MULTIDISCIPLINARY MULTIPOINT OPTIMIZATION OF A TRANSONIC TURBOCHARGER COMPRESSOR

A transonic centrifugal compressor for turbocharger applications has been redesigned by means of a multidisciplinary multipoint optimization system composed of: a 3D Navier-Stokes solver, a Finite Element stress Analyzer, a Genetic Algorithm and an Artificial Neural Network. The latter makes use of a database, containing the geometry and corresponding performance of previously analyzed impellers and allows a considerable reduction in computational effort. The performance of every new geometry is verified by a 3D Navier-Stokes solver. A Finite Element Analysis verifies the mechanical integrity of the impeller. The geometrical description of the impeller has been extended to better adapt the inducer part of the impeller to transonic flows. The splitters are no longer copies of the full blades but specially designed for minimum losses and equal mass flow on both sides. The blade thickness and number Of blades are unchanged because defined by robustness and inertia considerations. The operating range is guaranteed by a two-step optimization procedure. The first one provides information allowing a modification of the inlet section to guarantee the required choking mass flow and a more accurate prediction of the boundary conditions for the Navier-Stokes analysis of the modified impeller. The second one predicts the performance curve of the new geometry for which the choking mass flow is known. It is shown how these extensions of the optimization method have led to a considerable improvement of the efficiency and corresponding pressure ratio, while respecting the surge and choking limits without increase of the stress level.