Quantitative infrared thermography applied to blow moulding process : measurement of a heat transfer coefficient

Transcription

1 Quantitative infrared thermography applied to blow moulding process : measurement of a heat transfer coefficient By Serge Monteix*, Yannick Le Maoult 1 **, Fabrice Schmidt **, Jean Paul Arcens ** * Philips Special Lighting, Pont à Mousson Factory, Chemin de Montrichard, BP 149, 54705, Pont à Mousson Cedex, France. ** Cromep, Ecole des Mines d Albi Carmaux, Campus Jarlard, Route de Teillet, ALBI Cedex 09, France. ( 1) corresponding author. ABSTRACT This paper deals with an application of blow moulding process applied to PET bottles forming. The most important stage of this process is the radiative heating step which is realised with infrared ovens using powerful halogen lamps. To validate a 3D thermal control volume software, called Plastirad, developed in our laboratory, temperatures maps were needed on the plastic preforms as well as convective heat transfer coefficient inside the oven. This measurement has been performed with two different methods : IR thermography and hot wire anemometry. These two methods have been investigated and results are compared to focus on the interest of IR thermography. 1. Introduction : The industrial injection stretch-blow moulding process of Polyethylene Terephtalate (PET) bottles [1] requires a heating step prior forming. During this step, the amorphous preform corresponding with the form prefiguring the final bottle geometry is heated from the ambient temperature to the forming temperature between 105 to 115 C. Due to the weak PET thermal conductivity (0.25 W/m.K) and very high production rate (up to bottles per hour), the heating step is performed using infrared ovens. These oven are constituted with a row of special halogen-lamps, which permit a rapid heating with high flux intensities ( kw/m 2 ) and a cooling fan to control the external surface temperature, to avoid the thermal crystallisation which would make the transparent amorphous preform completely opaque and more fragile. Then, the ventilation appears as a key parameter controlling the preform temperature distribution. Our investigation deals with the computation of this heat transfer coefficient corresponding to the experimental set-up we reproduced from the industrial machine to the laboratory scale. Two different methods have been tested for this study : - Quantitative infrared thermography - Classical anemometry method Then these two methods have been compared and discussed. 1.1 The experimental set-up The experimental set-up developed is constituted with six halogen lamps (Philips 1000W-235V) fitted with white ceramic reflector settled above the rear quartz tube cylinder. A flat aluminium reflector is disposed behind the lamp panel and the cooling fan opposite the lamps (figure 1). F.2.1

2 Fig 1.shematic view of the infrared oven and geometric parameters The PET preform is a cylinder of height 100mm and diameter 25 mm which can rotate axially in the middle of the oven for a uniform heating through the preform circumference. 2. Heat transfer coefficient computation using an infrared camera. the heat flux received by the camera is related to the radiative properties of the material observed but any elements in the thermal scene contribute to the total heat flux integrated by the camera. Thus, precautions are needed to measure the real temperature of the preform with a good accuracy. To do so, the environment participation has been described with a classical radiometric equation involving ε pref and ρ pref, respectively the emissivity and reflectivity of the PET preform integrated over the spectral band of the camera chosen for the measurement. Due to the very short distance between the camera and the preform (# 0.5m), the atmospheric transmission is equal to one for our application. ε pref and ρ pref have been carefully investigated to choose the relevant spectral band to make infrared surface temperature measurements. 2.1 The choice of the optimal spectral band for the measurements. P.E.T spectral properties measurements (transmissivity, reflectivity) were obtained using a Perkin-Elmer FT-IR spectrometer for the spectral range 1 to 25 µm. 3 mm thick sheets were polished in order to obtain samples of 0.3 mm thickness. These properties are assumed to be independent of the temperature variation from [2]. From apparent transmission measurements T λ for thickness [0.3-3 mm] and [3], we get : Ka 1 T λ τ λ., and λ. d τ λ = e then 1+ ln( Tλ ) = K λ. d + ln( ) (1) a 1+ 1 These measurements lead to an intrinsic absorption coefficient K aλ independent of d (Beer s law is relevant) and reveal that PET presents successively transparent and opaque regions over large spectra between 1 to 25 µm and in the bandwidth [8-12 µm], so called long wave region (LW), we have computed the Planck s mean absorption coefficient : F.2.2

3 12µm 0 K aλ. L ( λ, T400K ) dλ (2) 8 ( T = 400K) = µm 0 L ( λ, T ) dλ K a K (2) and T =400 K (processing temperature of the preform) give Ka(T) = m -1, with the optical thickness K a (T).d, the photon mean free path of d = 1/Ka is minor than 30 µm which corresponds to an opaque material, then measurements using the [8-12µm] band are surface temperature measurements. For this reason, we have chosen an AGEMA 880 LW camera connected to a real time 12 bits thermography software among the three systems we get in the Ecole des Mines (VSW, SW and LW). Then the values of emissivity and reflectivity calculated for the radiometric equation in order to compute the real preform temperature are respectively ε pref = 0.93 and ρ pref =0.07. The camera was calibrated versus the temperature using a LAND blackbody between 30 C and 175 C corresponding to t he full window of the phenomenon we wanted to observe. Which leads to an discrepancy minor than 0.5 C in comparison with the blackbody reference tempe rature. 2.2 The influence of the oven environment on the IR measurements: a first approach. To control the temperature measurement of the preform inside the oven, we have firstly estimated the different irradiances Ei incoming to the lens of the camera from the different areas of the oven, mainly : E1 : irradiance received by the camera from the preform E2 : irradiance received by the camera from the quartz tube and reflected by the preform E3 : irradiance received from the ventilation support and reflected by the preform Assuming that all surfaces are diffuse, composing of grey material, E1,E2 and E3 terms have been calculated using a radiosity approach with integrated optical properties of the PET in [8,12µm] and different view factors in the oven computed using analytical formulation (the real geometry is replaced by an elementary one [3 ]). The computations of Ei and Ei/E total give (table 1) : Table 1 : Estimation of the different radiative irradiances received by the camera Radiative source Εi (W/m 2 ) Ei/Et (%) Preform % /direct the 6 halogen lamps system %, reflection the ventilation system 4.03 E-5 0.2%, reflection Table 1 leads us to conclude that the oven environment participation is negligible. Moreover, a simulation of the difference between the irradiance from a blackbody reference and the one incoming from the preform with optical parameters computed in 2.1 and a 25 C software environment temperature shows a discrepancy equal to 1 C around the 120 C preform temperature. This nume rical approach has been compared to experimental measurements; a sequence of images has been recorded when the preform was just entering inside the oven : it was possible to measure a new equivalent blackbody temperature related to the cumulated reflections on the preform just before heating. This temperature was close to 28 C. The difference between the new and default environment temperature T env (25 C) is 3 C. This value F.2.3

4 is slightly different from the computed value (1 C), but we assumed that a correction was not necessary. 2.4 Measurement of the heat transfer coefficient-methodology The computation of the heat transfer coefficient distribution through the preform height is based on the IR images acquisition. These images have been recorded with a frequency of 6 images per second for a preform in rotation inside the oven (figure 2) : Preform s thermogram Z Heating phase Z7 Zonal average temperature Z6 vs time Z5 Z4 Z3 Equilibrating phase Re-cooling phase for the ventilation analysis Z2 Z1 Spatial & temporal marks T(Z,t) Fig. 2. areas of interest and methodology We dealt with the spatial average temporal temperature variation over 7 circular zones regularly disposed along the preform height (figure 2), each circular zone had a diameter of 5 mm, large enough to check the homogeneity of temperatures(convenient also with the spatial resolution of the camera). Three steps are necessary for the measurement : - The first step : a 20s heating stage, with the ventilation system on, which enables to get a sufficient preform temperature. - The second step : an equilibrating phase of 10 s is applied to leave the preform temperature evolve freely due to conduction, radiation and natural convection in order to tend a null temperature gradient through its thickness [4]. - The third step : the ventilation is switched on back, we analysed more particularly this phase and the temperature distribution using a simple transient monodimensional semi-infinite analytical model [5] through the preform thickness by solving : 2 T 1 T T (4) =. and T ( x, t = 0) = T ; = ( (, = 0) ) ; (x =, t) = T 2 ini hg T x t T T ini x α t x y= 0 (4) leads to the temperature expression versus time and thickness (x) and h : h x h 2αt g g λ T ( x, t) T x λé + x hg αt ini (5) = erfc( ) [ e. erfc( + )] T Tini 2 αt 2 αt λ Where α is the thermal diffusivity of the material and λ its thermal conductivity, for P.E.T α = m 2.s -1 and λ = 0.25 W/m.K. The heat transfer coefficient h g is the amount of the radiative and convective transfer. To differentiate the convective part, we have to compute the radiative one with : h = h h = h c g r g 4 4 env ( T T ) - ε σ (6) T T env F.2.4

5 2.5 Accuracy of the mono-dimensional analytical model Since the preform geometry corresponds to a cylinder, the approach used is a rough assumption and to evaluate the discrepancy, we compared temperature computations from this 1D model to ones issued from a 3D ControlVolume Model (CVM) Plastirad software developed in our laboratory [6], dealing with a tubular geometry of 3 mm thickness close to the preform one. We compared temperature evolution from a uniform initial temperature, due to three different heat transfer coefficients: 5, 35 and 65 W m -2 K -1 and took into account the same PET properties, environment temperature equal to 30 o C, superimposed on figure 3 the outer surface temperature and the inner one at a 3 mm depth versus the time for the greatest value of h : Fig 3. Comparison of 1D computation and 3D CVM ; h g = 65 W m -2 K -1 Differences between the CVM and the 1D analytical method are low : the mean relative error is less than 2.3 %. We can notice that, for the internal temperature, a divergence is visible between the CVM and the 1D computation for t ~7s, this time corresponds, for a thickness E of 3 mm, when the surface perturbation is viewed at 1%, which leads to [7], t ~E 2 / α ~7s and to conclude to the interest of the simple 1D method. Then, using this approach and surface temperature values extracted from each interest zone, the heat transfer coefficient identification has been processed minimising the total relative difference between experimental temperatures and computations assuming a constant environment temperature. 4. Heat transfer computation from air flow velocity measurement This method is based on the air flow velocity measurement near the preform. Due to the narrowness of the oven (less than 10 cm), an ANS SNELCO hot wire anemometer with a speed range of 0-30 m/s was used and disposed inside the air flow at different points along the preform. Measurements were made with the lamps switched off at the ambient temperature due to the high radiative flux from the lamps during the heating stage. From the air velocity measurements, we have computed convective coefficient, with air properties at the ambient temperature~20 C. With a Reynolds in the interval 4000 < Re < 10000, it was assume that the air flow is turbulent. Then we used the Hilpert correlation [5] corresponding to our configuration and Reynolds scale : / 3 Nu = Re.Pr and Nu.λ air h = (7) d p F.2.5

6 5. Comparison between the two methods we have superimposed results of the heat transfer computations using the two different approaches (figure 5) : h-anemometric method H-infrared camera 60.0 Heat transfer coefficient (W m-2 K-1) preform height (mm) Fig 5. Comparison of hc along the preform height versus the two methods, Results by the IR thermography represent the heat transfer average value versus the air temperature and corresponding uncertainties. The mean relative difference between the two methods is close to 20 % on the hc value but on figure 5, we note that the two methods lead to a same heat transfer coefficient scale value. These results show, a high measurement dispersion due to the turbulent flow and we note a large incertitude for two methods due to the high non uniformity of the air flow and the temperature variation inside the air flow.. 5. Conclusion We wanted to estimate the heat transfer coefficient along the preform height and coupled with an air cooling. First, we analysed the environment participation in the oven and showed that, despite all the reflectors inside the oven, the participation of the preform environment was negligible. We developed a specific analysis using IR thermography with images recorded all along the preform height and focused this analysis on the cooling step in order to simplify the problem. A 1D analytical model was developed to compute the preform surface temperature distribution versus the heat transfer coefficient. To validate this computation, we compared the accuracy of the 1D model with a CVM. This comparison lead us to conclude that discrepancies are minor than 10%. Finally, we identified a local heat transfer coefficient h c(z). The comparison between the two methods provided heat transfer coefficients in the same scale. However, we noted that the method based on the thermography seems very sensible to the air temperature. But, due to the relative proximity of the heat transfer coefficient obtained by these methods, we will emphasize on the interest of the thermographic method which deals with all the preform height and limits the measurement time. The method was also tested successfully on industrial machines and lead to realistic values of h c in the range [35,60] W/m 2.K, range very close to our laboratory results. We are, now, working on a specific ray tracing method devoted to accurate radiative transfer computation inside the infrared oven and also to optimise of the position of the IR camera to be less sensitive to reflections in a complex environment. F.2.6

7 Acknowledgments : The Nestlé Waters Company has supported this work. REFERENCES [1] Rosato D., Blow molding handbook, Hanser Publishers, [2] Weinand D., Modellbildung zum Aufheizen und verstrecken beim thermoformen, Dissertation an der RWTH. [3] Siegel R., Howel J, Thermal Radiation Heat Transfer, Third Ed Hemisphere publishing Corporation, [4] Monteix S., Modélisation du chauffage convecto radiatif de préforme thermoplastique pour la réalisation de corps creux, thèse ENSMP, [5] Incropera F.P., DeWitt D.P, Fundamentals of Heat and Mass Transfer, fifth ed. John Wiley [6] Monteix S., Schmidt F., Le Maoult Y., Denis G., Vigny M., Comparison between a numerical model and an experimental approach of preform infrared heating - recent results, International Conference on Heat Transfer ASME-Australia-may [7]De Vriendt A.B., La transmission de la chaleur», vol.1, tome 1, Gaétan Morin, Chicoutimi Québec, F.2.7