PhD Thesis Proposal 

Matthew Yates

(Faculty Advisor: Professor Dimitri Mavris)

"Manufacturing Influenced Topology Optimization and Design of Truss Structures"

Tuesday, July 21

2:00 p.m.

Weber, CoVE

Abstract

It has been a common endeavor throughout history to build reliable structures with minimal resources. For hundreds of years now, trusses have been a common solution to fill that purpose since they require minimal resources and are simple to understand mathematically. Trusses are still in use today thanks to advances in technology that give them nearly unrivaled lightweighting ability. However, there is a consistent design problem that an understanding of trusses alone cannot fix. Many design projects have ended costing significantly more than the initial budget, which research suggests is due to flaws in design methodology. Due to their prevalence in the automotive and aerospace industries, for their lightweighting and aerodynamic properties, panel structures have received more focus in research in improved design methodology than trusses. Therefore, a new capability is necessary to incorporate truss structures into current and future design. 

System design research has shown that most incurred costs occurs in the early design stages, consisting of configuration selection and preliminary optimization, where little information is available. Traditionally, these steps have been completed using historical data or sequential optimization of multiple disciplines to reduce computational effort. However, modern systems design research has shown that a multidisciplinary, concurrent design approach is needed to prevent cost overrun. This entails coupled solvers, multi-objective optimizers, and model abstraction. Truss structures were one of the first structures we learned how to optimize, and the approach from Michell’s optimization to the member adding method has been relatively similar except for the near exponential growth in allowable topologies thanks to advancements in computer technologies. Recent research has even found computationally efficient ways to handle nonlinearities and stability, not traditionally included in truss optimization. To extend the research of modern systems design methods to truss structures, it is necessary to overcome the lack of ability to couple manufacturing with truss topology optimization and lack of ability to abstract the optimized truss topologies. This proposal presents questions and potential answers that may make manufacturing influenced truss design possible. The first investigation is to determine the best method to perform manufacturing influenced, multi-objective truss topology optimization. I posit that the domain is most likely convex and can be captured by a scalarization function. The hypothesis can be tested directly against the alternative hypothesis for verification. The second investigation is to determine how to best optimize the nonconvex objective function created by the manufacturing influenced objective. The proposed method is based on previous attempts to reduce topology complexity and can be tested for effectiveness against standard benchmark problems. The final investigation is to determine a methodology for truss metamodeling. The goal is to achieve a model that can accurately represent multiple domains and boundary conditions.
Using the result of the experiments, the improvement to the truss design methodology can be shown by using the proposed methodologies on representative test cases. The first test case is a simple domain to show how a trusses performance metrics can be used in atechnology impact matrix, which is used in early design. The next case represents the real
problem of designing launch vehicle insterstage architecture. The proposed methodologies should allow trusses to be included in configuration selection and preliminary optimization.
Both cases will show that a computationally efficient and a robust design approach can be
applied to truss structures.

Committee:
Dr. Dimitri Mavris (advisor), School of Aerospace Engineering
Dr. Graeme Kennedy, School of Aerospace Engineering
Dr. Claudio V. Di Leo, School of Aerospace Engineering
Dr. Adam Cox, School of Aerospace Engineering
Dr. Andrew Bergan, NASA Langley Research Center