The thermally optimized design of machine tools is a prerequisite for both high accuracy and high productivity. Numerical prediction of the temperature field is therefore an essential piece of the puzzle in the development and design process of the machine. Especially in the early stages of development, little data is available and only rough assembly dimensions are known. As the design process progresses, the level of detail increases and thus the model requirements change, but so does the information for more detailed modeling. Neither high-resolution FE models nor data-driven models can meet these evolving requirements. This article presents a hybrid approach based on a lumped parameter thermal network and the incremental integration of high-resolution FEM assemblies. The model approach allows both the qualified analysis and optimization of the temperature field of the machine tool in the early project phase by using simple pre-parameterized basic elements, and the consideration of fully designed components in the later course of the project. The model was verified experimentally on a rotary-tilt table and used for design considerations regarding the layout of cooling channels. The study showed that the steady-state temperature of the assembly could be reached much faster and reduced by up to 38%.