Center Segregation Analysis And Quality Prediction Research Of Casting Billet

Continuous casting is an important part of metallurgical production process, the quality of its production is also one focus people concern. This thesis mainly deals with No.65 steel of No.1 casting machine in Fujian MinGuang Steelworks Plant and has made a deeply research on the main mass of defect (center segregation) of No.65 steel. Based on kernel principal component analysis method (KPCA), the analysis and forecast module has been established. Used process data of No.65 steel were collected from the scene of the continuous casting to test the application of analysis and forecast module. And the work has been done above aims to provide advice to optimize the process values and dispose the billet with quality problem.Firstly, the center segregation was deeply researched in theory and main influence factors and their influence ways were found out. The data were obtained from factory in pre-process measurement and with them the KPCA module was established. As kernel function parameter was very important for application effect of the analysis module, particle swarm optimization method (PSO) was used to find the best kernel function parameter. According to PSO, the best kernel function parameter was determined 7.56. Based on the work having done above, the analysis module was established to determine the threshold value of SPE and Hotelling T2, which were the key indicators in determining whether center segregation would happen or not.Secondly, judge function was put forward to determine the quality grade of billet, using the ladder parameter combining with the weight of process parameters to achieve the prediction of center segregation grade. The link between parameters and segregation grade was determined by classifying the process parameters. Combination weighting method combining with No.65 steel’s characteristic was used and every process parameter’s weight was determined by relative literature. Based on the two portions above, the prediction of center segregation grade was achieved. For the billet with the center segregation, contribution figure was used to find the un-normal parameters to avoid segregation happening again.Finally, based on GUI of MATLAB,the applied platform of center segregation monitoring system was built, realized the visualization of center segregation analysis and forecast module. According to the test of the forty macroscopic sample data, the probability of segregation occurrence was 100 percent and the hit rate of the prediction module was 82.5 percent. That indicated that the center segregation prediction module was correct, which can be a good foundation of subsequent field application.