Method of Surrogate Model Construction which Leverages Lower-Fidelity Information using Space Mapping Techniques.

摘要

A new method of surrogate construction is developed and applied to a pair of computational tools used in the field of aircraft design. This new method involves the pairing of data sampled from the analytical model of interest with the execution of a similar analysis performed at a lower level of fidelity. This pairing is accomplished through the use of a space mapping technique, which is a process where the design space of a lower fidelity model is aligned a higher fidelity model. The intent of applying space mapping techniques to the field of surrogate construction is to leverage the information about a system's performance present at a lower fidelity level to bolster the predictive accuracy of a surrogate model based upon sampled data at a higher fidelity level. The results from the pairing of computational tools used in this research show modest gains in predictive accuracy for many of the cases investigated when compared to existing surrogate methodologies.