Solution of inverse heat problems in electrical machines with noninteger models

This paper presents a methodology for modeling interface problems occurring in heat diffusion processes with fractional models, characterized by their noninteger order. This particular model is used to solve inverse heat problems by a regularized inverse convolution. Fractional models are obtained from a noninteger integrator associated to a state space representation. Identification is performed in the time domain with an output error algorithm. This methodology is applied to the dynamics of heat transfer in the windings of electrical machines, between heat flux and temperature. Numerical simulations show that two types of problem can be solved by these techniques, like identification and inverse problems.