Comparison Of Different Fem Code Approaches In The Simulation Of The Die Deflection During Aluminium Extrusion.pdf

Int J Mater Form (2010) Vol. 3 Suppl 1:375–378 DOI 10.1007/s12289-010-0785-1 © Springer-Verlag France 2010

COMPARISON OF DIFFERENT FEM CODE APPROACHES IN THE SIMULATION OF THE DIE DEFLECTION DURING ALUMINIUM EXTRUSION L. Donati1, N. Ben Khalifa2, L. Tomesani1, A. E. Tekkaya2 1

2

University of Bologna – Dept. of Mechanical Engineering (DIEM)– Italy TU Dortmund – Institute of Forming Technology and Lightweight Construction (IUL) – Germany

ABSTRACT: In this paper a multihole die producing two overlapped U-shape profiles was used in order to analyze the die deflection during the extrusion process thus allowing the benchmarking of different FEM codes. The two die openings were differently designed in order to realize different die deflection (fully supported and partially supported). Profile lengths, die and profile temperatures, process load and die deflections were used as benchmarking parameters for FEM comparison. A summary of the different FEM codes approaches, of the computational times of the main outputs and their comparison with experimental results are presented. A detailed discussion for all output parameters is realized in order to understand the potentials and limits of each code. Finally a discussion on future perspective for FE code application and designing guides is reported. KEYWORDS: Extrusion, Process Simulation, Die deflection, FEM codes, AA6082

1 INTRODUCTION In the aluminium extrusion sector in the last years FEM codes are becoming the most important tools for process and product optimization. Nowadays, the simulation of the extrusion process by means of FE codes has been applied in a great number of papers available in literature [1,2,3] but its application in everyday production was limited due to several factors like computational times, user’s skills as well as prediction accuracy. Indeed, the inner complexity of the process, characterized by extremely high deformations, strain rates and heat exchange phenomena, has lead only in the last few years commercial FE codes to gain sufficient accurate solving capabilities. In order to clearly evidence the several code’ accuracy, reliability and computational times a conference series has been organized with the specific aim to benchmark the most diffused FEM codes by mean of a systematic comparison of the outputs with experimental trials performed under strict monitored conditions [4]. In this paper a summary of the outputs of the FEM codes participating to the 2009 edition of the conference are presented and discussed on the basis of the comparison with experimental results and with 2007 edition results.

2 EXPERIMENTS A very challenging experiment was set up, in order to clearly test the FEM code capabilities to simulate the relationship between material flow, die deflection and

profile distortions. In fact, one of the most critical features related to tool deflection for the extrusion dies are the ´tongues ´ that are necessarily adopted in the manufacturing of dies, for example, for U shape profile extrusion. In the experiments, a multiple hole die for the simultaneous extrusion of two U shape profiles was designed (Fig. 1, left) in order to produce an effective comparison for different tongue design strategies. Both profiles have the same dimension and feeder in order to balance material flow entrance. The two profiles were arranged one upon the other: the bottom one is characterized by a standard supporting shape while the supporting part of the upper profile was deeply shortened (26,95mm fig. 1, right) in order to achieve a measurable die deflection during the process. With this configuration, a higher deflection of the upper part of the die with the less supported inner profile contour is expected, thus producing a loss of contact in the bearing zones and an alteration of the thermal field at the two exits; as a consequence a difference also in material flow and profile distortions are expected for the two profiles. On the other side, the U-shape of the profiles allows the accessibility for the laser measuring devices to detect the deflection of the inner tongues along the extrusion direction, while the length of the U shape (59,3mm) allows an adequate magnification of the die deflection. The die was made of AISI H-13 hot-working tool steel tempered to 45 HRC hardness and built by WEFA, Germany. AA6082-O aluminum billets of 140 mm diameter and 300 mm length were used for the experiments.

____________________ * Corresponding author: Viale Risorgimento 2, +39-051-2090496, +39-051-2093412, [email protected]

376

Figure 1: Die design (section and 3D view).

The experiments were carried out on a 10 MN extrusion press at the laboratory of the Institute of Forming Technology and Lightweight Construction (IUL) at the TU Dortmund University at a ram speed of 10 mm/sec with a maximum ram stroke of 290 mm. The diameter of the container was 146 mm, so that an upsetting of the billets takes place at the beginning of the extrusion process. The die was heated to a target temperature of 400 °C inside the machine. The billets were heated up to 450 °C in a furnace, with the billet surface decreasing to 432 °C just before extrusion: the billet temperature reduction was caused by the billet loading procedure that took about 1 minute. The temperature of the container press can be considered as constant and equal to 430 °C due to its high thermal inertia, while the ram temperature was measured with a contact thermometer at the beginning and end of each cycle and it remained in the range 365-411°C. The profile temperature was continuously measured through a self-calibrating pyrometer for aluminum application only on the upper profile. The used pyrometer works with two different wavelengths so as to calculate the work piece temperature impartially from the surface property. The detection point was located 145 mm away from the die surface as reported in fig. 1 right. Two laser velocitymeters were used in order to continuous monitoring the profile speed of both profiles. The velocitymeters work contactless with lasers beams based on the Doppler principle. The die deflection was

measured with two laser beam distance sensors; the sensors operated with the triangulation method and had an accuracy of ±60 μm. The application of the laser sensor showed great advantages if compared to the use of strain gauges or tactile deflection sensors: the laser beam works without any direct contact with the hot die, they did not require any holes or joining procedure and nevertheless it provided a continuous measurement of the tool deformation. Both sensors were arranged at a small angle to the profile direction in the inner area of the profile shape. It was necessary to evaluate the exact angle between tool and sensors, to compensate for the difference between the diagonal path of the laser sensor and searched die deflection. The positions of the lasers were measured after heating up the extrusion press and the die to discard thermal effects. Three repetitions were required to show the possible scattering range of the measured parameters and to evaluate the accuracy of the results.

3 FEM SIMULATIONS Six FEM codes joined the 2009 benchmark: four were already present at the 2007 edition [5] (Deform, HyperXtrude, Dieka, QForm,) and two new entries partecipated to the 2009 edition (Simufact and MTD). The codes differ one from the other for different simulation approaches (lagrangian, eulerian or mixed) and consequently type of analysis (transient, steady state, mixed) thus generating different computational time or setting times. Table 1 summarizes the set-up and computational times for the different codes: seven codes are there reported because one presented also different

Table 1: Comparison of FEM codes set up times and computational time in 2009 and 2007 editions.

2009 Model set-up Calculation time 2007 Model set-up Calculation time

Code 1

Code 2

Code 3

Code 4

Code 5

Code 6

Code 7

70 minutes 61 hours Code 1-07

360 minutes 80 hours Code 2-07

120 minutes 150 hours Code 3-07

30 minutes 9.2 hours Code 4-07

10 minutes 3 hours Code 5-07

5-6 minutes 12-37 hours Code 6-07

55 minutes 157 hours Code 7-07

240 minutes 720 hours

30 minutes

6 minutes

10 hours

244 hours

200 minutes 1224 hours

(23.3 mm stroke)

(37 mm stroke)

200 minutes 629 hours

377

data depending on the adopted approach type (UL or ALE). By comparing the 2009 times with the 2007 ones, it’s remarkable that computational times extraordinarily decrease in just 2 years: in 2007 the computational times were in the range of 10-1200 hrs while now they are in the range of 3-150 hrs. For some codes the reduction of computational time is greater than 500% thus demonstrating the great effort that FEM codes developers invested in this direction.

AISI H-13 temperature dependent young modulus that was taken from literature and not tested on the specific tool steel used for die manufacturing [6]. Due to the die deflection during the stroke, different friction conditions acted in the bearing zones thus altering the material flow.

4 RESULTS AND DISCUSSION The profiles speeds, the load stroke curve, the die deflection, the temperature evolution on the profiles and in the dies were recorded during the process as reported in figures 2,3,4,5 and 6 together with the predictions of the different FEM codes. The load stroke diagram of the experimental trial with the confidence interval of the three repetitions is reported in figure 2: the typical direct extrusion load-stroke diagram was found with a maximum load of 8,00MN in test 4 (benchmark test) while 8,04 and 8,51 MN were obtained within trial repetitions. It’s possible to note that depending on the different simulation strategies some code are able to predict the load also in the die filling –i.e. before the peak load (codes 2,6 and 7)- while the others start the computation immediately after the peak load. A generally good prediction in term of diagram shape and quantitative values was found for all the codes except for the code 2 that deeply overestimated the prediction. Concerning the die deflection, the aluminum pressure acted almost equally at the beginning of the process on the two die tongues, but due to the different support of the two profiles the upper tongue (partially supported) significantly deflected under the aluminum pressure. The tongues displacements were recorded trough laser system as previously explained and the maximum recorded displacements are reported in figure 3. On the other hand also the die system is subjected to press forces so as to produce a displacement comparable to the fully supported tongue. In other words, the fully supported tongue resulted un-deformed respect to the other part of the die. Stating this situation the best parameter for evaluating the die deflection is to consider the difference in displacement between the upper and the lower tongue. Figure 4 reports the difference in displacement between the upper and the lower tongue recorded at different process strokes:

Figure 2: Load stroke comparison simulated vs. experimental.

Figure 3: Maximum die deflection in the two profiles.

Δdispl=disp. partially supported – disp. fully supported (1) A difference in deflection of 0,45 mm was found between upper partially supported tongue and bottom fully supported one, and it remained almost constant all along the ram stroke. Almost all the codes are able to properly predict the maximum displacement of the upper tongue (partially supported) while problem arises in the computation of the fully supported one. As a consequence the codes underestimate the deformation and they surprisingly predict the same delta deflection. The discrepancy could arise also on the definition of

Figure 4: Difference between tongues deflection at different extrusion strokes.

Here, the three repetitions showed exactly the same tendency in material flow: upper profile (partially supported) always ran slower than bottom one (fully supported) thus producing shorter profiles. Figure 5 reports the comparison between predicted and measured profile speed for the upper and lower profile: all the

378

codes properly estimate the mean production rates but only the code 6 properly predict that the fully supported profile will produce a longer profile. Such minor discrepancy is related to the complexity of this benchmark for FEM codes, in fact, in order to obtain reliable information, a double simulation should be carried out: in the first one the material flow and pressure on the die has to be computed; the analysis of die deflection should be made then the deformed die has to be used as ‘new tool’ and the simulation has to be run again in order to compute the difference in term of profile lengths and die temperature due to the loss of contact of the material in the bearing zone. Finally, figure 6 reports the profile temperature estimation: experimental data with the confidence interval based on the three repetitions are reported in red. All the codes properly predict the temperature increase related to stroke increment and also the absolute values; just code 4 seems slightly overestimating such measured temperatures. A more detailed description of experimental results and repetitions are available in [7].

joined the 2009 one. A generally high increase of codes accuracy was found in the prediction of process load, material flow and temperature evolution; in particular the prediction of the evolution of the thermal field can be now be considered as acquired knowledge. The codes were tested for the first time also on die deflection estimation: a general agreement with maximum deformation values was found but some more efforts have to be carried in modelling the die material behaviour. Of extraordinary interest is the deep reduction in computational time respect to the 2007 edition, reduction that raise up to 500% for same codes, and that seems promising for a further important reduction in the next years.

ACKNOWLEDGEMENT A special acknowledgement is tribute to Dr. El Mehtedi from Marche Polytechnic University, Italy for material flow stress characterization and to Wefa for die construction. The authors would also acknowledge the Collaborative Research Centers SFB/TR 10, SFB/TR 30 and the extrusion Research group FOR 992 of the German Research Foundation (DFG) and the MIUR (Italian Ministry for Research and Innovations) for supporting the event.

REFERENCES

Figure 5: Simulated and experimental profile speed for p.s. and f.s. profiles.

Figure 6: Simulated and experimental profile temperature (upper profile only).

5 CONCLUSIONS In the paper the current state of the numerical simulation for extrusion process analysis has been exploited by the benchmark: six codes were compared to experimental monitored data obtained by repeated trials; four of codes were present also at 2007 edition and two more codes

[1] Li, L., Zhou, J., Duszczyk, J. (2004) ‘Prediction of temperature evolution during the extrusion of 7075 aluminium alloy at various ram speeds by means of 3D FEM simulation’, Journal of Materials Processing. Technology, Vol. 145, pp. 360-370. [2] L. Donati, L. Tomesani, (2005) “The effect of die design on the production and seam weld quality of extruded aluminum profiles”, Journal of Materials Processing Technology 164-165 pp.1025-1031. [3] T. Kloppenborg, N. Ben Khalifa, A. E. Tekkaya, ‘Accurate Welding Line Prediction in Extrusion Processes’ Proceedings of the Extrusion Workshop and Benchmark, Key Engineering Materials Vol. 424 (2009) pp. 87-94 [4] L. Donati, L. Tomesani, M. Schikorra, A. E. Tekkaya “Extrusion Benchmark 2007 – Benchmark Experiments: Study on Material Flow Extrusion of a Flat Die”, Proceedings of the Extrusion Workshop and Benchmark, Key Engineering Materials Vol. 367 (2008) pp. 1-8 [5] L. Donati, L. Tomesani, M. Schikorra ”Extrusion Workshop and 2nd Extrusion Benchmark in Bologna” Light Metal Age, February 2008, vol. 66 n. 1, pagg. 40-43; [6] J. Sjostrom and J. Bergstrom, "Thermal fatigue in hot-working tools", Scandinavian Journal of Metallurgy vol. 34, (2005), pp. 221-231 [7] D. Pietzka, N. Ben Khalifa, L. Donati, L. Tomesani and A. E. Tekkaya “Extrusion Benchmark 2009Experimental analysis of deflection in extrusion dies”, Proceedings of the Extrusion Workshop and Benchmark, Key Engineering Materials Vol. 424 (2009) pp. 19-26