KRIGING-BASED MULTI-OBJECTIVE OPTIMIZATION OF FRANCIS TURBINE OPERATING UNDER 131.07 METERS OF HEAD AND ROTATING AT 1,000 RPM

• Main Author: Dunell Damanik, Denis
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77339
• Institution: Institut Teknologi Bandung

This bachelor’s thesis presents an optimization of a Francis turbine to contribute to alleviating the ever-increasing demand of electricity. The base design of the Francis turbine operates under 131.07 meters of elevation head and 1,000 RPM of rotation speed. The turbine is optimized for higher overall-efficiency and/or shaft power. Optimized parameters are geometries of Stay Vane, Guide Vane, and Runner. The optimization method is called surrogate optimization. The process involves iterative sampling. The surrogate function in question is a regression function whose shape is that of Kriging function which belongs to Radial Basis Function (RBF) family. Two successive stages of optimization processes are: initial sampling (by Latin Hypercube sample arrangement in the design space); followed by iterative selective-sampling which utilizes weighted multi-objective expected improvement criterion, which statistically evaluates the best location to sample. Each sample is evaluated in low fidelity CFD simulation. The search, i.e. iteration, goes on and on until stopping criterion is reached. After the stopping criterion is met, which were a time constraint, a set of optimal designs are obtained. Three optimal designs are then picked to be evaluated in higher-fidelity CFD simulation. Two of which are the following: the highest-efficiency design increases efficiency by 0.19%, shaft power increases by 4.44%, and the lowest static pressure is increased by 1.385 atm which reduces cavitation tendencies; whereas the best shaft-power design decreases the efficiency by 2.89%, increases shaft power by 24.93%, while decreasing the minimum static pressure by ?0.103 atm which worsens cavitation tendencies. This thesis also presents sensitivity analysis of the geometry parameters and lists optimum geometry parameters of the aforementioned three picked designs.