【作者】 彭晓波；

【导师】 桂卫华；

【作者基本信息】 中南大学 ， 控制科学与工程， 2008， 博士

【摘要】 铜闪速熔炼是提取铜的主要工艺方法,是一个非常复杂的高温、多相的物理化学变化过程,具有多变量、非线性、强耦合、大滞后、不确定性等特点,导致许多过程参量无法直接检测,并且难以采用精确的数学模型进行描述。目前,铜闪速熔炼过程的操作参数大多由生产操作决策人员凭经验主观确定,难以实现生产过程的持续稳定运行,生产过程的工艺指标波动较大。因此,研究铜闪速熔炼过程的操作参数优化,对于实现铜闪速熔炼过程的节能降耗、提高资源利用率以及充分发挥生产潜力、提高生产过程的技术经济指标,实现企业的可持续发展,都具有重大意义。本文在分析铜闪速熔炼机理的基础上,研究了铜闪速熔炼过程智能优化方法。论文首先介绍了铜闪速熔炼的机理和工艺,然后在总结铜闪速熔炼过程操作参数优化的特点和研究现状的基础上,提出了铜闪速熔炼智能优化方法,该方法主要包括冰铜品位、冰铜温度、渣中铁硅比三大质量指标的软测量与操作模式优化。该优化方法已成功应用于铜闪速熔炼过程,取得了明显的成效。主要研究成果如下:(1)提出铜闪速熔炼过程操作优化控制框架。框架包括软测量模型、工况评价、机理模型、操作模式优化、协调策略。框架首先由软测量模型对三大质量指标参数进行软测量,其软测量结果经过工况评价,工况非优时启动机理模型和操作模式优化,经过协调策略,获得综合输出;工况为优时保持操作参数不变。框架为铜闪速熔炼过程智能优化控制系统奠定了基础。(2)提出一种改进的T-S递归模糊神经网络(DTRFNN)方法进行软测量,解决了铜闪速熔炼过程中冰铜品位、冰铜温度、渣中铁硅比三大质量指标无法在线直接测量的问题。在DTRFNN结构确立以后,本文给出了其参数调整BP算法、收敛性证明,局部极小问题改进。改进的DTRFNN能较好地解决冰铜品位、冰铜温度、渣中铁硅比三大参数的软测量,精确度达到了97%。(3)在操作模式智能优化中,给出操作模式和操作模式优化的定义。并在此基础上提出了模式分解方法,该方法先采用模糊聚类的方法对操作模式集进行分解,再采用相似模式融合,将优化样本空间范围缩小,有效解决模式优化计算量大问题;在操作模式优化算法方面,深入研究弹性粒子群算法和遗传算法,提出了GARPSO算法。该算法分为两部分:1)采用一种自适应粒子群算法策略弹性地修正粒子速度的幅值,有效地避免了粒子群算法的早熟收敛问题。2)与遗传算法结合,用弹性粒子群算法模拟自然界个体成熟现象,使个体得到更大的提高。经过提高,交叉,变异三步,GARPSO算法能获得最优解,能为操作模式优化提供算法支撑。(4)开发了铜闪速熔炼过程操作模式优化系统。该系统实现了过程状态的可视化监控,三大参数软测量以及操作参数优化,数据库管理,打印、数据备份及帮助,数据采集与通讯五大功能。系统的应用实现了节能降耗,稳定了铜闪速熔炼生产过程。在获得基本相同的冰铜品位的情况下,在综合考虑风氧量成本的基础上,综合成本节约1.2%-1.5%,取得了很好的应用效果。更多还原

【Abstract】 Copper flash smelting is the main metallurgy technique to smelt copper.It is a very complex process with physical and chemical reactions of high temperature and multi-composition,which has such characteristics as multivariable,nonlinearity,strong coupling,large inertia,time varying and uncertainty.Many parameters in the process of copper flash smelting cannot be measured.It is almost impossible to build exactly mathematical model to optimize the process of copper flash smelting.At present,it is difficult to make technique indexes of the process of copper flash smelting steady and optimal in a long time because most of operational parameters are determined by human experience.So the research on intelligent optimization methods for operational patterns in the process of copper flash smelting is of very grand significance to the energy and material consumption reduction,the improvement of usage efficency of resource and equipment,the increment of production ability and technique indexes,and the continuous development of smelting enterprise.In this paper,intelligent optimization method of operational patterns is presented based on the mechanics of complex copper flash smelting. Firstly,this paper introduces the mechanics and technology of copper flash smelting in detail.Secondly,based on generalization of the characteristics and research conditions of operational parameters optimization in the process of copper flash smelting,the intelligent optimization method of operational patterns is suggested.This method mainly includes the optimization of operational patterns and the soft sensing of three parameters which are matte grade,matte temperature and the ratio of iron and silicon in slag.This method has been applied in the process of complex copper flash smelting successfully,and some excellent results were achieved.In this paper,the main results are as follows.(1)The framework of operational optimization control in the copper flash smelting process is presented in this paper.The framework includes the model of soft sensing,work condition estimation,the model of mechanics,the operational pattern optimization and coordination strategy.In the framework,firstly,the three qualitative indexs which are matte grade,matte temperature and the ratio of iron and silicon in slag are measured by use of the model of sofe sensing;and secondly the result of soft sensing is judged by use of the work condition estimation. If the work condition estimation is not well,the framework strats the model of mechanics and the model of operational pattern optimization. Through the coordination strategy,the results of two prats get synthetical output.If the work condition estimation is well,the operational parameters are not changed.The framework lays a foundation for the copper flash smelting intelligent optimization control system.(2)An improved dynamic T-S recurrent fuzzy neural network (DTRFNN)is presented in this paper to soft sensing,for the man-made obtain of the matte grade copper,matte temperature,and the ratio of iron and silicon in slag lags the smelting process.After the structure of DTRFNN is established,the parameters adjustment of BP algorithm,the convergence proof of DTRFNN,and the improvement of the local minima are given in the paper.The three parameters’ soft sensing can be settled very well with the improved DTRFNN and the accuracy of the output can get 97%.(3)In the operational patterns intelligent optimization,this paper gives the definitions of the operation patterns and the operation patterns optimization.Based on this,the method of operation patterns decomposition is given.Firstly the operational parameters sets are decomposed by the fuzzy clustering,and secondly the space of optimizational samples is dwindled by the method of similar patterns fusion.Then we can solve the compution burden promble.In the operational patterns optimization algorithm,by profound research of particle swarm algorithm and genetic algorithm,the GARPSO algorithm is presented which includes two parts:an adaptive scheme in particle swarm algorithmis is adopted in the first part to adjust the magnitude of the velocity resiliently,and the promble of premature convergence in particle swarm algorithm is avoided effectively.The second part, combining with genetic algorithm,a resilient particle swarm optimization algorithm to mimic the mature phenomenon is used in combination with genetic algorithm,and the individuals get the greater improvement than before.Through three steps of improvement, crossover,mutation,the GARPSO algorithm can get the optimum solution and supply the support for operational patterns optimization.(4)This paper develops the copper flash smelting intelligent optimization system.This system implements process status visual monitor,three parameters soft sensing and operation parameters optimization,data-base management,data printing,reserving and helping, and data collection and communication.The application of this system implements the energy conservation and stables the productive process.When the matte grade is probably same,the synthesis cost can save 1.2%-1.5%and the applied result is well.更多还原