MULTIOBJECTIVE OPTIMIZATION DESIGN OF A CENTRIFUGAL COMPRESSOR IMPELLER

• Main Author: Ikbar Syahputra, Reyhan
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77145
• Institution: Institut Teknologi Bandung

The centrifugal compressor is a prevalent type of turbo machinery with various applications. Among its components, the impeller plays a crucial role. Enhancing the performance of the centrifugal compressor impeller can trigger a cascading effect on the overall system performance. Currently, impeller designs for centrifugal compressors are often developed using both theoretical and empirical calculations. However, the created designs can still be further optimized due to design methods being unable to determine specific design variables. This research focuses on multi-objective optimization of centrifugal compressor impeller design. Four design variables are selected from the control points of a third-degree Bezier curve. The optimization aims to maximize the pressure ratio and isentropic efficiency concurrently. The Non-Dominated Sorting Genetic Algorithm II (NSGA-II) is employed for this optimization. The genetic algorithm utilizes a surrogate model as a fitness function, constructed through Latin hypercube sampling of centrifugal compressor impeller performance evaluated using Computational Fluid Dynamics (CFD) simulations. Three optimal designs are obtained, each surpassing the initial design. Under design conditions, the efficiencies of Design A, Design B, Design C, and the initial design are 91.61%, 93.34%, 93.63%, and 90.30%, respectively. Meanwhile, the pressure ratios for Design A, Design B, Design C, and the initial design are 1.862, 1.855, 1.840, and 1.792, respectively. This indicates the dominance of the optimal designs over the initial design, without mutual domination among the optimal designs.