2015

Komi, Erin; Hämäläinen, Juhani

This study sought to investigate the use of a high-performance computing environment to efficiently automate some of the early steps in the design process. The idea was to manipulate commercially available software to automate the running and results tabulation of a large number of topology optimizations. This was achieved by creating a finite element model of the chosen case study (genset base frame), creating superelements for large, complex components that were attached (engine and generator), preparing a topology optimization model, generating shape variables that altered the length, width and height of the design space, and launching a design of experiments on a computing cluster so that 64 variations of the topology optimization were run. Upon completion of each optimization run, the DOE automatically launched a program to interpret the results in several different ways, and then subsequently submitted the new design for reanalysis based on the initial design criteria. The results of the reanalysis were automatically tabulated for easy comparison of designs. The DOE set-up was also flexible enough to allow for additional topology optimization results interpretation and reanalysis after the initial runs were completed. The final result was hundreds of designs from which it was easy to select those which were most able to achieve the design goals.