This paper has a twofold objective: (a) optimizing the production process of individual cores using Taguchi methods, and (b) reducing the iron losses of assembled transformers, using neural networks. More specifically, we demonstrate the ability of the Taguchi technique accurately to characterize and successfully to optimize the transformer core production process with the minimum of experiments. Moreover, neural networks have been applied to predict iron losses of wound core distribution transformers at the early stages of core construction. The intelligent iron loss model is on-line applied in order to optimally combine the individual cores so that the iron losses of assembled transformers is reduced.