An optimization framework for design space reduction in early-stage design under uncertainty

Early stage design is marked by a low level of design definition, leading to high levels of uncertainty around design decisions made at this stage. Many techniques have been proposed to make better decisions given this uncertainty, including robust optimisation approaches and reliability-based design optimisation. Drawing inspiration from set-based design, this work presents a different approach. Instead of making a final decision with a margin for uncertainty, the procedure allows for a gradual reduction of the design space in a manner that maximises the designer's remaining flexibility. Two measures are first defined-the complexity of the remaining design space, and the regret, or potential loss of performance resulting from deciding at that time. The procedure solves for the Pareto front in the trade space between complexity and regret. To generate the Pareto front, the method uses two optimisers with one nested inside of the other; both the inner and outer optimisation problems are solved using a genetic algorithm. The outer optimiser is multi-objective with the complexity, or size, of the reduced design space and the resulting regret as the two objective functions. This method was used to solve a structural design problem, and the results are presented here.