GPU Enhancement of Weld Distortion Prediction

Project Description

Caterpillar is a leading producer of large mining, earthmoving and construction equipment. We use weld distortion simulation to optimize welding procedures, and to predict/correct manufacturing problems arising from weld distortion. This project would allow us to demonstrate the benefit of GPUs used in conjunction with HPC in order to greatly decrease analysis time. This research will directly benefit our company in several ways. By understanding and optimizing our processes virtually, we will be able to identify and eliminate non-value-added manufacturing processes, such as post-welding straightening of warped parts. In addition to the cost savings from increased factory efficiency and capacity, the ability to plan these processes in a virtual environment conserves valuable natural resources (material, and power) by eliminating much of the trial and error traditionally performed on physical prototypes to determine appropriate process parameters. Additionally, our HPC administrator is considering adding GPU computing to our current HPC system, and this project will allow us to demonstrate and quantify the potential value of this technology to Caterpillar.