The Online Pattern Recognition Research And Its Detection System Based On GAPSO-BP Algorithm For Higher Order Flatness

Iron and steel industry,as a pillar industry of the national economy, is a symbol of measuring a country’s comprehensive strength. The production of high precision strip is the focus of iron and steel industry’s development. Therefore, how to improve the technological level of strip material manufacturing has always been the focus of the attention of industrial enterprises. Flatness measurement and control is an important part of the cold-rolled strip production process. This paper has relied on the National Natural Science Foundation of China and taken the flatness, online pattern recognition and flatness control strategy as research subject. It also researches on flatness refinement control and signal acquisition and processing and gains certain achievements.This paper has taken the industrial pressure shape meter as the research subject and established the testing roller and force of the plate with the finite element model. Under different tension, changing thickness and angle, the effects of testing roller deformation for flatness measurement are researched. Through the comparison between the distribution of flatness internal tension and the distribution of the force for testing roller, it’s known that utilizing proper Compensation curve, the distribution trend of flatness internal tension and the pressure distribution for testing roller are almost the same. The way of using the pressure distribution of testing roller can judge the flatness problems online.For the multidimensional nonlinear problem of flatness fine control, it conducts high-order flatness recognition system modeling and researches the flatness defect formation law described by high-order nonlinear system. On the basis of ensuring orthogonality, it uses higher order Legendre polynomial as plate-shaped basic pattern recognition and recognize 6 times flatness defect as to adapt to flatness fine control. By studying the genetic algorithm and hybrid algorithm of particle swarm optimization, optimize the BP neural network structure and weights and thresholds and establish GAPSO- BP neural network on-line identification model. The problems, such as the slow convergence speed of BP neural network, the strong initial weights influence, easy to fall into local minimum and difficult to determine the set structure, are solved. Through experimental calculation, it’s shown that anti-interference ability and self-learning GAPSO-BP neural network adaptability are strong with high precision, which can provide reliable evidence for the formation of flatness control strategy.Using GAPSO-BP neural network model to conduct pattern recognition for the measured flatness, conduct classification research for flatness defect and develop appropriate control scheme. The analysis shows that the reorganization of this system is high, which can be used for online flatness loop feedback control.In the light of the characteristics of high flatness gauge piezoelectric sensor, it has taken use of cable signal acquisition and transmission technology, has designed a slip ring cable signal acquisition and transmission system, and developed flatness signal huge amounts of data acquisition and signal processing system and the corresponding module. The acquisition process is displayed on line for analyzing and processing.