PERBANDINGAN METODE OPTIMASI DAN PEMODELAN KETIDAKPASTIAN DALAM OPTIMASI TOPOLOGI UNTUK KONDUKSI PANAS

• Main Author: Limor, Frentsen
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/68425
• Institution: Institut Teknologi Bandung

This Thesis presents a comparison between different optimization methods of robust heat conduction topology optimization. In order to do so, different optimization method e.g. Optimality Criteria (OC), Method of Moving Asymptotes (MMA) and Stochastic Gradient Descent (SGD) are used. To obtain a robust optimization results, uncertainty is inputted in the problem. The uncertainty can be in the form of heat position, material conductivity and heat magnitude. The inclusion of this uncertainty will cause many deterministic results to be obtained, so it is necessary to find a way to include all the results of the deterministic optimization to update the design parameters. The statistical momentum approach is used in optimization which can be done in 2 ways, namely the Monte Carlo Simulations (MCS) and Polynomial Chaos Expansion (PCE) which also will be compared. This paper study the effect of pairing different optimization method with statistical moments and its effect to minimize compliance, convergence speed and constraints fulfillment speed. The results show that in terms of convergence speed and constraint, Optimality Criteria (OC) method stands out from the rest. For the minimization of the compliance, optimization methods such as Stochastic Gradient Descent (SGD) also its derivation (Adaptive Subgradient Methods) and Method of Moving Asymptotes yields better results in cases that has been done.