Optimal surrogates selection for embedded, hierarchical multilevel aircraft models

This study proposes a methodology that reduces the memory size of hierarchical multilevel embedded models while keeping its structure and satisfying constraints on accuracy and computation time. Based on a choice among surrogates (high dimensional model representation, neural networks, etc.) associated with each submodel, an overall hierarchical multilevel model that fulfills avionics systems requirements is provided via the resolution of an integer programming problem. This methodology is illustrated on a fuel model used for aircraft performance estimations.