Method Of Intelligent Recognition To The Hot Rolled Plate Shape Based On SVM

Since twenty-first Century, the rapid development of the global economy driven by thedemand for steel is greatly improved, and the progress of science and technology makes themost of the steel industry, the more and more high quality requirements. As two importantquality index of the strip, plate shape and thickness in recent years has been an important issueconcerned by researchers at home and abroad. With respect to the automatic gauge controlsystem (AGC) the improvement and extensive application, the flatness problem is particularlyprominent, recognition of shape and control technology has become the focus of currentresearch.As one of important parts of strip shape control, flatness pattern recognition results willdirectly influent the control effect of the shape quality. The shape principle defect recognition isidentified by a shape signal of flatness detection equipment acquisition, so as to determine theshape defect type which strip exist, and then to the back plate shape control links to reach thetarget shape required.Detection technology of hot rolled strip shape is the basis of flatness pattern recognition.Flatness detection method has many kinds, different from the cold rolling strip productionprocess, the hot strip production process of high temperature, poor environment, prone toinfluence the shape detection device on board, in addition to strip vibration, floating, there willbe interference, so the strip shape detection method for detecting often choose the non-contactshape. In this paper, to obtain the shape information we select multiple beam laser cross sectionmethod of flatness measurement technology.This paper introduces the shape recognition method of traditional as well as a variety ofshape intelligence recognition method. Because the shape control accuracy requirements aregetting higher and higher, the traditional method has been difficult to meet the controlrequirements, plus with products of hot rolled plate is thin, with the high-speed development ofhot rolled strip shape detection technology in recent years, the emergence of intelligentrecognition method of various shape, the shape pattern recognition to the development of highprecision, high speed, digital direction. Based on the traditional shape pattern recognition methods and intelligent shaperecognition methods, this paper presents a strip shape defect identification and classificationmethod based on support vector machine, using support vector machine (SVM) classificationmethod of a strategy to achieve the recognition of several common flatness defect, and theparameters optimization was carried out using the grid search method. Support vector machineis generally used to solve two problems, but the reality is often the multi-class classificationproblem, then support vector opportunities to take certain strategy to realize the multipleclassification, such as one to one and one to many strategies. The simulation results show that,the method is effective, combined with the flatness detection information, the simulationexperiment was made using LIBSVM toolbox of MATLAB, to achieve a correct classificationof six kinds of flatness defect.Taking the hot rolling mill flatness detection, flatness pattern recognition as the researchsubject, it has a certain practical significance for the development of shape theory, which hasimportant reference value to the application of support vector machine.