"Making A Mesh of Things," Feature Article, September 2003

making a mesh of things

It's easier than ever to overlay a CAD part with the mesh necessary for design analysis. That's good. And bad.

by Jean Thilmany,
Associate Editor

Finite element analysis software continues to morph, if not quite at light speed, then pretty darn fast. As FEA goes, so goes pre-processing. Pre-processing is the generation of the many-noded mesh that overlies a digital part and is the necessary prelude to analysis.

Processing capabilities are keeping pace with changes in FEA in increasing ease of use. But sometimes engineers trade simpler meshing and FEA applications for limited analysis capability.

From its inception in the 1940s until about a decade ago, FEA had been performed exclusively by specialized analysts who held Ph.D.s in the subject and had devoted their careers to the discipline. But the FEA field has seen great change over the past 10 years, with a jump in the number of computer technologies available to all levels and types of engineers, according to Bruce Jenkins, vice president of Daratech, a Cambridge, Mass., marketing and research firm.

FEA is the use of a complex system of points—called nodes—that form a grid, or mesh, across a model. The engineer assigns nodes at a particular density throughout the material, depending on many considerations, including the anticipated stress levels of a certain area and the detail wanted in the results.

The mesh contains the data on material and structural properties that define how the part will react to certain load conditions. In essence, finite element analysis is a numerical method used to solve engineering problems that involve stress analysis, heat transfer, electromagnetism, and fluid flow. If the analysis finds fault with a design, it is sent back for remodeling to correct those faults before a prototype is produced.

Pre-processing mesh technologies can now automatically generate the mesh necessary for analysis. Like most things, this is both good and bad. It means that expert analysts need not be involved in simple analysis, but it also means the software design engineers can run analyses without fully understanding FEA methodology, which could affect design.

And FEA is changing all the time. New developments include the stepped-up use of an actuator element that simulates the motion of hydraulic, pneumatic, and electric cylinders. This new element helps designers analyze complex mechanical events that would have been nearly impossible to analyze only a year or so ago, according to Bob Williams, development manager at analysis-software maker Algor Inc. of Pittsburgh. Algor and other FEA software developers continue to improve solver speed in order to reduce analysis time.

"They have to," Williams said. "The ever-increasing memory and disk space capacities on today's computers encourage finite element analysts to work with larger and larger models."

Makers of mesh technologies must keep up-to-date with stepped-up meshing times of their own. Frequently, third-party vendors make these pre-processing applications and FEA vendors incorporate the technologies into their applications. Due to the upgrades in meshing technologies, engineers can now create meshes directly from their computer-aided design geometry.

Today's analysts and engineers don't have to build separate, meshed geometric models based on their CAD files in order to run FEA analyses. When they need to run complex, high-end analyses, however, analysts can still choose to build their own file for mesh generation.

Today's closely integrated CAD, pre-processing, and FEA applications allow CAD and entry-level FEA technologies to work together within a common user interface and give design engineers a quick, easy way to see if their designs will meet specifications, Williams said. But often, even using the integrated technologies, mechanical engineers still have to transfer CAD geometry into an open file so it can be read by the pre-processing application and then by the FEA application. The translation process remains clunky, and vital information about part geometry can be lost in the translation, Williams said.

And integrating the technologies can cause other problems, he admits. Simplifying the FEA programs so design engineers—who aren't formally trained analysts—can use them limits the intricacy of the mesh as well as the depth of analysis. The number of pre-processing, or meshing, elements that engineers have to choose from is significantly reduced in these integrated products. And it limits the types of analyses that designers can run.

"Engineering analysts realized that the lack of element types, such as beam, truss, or brick elements, in a simple product kept them from analyzing even moderately complex parts of mechanisms," Williams said.

Models need to be meshed and made acceptable for analysis before FEA can be run. Software providers that make pre-processing applications must keep up with changes in FEA technology to remain competitive.

Some FEA companies find a better approach is separate CAD and FEA programs, while still allowing the direct transfer of geometry from CAD to FEA. Communication with the CAD solid modeler remains direct, but upgrades and expansions to the analysis program are easier that way.

Such programs are still compatible with CAD, they're just not as tightly integrated, so FEA meshes can be more complex with no loss of geometry information. An example is InCAD from Algor, which works directly with a number of major CAD packages and uses the same interface as the CAD system regardless of the type of analysis that is run.

No one can dispute that the mesh creation now happens faster than ever before.

"The job of creating grid points on a CAD file has gone from months to weeks to days to minutes," said Mike Kidder, corporate marketing director at Altair Engineering Inc. of Troy, Mich. His company makes HyperMesh, a pre- and post-processor for FEA. "That means the engineer has more time to do analysis than ever before."

Pre-processing jobs that took two hours only a few years ago are now commonly done in 12 minutes, Kidder added.

The Procter & Gamble Co. of Cincinnati has been using HyperMesh for just over two years for package analysis, said David Henning, section head in the package and device computer-aided engineering department. Henning's department analyzes product packaging—boxes and bottles for everything from health and beauty care products to pet food—to ensure that it protects the contents. Simulations extend to packaging damaged by a fall or a puncture.

With pre-processing software called HyperMorph from Altair Engineering, an engineer can interactively pull and morph the wing mesh of this airplane model without needing to change the CAD geometry of the original model.

HyperMesh prepares CAD geometries for analysis. The meshed geometries are then exported to Procter & Gamble's customized package analysis system, called Virtual Package Simulation. After Henning's department brought HyperMesh on board, structural analysts in other departments started using it because it prepares meshes for two of the main FEA programs the company uses: Abaqus and LS-Dyna. Abaqus is made by Abaqus Inc. of Warwick, R.I., and LS-Dyna by Ansys Inc. of Canonsburg, Pa.

The pre-processing software allows engineers to import geometry from many CAD sources for meshing, Henning said. It also helps standardize analysis.

"We now have an environment for creating targeted analysis in a very standard way," he said. "It's provided a clear process for getting geometry in and analysis out. All the analysis is created in the same fashion so we can rely on a standard analysis in which to compare bottle concepts. The analysis is set up the same way each time, so direct comparison between shapes is possible."

MESHING MOVEMENT

Today, engineers use mesh technologies and attendant FEA programs for an array of analyses. Some are related to manufacturing, but as often as not they've found their way into other industries. The Bioengineering Center of Wayne State University in Detroit, for instance, has been doing research in impact biomechanics and automotive safety for more than 60 years. The researchers hope to complement, maybe even replace, crash test dummies with complex models of the human body that will show what particular types of impact will do to different body types.

"The models can be used to better understand the injury mechanisms during an automotive impact and help to design countermeasures," said Philippe Beillas, who works on the project.

The human-body models today take the appearance of a series of body parts covered with a computerized FEA mesh. The researchers are also using them to study how the human skeleton moves. For pre-processing, the researchers use the ICEM CFD Hexa technology from ICEM CFD Engineering of Berkeley, Calif. They use Ansys software for analysis.

"The anatomical complexity and the irregularity of the shapes we mesh always make meshing a critical task in the development of our models," Beillas said.

The Center for Aerospace Structures at the University of Colorado in Boulder is also using meshing and analysis technologies from the same companies, but in a completely different way. The center's researchers aim to improve the technology of aerospace structures. They perform fluid-structure interaction analyses on models of planes and on parts of planes, though their analyses emphasize the aeroelasticity of complete aircraft. Researchers also perform flutter, vibration, and maneuver simulations on airplanes, said Philippe Geuzaine, a professor at the center.

The Center for Aerospace Structures uses meshing technologies and the analysis software to study airflow, which is a computational fluid dynamic analysis. Engineers frequently run analyses on computational fluid dynamic, or CFD, models to delineate the flow of fluid or other fluid-like substances through a structure. For instance, Airbus S.A.S., the maker of commercial airplanes in Blagnac, France, uses CFD analysis as a low-cost complement to classical wind tunnel testing, according to the company.

Whether engineers are analyzing a model of a wing or analyzing the complete airplane, CFD can produce reliable lift and drag data, as well as a detailed model of the airflow around or through a part. The analysis information helps engineers understand relationships between the shape and the aerodynamic performance of a part.

Like FEA technologies, CFD programs are available for the nonexpert as well as for the Ph.D. analyst. Also like FEA, the programs differ in the way users bring a solid model into a CFD analysis program. Engineers can perform a file-format translation, where they translate a CAD file into a standard language like initial graphics exchange specification or the standard for the exchange of product-model data. Or they can perform a direct-model transfer in which the CAD model is brought in to the CFD application without translation.

Finite element analysis, pre-processing, and post-processing software combine for easy evaluation of particular situations, such as this simulation of a rail being crushed. The simulation uses HyperMesh processing software and FEA software from Abaqus.

Pointwise of Bedford, Texas, which makes the CFD grid-generation program Gridgen, uses the format standard IGES to transfer geometry from a CAD system into a CFD analysis program. Essentially, the company's technology translates CAD geometry into IGES, meshes the model, and transfers it to a CFD program. It also performs a similar geometry transfer when analysts move files between different computer systems—when files are moved from a Windows to a Unix system, for example.

Although translation errors can occur, they're often caused by errors in the original CAD geometry and aren't introduced during the transfer process, said Rick Matus of Pointwise.

No matter how the CAD files are read into the CFD system, fluid analyses—like structural analyses—can run the gamut. For example, researchers at Brown University in Providence, R.I., used Gridgen to simulate and visualize airflow around a bat (the kind that flies). The researchers hope to more clearly comprehend the aerodynamics of bat flight and put those discoveries to use in better understanding certain principles of biomechanics, aerodynamics, and evolutionary biology. Researchers on the project are Rachel Weinstein, Igor Pivkin, S. Swartz, David H. Laidlaw, George Karniadakis, and K. Breuer.

But before they could simulate bat flight, the researchers needed to determine the shape of the bat in flight, which wasn't easy. You can't really see a bat's flying shape just by looking. When bats fly, their wings undergo large motions and deformations, according to the researchers. Plus, they fly fast, which makes taking measurements difficult. To get around the problem, researchers used small-bodied bats called Rhinolophus megaphyllus that weigh from three to five grams. They flew more than 20 bats through wind tunnels, capturing data of each bat's shape in flight with high-speed digital cameras. Software transformed the camera data into three-dimensional coordinates.

Using the coordinates, the researchers created geometric models for each of the 160 poses a bat holds during a single wingbeat. They used the Gridgen meshing technology to make a mesh of the air around the bat for each pose. The meshes create a time-lapse simulation of the airflow for the entire beat of a bat's wing. The researchers then imported the mesh into their own, homegrown CFD program to model the bats' static poses. They visualized the results, in order to understand them better and make them appear realistic. For this, they used a virtual reality environment called a computer-assisted virtual environment, or CAVE.

By plotting airflow, analyzing it, and visualizing the results, they found several interesting airflow features, including small vortices emanating from the trailing edges of the bat's wing. They'll be running more bat flight simulations in the future.

Because they can be used for an array of analyses, meshing technologies continue to constantly change FEA and CFD technologies, morphing—just like the mesh itself—in the face of user demands and needs.