Advanced Thermal Model Correlation Using Reduced-Order Models

Abstract

Uncorrelated thermal models are often based on an assortment of uncertain input parameters. Examples include interface conductance based on handbook values and estimated insulation e* values. Because of this, it is hard to know which combination of parameter values accurately predict reality. To overcome this, thermal model correlation uses test results to better estimate and validate these uncertain inputs. This process involves institutional knowledge and an iterative �guess and check� method that quickly becomes time-consuming and costly.

To overcome these challenges, reduced-order models (ROMs) were used. ROMs provide computationally efficient surrogates of high-fidelity thermal models (eg. Thermal Desktop� models). An approach for creating these surrogates using efficient sampling and data fitting was developed and successfully applied to a broad range of spacecraft applications. This approach provides numerous benefits including computational speed. Leveraging this speed, ROMs can be used to calibrate thermal model parameters to experimental test data using an automated, repeatable, and simple methodology.

This investigation will explore the use of ROMs for thermal model correlation. Details of the methodology behind this process will be provided. In addition, the computational efficiency and accuracy of this approach will be evaluated for a broad range of problems. Finally, several examples will illustrate how this method can be used in practice.