O.K. Engineering Cuts the Wait for Weight Optimization
O.K. Engineering has a computing environment made up of workstations and a small, five-year-old cluster with 12 single-core processors. The cluster is used mainly to run LS-DYNA* and Abaqus FEA* with 12 cores in parallel. The company would like to be able to offer automatic optimization with LS-Opt*, but runtimes of 16 to 18 hours per calculation make it impractical to carry out this optimization work, including shape optimization, weight optimization and modeling of contraction variants.
O.K. Engineering carried out a proof of concept to see how a cluster based on the Intel® Xeon® processor E5 product family could enable it to carry out more sophisticated engineering simulations more quickly, including running LS-Opt. It ran LS-Opt on a Fujitsu PRIMERGY* BX400 S1 Blade Server, which incorporates the Intel Xeon processor E5 product family. It features eight compute nodes with two sockets each and eight cores per socket, a total of 128 cores.
Using the new cluster, O.K. Engineering was able to carry out a weight optimization on a roll bar. Based on O.K. Engineering’s results, the calculation time of 16 to 18 hours was cut by 75 percent to four hours, making it practical to use LS-Opt for the first time. The faster performance creates opportunities for O.K. Engineering to broaden the range of services it offers to include optimization, as well as to accelerate the delivery of its existing services.