Coiling Temperature Prediction Of Strip Based On Wavelet Neural Network

With the development of hot rolled strip production technology,there is an increasing de mand of the market for the quality of the strip.As the last step of coiling the strip,laminar cool ing directly determines the microstructure of the strip and strip quality in terms of the controll ing the precision of the strip temperature.Based on the problem that,in hot rolling mill of a d omestic steel plant,the accuracy of strip coiling temperature control at the laminar cooling sta ge could not be improved,this paper researches the methods of improving of the accuracy of t he temperature prediction.After the analysis of strip cooling principle,there is an indepth analysis of mathematic m odel of laminar cooling process in this paper,researching various factors which impacts the ac curacy of temperature control at the laminar cooling stage.On the basis of the above,the pape r also studies the methods of improving the accuracy of coiling temperature control,with the main results shown as following:(1)The strip coiling temperature prediction model is used to describe the process of laminar cooling,during which the process is extremely complex and many factors could affect this.If the whole model were regarded as the input,it may lead to a large amount of calculation due to the input of overmuch sample dimension.As a result,the paper,firstly,simplifies and reduces the input sample data related to the coiling temperature prediction by using the method of rough set attribute reduction.Then,the paper uses the reduced data as the neural network input in order to get better model prediction.(2)Combining wavelet analysis with neural network can not only avoid the defects of BP neural network,but also reflect the advantages of wavelet analysis and the competence of generalization of neural network.After taking the attribute reduction,through repeated experiments and result analysis,the paper finds out that,compared with BP neural network,wavelet neural network has better approximating ability,fault-tolerant ability and faster convergence,which can effectively improve the performance of neural network.Through learning,training and simulating of the data of the laminar cooling stage in the hot rolled strip plant,the paper comes to the result that strip crimping temperature prediction,based on the intelligent model,can meet the factory’s high requirement of the precision of prediction,which proves that this method,used in strip coiling temperature prediction model is feasible.