Available online at www.sciencedirect.com
ScienceDirect
Procedia Engineering 132 (2015) 974 – 981
The Manufacturing Engineering Society International Conference, MESIC 2015
Simplified models for high pressure die casting simulation
E. Angladaa,*, A. Meléndeza, I. Vicarioa, E. Arratibelb, G. Cangasb
a
TECNALIA Industry and Transport Division, Mikeletegi Pasealekua 2, Donostia-San Sebastián E-20009, Spain
b
Fundiciones Inyectadas Alavesas, La Haya 12, Nanclares de la Oca E-01230, Spain
Abstract
The simulation of the High Pressure Die Casting (HPDC) process is a complex type of simulation. The industrial procedure is
based on consecutive manufacturing cycles that must be taken into account in the simulation. Moreover the part geometries use
to be complex and the alloy is injected at really high velocities. All of that usually implies long calculation times that in complex
cases can lead to several days.
Sometimes, the circumstances require to have available a fast solution despite involve a loss of accuracy. The work presented
hereafter discusses different possibilities to simplify the HPDC simulation models together with their benefits and drawbacks.
The simplified simulation models have been validated against a detailed 3D simulation model, previously correlated with
experimental results.
The comparative, shows that the use of simplified models may be a solution that makes possible a big reduction in calculation
times maintaining a reasonable level of accuracy.
© 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2016 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-reviewunder
underresponsibility
responsibility
of the
Scientific
Committee
of MESIC
Peer-review
of the
Scientific
Committee
of MESIC
2015 2015.
Keywords: finite elements; numerical simulation; heat transfer; HPDC; metal casting
1. Introduction
The numerical simulation of metal casting processes is a complex type of simulation. As Bonollo and Odorizzi
explains at reference [1], it typically requires to solve the filling of the cavity governed by the fluid-dynamics laws,
* Corresponding author. Tel.: +34-667-119-553; fax: +34-946-460-900.
E-mail address: eva.anglada@tecnalia.com
1877-7058 © 2015 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Peer-review under responsibility of the Scientific Committee of MESIC 2015
doi:10.1016/j.proeng.2015.12.585
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
the solidification and cooling according to the heat transfer laws and some additional effects related with the
physical metallurgy.
The fluid-dynamics simulation implies to solve the Navier-Stokes equations, see [2], which is always a very
complex calculation. The Fourier equation that governs the heat transfer, see [3], must be also solved. The solution
to both phenomena, fluid-dynamics and heat transfer, must be coupled, see [4]. In addition some physical effects
related with the physical metallurgy are included, typically the determination of the shrinkage defects and in more
advanced calculations, aspects as the microstructure types or the grain sizes as Dantzig and Rappaz explain at
reference [5]. The numerical complexity is increased by the fact that the solution must be calculated for transient
regime and usually complex geometries. This is especially true for the case of the High Pressure Die Casting
(HPDC) simulation, where part geometries are usually considerably complex and the alloy is injected at really high
velocities.
All of that usually imply long calculation times. In complex cases can lead to several days or indeed several
weeks, in spite of the advances experimented during last years in hardware and software technologies.
Despite these difficulties, at present the numerical simulation is widely used by the metal casting industry.
Sometimes the industry needs to predict the metal casting process as reliable as possible. So, it is needed to go to
detailed models able to reproduce accurately the situation at the industrial plant and wait for long time until the
solution is reached. But sometimes a fast solution is required although it implies a loss of accuracy. Typical
situations are related with the industrial necessities, for example during the offer phase or for the quick evaluation of
several alternatives. But there are other situations where this fast response is also interesting, as for example, during
the adjustment process of the simulation models.
The adjustment or correlation of the simulation models is referred to the adaptation of the simulation model to be
able to reproduce appropriately the data collected as reference during several controlled experimental tests. A useful
methodology is the use of inverse techniques. Temperatures at several points are registered during the experimental
manufacturing of a prototype and used later as reference to adjust the modelling. The simulation model is considered
adjusted when temperatures predicted by simulation agree well with those registered experimentally. This type of
methodology has been proved satisfactorily by different authors for different metal casting processes, as are the
cases of references [6–9] for the investment casting process and the references [10–13] for different processes with
metallic molds including HPDC.
One of the drawbacks of this methodology is its iterative nature. The model parameters are modified, the
simulation is executed, results are compared with experimental measurements and the parameters are modified again
until reach the adjustment. The number of simulations that is needed to perform until reaching the adjustment may
be very high, from several tens to several hundreds. So if each model calculation requires several hours or even
several days the mission is almost impossible. In these cases the use of a simplified model able to provide a fast
response is very interesting to perform first approximations. The final adjustment must be performed with a detailed
model able to reproduce the process with accurate enough, but the use of simplified models permits an important
reduction of the time needed to reach the adjustment.
This methodology, the use of a simplified model for first approximations during the adjustment process of a
HPDC, has been successfully used in other work of the authors verifying its applicability, see reference [13].
Although the results of that adjustment are shown in the cited reference, the use of the simplified model is only
succinctly mentioned.
It is felt that some aspects, advantages and limitations, of the use of simplified models are of general interest for
the research community. For this reason the work presented hereafter discusses different possibilities to simplify the
HPDC simulation models, evaluating the associated lack of accuracy versus the velocity improvement.
2. Methodology
The HPDC is a continuous production process based on consecutive manufacturing cycles. At beginning of the
manufacturing process the mold is usually too cold preventing the cavity filling. The alloy freezes partially during
the injection and does not fill completely the cavity. The successive cycles go increasing the mold temperature but
until it is not hot enough, the quality of the manufactured part will not be good, presenting surface defects and/or
internal porosity.
975
976
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
The detailed simulation of the HPDC process requires include the simulation of the preheating cycles until
reaching the mold thermal stabilization. Notice that thermal stabilization does not mean uniform temperature, as a
thermal gradient exists along the mold geometry which also varies during the cycle stages.
Several simplifications related with the inclusion of the preheating cycles, the phenomena to include in
calculation, as well as geometrical simplifications have been evaluated in terms of accuracy and calculation times.
The methodology followed to evaluate the accuracy of the simplified models has been the comparison of their
results against the results obtained from a detailed simulation model. This detailed model used as reference has been
previously correlated with experimental results, see [13]. The software used to perform the simulation is a
commercial finite element software called ProCAST. It is especially focused on metal casting processes simulation
combining several solvers that make possible the resolution of the computational fluid dynamics (CFD) coupled
with the heat transfer problem, including the prediction of shrinkage defects. It includes also additional modules able
to perform for example thermo-mechanical analysis or microstructural predictions. The alloy temperature
considered corresponds to a point located in the center of the transversal section of the cast part. The mold
temperatures have been collected at several points located on the contact face between the fixed and movable plates
of the mold, into and near the cavity.
2.1. Simplifications considered
x Simulation of the pre-heating cycles. A priori the inclusion of the pre-heating cycles in the simulation seems
highly advisable. It not only avoids the necessity to estimate the average temperature for the stabilized mold at
the cycle start but also it provides the temperature distribution along the mold. In other case, it is only possible to
assign a unique value of initial temperature for the whole mold. The number of cycles to simulate must be the
necessary to reach the mold thermal stabilization (typically between 5 and 20 cycles). Of course it implies a big
increment in time calculation as it is equivalent to perform 5 or 20 simulations instead of one. For this reason the
influence of the inclusion of the pre-heating cycles has been studied.
x Thermal and flow or only thermal analysis. To restrict the simulation only to the heat transfer analysis clearly
reduces the information provided by results. In this case no information is obtained about the flow behavior
during filling as velocities or flow paths. But in some cases the thermal results may be enough if they are
representative of the real thermal behavior. As thermal only analysis is usually much faster than the combined
thermal and fluid analysis, the differences between both cases have been analyzed.
x Geometrical simplifications. Finally the influence of the geometrical simplification has been considered. A
pseudo-2D model corresponding to one slice of the mold has been simulated and compared with the detailed
model results. In this case the reduction expected in calculation times is very significant but the question to solve
is its influence in the results accuracy.
2.2. Description of the detailed model
The detailed model used is a 3D finite element model formed by 495.999 tetrahedral elements and 100.593
nodes. The model geometry includes the cavity and the fixed and movable plates of the mold. In this case the cavity
geometry is quite simple (cylindrical shape 50 mm and 250 mm length). The reason is that it does not correspond
to an industrial part but with a prototype specifically designed to be used for the correlation of the simulation
modelling.
The injected alloy corresponds to AlSi9Cu3, widely used in aluminum die casting due to its good castability
(improved by the silicon content) combined with a medium strength (enhanced by the copper) and a good
machinability. The mold is manufactured in steel H13, a tool steel with excellent wear resistance and good thermal
shock resistance very used in manufacturing of die casting molds. The chemical composition of AlSi9Cu3 and H13
are included in Table 1 and Table 2 respectively. The material properties of the mold and alloy, all of them
temperature dependent, have been assigned to the corresponding volumes. The heat transfer coefficients assigned
between them are temperature dependent but also time dependent to take into account the cycle stages (mold plates
977
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
open, closed, etc.). Finally the boundary conditions have been assigned not only to define the cooling conditions but
also to define the alloy injection conditions.
Table 1. Chemical composition % of AlSi9Cu3
Si
Cu
Mn
Mg
Fe
Ni
Pb
Sn
Ti
Zn
Al
8.0-11.0
2.0-3.5
0.1-0.5
0.1-0.5
Max
0.80
Max
0.3
Max
0.2
Max
0.1
Max
0.15
Max
1.2
Rest
Table 2.Chemical composition % of H13 steel
C
Cr
Mo
Si
V
Fe
0.32–0.40
5.13–5.25
1.33–1.4
1.0
1.0
≥90.95
3. Results and discussion
3.1. Simulation of the pre-heating cycles
The time needed to solve only one cycle with the detailed model, considering the problem thermal and flow, is
equal to 6.9 hours. This calculation time has been obtained in a PC with 2 CPUs Intel ® Xeon ® E5335 quad core
@ 2.00 GHz and 8 Gb RAM. Notice that although the PC has 2 quad core CPUs, the simulation software only use
one of these eight cores.
The inclusion in the simulation of 5 or 20 preheating cycles is equivalent to perform 5 or 20 simulations of one
cycle, with the consequent increment in the calculation time. So, seems very interesting to study the influence of
including or not the pre-heating cycles as it could signify an important time reduction (5 to 20 times).
The evolution of the mold temperature during the pre-heating cycles simulation can be observed in Fig. 1. Fig. 2
shows the temperature results obtained including or not the pre-heating cycles in the simulation, where is observed
that the loss of accuracy is low. Nevertheless, the mold initial temperature used for the no pre-heating case has been
estimated based on previous results. In case of a worst estimation of the initial temperatures the differences are
bigger. Fig. 3 shows the results obtained considering different values for the initial temperature of the mold (±
20ºC). As can be observed the differences in results are more pronounced, although the variation applied to the
initial temperatures is only equal to 20ºC.
So, it seems that the advantage of including the pre-heating cycles in the simulation is more related with the
correct estimation of the average mold temperature, than with the mold temperature distribution. In cases where it is
possible to estimate correctly the average temperature of the mold plates interfaces, the pre-heating cycles
simulation may be omitted. Nevertheless, considering the difficulty of estimating this average temperature, it is
advisable to perform the simulation of the pre-heating cycles.
Fig. 1. Mold thermal evolution during pre-heating cycles simulation
978
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
Fig. 2. Temperature results including or not the pre-heating cycles in simulation
Fig. 3. Temperature results for different initial mold temperatures
3.2. Thermal and flow or only thermal analysis
The calculation time needed to solve the thermal and flow problem, that is, the fluid-dynamics and heat transfer
coupled analysis, is much higher than in the case restricted to only heat transfer calculation. The fluid-dynamics and
heat transfer coupled analysis needs 6.9 hours for the simulation of 1 single cycle with the detailed model, in the
previously mentioned PC, while the time needed for the case restricted to only heat transfer calculation is equal to
0.3 hours.
The drastic difference existing in calculation times makes patent the interest of limiting the simulation to the heat
transfer analysis when possible. Fig. 4 contains the temperature results for the case where only heat transfer analysis
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
has been performed and the case including the thermal and flow analysis. As can be observed, the temperature
results are similar and the lack of accuracy that implies the restriction to only thermal analysis is clearly
compensated with the important reduction in calculation times.
Fig. 4. Detailed model. Temperature results for only thermal vs thermal and flow.
To restrict the simulation to only heat transfer analysis is a good option in these cases where the interest is
focused on the thermal distribution and not in the flow behavior during filling (velocities, flow paths, etc.). One of
these cases is the simulation of the pre-heating cycles, as the objective of this analysis is to obtain the stabilized
temperature of the mold.
So, the methodology sometimes applied of limiting the pre-heating cycles simulation to the heat transfer analysis,
seems the best option to maintain the calculation times in reasonable values with a low loss of accuracy. Once the
mold thermal distribution is achieved, is possible to perform the thermal and flow simulation of the injection process
assigning the obtained thermal distribution as the mold initial temperature.
3.3. Geometrical simplifications
The geometrical simplification applied, is based on the concept that the heat transfer across the mold is mainly
two-dimensional. So, only a slice of the mold (see Fig. 5) has been selected to be used in the simplified simulation.
This section, the mold and the cavity have been discretized by means of a finite element mesh. This mesh (see
Fig. 5 right) is formed by 19172 nodes and 79965 tetrahedral elements forming a pseudo-2D model. The material
properties, heat transfer coefficients and boundary conditions applied are the same used in the detailed 3D model.
For people familiar with simulation models is evident that this pseudo-2D model is not completely representative
of the real case. In fact, it is not possible to perform the cavity filling simulation. But it can be useful in these cases
where the interest relies in the heat transfer analysis.
979
980
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
section
Fig. 5. Drawing of the HPDC mold and the section selected for the pseudo-2D model (left). Mesh corresponding to the pseudo-2D model (right)
The differences in calculation times for the detailed model and the pseudo-2D model are very significant (see
Table 3). In this case the lack of accuracy is more pronounced than in previous cases, as can be observed at Fig. 6.
But considering the drastic differences in calculation times, this option may be very advisable to some cases where
the accuracy requirements are not very high or when time restrictions are very important.
Table 3. Calculation times
Analysis type
CPU time (minutes)
Detailed model
1 cycle (heat transfer only)
20 cycles (heat transfer only)
21
425 (7.1 hours)
Pseudo-2D model
1 cycle (heat transfer only)
2
20 cycles (heat transfer only)
43
Fig. 6. Detailed vs pseudo-2D model temperature results (after 20 cycles)
E. Anglada et al. / Procedia Engineering 132 (2015) 974 – 981
4. Conclusions
The simplified simulation models have been able to reproduce appropriately the thermal behavior of the mold
and of the cast part, with significant reduction in calculation times. The model validations have been achieved by
comparison with the results of the detailed model previously correlated, with experimental data.
The inclusion of the pre-heating cycles in the simulation is highly advisable. The main reason is the difficulty to
estimate the mold average temperature in other case. Relating to restrict the simulation to the heat transfer analysis,
is a good alternative when the interest is only focused in thermal behavior. On his behalf, the use of pseudo-2D
models is advisable only in these cases where the accuracy must be sacrificed to reduce calculation times.
In base on the obtained results, the next procedure seems the most advisable option. Execute first a simulation of
the pre-heating cycles but limiting the analysis to the heat transfer. Next execute a second simulation including the
thermal and flow analysis. The mold temperatures obtained as result of the first simulation must be assigned as the
initial temperatures of the mold for this second simulation. In both cases the 3D detailed model must be used.
In the cases where time restrictions are important, the use of pseudo-2D models is a good alternative. If the loss
of accuracy in the results that it implies is acceptable, its use makes possible a drastic time reduction (10 times
compared with detailed one). This alternative has been validated by its use during the preliminary phases of the
adjustment process of a HPDC model, see [13].
Acknowledgments
This work has been developed as part of the UHSAC project (Ultra high strength aluminum cast systems for the
new generation of parts for the automobile industry), funded by the ETORGAI program of the Basque Government
(Department of Industry and Innovation).
References
[1] F. Bonollo, S. Odorizzi, P. Hansen, D.M. Lipinski, M. Schneider, I. Erauskin, I.L. Svensson, M. Wessen, E. Hepp, N. Gramegna. Numerical
simulation of Foundry Processes. Padova (Italy): SGE; 2001.
[2] C. Hirsch. Numerical computation of internal and external flows. Salisbury (UK): John Wiley and Sons; 1997.
[3] A.J. Chapman. Transmisión del calor. Madrid (Spain): Libreria Editorial Bellisco; 1984.
[4] M. Schäfer. Computational engineering. Berlin (Germany): Springer; 2006.
[5] J.A. Dantzig, M. Rappaz. Solidification. Lausanne (Switzerland): CRC Press; 2009.
[6] E. Anglada, A. Melendez, L. Maestro, I. Dominguez. Adjustment of Numerical Simulation Model to the Investment Casting Process.
Procedia Eng. 2013; 63: 75–83. doi:10.1016/j.proeng.2013.08.272.
[7] E. Anglada, A. Melendez, L. Maestro, I. Dominguez. Finite Element Model Correlation of an Investment Casting Process. Mater. Sci. Forum
2014; 797: 105–110. doi:10.4028/www.scientific.net/MSF.797.105.
[8] Y. Dong, K. Bu, Y. Dou, D. Zhang. Determination of interfacial heat-transfer coefficient during investment-casting process of single-crystal
blades. J. Mater. Process. Technol. 2011; 211: 2123–2131. doi:10.1016/j.jmatprotec.2011.07.012.
[9] H. Jin, J. Li, D. Pan. Application of inverse method to estimation of boundary conditions during investment casting simulation. Acta Metall.
Sin. Engl. 2009; 22: 429–434. doi:10.1016/S1006-7191(08)60119-2.
[10] J.-M. Drezet, M. Rappaz, G.-U. Grün, M. Gremaud. Determination of thermophysical properties and boundary conditions of direct chill-cast
aluminum alloys using inverse methods. Metall. Mater. Trans. A 2000; 31 A: 1627–1634. doi:10.1007/s11661-000-0172-5.
[11] A. Long, D. Thornhill, C. Armstrong, D. Watson. Determination of the heat transfer coefficient at the metal-die interface for high pressure
die cast AlSi9Cu3Fe. Appl. Therm. Eng. 2011; 31: 3996–4006. doi:10.1016/j.applthermaleng.2011.07.052.
[12] G.-X. Wang, E.F. Mathys. Experimental determination of the interfacial heat transfer during cooling and solidification of molten metal
droplets impacting on a metallic substrate: effect of roughness and superheat. Int. J. Heat Mass Transfer. 2002; 45: 4967–4981.
doi:10.1016/S0017-9310(02)00199-0.
[13] E. Anglada, A. Meléndez, I. Vicario, E. Arratibel, I. Aguillo. Adjustment of a high pressure die casting simulation model against
experimental data. In: M.A. Rosendo, G. Puig, I. Buj Corral, J. Minguella Canela, editors, The 6th Manufacturing Engineering Society
International Conference, MESIC 2015. Procedia Eng. In press.
981