Optimization of engineering systems under uncertainty often involves problems that have multiple objectives, constraints and subsystems. The main goal in these problems is to obtain solutions that are optimum and relatively insensitive to uncertainty. Such solutions are called robust optimum solutions. Two classes of such problems are considered in this dissertation. The first class involves Multi-Objective Robust Optimization (MORO) problems under interval uncertainty. In this class, an entire system optimization problem, which has multiple nonlinear objectives and constraints, is solved by a multiobjective optimizer at one level while robustness of trial alternatives generated by the optimizer is evaluated at the other level. This bi-level (or nested) MORO approach can become computationally prohibitive as the size of the problem grows. To address this difficulty, a new and improved MORO approach under interval uncertainty is developed. Unlike the previously reported bi-level MORO methods, the improved MORO performs robustness evaluation only for optimum solutions and uses this information to iteratively shrink the feasible domain and find the location of robust optimum solutions. Compared to the previous bi-level approach, the improved MORO significantly reduces the number of function calls needed to arrive at the solutions. To further improve the computational cost, the improved MORO is combined with an online approximation approach. This new approach is called Approximation-Assisted MORO or AA-MORO.

The second class involves Multiobjective collaborative Robust Optimization (McRO) problems. In this class, an entire system optimization problem is decomposed hierarchically along user-defined domain specific boundaries into system optimization problem and several subsystem optimization subproblems. The dissertation presents a new Approximation-Assisted McRO (AA-McRO) approach under interval uncertainty. AA-McRO uses a single-objective optimization problem to coordinate all system and subsystem optimization problems in a Collaborative Optimization (CO) framework. The approach converts the consistency constraints of CO into penalty terms which are integrated into the subsystem objective functions. In this way, AA-McRO is able to explore the design space and obtain optimum design solutions more efficiently compared to a previously reported McRO.

Both AA-MORO and AA-McRO approaches are demonstrated with a variety of numerical and engineering optimization examples. It is found that the solutions from both approaches compare well with the previously reported approaches but require a significantly less computational cost. Finally, the AA-MORO has been used in the development of a decision support system for a refinery case study in order to facilitate the integration of engineering and business decisions using an agent-based approach.