Research On A Strip Leveller Intelligent Control System Based On Laser Measurement

The flatness of plate is the symbol of country' industrialization.Levellers are the main equipments of improving plate flatness. The choice of technical parameters during the straightening process is the key factor to determine plate flatness after straightening. And the researches are concentrated on parameters choosing autumnally using intelligent methods. With the requirements increasing of the plate flatness and the automation of leveller control systems, there is an urgent need for a more precise and a higher degree of automation technical parameters selection method to replace the traditional manual parameter selection. This paper researches the methods of leveller intelligent control system based on laser measurement and its related technologies under the support of the national important technology products trial production scheme①. The main jobs of researches include the following sections.1) By comparing analysis of a wide range of the plate surface flatness measurement currently at home and abroad, researching the information of the plate surface flatness and a wide range of digital image processing technologies, select optical flatness plate based on the structure of optical detection techniques as a basis program.2) According to the features of straightening process, it abstract the three major characteristics from the empirical knowledge in the process of straightening samples-incremental, batch and online. During the comparise of interpolation method, artificial neural networks and support vector machine regression learning algorithm, the paper optimizes the sample learning process.3) The paper researches the technologies of leveller technical parameters selection using empirical knowledge. It abstracts the problem of technical parameters selection to searching for the optimal vector in a multi-dimensional state-space, searches for the best technical parameters using the good ability of global optimization in genetic algorithm. And the paper compares the performance of experiments on five technical parameters intelligent selection including, expert system, neural network, incremental SVR (Support Vector Machine Regression), batch SVR (Support Vector Machine Regression) and SVR-GA (Support Vector Machine-Genetic Algorithm).4) The paper develops the theory model of laser measurement equipment. It makes great sense to the model which includes mechanical shading devices, controllable optical imaging collection devices, wide power regulated supply devices, display control part.5) The paper develops leveller intelligent control system based on laser measurement. The system contains two sets of detecting equipments and technical parameters selection systems. The development of detecting equipments of import and export makes great sense to the optimizing technical parameters, it Lays the foundation for leveller technical parameters selecting.Through the researches, the main conclusions include:1) According to the theory of multi-source structure optical light, improves the laser triangular plate measurement by calculating the elongation according to the internal relationship between the lasers, eliminates the errors caused by plate vibration and bounce. Thus, a new steel plate obtain method has been put forward which is suited for the application of automate production. The method solves the certain errors brought by plate vibration and bounce and provides the basis and foundation for the system further design and development.2) The system includes the red and gray images mainly, in order to extract the red laser, in accordance with digital image processing technologies, the paper presents a image edge detection algorithm based on gray-scale which improves edge threshold and marginal of Canny algorithm. The results provide a precise edge for the elongation calculation.3) By solving the quadratic program, constructs a new mathematical model for sample learning based on support vector machine theory. The paper puts forward a new BIO-SVR (Batch Incremental Online-Support Vector Machine) algorithm which solves the problem of incremental, batch and online issues in the sample learning process.The method improves the automation of plate leveller production.4) According to BIO-SVR algorithm, constructs a new parameters selection model of key algorithm. That is, S:(entrance flatnessρin, technical parameters Ptec)→export flatnessρout. The paper puts forward to technical parameters selection model based on SVR-GA hybrid algorithm. The model makes BIO-SVR trained learning machine algorithm with a large number of samples as the fitness evaluation function of SVR-GA algorithm, plays the genetic algorithm global optimization capability and finds out the optimal technical parameters.The paper completes hardware and software systems of the plate detection system and has applied for patent protection. This system has been applied in the Tangshan iron and steel plant line straightening machine automation and control projects. The results shows that the system has good accuracy, practicality and scalability of high theoretical and practical value.