返回Google圖書搜尋

Pre-setting Improvement of Skin Passing Concerning Flatness and Reduction Combined with Defined Strip Surface Texture by Hybrid Modelling Techniques

其他書名	Final Report
出版	Publications Office of the European Union, 2004
URL	http://books.google.com.hk/books?id=moEgyQEACAAJ&hl=&source=gbs_api

註釋This project is aimed at improving the quality of the strips after the temper mill, by improving the set-up generated by the control computers before the material is rolled. ln this project, techniques of neural networks mixed with analytical models were used to develop hybrid models that are able ta take advantage of performances with the best results. Several aspects of skin-pass rolling, such as flatness, threading phase, mechanical characteristics (elongation and final roughness) were taken into account. Hybrid models were tested, and demonstrated their usefulness in improving existing models, especially when treating non-linearities present in these models. Several hybrid models based on the Ekelund and Maag models were used to establish the roll-force pre-set in order to obtain good flatness and save material out of elongation when threading. By using neural networks, a new model for pre-sets was developed in order to obtain tension and roll-force pre-sets. This is to maintain the most suitable pre-set for elongation and roughness. A model for roughness prediction was also obtained. These methods have some drawbacks: first of all, the need for considerable quantities of training data that are not always easy to obtain, especially for unusual rolling materials; secondly, to pre-process data and train the model, it is necessary ta supervise the conclusions with expertise. This drawback becomes more important when new materials appear and retraining is necessary to update the model.