Perpustakaan Digital ITB

This thesis presents a robust multi-objective approach for structural topology optimization. Performance and stability of a design are two conflicting objectives, hence, sometimes it is preferable to treat them as multiple objectives to gain important insights regarding various optimal topologies. The optimization problem is formulated as a simultaneous minimization of expectation and standard deviation of compliance in the presence of uncertainties in loading direction or loading magnitude. It is imperative to obtain a good representation of the Pareto front with as few finite element simulations as possible. This thesis proposes the combination of topology optimization with Polynomial Chaos Expansion (PCE) and Pareto Optimal Tracing (POT) to seek diverse Pareto optimal solutions efficiently. Three test cases, i.e., cantilever beam, simply supported beam and distributed load on carrier plate problem are used to test efficacy of the proposed method and the results are compared to the deterministic optimization. In distributed load problem, a new challange arises to deal with large number of random variables and therefore Gaussian random field is constructued by a Karhunen-Loeve Expansion paremeterization to reduce the random variables dimension. Thus, the impact of the PCE to reduce the number of realizations and POT to reduce the number of iterations are studied on the estimation of the Pareto front with various combinations of weights assigned to the expectation and the standard deviation. In addition, several non-dominating robust solutions are also validated using Monte Carlo Simulations (MCS). Results show that robust topology optimization with POT and PCE yielded better and more diverse topologies than the conventional approach, which restart the search from the baseline design.