Multidisciplinary design optimization （MDO） is widely employed to enhance turbomachinery compo- nents efficiency. The aim of this work is to describe a complete tool for the aero-mechanical design of a radial in- fl...Multidisciplinary design optimization （MDO） is widely employed to enhance turbomachinery compo- nents efficiency. The aim of this work is to describe a complete tool for the aero-mechanical design of a radial in- flow turbine and a centrifugal compressor. The high rotational speed of such machines and the high exhaust gas temperature （only for the turbine） expose blades to really high stresses and therefore the aerodynamics design has to be coupled with the mechanical one through an integrated procedure. The described approach employs a fuUy 3D Reynolds Averaged Navier-Stokes （RANS） solver for the aerodynamics and an open source Finite Element Analysis （FEA） solver for the mechanical integrity assessment. Due to the high computational cost of both these two solvers, a meta model, such as an artificial neural network （ANN）, is used to speed up the optimization design process. The interaction between two codes, the mesh genera- tion and the post processing of the results are achieved via in-house developed scripting modules. The obtained results are widely presented and discussed.展开更多
文摘Multidisciplinary design optimization （MDO） is widely employed to enhance turbomachinery compo- nents efficiency. The aim of this work is to describe a complete tool for the aero-mechanical design of a radial in- flow turbine and a centrifugal compressor. The high rotational speed of such machines and the high exhaust gas temperature （only for the turbine） expose blades to really high stresses and therefore the aerodynamics design has to be coupled with the mechanical one through an integrated procedure. The described approach employs a fuUy 3D Reynolds Averaged Navier-Stokes （RANS） solver for the aerodynamics and an open source Finite Element Analysis （FEA） solver for the mechanical integrity assessment. Due to the high computational cost of both these two solvers, a meta model, such as an artificial neural network （ANN）, is used to speed up the optimization design process. The interaction between two codes, the mesh genera- tion and the post processing of the results are achieved via in-house developed scripting modules. The obtained results are widely presented and discussed.
