| |
|
Description: |
In many industry branches, including automotive and aerospace, optimal system design problems appear in a decomposed form: design optimization tasks, associated with the elements that comprise the system, are distributed among departments or outsourced to suppliers. Smooth system integration can only be ensured when the interactions among elements are taken into account and suitable design speci cations are obtained for all subproblems. In this talk we present Analytical Target Cascading (ATC), a rigorous methodology for translating system design targets to proper element design speci cations while accounting for interactions. Using the multilevel hierarchy of the decomposed system, an optimization problem is formulated and solved iteratively for each element to minimize deviations of local responses from propagated targets. Solving the subproblems with appropriate coordination strategies yields consistency and system-level optimality. Using a supersonic jet design example, we will discuss recent ATC formulation extensions that enhance its applicability to multidisciplinary design optimization problems. We will also present extensions of the deterministic ATC formulation to account for uncertainties in the design optimization process using a probabilistic framework. Time permitting, we will also present an additional extension of the probabilistic ATC formulation that can be used for reliability target allocation.
|
|
Date: |
|
| Start Time: |
12:00 |
|
Speaker: |
Michael Kokkolaras (University of Michigan, USA)
|
|
Institution: |
University of Michigan, USA
|
|
Place: |
Room 5.5
|
| Research Groups: |
-Numerical Analysis and Optimization
|
|
See more:
|
<Main>
|
|
