Data interpolation based on neural network for optimal design of interior PM synchronous machine

This paper proposes Neural Network (NN) based data interpolation and design optimization algorithm for an Interior Permanent Magnet Synchronous Machine. Data interpolation using NN is suitable for estimating the performance of an electric machine, because NN is approximate function for representing nonlinear data. To utilize NN as an approximate function, training process is required. After training process, optimal design of an electric machine can be found by applying search algorithm, such as Particle Swarm Optimization(PSO), Mesh Adaptive Direct Search(MADS) etc. This procedure does not demand any additional numerical analysis of electric machine based on finite element method, allowing search of optimal model in such short computation time.