Development Of Machine Vision-Based Height Detection And Automatic Control Technology For The Head And Tail Cover Of Hot Rolled Slabs

In the roughing stage of hot strip rolling,the main shape defect of hot rolled slab is head warpage after rolling by roughing mill.The head warpage of slab will impact the work rolls and water-cooling device,and the buckling head will hit the stand rolls and rollway,which brings many disadvantages to the production.In the context of the manufacturing power strategy,intelligent steel production process has become an important way to promote the upgrading of traditional industries.To this end,by quantifying the slab warpage height online,then analyzing the factors affecting the slab end warpage,summarizing the law and establishing a mathematical model,we lay the foundation for the automation and intelligence of the hot rolling production process in the roughing section.To quantify the warpage height of slab end,the method of relying on human eye recognition cannot meet the requirements of production line stability and accuracy.The traditional machine vision inspection method is based on the edge of the slab for measurement,the shortcomings of the method is affected by the shape of the slab end,the measured warpage value is more downward buckling trend compared with the actual measured value,and even the actual warpage trend is the exact opposite of the situation.In this paper,based on the original visual inspection idea,we propose an improved method to indirectly measure the slab warpage value by fitting the curve with the laser curve as the reference,and complete the process of camera calibration,image acquisition and image processing with the help of Halcon software,and store the calculation results in the database to provide data support for the subsequent summary of factors affecting slab warpage and the establishment of a mathematical model based on neural network.The nine main influencing factors of slab head warpage are summarized,including entrance thickness,press down amount,lower roll line speed,upper roll line speed,upper roll diameter,lower roll diameter,upper surface temperature,lower surface temperature,and slab width,where the ratio of upper roll line speed to lower roll line speed is called roll speed ratio,denoted as SKI.based on the production big data,combined with the quantified warpage values,a data-driven 9-A 4-layer BP neural network model with10-6-1 is constructed and trained to predict the head warpage of hot rolled rough slabs.In order to facilitate the operation and visualization of the screen,the hot rolling roughing slab head warpage control software is written,which has the functions of cleaning and training production data,finding the slab history data,displaying the current slab and its warpage measurement value in real time,and giving the SKI suggestion value under the corresponding slab pass.The system has been successfully put into operation in a domestic hot rolling line,which greatly reduces the warpage of slabs after rolling in the roughing mill,creating conditions for online detection of warpage at the head and tail of slabs and unattended automatic steel rolling.