There is widespread use of computer models as tools in scientific research. As surrogates for physical or behavioral systems, such models can be subjected to experimentation, the goal being to predict how the corresponding real system would behave under certain conditions. For long-running expensive model codes, there may be a severe limitation on the number of experiments that can reasonably be done. This motivates the construction of a fast-running cheap approximation to the original code, for use in experiments where a large number of runs may be necessary. Here we discuss our approximation of a simulation model for the compression molding of sheet molding compound, applied to the manufacture of an automobile hood. The approximation was constructed using Bayesian interpolation methods for prediction of the movement of the flow front. The predictions were based on data generated by a sequence of computer experiments, using designs chosen according to a type of D-optimality criterion.
This article is from 'Computing Science and Statistics: Proceedings of the Symposium on Interface Critical Applications of Scientific Computing (23rd): Biology, Engineering, Medicine, Speech Held in Seattle, Washington on 21-24 April 1991,' AD-A252 938, p272-277.