Study And Application On Artificial Neural Network Based Rolling Force Model

Rolling process control technology of cold rolling plate and strip is the key to improve the quality and throughput of production. Mathematical models which describe the rolling process are the basic of process control, and their precision decides the precision of process control. Whereas the factors affected rolling process are complicated and change frequently, it's difficult to depict with mathematical models, the application of artificial intelligence to process control model research is a very effective way to promote the process control technology.The rolling force model of cold rolling is the core of cold rolling model system. Taking rolling force model as study object, and starting from rolling theory, this paper analyzes the characters of rolling force model on the cold rolling condition, and discusses Bland-Ford-Hill rolling force model that applied widely in engineering research. Moreover, the impacts of primary factors upon rolling force are labored.Then, based on the mathematical model study of rolling force, as well as the plant data of a 1450mm eight high tandem cold mills, a 7-12-1 topological neural network model of rolling force is established with BP (Back-Propagation) neural network. Aimed at the disadvantage that BP algorithm traps in local minimum facilely, ant colony algorithm is used to optimize the weights of BP neural network, and the predicted precision of rolling force is improved. In addition, the superiority of the ant colony algorithm based neural network model of rolling force is proved by learning the network offline.Lastly, we do a research on the application of the neural network model of rolling force to load distribution optimization for cold rolling. Instead of the mathematical model of rolling force, we use ant colony algorithm and take full advantage of its features including distributing and parallel computation, and global search to optimize the load distribution of a 1450mm eight high tandem cold rolling. Also, take equal relative load of each stand as objective function, and under the restriction of equipment qualification together with technology condition, the best load distribution is obtained.