Shape optimization of a sCO2 centrifugal compressor stage

The design of a centrifugal compressor for supercritical carbon dioxide (sCO2) power cycle must account for non-ideal gas effects and the possible occurrence of two-phase flows. Shape optimization techniques combined with computational fluid-dynamic (CFD) simulations can produce optimized designs while inherently coping with the peculiar flow characteristics near the thermodynamic critical point. This work presents the first shape-optimization attempt of such non-conventional compressors. The compressor stage includes the impeller and the vaneless diffuser and starts the compression close to the critical point. The impeller blade angle distributions and meridional channel are parameterized with Bezier control points, which grant a local shape control within the optimization routine. The pinch of the vaneless diffuser is optimized as well. The validated CFD solver accounts for both non-ideal effects and two-phase homogeneous flows under the assumption of thermodynamic equilibrium and barotropic fluid. The constrained optimization problem is solved with genetic algorithms. To reduce the computational cost, surrogates for the objective function and constraints are trained over a limited number of CFD results. The surrogate accuracy is improved throughout the optimization process by adding optimal stage geometries to the initial training samples. The optimized geometry shows an appreciable efficiency increase (0.7 percentage points) while delivering the design pressure ratio. Although performing better in the design condition, the operating range of the compressor is altered by optimization. This finding leaves the door open for future optimizations that include both design and off-design operating points in the definition of the objective function and constraints.