The Simulation Study Of Medium Density Fiberboard Glue System Based On Predictive Control Of Neural Network

With its excellent plate performance and efficient use of timber resources, medium density fibreboard (Medium Density Fiberboard) has become one of our most promising market wood-based panel products. Sizing is a very critical step in the process of Medium Density Fiberboard production, and the amount of raw materials(fiber, wood shavings and other production materials) and the amount of glue needed should always maintain an fixed ratio. Accurately control the amount of sizing will not only improve the performance and quality of the product sheet, but also can reduce the consumption of raw materials and the formaldehyde emission, which is important in energy saving and environmental protection as well.In the actual industrial production process, medium density fiberboard (MDF) glue system is a typical nonlinear controlled object with characteristics of time-dependent nature and coupling, coupled typical nonlinear controlled object. It is unable to use accurate mathematical model to describe this system. Furthermore, the conventional linear system control method is also very difficult to achieve the precise control of the amount of sizing.In the specific analysis on the basis of medium density fiberboard sizing process, this paper presents a neural network predictive control method. Then, according to the operating principle of sizing system, and establish a system of open-loop model. Secondly, designed two different control structures based on the open-loop model of sizing system. One is the neural network model predictive control structure, the other one is the generalized predictive control structure of the neural network. Both programs can achieve the structure and identification of sizing system predictive control structure model by neural network. It can overcome the difficulties of establishing a precise mathematical model, and make it easier for applying traditional predictive control into nonlinear systems.Finally, for both the feasibility and the control effect, the simulation is done by Matlab\Simulink, and compared the results with the traditional PID control method. Simulation results show that the design of the two control programs can achieved the glue fast track according to the dynamic change of fiber amount, compared with traditional PID control method, it has improved the control performance of the system. In addition, in applying interference simulation, neural network predictive control system has stronger ability of anti-noise, with faster speed of response of neural network generalized predictive control program.