Study On Intelligent Prediction Model And Optimization Design Of Hot Rolling Process Besed On FEM

In the steel industry,hot continuous rolling is one of the most important methods for producing steel plate and strip with high efficiency and good economic benefits,which occupies an important position in the nation's economy.Hot continuous rolling is a very complex process and needs to keep a continuous rolling constant for the tandem mill to produce the required shape,microstructures and performances.However,the temperature field and the stress-strain field of rolled piece change non-linearly during the rolling process.It is difficult to make the process parameters satisfy the rolling conditions.Therefore,reasonable process parameters are essential for the rolling production.It is not only crucial to ensure the quality of hot continuous rolling products but also important for improving productivity and reducing energy consumption.The purpose of this study is to establish a system for the intelligent analysis and optimization design of the hot rolling process,which makes the analysis quickly and efficiently.With this regard,FEM and artificial neural network were combined to establish the relationship between the rolling conditions and the temperature field,stress-strain field.The rolling processes for single-pass of steels such as 20,65Mn,304 and X80 etc.were thermo-mechanically coupling analized by using FEM.NALU algorithm was applied to establish the intelligent prediction model.In addition,continuous rolling process analysis and optimization design system was builded with dynamic programming optimization algorithm.By simulation of the above-mentioned four kinds of steels,it was analyzed the influence of such parameters as material property,rolling temperature,reduction ratio,front-back tension,initial stress field,friction,and rolling speed on temperature field and stress-strain fields.The predicted results are well consistent with the real preduction,which will provide reliable data for the intelligent prediction models.The established intelligent models includes three NALU layers.The input layer has 22 neurons corresponding to the material parameters,rolling conditions and positions,repectively.The output layer has 10 corresponding to the temperature field and stress strain field.The neural networks were trained for entrance zone,deformation zone and exit zone.Considering the simulation scale,the established models are suited for the prediction of temperature field and stress-strain field for plate rolling on larger sized rolling mills.In addition,they are universal and can be applied to other materials.Based on the intelligent prediction models,the rolling process of 304 steel on a five-stand continuous rolling mill with the roll diameter 1350 mm was optimized using dynamic programming.The reduction rate for each pass is 56.93,44,48,30.49,21.00 and 14 mm,respectively,and the rolling speed 22.4,31.7,44.4,61.8 and 84.8 rpm,respectively,which is consistent with the data for actual production line.This work shows that the combination of FEM and artificial neural network can avoid the calculation difficulties resulting from the geometric nonlinearity and the material nonlinearity in the simulation of multi-pass rolling.The established intelligent analysis and optimization design system is effective to analyze the continuous rolling process and of practical values for the development of new products.