Design Of Flatness Recognition Model Based On RBF And Implementation By FPGA

With the development of society,the demand for high-quality strip steel in various industries is becoming more and more urgent.Flatness is one of the important indicators to measure the production level of strip steel.In order to produce standard strip steel,the recognition and control of flatness must meet certain accuracy and real-time requirements.In recent years,with the rapid development of artificial intelligence and deep learning theory,new vitality has been injected into many traditional industries,and data-driven modeling methods have been widely used.Under the engineering background of flatness recognition,the conventional flatness recognition model based on radial basis function neural network is improved in this paper.And the realization technology of the flatness recognition model based on RBF neural network with field programmable gate array is further studied.Firstly,the flatness recognition model based on RBF neural network is deeply studied.Aiming at the limitation of its artificial feature extraction,the principal component analysis method is introduced to improve the flatness recognition model.So the flatness recognition model based on RBF neural network using PCA as feature extraction method was constructed.The feasibility of the method is verified by the simulated strip shape data and the measured strip shape data of a 900 HC reversible cold rolling mill respectively.The simulation results show the effectiveness of the method.Secondly,drawing lessons from the idea of reducing the modeling scope to improve the accuracy of multiple models control theory,a flatness recognition model based on multiple RBF neural networks is established.In this model,each sub-network is optimized by genetic algorithm to refine the effect of model input on each flatness characteristic parameter.The simulation results show that the recognition accuracy of each feature parameter has been improved,thus achieves the goal of improving the recognition accuracy on the whole.Finally,on the basis of theoretical research and practice,this paper maps the flatness recognition model based on RBF neural networks which established in MATLAB to ISE.The simulation results show that the accuracy and real-time performance can meet theactual engineering needs,and prove that the FPGA is a feasible and reliable model implementation technology,which lays a foundation for the design and application in actual production.