This paper presents a new methodology for identifying, by means of inverse methods, the convective heat exchange coefficients between an electric machine and ambient air. The inverse method that we use involves minimizing the residual between the machine’s internal losses and the total amount of heat extracted to ambient air in a thermal steady-state condition. First, we start by establishing the initial conditions for the optimization algorithm. Then, we introduce an optimal law of variation of heat exchange coefficients for different thermal phenomena. Finally, this approach is tested for different operating conditions to certify its validity using data extracted from tests performed on an internal permanent magnet synchronous machine.