axial strain values were then obtained for the free-blow process along with cavity pressure, cavity temperature, stretch-rod force and stretch-rod displacement which were gathered using an instrumented stretch-rod. 1.3 Constitutive Model for PET A critical component for an accurate and reliable simulation is the acquirement of a suitable material model that can capture the highly non-linear, viscoelastic behaviour of PET when it is stretched. Yan et al. [7] developed a constitutive model to represent the behaviour of PET during stretching; concentrating on the effects of polymer temperature, strain rates and mode of deformation. The model was implemented around the constitutive model developed by Buckley et

3 Key Engineering Materials Vols al. [8,9] that is widely used for thermosetting and thermoplastics near and above their glass transition temperature. The model consists of a parallel circuit with two parts; a bond stretch part representing the linear elastic deformation and the viscous flow, and a conformational part representing the non-linear elastic deformation of network stretch and the material viscosity. The experimental data used to construct the material model was attained from testing extruded PET specimens on a bi-axial testing machine by Queens University Belfast [10]. During material testing, operating temperature ranged from C, strain rate values 1/s to 32/s and modes of deformation included equal bi-axial, constant width and sequential. The material model was then applied to ABAQUS/Explicit using a VUMAT subroutine. 2.0 Experimental Trials The preform used for the experimental trials is detailed in Fig 1. This and was then blown on a semi-automatic lab scale SBM machine. 2.1 Preform Heating Fig. 1 Preform design The preforms were fully submerged in a 5 litre stainless steel oil bath along with a Grant GR-150 stirred thermostat that was filled with silicone oil and had a possible temperature range of C. The preforms were left submerged for a suitable time so that temperature equilibrium was obtained through the wall thickness and without any thermal crystallization present. The preform outside surface temperature profile was then recorded using a FLIR SC-640 thermal camera. This temperature profile is shown in Fig 2. Fig. 2 Experimental oil bath temperature profiles

4 1732 The Current State-of-the-Art on Material Forming From Fig. 2 it is evident that the temperature profile achieved from the oil bath is almost constant along the preform length. Although preforms heated using an IR oven achieve a temperature profile that can be highly variable along the length and through the thickness, a constant temperature profile is an attribute that is desirable when attempting to analyse the multiple parameter free-blow process. After heating in the oil bath, the preforms were then transported manually from the oil bath to the SBM machine with minimal transfer time in an effort to reduce the effect of cooling. This transportation time was on average 15s, and was considered when estimating the preform temperature profile when the stretch-blow process initiated. Cooling experiments were carried out by heating a preform at each oil bath temperature and allowing it to cool over time in the view of the FLIR SC-640. The rate of surface cooling was then recorded over the length of the preform to estimate the heat transfer coefficient. The heat transfer coefficient for the oil-covered preform surface to air was calculated to be 13W/m 2 K. Using a thermal conductivity of 0.24W/mK, the temperature profile over the entire preform, including through the thickness, could be determined. An average temperature drop of 8±2 C was calculated from the time taken to transport the preform from the oil bath to the SBM machine at an oil temperature of 100 C. The ±2 C error was applied to the preform temperature due to the fluctuations in the transportation time for each trial and the experimental error associated with the heating process and measurements. 2.2 Experimental Set-up The experiments take into account three pressure values (4, 6 and 8 bar), three flow valve settings (2, 4 and 6) and six oil bath temperature settings (95, 100, 105, 110, 115 and 120 C). The flow valve is a manually controlled valve located upstream of the preform cavity that controls the amount of air flow during the blowing process. The flow valve setting (flow index) on the ISBM machine used for this project ranges from 1 to 6. The effect of pressure timing with respect to stretch-rod position was also considered. The timing relationship between pre-blow pressure and stretch-rod activation can be manipulated by the ISBM machine operator can dictate the material distribution of the final bottle. The three timing settings used were -50ms advance where the pressure was activated 50ms before the stretch rod touched the hemispherical end of the preform, 0ms simultaneous where the pressure was activated at the same time as the stretch rod touched the hemispherical end of the preform and 50ms delay where the pressure was activated 50ms after the stretch rod touched the hemispherical end of the preform. This led to a full factorial DoE set-up of 162 individual experiments. Along with the high speed image captures to analyse the strain, the process outputs of cavity pressure, cavity air temperature, stretch-rod displacement and stretch-rod force were recorded using an instrumented stretch-rod [4]. Accounting for the large array of data gathered from the DoE, this paper will focus on those trials that emphasise the effect of different pressure rates; low rate and high rate. The conditions used are shown in Table 1. Table 1 Selected trial process parameters pressure rate Low Rate (DOE30) High Rate (DOE52) Line Pressure [bar] 6 10 Flow Index [1-6] 2 6 Temperature [ C] Timing [ms]

5 Key Engineering Materials Vols Free-blow Simulation 3.1 Simulation Set-up The free-blow simulation has been developed using ABAQUS/Explicit FEA software version The multiple inputs and process parameters required to model the bottle blowing process including preform geometry, stretch-rod geometry, preform temperature profile, stretch-rod displacement, supply pressure conditions and critically, the material model. Fig. 3 shows a simplified schematic of the SBM machine along with the ABAQUS numerical representation. Fig 3 (a) Schematic of SBM machine, (b) numerical representation The compressor supplies high pressure air to the preform through a pneumatic line where its magnitude is controlled by a pressure regulator. The air flow is then controlled by a flow control valve which can be finely tuned by the operator. Understanding this air flow is critical when forming a suitable FEA simulation. Schmidt et al. [12,13] used fixed volume experiments to ascertain the flow rate during at a given pressure value. The air flow rate was then calculated from the pressure gradient achieved in the rigid cavity volume. The true volume the air flow experiences, is in actual fact the preform volume and the volume of the pipework downstream of the control valve. This series of pipes, bends and orifices located between the flow control valve and the preform is defined as the dead volume and must be accounted for when supplying mass flow values to the simulation. Physically measuring this dead volume can be difficult as SBM machines can vary widely and technical drawings may not provide an accurate enough prediction. Salomeia [11] developed a method of determining the dead volume of any given machine from a simple two-stage pressure analysis. By performing rigid blow analyses with two fixed volumes i.e. rigid preform and rigid bottle, the dead volume can be calculated. For this project the dead volume was calculated to be approximately 85ml. Although in reality the dead volume is a complex series of pipes, bends and orifices, for the purposes of the simulation the dead volume is represented as a virtual added volume of 85ml with a non-specific geometry. The compressor is also represented like this with an extremely large volume so the pressure drop does not affect the simulation. The pressure build up within the preform is then modelled by applying an air flow from the compressor through the dead volume into the preform cavity. For low rate DOE30 (pressure 6 bar and flow index 2), the choked mass flow rate was calculated at g/s and for high rate DOE52 (pressure 10 bar and flow index 6) the choked mass flow rate was calculated at g/s. The air was assumed to be an ideal gas and any heat transfer from the cavity was neglected. The preform geometry shown in Fig. 1 was modelled in ABAQUS using deformable axisymmetric shell elements (SAX1), initial element size of 1mm and the preform thickness profile was applied. The stretch-rod was modelled using rigid elements (RAX2). Heat transfer between the preform and the stretch-rod was ignored and a simple Coulomb friction value of 0.05 was used.

6 1734 The Current State-of-the-Art on Material Forming 4.0 Results and Discussion The simulation results were compared to the experimental data for cavity pressure evolution, stretch-rod force and the strain evolution. 4.1 Cavity Pressure and Stretch-rod Force The cavity pressure evolution recorded by the instrumented stretch-rod shows similar trends for each of the selected trials. A linear pressure gradient for the initial part of the process under near constant volume conditions, then followed by a peak pressure value before the preform begins to yield. As the deformation continues and the volume expands, the polymer begins to strain harden and the pressure ceases reducing and begins to increase. This remains the case until the pressure is released at the end of the trial. Figure 8 shows the simulation results from both flow rate dependent trials compared to the experimental results. Note the 50ms pressure activation delay for the low flow rate DOE30, Fig. 8(a). It is also important to note that no stretch-rod force due to contact was predicted for the high rate trial during both the experiment and the simulation, Fig. 8(b). Fig. 8 Pressure and force comparison, simulation vs experiment, (a) low rate DOE30, (b) high rate DOE52 The initial linear pressure profiles from both trials display excellent correlation between the simulation and the experiment, indicating that the method of fluid exchange used is both accurate and robust. In the low rate trial, the predicted pressure is almost indistinguishable from the experimental data. In the high rate trial the peak pressure occurred at the same time however 9% was greater than the experimental reading. The magnitude of the predicted pressure compared to the experimental pressure at the latter stage of the freeblow process was around 20% larger. Both sets of data also reach constant magnitude, indicating that volume change was negligible. The predicted stretch-rod force for the low rate trial initially displays good correlation during the linear stretch deformation of the preform. There is then an unexpected drop in predicted stretch-rod force as the preform rapidly deforms, thus resulting in an under prediction of the peak force magnitude; 24% less than the experiment. This indicates that as the preform undergoes the rapid pressure-induced deformation, there appears to be over-dependency for the preform to deform axially; therefore reducing the amount of work required by the stretch-rod. The force gradient predicted after the stretch-rod stops moving correlates with that of the experiment. 4.2 Strain Data In order to compare the strain data from experiment and the simulation, certain key points on the preform were selected and the strain evolution over time recorded. These key points can be seen in Fig. 10.

7 Key Engineering Materials Vols Fig. 10 Strain selection points The true strain values at each point were recorded with respect to time, both the experimental VIC 3D data and the simulation ABAQUS data. Fig. 11 shows the true strain values at the three locations for the low rate DOE30 trial; Fig. 12 shows the same true strain values for the high rate DOE52 trial. Fig. 11 True strain data, low rate trial DOE30, VIC3D vs ABAQUS for (a) point 1, (b) point 2 and (c) point 3 Fig. 12 True strain data, high rate trial DOE52, VIC3D vs ABAQUS for (a) point 1, (b) point 2 and (c) point 3 The effect of air mass flow rate at constant temperature during the inflation process is clearly evident from the two flow rate dependent trials; the low rate trial taking approximately 1 second to reach maximum inflation while the high rate trial taking 0.25 seconds. It is also noticeable that during low flow rate inflation, the ratio of deformation is more favourable in the axial direction (ε yy ) while during high rate inflation the deformation tends to be more in the hoop direction (ε xx ). These results indicated that the overall shape of the inflating bottle is highly dependent on the flow rate. During the low rate trial, the significance of the linear stretch-rod deformation of the preform is highly evident. From 0-0.4s, point 2 and 3 clearly show linear axial stretching phase (blue) and negative hoop strain (red), an indication of uni-axial deformation. The overall trends of the axial

8 1736 The Current State-of-the-Art on Material Forming (ε yy ) and hoop (ε xx ) strains of the simulation compared to that of the experiment are very close, albeit having different magnitudes. It is also apparent from the strain results when the preform switches from predominantly uni-axial deformation (stretch-rod induced) to biaxial deformation (pressure induced); occurring at 0.25s for point 1, 0.4s for point 2 and 0.6s for point 3. The final strain values for the low rate trial indicates that the prediction of the axial strain is relatively close, possibly with too much deformation occurring in the simulation in both the hoop and axial direction at point 3 i.e. the base. Regarding the high rate trial, it is clear that there is very little uni-axial deformation for each of the points. Both the axial and hoop strain values follow a close trend relative to each other, signifying a highly biaxial deformation. For all three points the predicted strain values increase at a more rapid rate than the experimental VIC3D results suggesting that the material behaviour is too soft, either due to a preform temperature discrepancy or due to the material model. The predicted final strain values for both axial and hoop directions were comparatively close for point 1 with points 2 and 3 straying from the experimental values. 5.0 Conclusions A comprehensive set of experiments examining the free-blow deformation of PET preforms has been performed. The data gathered from these trials has been used to provide accurate input data for equivalent ABAQUS/Explicit simulation which has been validated using the experimental results. Free-blow simulations show good correlation with the experimental process outputs. The cavity pressure prediction is accurate up to the point of greatest preform deformation. Thereafter the pressure begins to drop sharply, more than the experiment dictates. The force tends also appear to be accurate although force magnitudes are under predicted. From these results, the preform material appears to be too soft Using the speckle patterns and high speed strain tracking software has proved to be a precise method of determining the axial and hoop strain values form the free-blow trials. Comparing these strain values to the predicted ABAQUS values shows that the strain trends are accurate and the simulation deformation modes are correct. The rate of deformation however appears to occur faster in the simulation, again indicating that the material behaviour is too soft. The overall analysis and subsequent simulation of the free-blow process appears accurate and robust. More insight into certain areas is required, such as the indication that the predicted final bottle shape appears to be bias towards the hoop direction, something not observed during the experiment. For further investigation, the application of a suitable mould using identical process input parameters seen here will highlight the effect contact, friction and heat exchange has on the final bottle shape and characteristics. References [1] H. X. Haung, Z. S. Yin and J. H. Liu, Visulisation study and analysis on preform growth in polyethylene terephthalate stretch blow moulding, Journal of Applied Polymer Science, vol. 103, no. 1, pp , 2007 [2] G. H. Menary, C. W. Tan, M. Picard, N. Billon, C. G. Armstrong and E. M. A. Harkin- Jones, Numerical simulation of Injection stretch blow moulding: comparison with experimental free blow trials, 10 th ESAFORM Conference on Material Forming, AIP Conf. Proc, vol. 907, pp , 2007 [3] C. C. Nagarjappa, PhD. Thesis, Developing and validating stretch blow moulding simulation through free blow experiments, 2012, Queens University Belfast, School of Mechanical & Aerospace Engineering

9 Key Engineering Materials Vols [4] N. Billon, A. Erner and E. Gorlier, Kinematics of stretch blow moulding and plug assisted thermoforming of polymers; experimental study, Polymer Processing Society Conference,21, Leipzig, 2005 [5] L. Chevalier and Y. Marco, Identification of a strain induced crystallisation model for PET under uni- and bi-axial loading: Influence of temperature dispersion, Mechanics of Materials, vol. 39, issue 6, pp , 2007 [6] S. Yan, J. Nixon and G. Menary, Understanding the behaviour of PET in Stretch Blow Moulding, 5 th International PMI Conference 2012 [7] S. Yan, G. Menary, Modelling the constitutive behaviour of PET [8] C. P. Buckley and D. P. Jones, Glass-rubber constitutive model for amorphous polymers near the glass transition temperature, Polymer, vol. 36, no. 17, pp , [9] C. P. Buckley and D. P. Jones, Hot-drawing of poly(ethylene terephthalate) under biaxial stress: application of a three-dimensional glass-rubber constitutive model, Polymer, vol. 37, no. 12, pp , [10] P. J. Martin, C. W. Tan, K. Y. Tshai, R. McCool, G. Menary, C. G. Armstrong and E. M. A Harkin-Jones, Plastic, Rubber and Composites, 2005, vol. 34, no. 5/6, pp [11] Salomeia, Y.M., PhD Thesis, Improved Understanding of Injection Stretch Blow Moulding through Instrumentation, Process Monitoring and Modelling, 2009, Queen's University of Belfast, School of mechanical & Aerospace Engineering [12] F. M. Schmidt, J. F. Agassant and M. Bellet, Experimental study and numerical simulation of the injection stretch/blow moulding process, Polymer Engineering & Science, vol. 38, no. 9, , Sept 1998 [13] M. Bordival, F. M. Schmidt, Y. Le Maoult and V. Velay, Optimization of preform temperature distribution for stretch-blow moulding of PET bottles: Infrared heating and blowing modelling, Polymer Engineering & Science, vol. 49, no. 4, pp , [14] ABAQUS Documentation, Simulia

10 The Current State-of-the-Art on Material Forming / Analysis and Simulation of the Free-Stretch-Blow Process of PET /