Mold-filling simulation has never worked equally well with all parts. For a relatively thin, shell-like part with an easily defined midplane, any of several simulation software packages can predict how it will fill with acceptable accuracy. Try the same with a thick part or one without an obvious midplane, and commercial filling-analysis software will likely return less accurate results, a shortcoming conceded by all the leading simulation vendors. "The more you get away from thin, sheet-metal-like components, the less value simulation has traditionally had," acknowledges C-Mold v.p. Peter Medina.
These simulation-resistant parts represent only a small fraction of molding jobs--"probably less than 10%," estimates Moldflow v.p. Ken Welch. Yet this fraction encompasses many high-value, high-precision parts--electrical connectors being a prominent example. The breakthrough came at the end of last year with the introduction of commercial software for "true 3D" mold-filling simulation. "3D simulation solves classes of problems heretofore unsolvable," notes Welch.
Unlike midplane--or so-called "21/2 D"--representations that currently prevail in finite-element mold-filling analysis, "true 3D" simulations employ a mesh of volumetric elements as the basis for their calculations. Three such products are now available to any molder with a high-end personal computer or CAD workstation: From Moldflow comes MF/Flow3D. Plastics & Computer has introduced faSolid. And C-Mold has an as-yet unnamed package that it is releasing only to members of a user consortium.
A fourth 3D simulation package has been developed by the Industrial Materials Institute of Canada's National Research Council. But it presently runs only on expensive computers with arrays of parallel-processors. While the software isn't directly marketed to molders, it is available on a consulting basis, according to its developer, Georges Salloum.
Despite its promise for some types of problems, 3D analysis won't replace midplane-based simulation for the majority of molding jobs. "There's a lot of 3D euphoria right now, but it's not for everybody," says C-Mold's Medina. "The market needs to be educated about where 3D fits in and where it doesn't."
3D for the PC
One important aspect of the new 3D simulation products is their relative accessibility in terms of computer requirements, model preparation, "solve times," and cost.
The computing limitations of years past posed an effectively insurmountable barrier to true 3D analysis. According to Giorgio Bertacchi, co-founder of Plastics & Computer, a run-of-the-mill molding simulation in 3D formerly required a supercomputer and hours of computation time. Medina agrees: "We were in a position of waiting for computing power to catch up." Now it has, and all three commercial 3D packages run on a high-end PC or CAD workstation. "To make the technology viable, you had to be able do this with a PC," Medina says.
3D analysis times can vary widely, depending on the complexity of the model. For example, Bertacchi says a 3D analysis takes anywhere from 15 minutes to a few hours to set up and solve. Moldflow customer-service engineer Peter Rucinski reports that a relatively simple part--a thick valve fitting--required 50,000 elements to model and was solved in less than an hour on a Pentium PC with a couple of hundred megabytes of RAM. An electrical connector with about 440,000 elements took several hours to solve. "The amount of time is related to the number of features and the number of transitions between thick and thin sections," Rucinski says.
Making the mesh
All three of these packages have provisions for quickly generating a 3D mesh directly from a solid CAD model. However, they differ from earlier flow-analysis products--such as Moldflow's Part Adviser or C-Mold's 3D Quickfill--that also work directly from a CAD solids model and do away with manual meshing tasks. These earlier packages are not the same as true 3D, says Bertacchi. He explains that previous solids-based simulations employ essentially-conventional 21/2 D analysis techniques and simply keep the midplane modeling hidden from the user.
Beyond the shared capability to work from a solid model, differences between the new 3D simulation products start to emerge. Here's a closer look at the three products:
MF/Flow3D, which uses tetrahedral volumetric elements, relies on Patran or SDRC I-DEAS Master Series to provide pre- and post-processing capabilities for mesh generation and results presentation. (Moldflow plans to add its own pre- and post-processing environment later.) Moldflow's Welch says the software has an "adaptive mesh" feature that eliminates the distortion of elements as the mesh is applied to the CAD model. Adaptive meshing cuts down on the need for mesh repair by the user. Right now, MF/Flow3D provides only filling and packing analysis and thus has outputs such as fill time, velocity, density, pressure, and viscosity distribution. Welch says the company will add warpage analysis to future versions. According to Moldflow CEO Marc Dulude, the software costs between $15,000 and $18,000.
C-Mold, too, uses tetrahedral elements and the commercial pre- and post-processing capabilities of Patran and similar software. Its current implementation provides only filling and packing analysis, but C-Mold's Medina says the company will extend its 3D technology to shrink and warp analysis and will have a version suitable for thermosets--a requirement for chip packaging, which is a potentially important application for 3D simulation.
At press time, Medina reported that the non-isothermal version of its 3D code was just being completed, though a prototype isothermal version was already in use. Because Medina believes that 3D simulation will require more user support and feedback than its other simulation products, the company is initially releasing its 3D software to the members of a user consortium. "Although we could release a working code tomorrow, we don't believe 3D simulation is a production technology yet," he explains. The cost of joining the group will be $15,000/yr for two years. That sum includes a working copy of C-Mold's 3D software. Molders interested in joining the group can contact Medina at the company's Louisville, Ky., offices.
Plastics & Computer tackles 3D problems differently. For one thing, faSolid uses brick-shaped hexahedral elements. It also has its own pre- and post-processing capabilities. 3D faSolid starts with a CAD solid model in .STL format, in which the model is represented in slices. You decide how many planes to slice the model into, Bertacchi says. The brick elements are applied to these planes with the thickness of the plane becoming the thickness of the elements. Because fewer planes add up to fewer elements, users can thus balance their accuracy requirements against modeling complexity. Bertacchi adds that faSolid has built-in provisions to "relimit" the mesh--that is, "push back" the brick-shaped elements where needed to make the final mesh conform to the shape of the part.
Aside from filling and packing analysis, faSolid offers cooling analysis as well. Bertacchi says the product will soon be extended to include warpage. Price for faSolid is $25,000, and the company offers 10 free benchmarking studies to customers willing to share validation data.
Given the strides made by conventional flow simulation in accuracy and ease of use, it's likely that 3D simulation will be used only on parts where a midplane representation won't work. In fact, on thin parts, 3D might even be considered an inefficient solution. "You end up with an excessive number of elements because the 3D mesh places many elements across the thickness of the part, says Leonid Antanovskii, a C-Mold research engineer.
A few oddly shaped thin parts might be suitable for 3D analysis because they have no obvious midplane. But vendors say it is mainly thick parts that are not well represented by a midplane. Coming up with a precise definition of "thick" may take time and experience using 3D software, but C-Mold's Medina offers a rough rule of thumb: Parts with an aspect ratio greater than 1:4 (thickness to length) are likely candidates for 3D simulation. Yet Medina notes that even some thick parts will continue to be well-served by a midplane model if they are long enough to escape excessive sidewall effects from heat and friction. He notes that simulation by 21/2 D considers only the top and bottom surfaces of the cavity as sites where heat conductivity and drag forces occur. "If a thick part is long enough, sidewall impact is minimal," Medina says.
Thickness considerations don't apply just to the overall part but also to sections of the part. In fact, vendors point out that parts with transitions between thin and thick geometry should be considered prime candidates for 3D simulation. "Junction loss from thick to thin sections is an issue. When it's repeated over several transitions, the accumulated error can be significant," explains C-Mold's Antanovskii.
Of all the parts that may be deemed "too thick" for conventional analysis, electrical connectors are the example cited most often by vendors, who report considerable interest in 3D flow analysis from that industry. A number of other parts will fall into this category as well. Bertacchi believes that many micro-molded parts exhibit an aspect ratio suitable for 3D simulation. "For some of these parts, it's the only way," he says.
3D simulation may help another class of parts--some excessively thick and some not. "Wherever you have geometry where the flow changes direction, 21/2 D will force it to take a path that might not be correct. The prediction of properties will be off. So you might see higher precision with 3D and a better prediction of properties," says Antanovskii. He cites threaded parts as an example. "Parts with threads have been challenging in 21/2 D because it assumes flow only in the midplane direction," he says.