Development and Application of Robust Design Optimization Methods for Turbomachinery

Noaparast, Michel

4. term

Energy Engineering, Master

2001

2001-06-30

63

Små variationer i produktionen kan ændre et produkts ydeevne, og disse usikkerheder bør derfor håndteres allerede i udviklingsfasen. En måde er Robust Design Optimization (RDO): at kombinere numerisk optimering med følsomhedsanalyse for at finde designs, der holder ydeevnen stabil, når input varierer. Dette projekt udvikler og anvender RDO til turbomaskineri, der bruger invers design (man tager udgangspunkt i ønsket ydeevne og beregner bladets form). Der blev opbygget en CFD-model (computational fluid dynamics) for at muliggøre RDO-processen, og nye tilgange inspireret af eksisterende RDO-metoder blev testet på en repræsentativ case. I denne case gav ingen af de testede RDO-tilgange et mere robust design end konventionel optimering. Resultaterne viser dog, at designs med samme gennemsnitlige ydeevne kan have forskellig variationsbredde og dermed følsomhed. En separat undersøgelse af impellerbladets hydrauliske følsomhed peger på, at bagkanten (trailing edge) er det mest følsomme område, og derfor anbefales strammere produktionstolerancer dér for dette design. Da følsomhedsanalysen er udført på ét nominelt design, kan resultaterne ikke generaliseres til alle impellere. Indtil bredere tendenser er identificeret, bør hver konstruktion have sin hydrauliske følsomhed undersøgt, før tolerancer fastlægges.

Small variations in manufacturing can change how a product performs, so these uncertainties should be addressed already during development. One approach is Robust Design Optimization (RDO): combining numerical optimization with sensitivity analysis to find designs that keep performance stable when inputs vary. This project develops and applies RDO for turbomachinery that uses inverse design methods (starting from a target performance to compute the blade shape). A CFD (computational fluid dynamics) model was built to enable the RDO process, and new approaches inspired by existing RDO methods were tested on a representative case. In this case, none of the tested RDO approaches produced a design that was more robust than conventional optimization. However, the results show that designs with similar average performance can differ in their performance variation and sensitivity. A separate hydraulic sensitivity study of the impeller blade indicates that the trailing edge is the most sensitive region, so tighter manufacturing tolerances are recommended there for this design. Because the sensitivity analysis was performed on a single nominal design, these findings cannot be generalized to all impellers. Until broader patterns are identified, each design should have its hydraulic sensitivity assessed before tolerances are specified.

[This abstract was generated with the help of AI]

Download
View record in AAU Student Projects