Strip Coiling Temperature Accuracy Of Prediction Method

The control accuracy of the coiling temperature of the hot strip laminar cooling system affects the strip's structure and mechanic properties directly, which is the crucial factor to guarantee the better quality and flatness of strip. Today, many countries also use the hot strip laminar cooling system to control the coiling temperature. Therefore, it has a profound realistic meaning towards the analysis and research on hot strip laminar cooling system. The control accuracy of the coiling temperature of the hot strip laminar is main due to the accuracy of the predictive sections.This paper takes a laminar cooling system of a domestic hot strip factory for background, and has a deep study for how to raise the accuracy of cooling temperature of the hot rolled strip.As the mathematic model of hot strip laminar cooling system is based on the theory of calorific, so the paper first analyze the models of heat transfer, and particular analyze the mathematic model of the process of this hot strip laminar cooling system. This mathematic model mainly includes water cooling model and air cooling model, its calculated precision influences the last effect of control cooling. So the proper heat transfer model is very important thing which can improve the accuracy of the coiling temperature.The controlling process of the hot rolled strip is nonlinear, strong coupled and indeterminate. So the traditional modeling technology could hardly improve the precision of hot rolling model and more. In order to satisfy the demands of high precision of the coiling temperature, a genetic-neural network method to predict coiling temperature based on Data Mining was put forward. To make full use of the association-analysis capability of Data mining, the approximation capability of neural network and the liability to fall into the global optimum solution of genetic algorithm, and newly used the association rule mining method to establish the model of the neural network, combined the traditional model with BP algorithm which optimized by modified genetic algorithm. The overall-find capability of the algorithm can avoid the weak point of the BP network and also make the nonlinear approaching ability of multi-layer feed-forward network more effective. This algorithm improved the performance of neutral network by applying it to neutral network. Some referential thoughts and ideas have been put forward for BP neutral network. The simulated results show that the integrated model can satisfy the high accuracy demand of the prediction of the coiling temperature and laying a favorable base for furthering application online.