【作者】 张怡卓；

【导师】 曹军；

【作者基本信息】 东北林业大学 ， 机械设计及理论， 2008， 博士

【摘要】 中密度纤维板(MDF)是国内外人造板生产和市场需求的主流产品,是节约与有效利用生物质材料资源的主要方法和途径。中密度纤维板的施胶控制过程是其生产工艺的关键技术环节,决定了板材的生产成本与产品质量。我国中密度纤维板生产线的自动化控制水平较国外存在较大差距,使得纤维板生产普遍存在达不到环保要求,原材料消耗较大,生产成本较高的问题。本学位论文结合国家“948”项目“人造板施胶控制关键技术引进课题”(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),针对中密度纤维板生产的施胶控制过程进行了研究,在讨论了支持向量理论及其应用的基础上,提出了基于支持向量辨识的中密度纤维板施胶过程模糊自适应控制方法,主要研究内容如下:首先针对影响中密度纤维板生产质量及板材性能指标的施胶比例参数进行了模型辨识研究。为了提高施胶比例辨识模型的准确性,本文提出了融合特征选择技术的施胶比例辨识模型输入参数的优选方法,在优选参数的基础上运用支持向量回归技术构建施胶比例辨识模型。针对施胶比例辨识模型的输入参数与支持向量回归参数的优化问题,引入了自适应遗传算法,设计了施胶比例辨识模型的自适应GA-SVR算法,试验结果表明该算法实现了辨识模型输入参数的合理优选及支持向量回归参数的有效优化,辨识模型实现了施胶比例的准确预测。在中密度纤维板施胶控制过程中,施胶量应实现在施胶比例参数作用下与纤维量进行随动。本文针对中密度纤维板施胶过程中的被控对象(电机转速与泵流量)进行了状态空间辨识研究。由于施胶泵流量除受电机转速外还受工作压力、介质黏度及泵自身等多因素的影响,本文提出了基于SVC多模型的中密度纤维板施胶流量辨识方法。该辨识方法利用支持向量聚类技术实现电机转速与泵流量的状态空间分解,应用支持向量分类技术设计了多模型切换策略,并运用支持向量回归技术构建了状态空间内的转速与流量辨识模型。在聚类过程中,针对传统的关联矩阵运算复杂度高的问题,给出了适用于中密度纤维板的电机转速与泵流量的动态核参数邻近图聚类方法。实验结果表明,动态核参数邻近图聚类方法实现了较低时间复杂度下的电机转速与泵流量准确自聚类,基于SVC多模型的施胶流量预测具有较高的准确性和鲁棒性。针对MDF生产中的施胶系统的动态特性差异,本文提出了中密度纤维板施胶过程的模糊自适应控制策略。该策略建立在电机转速与泵流量状态辨识的基础上,针对不同状态空间建立相应的模糊控制规则,通过判断电机转速与泵流量的状态空间,实现模糊规则的自动切换,进而完成状态空间内的模糊自适应控制,实验表明该模糊自适应控制策略实现了对中密度纤维板施胶过程的稳定、可靠控制。针对中密度纤维板施胶过程的实际问题,在上述理论研究基础上,研发了“MDF-1型施胶控制系统”,在本文最后简要地介绍了该系统的软件、硬件及通信子系统。系统运行结果表明,本文提出的“基于支持向量辨识的中密度纤维板模糊自适应控制方法”有效地提高了施胶量控制精度与板材生产质量,降低了生产成本,同时该系统故障率低、稳定性强。更多还原

【Abstract】 Middle density fiberboard is the mainstream of panel production as well as the major need of domestic and overseas panel market.It is an effective way of saving biological material resource.Gluing control is the key of fiberboard production,and it decides the cost and quality of the products.As to the automatic control of MDF product line in our country, there is still distance to the technique of foreign countries’.The problem of being unable to meet the need of environmental requirements,the high consumption of raw materials and the high cost of production are obvious.In this dissertation,the author made an in-depth study on the factors that influence the gluing process of MDF.On the basis of systematic study on the support vector machine theory,the author proposed a fussy self-adaptive method of MDF gluing control based on support vector identification.The content of this dissertation is as follows:In this dissertation,the author first made a model identification research on the indexes that influence the quality of MDF and gluing proportion parameters.In order to improve the accuracy of gluing proportion identification model,the author proposed a method that merges the feature selection technique to optimize the input parameters of the gluing proportion model, and uses the support vector regression technique to construct gluing proportion identification model with the help of optimized parameters.Aiming at the optimization of gluing proportion model input parameters and support vector regressive parameters,the author introduced selfadaptive genetic algrothm into the research and designed a self-adaptive GA-SVR algorithm for gluing proportion identification model.The results of experimentation showed that this algorithm has realized the optimized selection of identification model input parameters and support vector regressive parameters.The model can make exact prognostication on the proportion of gluing.In the process of MDF gluing control,the servo control of the gluing ratio should be realized under the control of gluing proportion parameters.Aiming at this,the author made a state space identification research on the rotational speed of motor and the flux of glue pump. As the flux of the glue pump is influenced by working pressure,medium viscosity and various factors related to the pump apart from the motor speed,the author proposed a SVC-based multi-model MDF gluing flux identification method.This method has realized the state space decomposition of rotational speed and flux with the help of support vector clustering technique. In the process of clustering,the author proposed a proximity graph-based dynamic kernel parameter clustering method that is suitable for the control of motor rotational speed and glue pump flux in MDF production to solve the problem of high complexity of relative matrix calculation.The result of experiments showed that this method has realized the accurate selfclustering of motor speed and pump flux under a low time complexity,the SVC-based multimodel gluing flux prognostication has a high accuracy and robust.Aiming at the characteristic difference of the gluing system in diversified MDF production,in this dissertation,the author proposed a self-adaptive controlling strategy for MDF gluing process.This strategy is based on the state space identification of the motor speed and pump flux,corresponding fuzzy control rules are set for different state spaces.By judging the state space of motor speed and pump flux,the switch of fuzzy rules is realized,thus the fuzzy self-adaptive control in the state space is realized.The result of experiments has showed that this strategy has realized the stable and reliable control of MDF gluing process.Taking the actual situations of MDF gluing into consideration,the author designed a "MDF-1 Parallel-Line-gluing control system" on the basis of the above theories.In last section of this dissertation,the author briefly introduced its software,hardware and communication subsystem.The result of system operation showed that the support vector identification-based MDF fuzzy self-adaptive control method could effectively increase the accuracy of gluing proportion control and the quality of production.The production costs as well as system failure are reduced,but the stability is enhanced.更多还原