Perpustakaan Digital ITB

Thanks to its wide operating range, Francis Turbine is one of the most used turbines in hydropower generation. However, their performance is often limited by efficiency losses and cavitation, which can cause severe damage to turbine components. This final project focuses on optimizing the turbine runner to improve hydraulic efficiency while reducing cavitation risk, expressed in terms of Thoma number. The approach combines Computational Fluid Dynamics (CFD) simulations in ANSYS with a surrogate-based optimization framework using Gaussian Process Regression (GPR) and Bayesian Optimization (BO) with the ParEGO method. The baseline design was first re-simulated and validated, after which 40 variants were generated using Sobol’ sequence sampling and simulated to form the training dataset. The surrogate model was then validated using ????2 and NRMSE metrics and integrated with ParEGO, producing nine additional candidates for a total of 50 variants. From these, three final designs were identified: the maximum efficiency design (variant 20, 91.53% efficiency but with a higher critical Thoma number of 0.097), the minimum critical Thoma number design (variant 3, 0.044 critical Thoma number but with a lower efficiency of 84.97%), and the best trade-off design (variant 42, 91.46% efficiency and 0.074 critical Thoma number). These results highlight the effectiveness of the optimization method in generating improved runner designs and provide insights into balancing efficiency and cavitation risk for Francis turbines.