Automatic learning techniques for on-line control and optimization of transformer core manufacturing process

In this paper, a novel computer based learning framework that has been developed and applied for the online control and optimization of transformer core manufacturing process is presented. The proposed framework aims at predicting core losses of wound core distribution transformers at the early stages of transformer construction. Moreover, it is used to improve the grouping process of the individual cores by reducing iron losses of assembled transformers. Three different automatic learning techniques (namely decision trees, artificial neural networks and genetic algorithms) are combined and their relevant features are exploited.