The use of neural networks combined with FEM to optimize the coil geometry and structure of transverse flux induction equipments

A method is presented to optimize transverse flux induction heating (TFIH) inductor for a uniform temperature distribution. There were two neural networks used for eddy current and temperature field prediction respectively. The trained networks used for tested examples show a reasonable accuracy for the prediction, and then can be used for two purposes. One is to provide a good guessed value of the temperature dependent parameters for each finite element and an initial value for temperature field solution, which speeds up the iterative solution process for the nonlinear coupled electromagnetic thermal problems. The other is to be used in the optimization process.