EXPERIMENTAL STUDY AND NUMERICAL SIMULATION OF PREFORM INFRARED RADIATIVE HEATING Serge Monteix (*), F. Schmidt (*), Y. LeMaoult (*), G. Denis (**), M. Vigny (**) (*) Ecole des Mines d Albi-Carmaux Campus Jarlard, Route de Teillet, 81013 ALBI CT Cedex 09 (France) (**) Perrier-Vittel M.T., BP 43 88805 Vittel Cedex (France) ABSTRACT The injection stretch-blow moulding process o thermoplastic bottles requires an heating step beore orming. An amorphous preorm is heated above the glass transition temperature (80 C or the P.E.T.) using an inrared oven. This step is undamental in order to determine the thickness distribution along the preorm height and then insure high quality bottles. Thus, the optimisation o the inrared oven is necessary. Various experiments have been conducted to characterise the heat source and the semi-transparent properties o the P.E.T. Measurements o air temperature inside the inrared oven and air cooling speed have been processed. These parameters have been implemented in control volume sotware that simulates the heating step. The surace temperature distribution o the preorm has been measured using an inrared camera. Comparisons between experimental and numerical results or a rotating preorm are presented. 1. Introduction The injection stretch-blow moulding process o thermoplastic bottles requires a heating step beore orming. An amorphous tube-shaped preorm is heated above the glass transition temperature (80 C or the P.E.T.) using an inrared oven [1]. As the P.E.T. conductivity is low (0.29 W/m/K), halogen lamps permit a rapid heating with a high lux intensity. In addition, a cooling an is necessary in order to prevent P.E.T. thermal crystallisation that will induce nonamorphous preorm. However, there are plenty o parameters such as number o lamps, distance between halogen lamps and preorm, electrical power o the lamps, air cooling speed, relectors, rotation speed,. Thereore optimising inrared oven in order to design new plastic bottles is still a long and costly work. The main objective o this work was to develop numerical sotware using control volume method that simulates the heating step o the preorm. For that, we have to take into account inluent parameters o the heating step such as halogen lamps characterisation and P.E.T. spectral properties. In addition, we need to measure the preorm temperature. We have used an 880 LW AGEMA inrared camera (connected to real time sotware) in order to measure external and internal preorm surace temperature distribution, heat transer coeicient along preorm height. The air temperature surrounding the preorm and inside the preorm is measured using thermocouples. 2. Heat source characterisation Halogen lamps commonly used or the heating step are presented in igure 1. They are composed o a coiled tungsten ilament, contained in a quartz tubular enclosure coated with ceramic in its back (ceramic relector). Between the concentric cylinders, there is a neutral gas (Argon) that allows high temperature o the ilament without important evaporation o the tungsten.
Spectral properties o these elements have already been measured in previous paper [2]. Using these data, the tungsten ilament and quartz tube temperatures are calculated using a net radiation method [3]. In a irst approach, the relectivity o tungsten and quartz elements is neglected. Finally, the steady state heat balance equation leads to the ollowing system o equations (1), solving using a Newton-Raphson method: P S S T T 4 4 2 L 2 L k k g g T T d ln * d T T d ln * d 0 S T 4 hs T T 0 (1) Where P is the electrical power o the lamps; the indices and q reer respectively to the tungsten and the quartz tube; T is the temperature, L the lamp length, d the ilament diameter, and the average emissivity and absorptivity. At the nominal lamp power o 1 kw, this model permits to calculate a ilament temperature at 2500 K and quartz tube temperature at 1000 K. In order to measure the ceramic relector eiciency, experiments were perormed using a thermopile detector with a large spectral bandwidth [0.25-26 m]. This experimental set-up is shown in igure 2. We measured the spatial directivity o the lamps with and without ceramic relector, moving the lux sensor around the ilament. The eiciency o the relector K e or each angle is calculated using: K intensity irradiaterom the lampwithrelector e ( (2) 2 * intensity irradiaterom the lampwithoutrelector) The eiciency coeicient is plotted versus angle in igure 3. The average value is about 68 %. In addition, we note that the temperature o the tungsten ilament has no eect. 3.Characterisation o P.E.T. spectral properties P.E.T. spectral properties measurements were perormed using a 3-mm thick sheet. These sheets were processed by injection moulding in the same conditions than tubular preorms. Thermal analysis experiments were carried out in order to measure the sheets crystallinity. Samples o 10-15 mg weight were cut away rom the thick sheet. The DSC experiments were perormed on a Mettler calorimeter at a heating rate o 10 C/min. The average crystallinity value o these measurements was about 5 %, corresponding to an amorphous P.E.T., which means that diusion o the radiation can be neglected. Then we have proceeded to P.E.T. spectral properties measurements (transmitivity, relectivity) using a Perkin-Elmer FT-IR spectrometer. The previous thick sheets were polished in order to obtain samples o 0.3-mm thickness. The measured absorbtivity is plotted versus wavelength (igure 4). The spectral distribution radiated by the tungsten ilament at temperature o 2200 K is superimposed. It appears rom these experimental data that the material is transparent at short wavelengths where the intensity o the radiation is maximum. In addition, the polymer is opaque or wavelengths in the range o 8 to 12 m with an average emissivity o 0.93. This bandwidth corresponds to the one o the inrared camera.
4. Experimental data Figure 5.a shows the experimental set-up developed on the basis o industrial inrared oven using in injection blow moulding machines. This device consists o six halogen lamps (Philips 1000 W-235 V SK15) describes previously, monitored by power device. The preorm is ixed on a rotating support to provide homogeneous hoop temperature. Behind the preorm, there is a cooling an controlled by a DC supply. A pneumatic system allows the preorm displacement inside and outside the oven. 4.1 Measurement o the air temperature inside the oven For the measurement o the air temperature inside the oven, J type thermocouples were used. They are connected to a Keithley recorder instrument. All thermocouples are shielded with aluminium adhesive tape. They are mounted in the oven as presented in igure 5.b. In order to place a thermocouple inside the preorm, the top o the preorm is drilled. Average data issued rom a set o measures are plotted in igure 6. In particularly, we note that the temperature o the air inside the preorm, still increase ater the heating stop. 4.2 Measurement o surace temperature distribution Typical inrared thermograms or external surace temperature distribution o the preorm are presented in igure 7. However, the internal surace temperature measurement is more tedious and necessitates developing a speciic process. First, we have to cut the preorm in its longitudinalsection (igure 8). Then, the two hal-parts o the preorm are sticking together using a conductive grease to insure joining during the heating time. All the lamps are set to the same temperature, coupled with the cooling an, during 20 seconds. Then the rebuild preorm exits rom the oven in ront o the camera. In order to make possible the internal surace temperature measurement, one hal-parts is removed (immediately ater preorm recovering or ater a characteristic inversion time). Comparison o external and internal surace temperature measurements versus time (igure 9) provides the temperature inversion throughout preorm thickness (requency o analysis is equal to 1 rame per second). 4.3 Measurement o air orced convection heat transer coeicient In this section, we want to determine the air orced convection heat transer coeicient, generated by the cooling an along the preorm height. Two dierent methods are developed and compared. The irst one consists in measuring air-cooling speed using a hot wire anemometer (Ans Snelco). The calculated Reynolds numbers deduced rom these data are in the range between 4000 and 10000 that is to say a turbulent low. Then, we identiy rom the literature [5,6,7] dierent classical relationships that are suitable or a cylinder in cross low, i.e. Nu Re, Pr. The second method is based upon the preorm surace temperature measured in presence o cooling low. The ollowing speciic experimental procedure has to be observed in order to compute the heat transer coeicient: irst, a 20 seconds heating time coupled with cooling; then turn o heating and cooling in order to insure an uniorm temperature throughout the preorm thickness; then restart the cooling an. Thus, the last step o the procedure allows solving analytically the energy balance integrated over the thickness i we assume that the main part o the preorm is a hollow tube. For a given height, the relationship or the heat transer coeicient h yields:
External surace temperature distribution versus time is presented in igure 12. In addition, numerical computations allow computing temperature distribution in the preorm crosshttp://dx.doi.org/10.21611/qirt.2000.026 dt Cp dt h 0 Re Ri T T Re o t 2 2 (3) Where T is the thickness-average temperature, C p the P.E.T. heat capacity, the P.E.T. speciic mass, T the thickness-average ambient temperature, R e and R i are respectively the external and internal preorm radii. R i is computed using the depth penetration throughout the preorm (4) over a characteristic time t, or which we have suicient temperature data to calculate the initial slope dt o the temperature-time plot (igure10). dt t 0 Ri Re e Re 2 t (4) Where is the P.E.T. diusivity and the argument o the er unction, computing assuming the surace temperature has decreased o 90% at the depth thickness. In igure 11, we present a comparison between the heat coeicient measured using both methods. The agreement between the two dierent methods is air. It permits to calculate an average air cooling heat transer coeicient o 45 W/m 2. K± 5. All these process parameters will be useul or the numerical model assessment. 5. Modelling o inrared PET preorm heating 5.1 Control-volume model 3D control-volume sotware, called PLASTIRAD, has been developed in order to compute heat transer during the inrared-heating step. The preorm is meshed using hexahedral elements so-called control volumes [8]. The temperature balance equation including radiative transer (thermal power absorbed by the semitransparent P.E.T. preorm radiates rom halogen lamps) is integrated over each control volume and over the time interval rom t to t+t (5): t e k T. nddt r. n T c p d dt ddt (5) t te Where e is the control-volume, q c the Fourier s lux and q r the radiative heat lux. The unknown temperatures are computed at the cell centres. Further details on the numerical procedure (view actor computation, volume source) are given in previous paper [9]. 5.2 Numerical simulation We perormed simulation o the heating step or a P.E.T. preorm, taking into account processes parameters introduced previously. The values o the main parameters are summarised in the table below. Heating time (s) Cooling time (s) Number o lamps Distance lampspreorm (mm) t e Heat transer coeicient (W/m 2. K) Lamps Temperature (K) Rotation Speed r s- 1 ) 20 20 6 20 45 2400 1.2
section (igure 13). Finally, a comparison between numerical and experimental data is presented in igure 14. Preliminary comparison shows a discrepancy, between computed and measured temperature. We observed that the computed temperature decreases more rapidly during the cooling step. We have to introduce a more accurate heat transer coeicient and time dependent air temperature in order to improve the simulation. Conclusion Numerical simulations o P.E.T. preorm heating stage have been perormed successully. Characterisations o halogen lamps and P.E.T. spectral properties have been processed in order to deine assumptions necessary or describing the heat transers inside the oven. An original method was developed to measure the internal surace temperature o the rotating preorm, using a cut-preorm technique. This procedure allows measuring the so-called inversion temperature phenomenon. The orced convection coeicient due to the cooling an was measured along the preorm height using an inrared camera. In addition, the inrared oven was instrumented using thermocouples that provide the air temperature evolution inside the oven and inside the preorm. The inrared camera is a powerul device in order to make quantitative temperature measurements and to assess the numerical simulations All the measured parameters have been included in the numerical model. The agreement between numerical simulations and experimental data is air considering the simple model used or absorption (Beer-Lambert law). Future developments will include coupling between inrared heating simulations and stretch-blow moulding simulations. Acknowledgements Perrier-Vittel Company supports this work. Special thanks go to J-P. Arcens, C. Miquel and M. Farasse Reerence [1] Rosato D. V., Blow Moulding Handbook. New York: Carl Hauser, 1989 [2] Le Maoult Y., F. Schmidt, El Hai M., Lebaudy P., «Measurement and Calculation o Preorm Inrared Heating», 4th International Workshop on Advanced Inrared Technology and Applications, Firenze, pp. 321-331, Sept. 15-16, 1997 [3] Petterson M., Heat Transer and Energy Eiciency in Inrared Papers Dryers, Ph. D. Thesis, Lund University, Sweden [4] Özisik M.N., Finite Dierence Methods in Heat Transer,CRC Press,1994 [5] Holman J. P., Heat Transer, Seventh Ed., Metric,1992 [6] Incropera F. De Witt D.P., Fundamentals o heat and mass transer, Third Ed. 1990 [7] Özisik N. M., Heat Transer a Basic Approach, Int. Ed., 1985 [8] Patankar S.V., «Numerical Heat Transer and Fluid Flow», Ed. Mc Graw Hill, New-York, 1980 [9] Monteix S., Experimental Study and Numerical Simulation o Preorm or Sheet Inrared Radiative Heating, AMPT 99, Dublin, 3-6 August 1999
Ceramic relector uartz Tube Tungsten ilament Fig. 1 Inrared electric heaters Inrared heater Thermopile Fig. 2 Experimental set-up or relector inrared heaters characterisation. Fig. 3 Ceramic relector eiciency Heaters at 2200K PET 0.3mm Fig. 4 PET spectral absorbtivity / Heaters Intensity Fig. 5 Experimental inrared Oven Set-up Fig. 6 Thermocouples oven s instrumentation Z Experimental Proile Fig. 7 Air ambient temperature environment inner surace + Temporal marking Fig. 8 Temperature thermograms
Fig. 9 Preorm longitudinally cutted Fig. 10 Preorm surace temperature using inrared Camera (LW 880) Fig. 11 Surace temperature versus time /Cooling restart Fig. 12 Forced convection coeicient Fig. 13 Numerical simulation o preorm surace temperature versus time Fig. 14 Numerical simulation o preorm temperature versus time across thickness Fig. 15 Experimental and Numerical Surace Temperature Results versus time