Helping design and engineering professionals discover, evaluate and specify technologies and processes that shorten the design cycle and enable success.
May 2026 Special Focus: Artificial Intelligence in Design and Simulation
In this Special Focus Issue, learn about the latest developments in the integration of artificial intelligence into engineering workflows.
Design & Simulation Software Guide 2025
In this Special Issue, Digital Engineering presents its second annual guide to design and simulation software vendors.
Satisfying customers is job one for a consultant. Seeing the positive results of your work at the supermarket is the unexpected bonus. That’s what happened to Matt Hyre, a Ph.D. in computational modeling and a consultant for Emhart Glass in Enfield, CT, after he modeled the glass-making process for a major spaghetti-sauce manufacturer.
The ProblemAlmost 30 percent of the bottles made by the manufacturer were plagued by "belly bands," a defective rippling in the glass around the middle of the bottle. The company asked Hyre to determine the source of the problem. To do so, Hyre modeled the process using a combination of traditional CFD tools and advanced 3D visualization software.
The Solution
Hyre started the project by requesting the CAD models for the glass-forming operation, including the forehearth or cooling equipment; the feeder that drops the liquefied gobs of glass into the mold; the mold itself; the plunger; and the preform or parison, which expands inside the mold to produce the bottle. The models, created with Pro/Engineer, were exported in STEP and assembled into files for CFD analysis.
"The geometry requirements for CAD and CFD models are different," says Hyre. "In a CAD model, if there’s a gap so small that you can’t see it with the human eye, it doesn’t matter. But in the CFD world, that gap will kill you. You won’t be able to run the model because you have a tiny gap between lines in the geometry."
Before Hyre could finish the forming simulation of his problematic spaghetti-sauce bottle, he had to determine how much heat was going out of the glass and into the mold. He needed to know the mold temperature, so he created another model in FLUENT to show the heat removal from the mold itself. He then coupled the forming simulation in POLYFLOW with the simulation of the heat removal from the mold in FLUENT so that boundary conditions could be passed between the two.
"You don’t know what the mold temperature is, so you pick a reasonable value," he says. "You run your simulation and get a flux profile over time from the glass to the mold, based on what you assumed. You then take that flux profile and apply it to where the glass would be in your FLUENT model, and run that model over and over until you get a mold temperature distribution.
"You then go back and start your POLYFLOW simulation again, using the mold temperature distribution, and calculate a new flux between the glass and the mold. You’ll get a different flux profile, which you put back into the FLUENT model. You do this until the temperature in the mold doesn’t change much."
Hyre admits it’s a complicated and lengthy process—lasting from three to five days—but at the end he not only knows the glass temperature, but the temperature of all the equipment around the glass. That’s an advantage because he has the necessary data should he need to examine problems with mold cooling, size of the mold, or the mold temperature distribution.
The simulation generated about 70,000 time steps, each of which was analyzed to see what happened to the bottle over the length of the forming process. The time steps represented roughly 100GB of data. Some of those were graphics files for postprocessing, but most were boundary conditions and the geometries needed for subsequent time steps.
Hyre created a file to define what he wanted to process and analyze—the geometry, temperature, viscosity, pressure—a process that took about half an hour. He then imported the files into CEI’s EnSight to take advantage of the software’s high-end visualization capabilities. EnSight read Hyre’s definition file and retrieved the data in less than an hour.
