Perpustakaan Digital ITB

Centrifugal compressors are widely used in energy, petrochemical, and process industries, where performance directly governs overall efficiency. To form the design variables, this study applies a surrogate-based multi-objective optimization framework to modify impeller geometry and achieve higher pressure ratio and efficiency; relying solely on CFD can be computationally expensive (time-consuming and requiring high computing resources). The impeller geometry was parameterized using six Bézier control points on the meridional plane. Fifty different boundary-variable samples were added using Latin Hypercube Sampling (LHS) and simulated in ANSYS CFX to obtain performance data. Kriging surrogate models were trained (R² = 0.7525 for pressure ratio; 0.7618 for efficiency) and validated with iterative retraining. NSGA-II was then executed with population = 100, generations = 200, tournament selection, scattered crossover, adaptive-feasible mutation, and Pareto fraction = 0.35. After three retraining cycles with CFD-validated candidates, the optimized designs showed measurable gains: the efficiency-focused case raised efficiency by 1.944% with pressure ratio up 1.011%; the pressure-ratio-focused case increased pressure ratio by 1.468% with efficiency up 0.518%; and the balanced case improved pressure ratio by 1.308% and efficiency by 1.483%. Flow analysis confirmed smoother Mach number distributions, more favorable pressure gradients, and reduced secondary flows. In conclusion, the surrogate-based optimization effectively enhanced impeller performance while reducing computational cost, and it can be extended with additional variables and objectives; it also provides a solid reference grounded in Akhmad Fathoni’s work for future turbomachinery design exploration.