Sensitivity analysis of coupled variables in integrated STOP models

The computational complexity of integrated structural-thermal-optical performance models, particularly for one-of-a-kind telescope missions, often limits the maximum feasible number of analysis cycles. Applications requiring a large number of model evaluations, such as uncertainty quantification and sensitivity analysis, can become infeasible if the number of required iterations for the analysis exceeds the allocated computation time or budget. Another major obstacle to the sensitivity analysis of integrated models is the dependence between variables, as correlation between the variables changes the contribution they have on the variance in output model predictions. This paper presents a multifidelity integrated modeling methodology that incorporates both high-fidelity and low-fidelity models to reduce the model’s computational cost while retaining sufficient accuracy for sensitivity analysis. Additionally, an estimation of the closed sensitivity indices is presented for the fully integrated model, taking into account the dependence between variables in structural-thermal-optical performance models.