A Practicing Engineer's View of Benchmark Problems in Structural Optimization

Abstract

What is structural optimization? The answer depends on who you ask. One practicing engineer will likely have a different answer than an engineer in academia or a representative in the structural engineering software community. That said, two different practicing engineers will likely have somewhat different views, as would two different professors or software vendors. The authors of this paper represent private practice and the software industry. Several benchmark problems and design tools in structural optimization (some existing in the literature and a couple of new ones) will be reviewed and discussed. The paper will discuss how a practicing engineer would formulate a benchmark problem in optimization, from choosing the appropriate design variables (discrete, continuous, etc.), determining the constraints (such as limits on strength and/or deflection), to defining the desired objectives (least weight, least cost, etc.). There are numerous optimization methods well known for solving benchmark problems, i.e., gradient based methods (such as Newton's Method), population based methods (such as genetic algorithms) as well as other search methods. A practicing engineer, however, may choose to use more conventional software to solve the problem. The paper will investigate the solutions found using conventional structural software and discuss some of the capabilities of existing software to truly perform optimal design.