Speeding Up Model Correlation With Breakout Models: the VIPER Integrated Thermal Model Correlation Story

Correlated integrated thermal models (ITMs) of spacecraft are needed to verify that hardware will stay within their Allowable Flight Temperature (AFT) limits. Unfortunately, the model correlation process is slow. Large models with hundreds of sensors can take several months to correlate. To speed up this process, many have pursued developing and using algorithms that find an optimal set of specified model parameters that minimize error. However, these algorithms still require models to run for many iterations and cannot address model deviations due to missing contacts or oversimplified geometries. Additionally, past projects have used engineering judgment to divide up correlation activities between multiple analysts. However, the problem of ‘how to split up the correlation activities’ is challenging. The Volatiles Investigating Polar Exploration Rover (VIPER) thermal team was recently faced with this problem. In this paper, we present Veronica, a new approach developed by the VIPER thermal team to speed up spacecraft ITM correlation by parallelizing the process as much as possible. Veronica leverages the fact that vehicles often have different thermal zones that are mostly independent of each other. The key to the Veronica approach is splitting up the ITM, according to a specific criterion, into many faster running “breakout models”. By correlating these “breakout models” in parallel, the overall correlation effort can be sped up. Using Veronica, the VIPER thermal team was able to use 14 analysts to correlate the ITM within 22 weeks (spanning Thanksgiving, Christmas, and New Years). After only updating the VIPER ITM with changes used to correlate the “breakout models”, we were able to achieve an overall Root Mean Square (RMS) error of 7.4C for hot thermal balance and 10C for cold thermal balance. Out of 37total key components, 54% had RMS errors within 5C and 81% had errors within 10C.