【作者】 赵化启；

【导师】 曹军；

【作者基本信息】 东北林业大学 ， 机械设计及理论， 2009， 博士

【摘要】 刨花板生产线调施胶过程控制的水平,直接影响刨花板的物理力学性能和甲醛释放量,直接决定调胶的配比精度和施胶的实时准确性。本文结合国家“948”项目“人造板施胶智能诊断控制关键技术引进”(2006-4-109)与黑龙江省科技攻关项目“刨花板施胶、混胶智能数控系统研究”(GB06A505),在详细分析刨花板调施胶过程生产工艺和控制机理的基础上,通过对预测控制理论和技术的深入研究,提出了刨花板调施胶过程预测控制方法,主要研究内容如下:针对调胶过程中各原料的动态称量精度问题,分析了动态称量调胶过程的控制机理,建立了调胶过程动态称量模型,在此基础上,设计了调胶迭代学习控制方法。该方法通过对调胶系统进行控制尝试,以输出轨迹与给定轨迹的偏差修正不理想的控制信号,产生新的控制信号,使得系统能够用非常简单的方式和较少的先验知识处理不正确程度较高的调胶控制系统。针对刨花板施胶过程中存在的胶液流量实时跟踪刨花流量延迟、施胶量不准确等问题,提出一种改进的广义预测控制方法。该方法利用广义预测控制的思想,将刨花板施胶过程流量控制预测模型用受控自回归积分滑动平均模型CARIMA来描述,并通过对广义预测控制与动态矩阵控制的控制律等价性分析,来推求最优控制律的参数。仿真结果表明,该方法对刨花板施胶控制系统的时延和参数变化具有很强的鲁棒性,有效提高了施胶量的实时准确性。针对刨花板施胶过程存在的非线性特性,对典型的非线性Hammerstein模型进行了辨识方法研究。在研究中对通常算法存在的线性参数估计的一致性无法得到保证的问题,提出一种新的思路,将模型的非线性静态模块用最小二乘支持向量机模型描述,线性动态模块用状态空间模型结构描述进行辨识,并解决了最小二乘支持向量机因将二次规划问题转换成线性方程组而失去了支持向量机解的稀疏性问题。仿真结果表明,该方法增强了Hammerstein模型的鲁棒性,提高了模型辨识精度,减少了模型辨识时间。针对刨花板施胶过程中存在的胶液流量响应刨花流量慢、滞后性强和非线性明显的现象,提出了基于改进的Hammerstein模型的预测函数控制(PFC)方法。该方法将施胶控制输入结构化,把每一时刻加入的控制输入看作若干事先选定的基函数的线性组合,使控制量的输入规律性更加明显,并利用Hammerstein模型,把施胶非线性系统的控制问题分解为线性模块的动态优化问题和非线性模块的静态求根问题。仿真结果表明,该方法可提高施胶响应的快速性,增强刨花板施胶控制系统的鲁棒性。针对刨花板施胶过程控制系统的抗干扰问题,进行了深入研究。设计了基于KPFC-PI的刨花板施胶控制方法,将比值器和PFC作为系统外环,用以提高胶液跟踪刨花的速度和鲁棒性,系统内环路采用PI控制器,增强系统抗干扰性,并通过调节PI参数,拟合简化整个内环路,作为外环PFC的广义处理对象,用来减少干扰引起的偏差。仿真结果表明,该方法可有效提高刨花板施胶精度、保持施胶量稳定性和具有较强抗干扰性能。结合本文上述理论研究,对刨花板调施胶控制进行了仿真实验研究。调胶仿真结果表明,采用迭代学习方法构造的迭代学习因子q可以使提前关闭量u快速调整到理想整定值;调胶实验结果表明,采用迭代学习控制可以有效提高调胶精度,保证调胶配比精度。施胶仿真实验结果表明,本文研究的三种施胶控制方法中,KPFC-PID控制方法的抗干扰能力最强,广义预测控制方法的响应效果较好,基于Hammerstein模型的预测函数控制方法对非线性情况处理最好。更多还原

【Abstract】 During the process of producing particleboard,control level of mixed and used glue process has a direct influence on the physical and chemical properties of particleboard and formaldehyde emission.This paper combines with National 948 project,the title of which is“Key technology of diagnosis and control for wood-based board used glue”(2006-4-109) and HeiLongJiang Province scientific and technological project,the title of which is“Research on intelligent NC system of mixed and used glue for particleboard”(GB06A505). Through studying predictive control theory and related technology,put forward the process of particleboard an advanced predictive control technology is proposed basis on deeply analyzing production technology control mechanism of mixed and used glue for particleboard,Main research contents are as follows:For dynamic weighing accuracy problem of various raw materials in mixed glue process, dynamic weighing model is established by the means of analyzing control mechanism of this process.Then an kind of iterative learning control method for used glue is designed.Through controlling mixed glue system,the deviation between output track and given track is used to rectify the non-ideal control signal.By doing so,new control signals will be generated.With these new signals,system could be able to deal with mixed glue control system with a higher incorrect degree in a simple approach.And it only depends on less prior knowledge.During the process of used glue,problems such as delay of tracking in real-time and inaccuracy of used glue are generated.An improved generalized predictive control method is proposed.This method based on theory of generalized predictive control.And the method describes predictive control model of the glue flow used on surface layer by controlled autoregressive moving average model CARIMA.And through analyzing equivalence of GPC and DMC control laws,parameters of optimal control law could be inquired.The simulation results show that the method could effectively improve the accuracy of controlling used glue’ s volume and exhibits strong robustness to problems exited in used glue on particleboard surface layer control system,such as delay,variation of paramaters.For the nonlinear property existed in the process of used glue,a research is carried on typical nonlinear Hammerstein model with identification method.However,consistency of linear parameters estimation can not be guaranteed by common algorithm.In these circumstances,this paper presents a new idea.The nonlinear static module of model is described by using LS-SVM model.And the linear dynamic module is identified by structure of state space model.The sparse problem caused by converting problem of quadratic planning into problem of linear equations is also solved.The simulation results show that the method enhances the robustness of Hammerstein model,improves the accuracy of the model identification,and reduces the time consumption of model identification.In the process of used glue,there are problems,such as the response of glue flow to core layer particle flow is slow,significantly delay and obvious nonlinear.The predictive founction control(PFC)method based on improved Hammerstein model is proposed.The method will be used to structure glue control input,and to consider every control input added in every moment as a series of pre-selected linear combination of basis functions.Therefore,the input regularity of control value is more evident.And through utilizing Hammerstein model,nonlinear system’ s control of used glue on core layer is divided into two parts.The one is dynamic optimization of linear system’ s module.The other is statically finding roots of nonlinear module.Simulation results show that the method can improve the rapidity which is used to measure the response of used glue on core layer,and enhance robustness of used glue control system.Research on anti-interference problem of used glue process control system for particleboard.Used glue control system is designed based on KPFC-PI.In system,ratio device and PFC work as system’ s outer loop used to improve tracking speed and robustness.PI controller is worked as system’s inner loop and used to improve anti-interference of system. Through adjusting PI parameters and fitting simplified inner loop,the entire inner loop worked as generalized object of the outer loop.By doing so,the deviation caused by interference will be reduced.The simulation results show that the method can effectively improve the accuracy of used glue on core layer for particleboard,maintain stability of used glue volume and possess a strong performance of anti-interference.Combined with theoretical researches mentioned above,a simulation experiment is carried on.Mixed glue simulation results show that the use of iterative learning method for constructing iterative learning factor q can adjust advance closure volume u to ideal setting value with rapid speed;Mixed glue experimental results show that the use of iterative learning control can effectively improve the accuracy of mixed glue,and ensure the accuracy of additive ratio.Used glue simulation results show that among three types of control methods mentioned in this paper,KPFC-PID control method has the strongest anti-interference ability, generalized predictive control method responds better,and predictive functional control method based on the H model deals with nonlinear situation much better.更多还原