Partners and International Organizations
(English)
|
KTH (S), LUTAB (S), ABB (S), CADOE (F), CERFACS (FR), Liebherr Aerospace (F), Ratier-Figeac (F), EPFL (CH)
|
Abstract
(English)
|
Parametric CFD analysis computes the flow for a whole range of conditions and shapes in one shot. The velocity field and aerodynamic forces are represented as multi-variate polynomials which can be evaluated instantly in optimization loops or in performance evaluation of a large number of design alternatives. A parametric CFD capability will therefore give faster design cycles and improved designs because more alternatives can be investigated more easily.
The OPTFLO package developed in the ESPRIT IV project OPTIBLADE is the first parametric CFD system built from industrial strength components and demonstrates feasibility of the approach. It is parallelised and can exploit multi-processor computers or networks. Such HPC soft- and hard-ware can provide an order of magnitude reduction in turnaround times for solutions of current problems, and of course capacity for solving larger problems. An investigation of a particular shape optimization task using only standard single-case flow solutions on a high-performance computer showed a reduction in total time from 22 days to 10 days, essentially removing all waiting time because the runs could be performed over night.
The test cases were chosen by the industrial participants to be representative of their respective applications: a novel S-shaped symmetric airfoil profile for a reversible fan, a compressor cascade, and a NACA16707 profile for a high-performance propeller. The computational models were built in the I-DEAS interface to OPTFLO. Parametric solutions were then compared to single-case solutions for validation. The results of the analysis of, say, lift force, can be presented in the form of sensitivity curves, xy-plots of lift force vs. a chosen parameter, or as parametric plots showing the variation of lift force with two parameters. Such tools offer quick assessments of which shape parameters are the most important and which could initially be kept fixed in the search for an optimal design.
The software automatically estimates the allowable parameter ranges for a given accuracy requirement. In some cases they were smaller than anticipated, showing a need for automatic construction of one polynomial per sub-range in a splitting of the user specified range.
The experiments could not be carried all the way to generation of an optimal airfoil profile, because OPTFLO is limited to inviscid flow models so efficiency losses due to skin friction and turbulence could not be calculated. A fully general 3D multi-block interface with the type of mesh control required for viscous, high Reynolds number flows, requires no theoretical development but needs a substantial programming effort.
In conclusion, the project has demonstrated feasibility of the parametric CFD approach for industrial problems, and clarified the further development needed in theories, computational models, and software.
|
