Double Roll Welding And Brazing Process Neural Network Control Research

The brazing process of double layer welded pipe temperature control system is anonlinear process control system with quickly response, it is difficult to get satisfactoryresults by using conventional PID control method, and the artificial debugging is moredifficult when disturbance is relatively large, in addition, it is easily far away from the setvalue. The neural network controller is widely used for its advantages of simple structure,strong robustness, not mathematics model of the controlled object, coarse quickly and soon. However, for the single neural network controller, it also has problems such as lowcontrol accuracy,"adjusting dead zone" and so on. At the same time, as the brazingprocess of welded pipe cross-sectional area, the factor of electrode contact resistancehave great influence on the temperature, we apply neural network control technology andBP algorithm with tutor rules, and improve BP algorithm, then design BP neural networkcontroller to improve the precision on the premise of not affect computing speed.Based on the familiar with production process of double layer welded pipe, itmathematically analyzes the mechanism of brazing process, gets the steady statemathematical model of the brazing temperature change, provids the theory basis for theresearch of control algorithm; through the research of control mechanism of neuralnetwork controller and algorithm, it proved the forward and reverse process of algorithmin controller and designed BP neural network controller. Through the steady-state analysisand simulation verification, it realized optimization and control of the main technologicalparameters in brazing process, like the welding temperature, welding pressure, weldingspeed and so on.In this paper, we design production line of welded pipe experiment, at the same time,we write temperature measurement and control software combined with MATLAB andKingview software. Using object-oriented animated graphics, open database format andenhanced trend analysis which provided by Kingview, it completes variable configurationand monitoring picture, realized the real-time curve display and can get previously data atany time through the format of query database to visit historical report forms anddisplayed on the screen. It is easy to print, save, query and as a reliable basis for furtherresearch, at the same time, Kingview as the interface with peripheral hardware, it realizesdata transmission with hardware peripheral and drives an external actuator; internalcontrol layer realized MATLAB and Kingview data exchange by OPC, and realizes advanced control algorithm by using powerful data processing of MATLAB and controlsystem toolbox.Experiment results show that, using BP neural network controller to control thewelding process not only has smaller steady-state motion, but also has strong ability toadapt the fluctuation of cross-sectional area, electrode contact resistance and other factors.