Active Manifold and Model Order Reduction to Accelerate Multidisciplinary Analysis and Optimization - presented by Prof. Charbel Farhat

Active Manifold and Model Order Reduction to Accelerate Multidisciplinary Analysis and Optimization

Prof. Charbel Farhat

Prof. Charbel Farhat
Slide at 57:33
Charbel Multiobjective DAO of Rigid CRM Configuration Under Box Constraints Performance of the proposed active manifold (subspace) approach using the database of piecewise global, nonlinear, CFD-based HPROMs offline wall clock time on 256 cores of a Linux cluster online wall clock time on 16 cores of the same Linux cluster Computational task HPROMs Speedup (w clock time) (w clock time) (w clock / CPU) Construction AS 38.70 hrs Construction database HPROMs 44.73 hrs Optimization problem (online) 776.70 hrs 0.76 hrs 1,022.0/6,132 Optimization problem (offline + online) 776.70 hrs 84.19 hrs 9.2/55.2 for a two-objective (only) realistic MDAO problem grounded in high-fidelity computational models, the proposed approach delivers one to two orders of magnitude speedup factors - even when all investment costs are accounted for In the presence of viscous flows, higher speedup factors can be reasonably expected
Share slide