Cutting the costs and increasing the added value of steel products using new production methods and advanced control systems are the key factors in competitiveness of the European steel producers. In order to meet the challenge of the steadily growing pressure to improve the product quality, rolling mills employ extensive automation and sophisticated on-line data sampling techniques. Since the number of factors involved in the processes is very large, it takes time to discover and analyse their quantified influence. The paper gives a survey about the knowledge processing, using neural networks in rolling. The two main streamlines are shown by exemplary case studies: Self Organizing Maps as Data Mining tool for discovering the hidden dependencies among the influencing factors, finding the relevant and irrelevant factors, as well as application of different types of neural networks for optimisation of the draft schedule.