Monitoring And Quality Prediction For Continuous Rolling Production Process

With the development of modernization,the applying field of seamless steel tube is enlarging.The demand of seamless steel tube is becoming much larger.The continuous rolling process as the second deformation process in the production of seamless steel tube plays an important role.At the same time,the quality of the rolled tube is also very important to the final production.The rolling process is very complex and has many characteristics,including multi-period,strong nonlinearity etc.The relationship between the continuous rolling process and process variables is very difficult to be determined by mechanism model.In this thesis,we use the methods of data modeling to build the correlation model between the continuous rolling process and process variables.The continuous rolling process belongs to the typical batch process,so we can use the modeling methods of batch process to build the process monitoring and fault diagnosis model.In this thesis we divide production data into several sub-periods by K-means clustering algorithm combined with production process,then we establish a continuous rolling production monitoring and fault diagnosis model based on multi-stage MPCA method.Simulation experiments show that the rolling production process monitoring and fault diagnosis model based on multi-stage MPCA method is effective.In order to predict the quality of the rolled tube,in this thesis we use SVM method to establish a quality prediction model.By the method of mixing the polynomial kernel with the RBF kernel we build a quality prediction model having a good learning ability and a good predict ability.Then we make the parameter optimization through cross validation.Simulation experiments show that quality prediction model is able to predict the transverse wall thickness of the rolled tube.Finally this thesis builds three-dimensional finite element model for the continuous rolling process of seamless steel tube.Through this finite element model we can emulate practical condition of field production and test the parameters obtained from soft measurement model.