【作者】 杜玉晓；

【导师】 吴敏；

【作者基本信息】 中南大学 ， 控制理论与控制工程， 2004， 博士

【摘要】 铅锌冶炼ISP工艺是近代火法炼铅锌的先进方法之一，密闭鼓风烧结过程作为其中的一个重要流程，直接影响到熔炼炉炉况和铅锌产量。目前铅锌烧结过程控制技术落后、自动化水平低，过程优化控制成为制约烧结矿产量质量的一个瓶颈。针对具有强非线性、强耦合性、不确定性、时变、大滞后、多约束特点的铅锌烧结过程，采用传统的控制理论或单一的智能化技术难以满足多目标全局优化控制要求，本文提出一种智能集成优化控制技术，有效解决了铅锌烧结过程状态优化控制和产量质量优化控制问题。论文的主要研究成果包括： (1) 智能集成优化控制技术 基于铅锌烧结过程特性和控制问题分析，提出了智能集成优化控制技术基本框架，包括基本概念、集成结构和系统实现三部分内容。从集成形式上划分为串联智能集成、并联智能集成和嵌套智能集成三种基本结构，提出了建立智能集成优化控制系统的信息化、模型化、控制优化和工业化关键问题与技术。 (2) 模糊专家聚类网格状态优化控制技术 针对状态实时检测问题，建立了烧穿点软测量模型、烟气温度场分布模型；针对过程大滞后特性，采用模糊分类方法，将工艺参数神经网络模型与时间序列神经网络模型集成，建立了综合透气性预测模型，将工艺参数神经网络模型与时间序列灰色理论模型集成，建立了烧穿点预测模型，有效地提高了状态预测精度。 在铅锌烧结过程中存在有大量模糊的不确定性信息，专家操作经验往往以一种定性描述形式出现。采用模糊专家优化规则将混合料分为高品位矿、中品位矿和低品位矿，针对不同品位的混合料进行不同的状态优化值设定。采用模糊专家控制结合聚类网格控制的策略，根据超前预测误差获取状态优化控制参数，模糊专家控制可以模拟人类专家的优化操作，聚类网格控制则是基于状态集成预测模型的精确控制，整体优化控制算法具有工业有效性和较高的控制精度，解决了铅锌烧结过程中具有多约束、不确定性和非线性特点的状态优化控制问题。 (3) 聚类搜索遗传混沌产量质量优化控制技术 针对过程大滞后和产量质量测量问题，采用改进的BP神经网络，中南大学博士学位论文摘要建立了铅锌烧结矿产量、含铅量、含锌量、含硫量、二氧化硅含量、氧化钙含量预测模型。 用罚函数法将多目标约束优化问题转化为无目标约束优化问题，采用基于模糊聚类的并行搜索进行粗优化，采用最优保存简单遗传混沌优化算法进行精确优化，整体优化控制算法具有全局收敛性和工业有效性，解决了铅锌烧结过程中具有多目标约束强非线性特点的产量质量优化控制问题。 (4)智能集成优化控制系统 采用p XL集散控制系统和管理信息系统实现了底层自动化和高层信息管理，在信息化、模型化和控制优化的基础上建立了智能集成优化控制系统，通过对状态短期优化控制策略和产量质量长期优化控制策略的智能协调，实现了铅锌烧结过程多目标全局优化控制。 通过应用铅锌烧结过程智能集成优化控制技术，从整体上提高了铅锌烧结工艺过程优化控制水平，有效地抑制了烧穿点和透气性波动，产量质量得到了提高，同时加强了信息管理能力，降低了工人的劳动强度，取得了显著的经济效益和社会效益。同时也为复杂工业过程优化控制提供一套实用的、值得借鉴的工业化实现方法。更多还原

【Abstract】 The Lead-Zinc Imperial Smelting Process is one of the modern advanced methods of the Lead-Zinc smelting utilizing neoteric burning technology. As an important loop of ISP, imperial updrafted-sintering process has a direct influence on state of the smeltering furnace and output of Lead-Zinc. At present, for the laggard control techniques and the low-level automatization, the process optimal control has become the key problem to restrict the output and quality of the sinter. Since Lead-Zinc sintering process possesses the characteristics such as strong nonlinear, strong coupling, uncertain, time-varying, long time-delay, multi-constrained, neither the traditional control theory not simple intelligent technique can satisfy the demands of multi-target global optimal control. Thus, the thesis proposes an intelligent integrated optimal control technique, which solves the optimal control of state and output-quality in Lead-Zinc sintering process. The main study achievements include:(1) Intelligent integrated optimal control techniqueBased on the analysis of the characteristics of the Lead-Zinc sintering process and control problems, the basic frame of the intelligent integrated optimal control is proposed, which includes the three part of basic concept, integrated structure and system building. It is divided into three basic structures of series intelligent integrated, parallel intelligent integrated and nesting intelligent integrated. Thus, the key problems and techniques of the informatization, modelizaion, control optimization and industrialization to establish the intelligent integrated optimal control system are proposed.(2) State optimal control technique of fuzzy expert control and clustering grid algorithmAimed at the state real-time measure problem, the model of BTP and the temperature distribution model of waste gas are established. Aimed at the character of long time-delay, with the adoption of fuzzy clustering method, the integrated predictive model of synthetical permeability isestablished with combination of the neural network model of technics parameters and the neural network model of time series, and the integrated BTP predictive model is established with combination of the neural network model of technics parameters and the grey theory model of time series, which improves the precision of state predictive model effectively.The Lead-Zinc sintering process has a large number of fuzzy and uncertain informations and qualitative expert operation rules. The sintering materials are divided into three types of high quality, mid quality, low quality , and for the different-quality materials the state optimal values are decided respectively, by using fuzzy expert optimization rules. The strategy of fuzzy expert and clustering grid are adopted in order to find the state optimal control parameters, according to ahead predictive errors. The fuzzy expert control has the function to simulate the human experts optimization control, while the clustering grid control is an accurate strategy based on state predictive models. The whole optimization control algorithm possesses industrial validity and higher control precision, which resolves the state optimization control problem with multi-constrained, uncertain, nonlinear, characteristics.(3) output-quality optimal control technique of clustering searching , genetic algorithm and chaos optimizationAimed at the long time-delay characteristic and the measure problem of quantity and quality of sinters, the predictive models of quantity, Pb content, Zn content, S content, SiO2 content, and CaO content of sinter are proposed, by using the improved BP neural network.The penalty function method is used to transform the muti-target-constrained optimization problem to unlimited optimization problem. The parallel searching based on fuzzy clustering is used to realize the raw optimization, while the elitist preserved simple genetic algorithm and chaos optimization are used to realize the accurate optimization. The whole optimization control algorithm possesses the gl更多还原