Development of a Multidisciplinary Design Optimization Process for Automotive Components: Metamodel-Based Optimization of a Center Stack Display Bracket
Publicerad
Författare
Typ
Examensarbete för masterexamen
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
The design of automotive components requires balancing requirements from mul tiple engineering disciplines to arrive at a satisfactory design in increasingly short
development cycles. Multidisciplinary Design Optimization (MDO) can facilitate
learning and discovery of high-performing designs with conflicting requirements in
a time-efficient manner.
This thesis presents an MDO process for interior structural components subject
to crash and noise, vibration, and harshness (NVH) requirements. Internal needs
at Volvo Car Corporation for such a method are identified using semi-structured
interviews. Relying on previous findings in the field of metamodel-based MDO, a
sizing optimization workflow is developed in the software modeFRONTIER using
a simple geometry subject to a head impact load case and a modal requirement.
The presented process involves variable selection by parameterization and statisti cal variable screening, sampling using an experimental Uniform Latin Hypercube
design, metamodeling with multiple metamodel algorithms, and subsequent global,
nongradient optimization with the genetic algorithm MOGA-II.
The MDO process is applied to a Center Stack Display bracket design problem
subject to three head impact load cases and an NVH-related modal requirement.
Significant performance increases for critical objectives are shown compared to a
pre-optimized design; however, the optimization does not reach all performance tar gets, thus prompting a discussion about process limitations, including the need for
global information and implications of geometry lock-in. Organizational, societal,
ethical, and ecological aspects related to MDO in the automotive industry are briefly
considered. Finally, recommendations regarding future work are given, including the
implementation of finite element mesh morphing to allow for more complex design
variables and method improvements for variable screening.
Beskrivning
Ämne/nyckelord
Multidisciplinary Design Optimization, Variable Screening, Metamodel-Based Optimization, Design of Experiments, Latin Hypercube Sampling