The objective here is to develop the technology to efficiently solve structural optimization problems with very large numbers of design variables and constraints. Present numerical optimization algorithms, which are highly efficient and reliable, require storage of sensitivity information for large numbers of responses. Also, the direction finding problem requires solution of a large sub-problem, often requiring significant computational resources. To overcome these limitations, the effort proposed here will be devoted to 1) developing one or more algorithms which do not require storing large amounts of sensitivity information in central memory or using "spill logic" to store and manipulate this information and 2) developing algorithms that do not require solution of a large and time consuming sub-problem to find a search direction. Once candidate algorithms are identified, a prototype computer program will be created to study the proposed algorithm(s). The GENESIS program from VMA Engineering will be used as a test program to generate design problems with many thousands of design variables and constraints. The goal is to create and market commercial software that allows the engineer to solve structural optimization problems on the order of tens of thousands of design variables and constraints.
Potential Commercial Applications:Commercial applications are immediately insured by adding this technology to existing structural optimization software such as GENESIS. This will allow engineers to consider much larger numbers of design variables than are now possible for conventional structures such as aircraft wings and fuselages. Also, this will be directly applicable to topology optimization with general nonlinear constriants.
