FEA makes manufacturing outcomes more certain
Engineers at the National Nuclear Security Administration's Kansas City Plant work with a number of joining processes, such as inertia welding, press fits, and solder joints. To get strong and reliable joints when faced with new manufacturing processes, engineers often turned to the analysis department. But simulation requests soon turned into a backlog.
To counter that problem, engineering and analysis groups devised a way to combine FEA with everyday engineering through online Simulation Advisors. "These are customized Web interfaces that manufacturing engineers use to perform virtual testing and prototyping," says Aaron Seaholm, an analyst at Honeywell Federal Manufacturing & Technologies LLC. Honeywell manages and operates the NNSA's Kansas City Plant. "Engineers with little or no FEA experience can type in a range of loads for a particular manufacturing process, run an analysis, and get back results that predict the outcome of the process. A supercomputer running Abaqus FEA software does the heavy lifting. Meshing and solving routines are automatic," he says.
Honeywell engineers devised the Simulation Advisor program to help engineers study how proposed designs will behave during manufacturing. Designers can simulate dozens of joining operations in a fraction of the time it takes to cut and test a single physical prototype. Simulation helps answer questions about how changes to the design or to the manufacturing process will affect the final product
Seaholm says the Simulation Advisor for inertia welding has been particularly effective. Inertia welding works like this: A machine holds one part and spins another at a high rpm. The two parts are then forced into contact where friction heats and softens the metal to form a weld.
Inertia welding has several advantages as a production process. It's a faster way to join axisymmetric parts than tungsten inert-gas or gas-tungsten arc welding. "Weld quality is another plus," says Seaholm. "Inertia welding requires no flux or fillers, and the heat-affected zone is smaller than in other methods. This helps preserve metal strength," he adds.
Inertia welding was a good candidate for an advisor because the facility had purchased new equipment to make sealed pressure containers of stainless steel. Engineers planned on running numerous tests to get a feel for the welder before beginning production. Because such benchmarking would be time consuming and expensive, the analysis staff suggested that a Simulation Advisor might reduce the number of test specimens needed to map the machine's capabilities.
The Advisor project had two steps. First, the welding group and analysts determined the relationships between rpm, force, and temperature variables.
Inertia-welding simulation is based on a coupled thermal-structural solution using Abaqus/Standard software. The team customized the friction subroutine in Abaqus to relate heat transfer from friction to temperature changes over time in the metal. "Once we knew how the metal would heat up, and how heat would dissipate as friction slowed and stopped the free-spinning part, we could determine the changes in material properties. These were the variables the welding engineers had to control to get good welds," says Seaholm.
The second step created a simple interface with HTML and an open-source scripting language called Python so welding engineers could run their own simulations. " Parameter fields on the input screen correspond to speed and preload controls on the equipment. Other fields handle parameters for part dimensions such as thickness, radius, height, and angle," says Seaholm. "As a reality check on the range of parameters, we correlated simulations to sections taken from actual parts made with the new machine."
Because the Abaqus pre and postprocessors are also written in Python, Seaholm's team could readily script an automated front end that drives the inertia-welding analysis. "The Advisor starts when an engineer pulls up an HTML page on a Web browser. The user enters simulation parameters and submits them to the software. When the job finishes, the system calls the postprocessor and transfers results in the form of a jpeg or animation file to the Web server where the engineer can see them. Posted results update regularly every 30 sec or so. No analysis takes longer than a few hours, so the welding group can quickly focus on the parameters most likely to yield quality physical prototypes," he says.
The Kansas City Plant now has nearly a dozen such Advisors to simulate springs, kinematic sleds and high-impact-tests, and thermal expansion. "Every Advisor we build captures more expertise and knowledge, leverages the investment in technology, and trims costs," adds Seaholm.