Research On Heat Transfer And Friction Signal Characteristics And Prediction Model Of Surface Longitudinal Crack Of Continuous Casting Slab

As the key link of iron and steel metallurgy,With the advantages of high production efficiency,good product quality and low energy consumption,the proportion of continuous casting in iron and steel metallurgy industry is increasing gradually every year.High quality,high efficiency and low energy consumption are the goals pursued by continuous casting technology,but the appearance of longitudinal cracks on the surface of slabs will affect the smooth production of continuous casting,and has become the main factor restricting the efficient production of high-quality slabs.Sometimes,a serious surface longitudinal crack can cause a serious breakout accident.Even though a small longitudinal crack appears on the slab surface,finishing treatment or some repair methods are also needed.Therefore,it is of great significance to investigate the characteristics of mould heat transfer and friction behavior and develop an identification model when the longitudinal cracks occur,which is of great help to improve the quality of slabs.So this paper researched from the following four aspects,as follows:First of all,the formation mechanism and main influencing factors of surface longitudinal cracks were analyzed,the heat transfer and friction signals during the occurrence of longitudinal cracks were obtained and their variation characteristics were investigated based on the on-line mould monitoring system.Secondly,based on the pouring data of a domestic steel plant,the temperature,heat flux,friction tension,friction pressure and transient friction fluctuation of 30 samples of surface longitudinal cracks were obtaiined and analyzed,the characteristics of heat transfer and friction behavior of surface longitudinal cracks were revealed.The results show that when the longitudinal crack occurs,the heat transfer and friction signals will have a response,and the distribution range of relevant characteristic parameters is obvious.Again,based on the pouring data of a domestic steel plant,we selected 30 cases of surface longitudinal crack samples and 90 cases of non-surface longitudinal crack samples.And a total of 42 features are extracted from three adjacent rows of thermocouples,which includ the temperature amplitude,the temperature rate,the temperature velocity and temperature standard deviation.We use these data to build the initial feature database and the PCA dimensionality reduction algorithm was used to reduce the initial database from 42 to 7 dimensions,which was used as a basis for judging the longitudinal cracks on the surface of continuous casting slab.Finally,based on Python language,the surface longitudinal crack identification model basede on PSO-XGBOOST were established respectively,and the model was tested after training.The test results are compared with the Decision Tree algorithm,XGBOOST algorithm and GBDT algorithm,The results show that the Accuracy of PSO-XGBOOST is 95.83 %,the f1＿score is 92.31 %,the False Alarm Rate is as low as 5.56 %,and the Alarm Rate is 100 %.And the evaluation indexes of the model are better than those of the other three models,which provides an accurate method for identifying the longitudinal cracks on the surface of continuous casting billet.