Towards a Digital Twin Enabled Multifidelity Framework for Small Satellites

Abstract

In this work, a multi-fidelity framework for the simulation ofsmall satellites is investigated. Taking into account the conceptof digital twin, our work focuses on handling a constantstream of live data. Towards this end, current multi-fidelitymodelling methods and low fidelity surrogate models for timeseries were surveyed. A multi-fidelity approach is used tocombine a low fidelity surrogate model with a high fidelitymodel. As a high fidelity model, a previously investigatedfinite element model is assumed. As a low fidelity model,auto-regressive and recurrent neural network-based modelsare investigated. Through cokriging, the low fidelity data iscorrected by the high fidelity data through a comprehensivecorrection, where the parameters are given through Gaussianprocesses in order to perform uncertainty quantification. Asan application, the thermal simulation of a small satellite, andthe use of this framework in conjunction with sparse telemetrydata is proposed. This online, statistical approach aims toprovide a tool for performing fault detection.