Board-level simulation has to consider, at the earliest stage of the conception, the impact of the vicinity of numerous high and medium powered devices. In 1996, the concept of a surrogate, or Compact Thermal Model wass defined by the European consortium DELPHI to minimize the computation times, from days to minutes. A DELPHI-style CTM resumes an electronic component as a simple cuboid form and a network of resistors that link a single temperature-sensitive node to major surfaces of heat extraction. Our latest improvements made to generate steady-state models for System-In-Package devices showed that the number of boundary condition scenarios, such as JEDEC 38-set, is quite prohibitive when several sensitive nodes need to be monitored. Moreover when the package topology is asymmetrical, a clever definition of a set of isothermal major surfaces of the derived surrogate model is henceforth mandatory for reaching a good model agreement. Thus the present work investigates the use of fractional factorial experiment to define the lowest number of numerical simulations as well as a smart external-surface subdivision based on Genetic algorithm fitting technique. Then the coupled approaches are applied on a realistic case with the aim to generate the most compliant Boundary-Condition-Independent thermal network. The comparison with experiment highlights the performances of the deducted steady-state Compact Thermal Model.