【作者】 王颖；

【导师】 孙丽萍；

【作者基本信息】 东北林业大学 ， 农业电气化与自动化， 2013， 硕士

【摘要】 中密度纤维板(Medium Density Fiberboard)凭借其优良的板材性能和对木材资源的高效利用,已成为目前我国市面上最具发展前途的人造板产品之一,而其生产过程中的施胶环节是一道十分关键的工序,原料(纤维、刨花等生产材料)量与胶液量需要始终保持一定的比例不变。准确控制施胶量不仅会提高产品的板材性能和品质,而且在减少原料消耗、降低甲醛释放量、实现节能环保等方面也有重要意义。在实际工业生产过程中,中密度纤维板施胶系统是一个具有时变性、耦合性的典型非线性被控对象,无法用准确的数学模型进行描述,而且传统的线性系统控制方法也很难实现对施胶量的精准控制。在针对中密度纤维板施胶工艺过程具体分析的基础上,本文提出了神经网络预测控制方法。然后,根据施胶系统的工作原理,建立起系统的开环模型。其次,又对施胶系统的开环模型,设计出两种不同的控制结构,一种是神经网络的模型预测控制结构,另一种是神经网络的广义预测控制结构。这两种方案均可通过神经网络实现对施胶系统预测模型的构建与辨识,克服了施胶系统难以建立精确数学模型的困难,同时解决了难以将传统预测控制应用于非线性系统的问题。最后,针对两种方案的可行性以及控制效果,本文通过Matlab\Simulink进行了仿真验证,并与传统PID控制方法进行了比较。仿真结果表明,所设计的两种控制方案均能够实现胶液量快速跟踪纤维量的动态变化,与传统PID控制相比,提高了系统的控制性能。另外,在施加干扰的仿真试验中,神经网络预测控制方案下的系统抗干扰能力较强,神经网络广义预测控制方案下系统的响应速度较快。更多还原

【Abstract】 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.更多还原