Estimation of mixed heat transfer coefficients on complex CFD-simulation models of 5-axis machine tools/ machining centers poses a whole new set of challenges. The enclosure around the machine which protects the user from chips and machining-process is also responsible for creation of an additional environment which has coolant supply, chips and moving air due to temperature changes and oil-emulsion exhaust. This results in a highly complex CFD model which has inner and outer environments along with multiphase fluid domains. The computational effort required is extremely large. The split-simulation approach divides the simulation model into internal and external environments. The natural convection is emphasized for external environment model and forced convection for internal multiphase environment model. This division significantly reduces the computational efforts involved by neglecting uninfluential solid bodies, cavities etc. and thereby the model size. The external and internal influences are mapped separately onto natural and forced convection coefficients using neural networks, which can be used to predict the combined heat transfer coefficients for any load cases within the training boundaries. The temperature fields obtained from split-simulation model are validated with thermography and temperature sensor readings taken from the machine in the form of a use-case study.