【作者】 刘建华；

【导师】 桂卫华； 谢永芳；

【作者基本信息】 中南大学 ， 控制科学与工程， 2013， 博士

【摘要】 摘要：闪速熔炼是提取金属铜的主要工艺,我国铜精矿资源复杂、变化频繁,需要迁移工况以保证熔炼性能。目前主要是基于人工经验调整操作参数进行工况迁移,导致迁移时间长、工况波动大、能源消耗高。铜闪速熔炼生产过程积累了大量的工业运行数据,研究如何从这些数据中挖掘出指导工况迁移的操作参数调整规则,完成熔炼工况的高效迁移,对实现熔炼企业的节能降耗具有重要意义。本文研究并提出了基于操作模式的铜闪速熔炼过程工况迁移策略,主要解决“目标工况设定”和“工况迁移路径选择”两大难题,提出了基于操作模式匹配、基于关键工艺指标预测的操作模式优化两种目标工况设定方法,以及面向调整代价最优的铜闪速熔炼过程工况迁移路径优化选择方法,开发了基于操作模式的铜闪速熔炼过程监控与优化系统。主要研究工作和创新性成果如下：(1)针对操作模式库庞大引起最优操作模式搜索效率低的问题,提出了基于柯西不等式的操作模式分级快速匹配方法。首先采用主元分析法设定属性权重,提高了匹配可靠性；然后将柯西不等式引入到初级匹配的相似性度量准则中,减少计算复杂度,快速获取相似操作模式集,使次级匹配只在相似操作模式集中进行,加快了匹配速度。最后利用UCI数据集与实际数据应用结果进行验证和分析。(2)针对铜闪速熔炼过程冰铜温度、冰铜品位和渣中铁硅比三大工艺指标难以在线检测的问题,提出了基于投影寻踪回归的铜闪速熔炼过程关键工艺指标预测方法。利用匹配获得的相似操作模式集,分别建立了预测三大工艺指标的投影寻踪回归子模型。为求解模型参数,以各子模型投影指标为目标,采用实数编码混沌伪并行遗传算法优化求解投影方向和多项式系数；以综合投影指标为目标,通过选择合适的岭函数个数,协调各子模型预测精度与模型参数更新速度,改善了模型整体性能。数据验证结果表明三大工艺指标的最大相对误差分别为6.08%、3.79%和6.93%,预测模型能满足实际生产需求。(3)针对优良操作模式库中不存在与铜精矿资源相适应最优操作模式的问题,提出了基于实数编码混沌伪并行遗传算法的操作模式优化方法,通过优化求解稳态工况下的最优操作参数,获取最优操作模式。综合多个工艺指标,构建了综合工况评价函数和操作模式优化模型；将混沌信息交换机制引入到“独立进化、信息交换”的伪并行遗传算法中,并采用实数编码形式完成各子种群的独立进化,从而设计了一种基于实数编码的混沌伪并行遗传算法。并将该方法应用于求解组成操作模式的最优操作参数分量,进而提取所需最优操作模式。最后通过算法性能测试和实际数据应用结果进行验证和分析。(4)针对基于经验的操作参数调整引起工况迁移路径选择盲目的问题,提出了面向调整代价最优的铜闪速熔炼过程工况迁移路径优化方法。综合考虑工况迁移时间、能源消耗,构造了工况迁移性能评价指标,建立了反映熔炼过程工况迁移路径与操作参数调整过程的动态关系模型,结合产品质量、工况波动、以及由当前工况和目标工况组成的端点等约束条件,将工况迁移路径优化选择问题转化为具有多约束和端点固定的优化问题,并利用Legendre伪谱法进行求解,获取最优工况迁移路径。工业应用结果表明所提方法与人工调整过程相比迁移时间缩短了1.5小时,氧量节省22879m3,冰铜温度、冰铜品位和渣中铁硅比三大工艺指标波动的平均相对误差分别减少了1.28%、0.65%和0.49%。(5)结合本文研究成果,开发了基于操作模式的铜闪速熔炼过程监控与优化系统,并应用于某冶炼厂实际生产过程。现场应用结果表明,本文提出的技术路线和方法具有较强的实用性。更多还原

【Abstract】 Abstract:Copper flash smelting process is the main production method employed for extracting copper. To ensure smelting performance, the smelting working-condition should be transited to adapt complex and changeable concentrate in China. Now, the transient procedure was mainly realized by adjusting operational parameters with artificial experience, resulting in excessive consumption of time and resources, as well as the drastic fluctuation. A large number of industrial data has been accumulated in copper flash smelting process. The research how to mine the rule of operational parameters adjustment for efficient transition, which are used to guide the working-condition transition, is important to achieve energy-saving and cost-reducing for Smelter.In this paper, operationa patter-based working-condition transition strategy for copper flash smelting process was proposed to solve the two issues, the setting of expected working-condition and the selection of working-condition transient trajectory. For setting the expected working-condition, two methods were present, the one is the operational pattern matching, and another one is the operational pattern optimization based on key process indicators prediction. Meanwhile, the adjustment-cost oriented optimization strategy of working-condition transient trajectory for copper flash smelting process was also present. The the monitoring and optimization system based on operational pattern for copper flash smelting process was developed. The main research results are as follows:(1) Considering the rawback of low efficiency of pattern matching caused by huge pattern case, Cauchy-Schwarz inequality based operational pattern two-grade fast matching strategy was developed. Using PCA, the attributes weights were set to improve the reliability of matching process. In order to improve the pattern matching speed, the Cauchy-Schwarz inequality was introduced to the similarity criteria of first matching process, which is used to reduce the computational complexity and extract the similar operational patterns quickly. The the secondary matching process was completed only in the similar operational patterns, which can be used to accelerate the speed. The application results of UCI data set and the actual data were given to verify the effectiveness.(2) Aim to the difficult problem of detecting the matte temperature, matte grade and the ratio of Fe to SiO2in slag, projection pursuit regression-based key process indictors predicting for copper flash smelting process was proposed. With the similar samples set retrieved from the operational patterns base, the three sub-models based on projection pursuit regression were developed, respectively. For solving the model parameters, with the object of prejection indicator function of each sub-model, the projection direction and polynomial coefficients were optimized with real-coding chaos pseudo-parallel genetic algorithm. With the target of comprehensive projection indicator, by secletign the appropriate number of ridge funcition to coordinating the balance between the accuracy of sub-model and learning speed of model paramters, which is used to the improve the model overall performance. The results of actual data shown that the maximum relative errors of three key process indictors are6.08%,3.79%and6.93%, respectively, which demonstrated that the model can meet the need of actual production.(3) Aim to the problem of non-matching the optimal operational pattern from excellent operational pattern base, real-coding chaos pseudo-parallel genetic algorithm-based operational pattern optimization method was proposed, and the optimal operational pattern can be achieved by optimizing the optimal operational parameters of steady-state working-conditon for copper flash smelting process. The comprehensive working-condition evaluation indicator was formulated by integrating several quality indicators in advance, and then the optimization model of operational pattern was also present. The chaotic information exchange mechanism was introduced in the pseudo-parallel genetic algorithm with "independent evolution, information exchange" strategy, and the coding form with real number was also employed to complete the evolution procedure of sub-population. Then the real-coding based chaos pseudo-parallel genetic algorithm was developed to optimize the optimal paramters, and the optimal operational pattern was formed by integrating the working-condition parameters. The algorithm performance test and the actual data applications results were given to verify the proposed method.(4) Aim to the blind problem of seletting the working-condition transient trajectory based on artifical experience, the adjustment-cost oriented working-condition transition strategy for copper flash smelting process was proposed. Firstly, by integrating transient time and transient raw materials consumption, an adjusting cost function was present to evaluate the transient performance. The dynamic modle was employed to reveal the relationship between working-condition transient trajectory and adjustment procedure of operational parameters. Then according to the constraints about product quality, working-condition fluctuation, and the two-point consist of current point and the endpoint, the working-condition transient trajectory optimization problem was converted into an optimization control problem where subject to mul-constraints and fixed two-point. Finally, the Legendre pseudospectral method was employed to optimize the working-condition transient trajectory. Compared with the manual adjustment process, the application results of actual data shown that the transient time can be reduced about1.5hours, and savinig the oxygen22879m3, and the average relative errors of three key process indicators were reduced by1.28%,0.65%and0.49%, respectively.(5) Operational pattern based monitoring and optimization system for copper flash smelting process was developed with the research results in this paper, and it was also applied to a smelter. The result shown the strong practicability of technical route and methods proposed in this paper.更多还原