This paper compares two different methods for robust design improvement. The first method, called stochastic simulation, combines traditional Computer-Aided Engineering (CAE) simulation tools with variation in the simulation model parameters in order to estimate the resulting uncertainty in system behaviour for design improvement. The second method, called DACE, employs traditional Design of Experiments (DOE) methodologies to build statistical models of CAE simulation tools, called emulators because they emulate the behaviour of the simulator. The emulators are much faster to compute than the corresponding simulation model and can therefore be used to search the design space for robust solutions in an efficient way. The two methods can therefore be characterized by their computational cost, flexibility and accuracy. Two example problems are used to highlight the methods and their advantages. The use of measures of variation in responses is carried forward to be included in multi-objective optimization, so that robustness is naturally considered as a design objective.
展开▼
