Finite-element analysis has let aerospace and automotive OEMs pare costs off designs for years. Affordable computers and userfriendly FEA software now let even small and midsized companies cut costs by numerically simulating manufacturing processes including casting, stamping, and cutting. Such simulation lets users "see" how components will respond to loads, pinpointing stresses associated with shock, temperature, vibration, and fluid flow. Users can test components in the digital world and iron out design flaws early in development before cutting chips.
Manufacturing-simulation software based on FE solvers is typically nonlinear, or a combination of nonlinear and linear. Linear analysis involves induced displacements that are proportional to applied loads, such as making sure a product can withstand its own weight. Linear analysis alone can't handle simulating dynamic manufacturing processes with exceedingly large strains and displacements.
The simulations here belong to what's often called the virtual manufacturing world. The goal is to get parts right the first time.
CASTING NEAR-NET PARTS
Foundries tasked with producing near-net-shaped parts use FE-based software such as ProCast from ESI Group, Paris, which models the underlying physics of metal casting. "Be-cause the software is intended for casting simulation, it has CFD capabilities beyond that of a general-purpose stress-analysis package," says Mark Samonds, technical director, ESI U.S. R&D Inc., Columbia, Md. "For example, the software couples thermal, flow, and stress calculations so users can predict part deformation and residual stresses during metal solidification and cooling. This lets foundries modify mold shapes up front. And for investment casting, where radiation is the dominant mechanism, the program predicts a casting's self-radiative effects from one region of the shell to another region, and also from the casting to the furnace walls."
The vast calculations such programs provide are also important for applications such as single-crystal casting. Here an industrial oven heats one end of a mold while a chiller cools the other end. Metal cast in the mold starts solidifying at the cool end while the dendrites of the crystal grow towards the heated end. "Parts are pulled out slowly, so view factors, or fractions of thermal energy, must be recalculated continuously. This is so time consuming, it's almost impossible in a package not targeted for this application," says Samonds.
In fact, calculating thermal energy is paramount and absolutely critical says Samonds, so it's necessary to predict the temperature field at all times, including heat transfer from metal to mold, and from mold to air or cooling lines. Therefore, the software includes such features as one that couples the thermal history at any location in the casting with the nucleation and growth of microstructures. This lets users, for instance, predict grain structure formation to gauge part strength, which is especially important for critical components such as in automotive structures.
"Users over the last 20 years have gained confidence in FE software and find its predictions agree well with what they see on the shop floor," says Samonds. "Simulation is especially handy for eliminating or reducing trial-and-error experiments. For example a casting process that used to take 10 to 20 iterations might now take only two or three. For expensive parts, a few foundries are just running simulations and immediately going to production shots, without any trials."
Samonds says that parallel computing, which lets users bring many CPUs to bear on a problem at once, is a continuing trend. Jobs that once took weeks to solve now take only hours. And 64-bit processors make calculations faster, provide more memory, and allow larger file sizes, important as models get bigger and bigger. Companies now regularly solve models with 5 million nodes.
Simulating the stamping of parts such as automotive fend-ers also requires nonlinear FE software, says Olivier Morisot, marketing manager, ESI North America., Bloomfield Hills, Mich. "Fenders are stamped on a huge press called a line die. Stamping thins parts in some areas and thickens them in others, building in a lot of stresses and strains. Pam-Stamp 2 software provides sheet-metal-stamping simulation and accounts for process parameters such as die geometry, material behavior, blankholder force, and draw-bead shape and position."
Such FE software, according to Morisot, is not yet artificial intelligence because it doesn't tell users how to fix design problems. But it's a vast improvement over the old method, which had production engineering pulling out their hair guessing as to tooling and process, building the tooling, seeing where parts split, broke, or wrinkled, then changing the tooling and trying again.
In contrast, a shop now imports a 3D model of the fender and generates a mesh. The user types in the die-surface finish requirement, process parameters such as press tonnage and the lube to be used on the die, and the material. The software then predicts thick and thin areas, high stress areas, and so on. Based on these virtual results, the designer can change the material on the fender model if need be so parts meet specifications for performance and durability.
"Tool and die shops can now afford enough computer power to run simulations," says Morisot. For example, Advanced Cam Inc., Utica, Mich., targets aircraft and automotive work. It often designs line and progressive dies for automotive components including inner reinforcements, rails, inner doors, and deck lids. The FE software lets the company run formability tests to modify tool geometry up front and compensate for the effects of springback. This is a big problem with high-strength steel. It's a function of stress distribution at the end of forming.
Chatter is, of course, wavy marks on finished parts that result from the part and tool bouncing against each due to excessive vibration in the tool, toolholder, spindle, and workpiece. To develop machiningsimulation software that lets users cut first parts correctly and efficiently, it was necessary to model both the cutting process, or interaction between the tool and the workpiece, and the machine tool, says Yusuf Altintas, professor in virtual machining at the University of British Columbia's Manufacturing Automation Laboratories (MAL) Inc., a spin-off company from the university research lab. It distributes simulation software developed at the lab through researching problems that originate from factories and working with machine tool, cutting tool, and spindle builders.
"We developed and used a finite-element-based program to analyze chip removal, deformation, and temperature zones on cutting tools with arbitrary shapes. This research was primarily for tool design and optimization of the cutting edge in machining hardened die steel and aerospace alloys such as titanium. We spent many hours of research to model cutting as a function of work material, tool geometry and material, chip load, cutting speed, and dynamic flexibilities of the machine tool and workpiece. This let us predict cutting forces, torque, power, and chat-ter-free speeds and cutting conditions for maximum productivity without damaging the tool, spindle, and workpiece," says Altintas. "This re-search became the basis for our Cut-Pro software, which simulates machining operations including milling, boring, and turning."
Weiss Spindle Technology Inc., Boeing, Pratt & Whitney Canada, and others worked with the lab to design a spindle that would not vibrate at required cutting conditions. Eliminating spindle vibration would help eliminate chatter, a problem common to most shops.
"Weiss built us a prototype spindle, which had thermocouples and displacement sensors inside, for experimentally verifying the finite-element spindle model we developed. Eventually, this led to SpindlePro, linear and nonlinear FE software for spindle performance analysis and optimization. The software uses the Timoshenko beam theory in the FE model, which includes axial, bending, and torsional behavior of the spindle system by considering bearing preload and speed effects.
According to Altintas, the cutting problems in most shops are with milling, not turning. To address the problem, the milling module in the cutting software makes predictions including the surface finish caused by feed marks, chatter-free axial and radial depth-of-cut, chat-ter-free spindle speeds, and design and analysis of inserted, indexable cutters. "On average, cycle-time reduction is 50%," says Altintas, "In machining jet-engine compressors, however, we went as far as 83% in one application."
The thick and thin of sheet metal
"Thinning" in sheet-metal parlance is actually a "loss of thickness." Thinning is the initial material thickness minus the thickness after stamping. It's a different way to look at thickness distribution and important because sheet-metal parts sometimes must meet a maximum allowable thinning requirement, for example 15%. This dictates for a sheet 1-mm thick, the final part can't be less than 0.85 mm anywhere. Too much thinning causes a part to go out of spec from a strength standpoint, while too much springback does so from a geometry standpoint.
Advanced Cam Inc., advcam.com
ESI Group, www.esi-group.com
ESI North America, (248) 203-0642,
ESI U.S. R&D Inc., (410) 480-7132
MAL Inc., The University of British Columbia, malinc.com
Weiss Spindle Technology Inc., (440) 946-4003, weissgmbh.de