Latest from Molds/Cores

Franklin Precision Castings
Wisconsin Aluminum Foundry
Ryobi Die Casting
GF Casting Solutions
Messe Dusseldorf
Thongchaipeun | Dreamstime

Predicting Castability for Thin-Walled HPDC Parts

Dec. 15, 2014
Current simulation tools account for the physics of fluidity, and should be capable of predicting castability too. Investigatng thermal resistance Shot sleeve heat loss Effect of conventional HTC values

The term "thin-wall castings" for the HPDC process has been investigated for well over two decades.  However, the perception of what is considered a thin-walled part has changed over the years and continues to change.  Historically, 3-mm wall thickness was considered a thin HPDC part and that number decreased over the years to 2 mm, and further to 1 mm. Currently, there is a push for even thinner walled parts in the electronics industry, with thickness demands as low as 0.6 mm.  The automotive industry has been looking to make ultra large castings (ULC) using various processes, including semi-solid, permanent mold, and HPDC processes, with target thickness in the range of 1 to 2 mm[2].

Thin-walled parts are difficult to cast because the melt cools rapidly upon contact with the relatively cold die steel and can solidify quickly before die filling is complete.  The distance a given molten material travels before it freezes and stops moving is commonly referred to as "fluidity".  The dominant variables affecting fluidity are:  thermophysical properties of the melt; the temperature of the melt above liquidus (superheat); and mold coating[3,4].  Much of this work on quantifying fluidity was based on experiments under low pressures, generally not exceeding 15 psi to force the liquid metal through a passage. 

Further, die temperature and the heat-transfer coefficient at the die surface do not seem to be considered properties that define fluidity, though Dewhirst[4] acknowledged that the heat-transfer coefficient at the mold surface does play a significant role in the measured flow lengths for a given alloy type under given test conditions.

The focus of this work is to determine if computer models can realistically predict the castability of a given ultra thin-walled part and to test the sensitivity of dominant variables that affect this castability prediction.  We confined our investigation to one cast material (A383) in a die made of H13 steel, for the HPDC process.  Our goal is to determine if we can use simulation at the concept phase of product design to specify the foundry settings necessary to successfully cast that given product.  The variables accounted for in this study are: superheat; die temperature; fill time; and die surface condition, including any "contaminants."

Thermal resistance melt/die interface. When the melt suddenly contacts the relatively cold die surface, its initial rate of heat loss is strongly controlled by this "thermal contact resistance" (Rc) at the microscopic interface that separates the melt from the die.  The value of this thermal resistance depends on the die surface roughness, die lube deposits, oxide layer on the die and oxide film on the liquid metal itself[5,6].  The die casting industry commonly refers to a heat-transfer coefficient (HTC) instead of a thermal resistance (Rc).  This HTC is simply the inverse of thermal resistance (HTC = 1/ Rc). 

Since we are exploring heat loss during filling, we are concerned with the insulating characteristics at the die surface for time scales ranging from (0.001s to 0.1s).  The perception of what the heat-transfer coefficient is at the die surface when contacted by the melt varies significantly in HPDC, with values seen to range from ~5,000 W/m2K to ~120,000 W/m2K.  We are careful to select this value as accurately as possible because this HTC strongly affects the prediction of the heat loss during fast shot. 

Measuring the contact resistance of the melt suddenly contacting the die surface is a challenge because the surface temperature sensor must have a response time of less than 1/10th the timescale that we wish to know this value for [6].  So if we are concerned with knowing this value during die filling, we must have sensors with response times in the sub-millisecond range. 

A detailed investigation of the influence of surface roughness, die lube deposit, and melt oxide was conducted by Y. Hichal for A383 alloy.  We selected measurements from Hichal's work because he used a surface temperature sensor with a response time of 40 nanoseconds!  This sensor made it possible to determine the thermal resistance during millisecond timescales that occurs in the die casting process.

Thermal resistance selected for this investigation. Figure 1 plots thermal-resistance measurements for different surface conditions[6].  The baseline resistance is that of a smooth H13 surface polished to a mirror finish; this corresponds to a surface roughness of 0.5 µm.  The measurement for this baseline value is the thermal resistance measured under an inert atmosphere with no oxidation on the melt or the steel surface.  The other three values are measurements of the additional effect on thermal resistance: surface roughness (3 µm; typical die casting die surface finish), die lube deposit, and oxide formation on the melt surface.  We calculated the percent increase for each additional component of Rc and summed them up to arrive at a thermal resistance of Rc= 2.4 x 10-6 m2K/W. We took this to represent the typical condition for the HPDC process.  This value corresponds to a HTC of 417,000 W/m2K.

What Was Simulated?

In this study we selected the geometry of the housing of a laptop with an average thickness of 0.8 mm, cast in A383 alloy.  This serves as a typical geometry with the challenges seen in the diecasting industry for electronic housings (Figure 2.)  Kim et al.[1] selected this geometry to analyze, and focused their work on gating optimization with trials on the diecasting machine to determine the process parameters that made good castings.  Their foundry trials determined the shot speed that began to make a good casting (for a given gate design).  We compared our simulations with their foundry experience so as to test the validity of our boundary condition.

Shot sleeve heat loss.  We simulated the pouring of the melt into the shot sleeve (3.0 sec pour time) with the same settings selected by Kim et. al., along with the slow and fast shot using FLOW-3D® CFD software.  Figure 3 shows the temperature loss from pouring, along with the slow and fast shot stages.  We determined that the actual metal temperature entering the biscuit is in the range of 595°C; a total temperature drop of ~75°C. 

Heat loss during fast shot.  Results of the pouring and slow shot are taken from above and the simulation is allowed to continue with fast shot starting well before the melt reaches the gates.  Figure 4 displays the results for the filling of the part (fast shot well before melt hits the gate, furnace temperature 670°C, 0.35m/s slow shot, 2.2 m/s fast shot).  This clearly shows that we cannot make a good part with this shot speed.  There are numerous areas with visible defects due to solidification well before the part is completely filled.  Kim et al. found that they could not make a good casting at this shot speed and therefore our results agreed with their foundry findings. 

We then repeated the above simulation with a fast shot of 4.5 m/s starting before the melt arrives at the gate, with all other settings remaining the same (Figure 5).  We still had splashes during the shot, but there was not enough time for these splashes to solidify before the die was completely filled.  These results predict that we will make a good casting at this setting. 

Effect of Fast Shot Position

The previous simulation was done with a fast shot of 4.5 m/s, which started well before the runner was filled.  We repeated this simulation, but with the fast shot starting when the runner is nearly filled. This is common practice in diecasting, where it is assumed that the runner fills completely before the part begins to fill.  Figure 6 shows the fill sequence with this setting.

We can see that because of the shape of the runner system and the gate inlet design, the melt enters the cavity geometry before the runner is filled.  This may seem like a very minor pre-fill condition, but it was enough to cause significant heat loss from the melt that first entered the cavity and ultimately resulted in very visible defects in the casting.

Effect of Using Conventional HTC Values.  It is known that a lower heat-transfer coefficient equates to a more insulating die surface.  We also know that Kim et al. did not make a good casting for this gating design with a shot speed of 2.2 m/s.  We decided to test the validity of our claims testing the highest HTC value seen in the industry (120,000 W/m2K).  We wish to test the ability of this boundary condition  to predict the no-fill defects experienced in the foundry with this 2.2 m/s fast shot speed.

Figure 7 shows the fill sequence for this lower HTC.  The simulation predicted that we will be able to cast this part successfully with no area in the part having cold-low defects.  This contradicts the foundry experience and therefore calls for a higher HTC in order to predict the no-fill defects that were observed.


This work shows that simulation can explicitly predict the castability of thin-walled parts, and the type of defects that will form.  By using a HTC of 417,000 W/m2K we were able to specify the shot speed, the furnace temperature, and die temperature required to successfully make this casting because:

a)  We correctly predicted that we cannot make a good part with a shot speed of 2.2 m/s;

b)  We correctly predicted that we will still produce scrap castings at 4.5 m/s unless we start the fast shot very early;

c)  When we used the highest HTC value found in the die casting industry, we still failed to predict no-fill defects found at 2.2 m/s in the foundry trials.

We also concluded that in order to predict castability with this level of certainty, we must simulate all stages of the casting process so that heat loss is quantified at each of these stages.

Rabi Bhola is the president of Bholster Technologies, a Toronto research and consulting agency that  provides its clients expertise in casting simulation using FLOW-3D® product and process development. Contact him at [email protected], or visit  

Prof. S. Chandra holds the Wallace G. Chalmers Chair of Engineering Design at the University of Toronto, in Ontario. Contact him at [email protected]


1.  Young-Chan Kim, et al, “Die Casting Mold Design of the Thin-walled Aluminum Case by Computational Solidification Simulation”, Journal of Material Science & Technology, Vol. 24 No.3, 2008
2.  Maj, Michael. H., “Ultra-Large Castings of Aluminum and Magnesium”, FY 2005 Progress Report, Automotive Lightweighting Materials.
3.  Di Sabatino M., “Fluidity of Aluminum Foundry Alloys”, PhD Thesis, Norwegian University of Science and Technology, 2005.
4.  Dewhirst, B.A., “Castability Control in Metal Casting via Fluidity Measures: Application of Error Analysis to Variations in Fluidity Testing”, PhD Thesis, WPI, 2008.
5.  Bhola, R. and S. Chandra* (1997). “Deformation of molten metal droplets during impact on a solid substrate”, Proceedings of the 10th Annual Conference on Liquid Atomization and Spray Systems, PP 160- 164, Ottawa. Ontario.
6.  Heichal, Y., “Measuring Thermal Contact Resistance Under an Impacting Droplet of Molten Metal”, MASc. Thesis, University of Toronto, 2005.
7. FLOW-3D User Manual, © Flow Science Inc. 2014