An automated procedure has been developed for correlating thermal models with test or flight data. The procedure achieves model correlation by determining the best combination of changes to a model's linear and non-linear conductors in order to minimize the differences between model temperature predictions and measured data. This procedure has been developed for use with the SINDA Thermal analyzer program, but the methods are general and apply equally to other thermal programs, such as ESATAN. The method is based on Design Sensitivity and Optimization techniques and utilizes the solution sequences available in MSC/NASTRAN. The thermal model is reformulated into an equivalent structural model, which is then optimized using built-in MSC/NASTRAN routines to match specified results.