The aim of this study is to obtain quickly thermal information on some sensitive components of an electrical motor in a short lapse of time. The thermal behavior of this complex system is modeled and validated. The corresponding detailed model leads to the resolution of a linear system of differential equations of high order, but is not easy to handle for control purposes, due to long CPU time, In order to compute faster, two model reduction techniques are used: the Eitelberg method and the reduction by modal identification. The reduced models effectively give good temperature accuracy on the critical points of the motor with very fast computation time.