According to the world wide safety standards, a classification for transport and handling is required for every single mixture. To reduce the experimental effort to determine physical properties for the classification of multicomponent mixtures of polyurethane raw materials, existing predictive methods for the initial normal boiling point and for the flash point have been extended based on group contribution methods like UNIFAC and modified UNIFAC (Dortmund). A new predictive method for the auto ignition temperature has been developed. The calculation results were evaluated based on experimental data for more than 1500 mixtures. It could be shown that these approaches can be reliably applied for the classification of multicomponent mixtures of polyurethane raw materials. (C) 2015 Elsevier B.V. All rights reserved.