Strip Flatness Recognition And Control Based On Intelligent Methods

In the process of industrial modernization, iron and steel industries are believed as one of the fundamental industries, while the development of production of cold rolled strip industry is also an important issue. Be able to produce the high-quality cold-rolled steel strip has become an essential conditions of national defense industry, automotive manufacturing and processing of high precision instruments. The shape is one of the important quality indicators of strip materials, with the increasing demand of high quality cold-rolled sheet and strip worldwide, the problem of shape controlling has become a worldwide issue. Solution is mainly involves three aspects of technology: First, the line shape flatness detection; second, shape data processing and defect recognition; Third, control theory, control methods and the corresponding control system to achieve real-time control of the shape.This article focuses on shape data processing regarding pattern recognition research and discusses the main problems existing at home or abroad, and then pointed out the dynamics and development trends in this field. Concerning different types of plate-shaped defects, the different shape defects classifiers are designed and do classification and identification. Using the section shape detector from Swedish company named ABB. we can do monitoring of the data line and simulation for roller cold rolling mill in the strip-line. Finally, we use the designed fuzzy controller to control the regulation of the identified amount of complete shape.The paper pointed out that in the shape recognition process, the orthogonal polynomial method and the least squares polynomial regression decomposition method show poor ability in anti-interference, can not determine the approximate size and the approximation order n accuracy is limited, can not meet the needs of high precision real-time control. Therefore, for producing the thinner hot-rolled strip, doing hot rolled strip shape detection and rapid development of cold-rolled strip flatness characteristics of high precision, the intelligent pattern recognition is introduced and let the recognition of shape to be high-precious, speedy, and achieving digital direction.For single strip flatness pattern recognition a Multi-Classification SVMs classifier in terms of the theory of SVM is presented and which can tell the various properties of panel surface. Theoretically, the supervised method of"one-class-against-the-rest"can be applied for training the Multi-Classification SVMs classifier and the sample data is obtained by preprocessing the data which is collected through the flatness detector in the cold-rolled operation. Finally, the performance of classifier should be checked by standard test data. According to the results of simulation process, the proposed method performs a very high recognition rate in processing of little sample data and also has a good ability of generalization. The classifier could obtain completely correct classified results so that it can achieve a successful application in recognizing the real shape. Basically, the method based on Multi-Classification SVMs provides an possible option in this case.For complex strip flatness pattern recognition, this paper presents a method which improves the BP neural network to preprocess strip defecation data. At the same time, it builds up a double hidden-layer BP neural network model to adjust the shape of sigmoid activation function. By comparing the data of this method with the formula Levenberg-Marquardt preprocessing method, it apparently can be seen that not only can the time of learning BP neural network be effectively reduced, but also the network generalization ability can be improved by this method, which is good to on-line identification of the strip defects.Combination of these two identification methods, we can do controlling towards the parameters with fuzzy controller. Also, using Matlab software for simulation purpose. The results show that the control system is satisfied in terms of recognition process and controlling issues. These methods are able to control the shape of cold-rolled sheet and strip shape accurately and efficiently. At the same time, it also provides a theoretical basis and practical method for on-line adjustment.In this paper, we choose the cold strip mill flatness pattern recognition and control subjects for research which have theoretical and engineering significance. It has the great value for development of the theory of shape control and the promotion of application of intelligent technology of pattern recognition.