Machine Design

FEA For The Shop Floor

Finite-element programs can show how molten-metal flows into molds, predict relaxation rates for residual strain in welded parts, and calculate springback in sheet metal.

An actual part on the left shows a few surface flaws that were predicted by Flow3D in the two right-most images. Red indicates flaws in the middle image, and blue shows flaws on the mesh. Predicted locations for porosity are in good agreement with observations.

MSC.Marc and MSC.Dytran, nonlinear FEA programs developed by MSC.Software Corp., analyze manufacturing processes such as die casting, welding, bulk forming, profile rolling, hydroforming, and stretchforming. Stretchforming and hydroforming can be simulated to accurately predict spring-back before building a die. In stretchforming, panels are stretched over a mandrel and then released, whereupon they spring back. These simulations allow properly forming panels once so a final shape matches the design without rework.

The material and drilling model worked well for two of the test feed rates. The values are average height because real world operations generate scatter about the calculated heights. At lower rotational speeds, the burr forms more by the tool catching and tearing material, creating a larger burr in a process not captured in the current model. These so-called unstable burrs are randomly located about the hole and will be the focus of later studies.

The drawing shows intermediate and finishing positions of a drill as it turns through a five-mm thick aluminum plate. Burrs are the last bit of material not quite removed by the drill. Some production applications give workers no way to remove burrs, making it more important to avoid generating them.

Saunders model calculates the geometry and stresses that generate burrs. Later studies might experiment with different drill geometries and process conditions to minimize burr height.

The Algor FEA model shows tube and membrane components that Scandroli analyzed. Welds form the membranes between the tubes. Coal-fired boilers measure some 40 X 100 ft. The boiler runs at 800°F with an internal pressure of 4,200 psi.

The tubes form what's called a waterwall, as shown in this Algor model. Compounds in the coal create corrosive conditions that erode the welds between the tubes.

The results from HyperForm from Altair Engineering show that the trunk-lid support will probably wrinkle, as indicated by the blue areas. The software also provides the more traditional forminglimit diagram. The software includes direct CAD readers for most geometry data including Catia and Unigraphics.

Trial-and-error optimization of any manufacturing operation is fine — if you have unlimited time and money. But with design cycles shrinking and manufacturers forced to get presses and die casters up and running without weeks of fine-tuning, everyone is on a strict schedule and tight budget. Fortunately, software companies are developing new types of predictive technology that guide machine-tool setups for several operations.

A few manufacturers are reporting as much as 33 to 50% of the total cost of manufacturing per part can be eliminated with good process prediction and an accurate shape of the preform, which then makes parts with less excess material. What's more, shorter development periods call for finding the right tool shape without lengthy trial-and-error periods.

For example, scientists with MSC.Software, Los Angeles, have combined features in the company's MSC.Marc, a general-purpose nonlinear program, and MSC.Dytran, software for modeling short-duration events, to produce a package that predicts springback in hydroformed sheets. Fluid-flow software lets engineers see inside die-casting molds. In addition, researchers at Penn State Erie, one of many universities conducting manufacturing research, are studying the dynamics of drilling.

Software that simulates injecting molten metal at high pressures into a die-cast mold lets engineers at Visteon, a supplier to Ford Motor Co., "look" inside the tool as it fills. The software shows inclusions, predicts porosity, and helps size and locate overflow reservoirs commonly called pads. Engineers can try different gate and processing conditions to come up with the best-performing die.

"Finding the best die design can take several months of expensive trial and error," says Brad Guthrie, a technical specialist with Visteon. "Now it's done in a matter of days with software," he adds.

Visteon engineers have conducted simulations with several programs. One called Flow-3D CFD, from Flow Science Inc., Los Alamos, N.M., is a general-purpose fluidflow program that predicts flows with free surfaces during die filling. Algorithms in the software track sharp liquid interfaces (surface of the advancing wave) through arbitrary deformations, and apply correct normal and tangential stress boundary conditions, an accuracy feature that distinguishes it from other fluid-flow programs.

Filling a die with high-pressure metal in less than half a second creates several problems, catching excess material, for example. Overflow pads must be properly located before production. Pads also capture oxides and other nonmetallic inclusions that form when molten metal enters a relatively cold die. Every ounce caught by the overflow gets remelted, filtered, and recast in another part, so pads save material.

"Porosity, another defect, comes from a flow front that breaks and reforms, trapping gas," says Guthrie. Porosity from poor filling can create blisters in 2 to 4-mm thick walls, or provide a leak path when they are machined. Die changes are expensive, and involve downtime and almost always require welding, which cuts mold life by about 50%.

A die's model geometry, however, can be changed quickly on the computer and then reanalyzed to see the effects of the change. As a result, engineers eliminate porosity problems in days rather than weeks or months. Visteon engineers have used the technology to eliminate problems for in-house die-casting operations as well as for suppliers.

Recently developed finite-element-based manufacturing programs predict the amount of springback in sheet metal that results from tool shape, press conditions, and material grade. The software works with hydroforming and stretchforming processes. The advantage of the processes is that a single hydroformed or stretchformed part can potentially replace several stamped pieces that must be welded into an assembly. The recent capability is part of Marc 2000, a nonlinear FEA program from MSC.Software, Los Angeles.

Hydroforming uses a bladder filled with pressurized oil to force metal parts against a die. Should wrinkles appear, production engineers can try different bladder shapes, pressures, or pressure schedules. The aluminum skin on most aircraft is hydroformed, as are several automobile suspension components.

The challenge in the process is that removing hydrostatic pressure from a formed sheet releases elastic strain-energy, letting the sheet spring back to a different shape, one that often deviates from design specifications. Accurately predicting springback would let toolmakers account for it in the die. Simulating the entire process helps evaluate springback as well as worksheet behavior during intermediate forming stages.

When a drill exits material, it often generates burrs that must be manually removed. "This can amount to 30% of the part's cost," says L. Ken Lauderbaugh Saunders, professor of mechanical engineering, Pennsylvania State in Erie, Pa. "Our analyses shows that drilling operations can be modeled in FEA to accurately predict burr size from the process conditions." Although the study does not show how to eliminate burrs, it's certainly a step in that direction. In addition, it may be useful for manufacturers who often drill holes where only one side of the material is accessible, making rework impossible.

Saunders drilling model highlights the challenge of simulating even simple manufacturing operations. For instance, drilling in FEA requires modeling-cutting forces, finding stresses, evaluating failure criteria, and simulating material removal. "We start with a meshed model of material at an initial drill location," explains Saunders. "If stresses after a solution are not high enough to cause failure, then element thickness is reduced and a new model generated and solved." The process repeats until element stress exceeds its ultimate failure stress.

Modeling forces require calculating a combination of torque, power, and thrust. Details for the calculation are beyond the scope of this article but depend on cutting mechanics and process parameters such as speeds and feeds. The stress state is based on forces and material remaining in front of the drill. A burr forms in the model when stress (maximum normal stress) exceeds the ultimate strength of the material in tension.

Working with FEA software from Ansys Inc., Canonsburg, Pa., let Saunders automate model building, meshing, and solving with a feature called log files. These make it easier to change feedrates, spindle speeds, material properties, and tool geometry. In addition, elements are layered in half-revolution-feed increments. As the drill goes through the workpiece, each layer is removed by an "ekill" command in the program. It turns off elements "removed" by the drill.

Saunders ran this nonlinear model with each step generating boundary conditions for the next. The time for each step is the amount needed for a half-revolution of the drill. "The model produced burr heights that agree with two of the three tested feedrates," he say. "The process is quite temperature dependent, so keeping materials cool minimizes burr height." Next, suggests Saunders, might be to experiment with different tool designs to see how burrs can be minimized.

The recent viscoplastic material model in a finite-element analysis program let a welding repair company predict residual stress in a weld overlay called Unifuse. Tony Scandroli, project manager with Welding Services Inc., Norcross, Ga., devised a method of predicting weld distortions, and used the FE software to show how stresses relax with time and temperature in the boiler.

The corrosive environment inside a coal-fired boiler can erode waterwalls, the close-packed tubes that carry water and steam. One way to prevent corrosion involves welding a protective cladding over the eroding waterwall using materials such as 312 duplex stainless steel. "Welding temperatures produce residual stresses close to that for yielding in the weld zone," says Scandroli. "Stresses also come from differences in the weld and base material expansion coefficients, and from thermal gradients that produce distortion." The power-plant's owners wanted to know if high residual stresses would decrease the duty cycle of the boilers, and would the stress relax at high operating temperatures.

FEA-developer Algor, Inc., Pittsburgh, had recently added the viscoplastic material model with creep characteristics, to its package. Developing a model to predict a stress-relaxation level due to creep required the recent material model. "A boiler at 800°F is like a low-temperature heat-treatment," says Scandroli. "But running for 4,000 hr at a time is enough to let residual stresses relax."

A viscoplastic material model requires at least three material coefficients for an algorithm called the creep power law. This material model also requires the modulus of elasticity, yield stress of the weld material, and its tangent modulus.

To validate the analysis, Scandroli's team measured distortion profiles on repaired boiler sections over several months. "I modeled residual stresses caused by welding and then applied the boiler operating conditions for 4,000 simulated hours. The measured distortion profiles were then compared to the FEA model's predicted results. The numbers compared favorably, thus validating the FEA model. So operating conditions do follow the theory that says thermal treatments produce a stress relaxation in a Unifuse weldoverlaid repair application," he says.

A recently added module to HyperForm from Altair Engineering Inc., Troy, Mich., lets die designers produce tooling and confirm it works as predicted. Until recently, the software only let designers see where dies turn out parts with wrinkles or tears. Users would then tinker with the stamping machine's settings and inputs to eliminate flaws. For example, wrinkling might be taken care of by adding extra material to the part. This puts strain conditions around the boundary of the part, creating extra friction so the material cannot move. The technique moves defects outside the trim line, and therefore produces a useful part.

V4.0 of the software more clearly points out areas of formed sheet-metal parts that will wrinkle or tear, as well as other questionable areas, and those that will form well. The software unbends flanges in the final formed part to better calculate a blank size.

"The software also accounts for planar anisotropy", says Subir Roy, project manager with Altair. "It does this using so-called R values that describe material conditions along the axis on which the sheet material is rolled (0°) as well as at 45° and 90° to the axis." Minor imperfections and rolling conditions alter these values from batch to batch. R values also help predict sheet-metal earing (variations along a stretched edge) during deep drawing, as might occur when forming beverage cans. Earing is excess material that must be trimmed.

"The software also examines parts welded with different material properties and thicknesses on either side of the weld line," says Roy. An updated geometry-cleanup panel, he says, provides interactive control for preparing CAD model surfaces for automeshing. Other new capabilities include automatic surface organization by fillet and break angle, common edge suppression, and detection and suppression of holes. "Users will be glad to hear that V4.0 runs most models in minutes, thanks to new code that provides accurate results for any shape," he says.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.