During conceptual design, designers need tools to help improve design decisions and reduce design times. We are working to develop techniques to create Bayesian surrogate models that respond to designers' needs during conceptual stages of the design process. Bayesian surrogate models give analytical form to the overall performance of a system and can evolve along with the design. Bayesian surrogate models provide a mathematically rigorous framework in which computational models can be updated based on previous outcomes. In this paper, we present techniques that allow the addition or suppression of parameters without discarding previously obtained information. We also present a case study that illustrates how a surrogate model is constructed in stages when parameters are added or suppressed during the design process. Visualization tools, such as plots of the main effects of parameters, can be derived from surrogate models. These tools can be used to provide knowledge about the parameters that influence the design. Finally, a design problem is used to illustrate how Bayesian surrogate models can inform the designer about tradeoffs that would not be apparent from simulation data alone.
展开▼