【作者】 王春生；

【导师】 吴敏；

【作者基本信息】 中南大学 ， 控制理论与控制工程， 2008， 博士

【摘要】 密闭鼓风烧结是铅锌冶炼ISP工艺中的一个重要流程,配料过程作为其首道工序,直接影响到铅锌烧结生产的成本、质量产量和能源消耗。目前,铅锌烧结配料过程作为稳定和优化生产的首要环节,其作用尚未得到充分发挥,存在着配料准确率和经济性不高的问题。同时由于铅锌烧结配料过程控制水平较低,导致了生产成本高、烧结块质量差、产量低,一方面造成了能源浪费,另一方面造成了环境污染。针对上述问题,本文主要围绕铅锌烧结配料过程智能集成建模与优化控制策略开展研究,取得的研究成果主要包括以下五个方面:(1)烧结块成分智能集成预测模型针对复杂的烧结块成分预测问题,提出一种基于过程神经网络和灰色系统理论的烧结块成分智能集成预测模型。该模型首先利用过程神经网络可充分表达时间序列中时间累积效应、灰色系统可弱化数据序列波动性的优点,分别对铅锌烧结块成分进行预测,然后从信息论的观点出发,提出一种熵值方法,重新定义预测误差序列的变异程度,从而获得各个预测模型的加权系数,通过对两个预测模型的预测结果进行加权集成,获得更加准确的铅锌烧结块成分预测结果。结果表明,智能集成模型的预测精度高于单一预测模型,能有效地对烧结块成分进行预测,满足了配料计算对预测精度和数据完备性的要求。(2)烧结返粉量智能集成预测模型针对烧结返粉量变化趋势复杂,受多个因素影响,难以用单一预测模型进行有效预测的问题,提出一种基于改进灰色系统和支持向量机的智能集成预测模型。首先利用改进灰色系统和支持向量机两个单一预测模型分别对烧结返粉量进行预测;然后基于预测精度的数学期望和标准差,以其数学期望最大化和标准差最小化为目标函数,通过求取最优加权系数,建立烧结返粉量智能集成预测模型进行返粉量预测。结果表明,该集成预测模型能够获得更高的的预测精度,能有效地对返粉量进行预测,预测结果为确定烧结返粉配比提供了数据支持。(3)一次配料定性定量智能集成优化算法针对传统配料方法中存在的成本高和准确率低的问题,提出一种定性定量智能集成优化算法。在对烧结主要原料特性和经济性进行分析以及建立烧结块成分集成预测模型的基础上,首先以成本最小为目标建立烧结配料优化模型,分别采用专家推理策略和免疫遗传算法对烧结配料进行优化;然后,在对免疫遗传算法进行改进的基础上,从系统论的观点出发,采用定性定量综合集成方法,把过程神经网络技术、灰色系统理论与专家推理策略、改进免疫遗传算法有机结合,实现了烧结配料的进一步优化,提高了配料的准确率,降低了烧结成本,取得了可观的经济效益。(4)基于烧结工况综合评价的二次配料智能优化策略在对烧结生产全流程各参数间关系进行分析的基础上,提出了基于烧结工况综合评价的二次配料智能优化策略,建立了烧结生产工况综合评价模型,并提出了基于聚类分析的操作参数匹配优化算法。首先,通过建立烧结返粉量、烧结块含铅量、含锌量以及含硫量预测模型,将这些模型的输出作为烧结生产工况优劣的综合评价因素,利用烧结生产工况综合评价模型,采用模糊综合评价法,实现对烧结生产工况的综合评价;其次,根据对烧结生产工况综合评价的结果,在利用加权模糊C均值聚类算法对优化样本数据集进行聚类的基础上,通过操作参数匹配优化算法,获得二次配料过程具体的操作参数优化值,作为实现二次配料过程过程优化控制的操作指导。结果表明:该方法可显著改善工况波动,减少了由于操作盲目性造成的生产工况不稳定,进而提高了烧结块的产量和质量。(5)烧结配料过程智能集成控制策略由于烧结配料过程中的物料流量受许多不确定因素的影响而波动很大,具有很强的非线性和大滞后等特性,难以建立确切的数学模型,其控制问题很难用传统的控制理论和方法解决。为了提高配料的准确度和稳定性,结合模糊控制和PID控制的特点,提出一种基于加权因子的烧结配料模糊自适应PID智能集成控制策略,分别设计了模糊控制器和自适应PID控制器。利用加权因子将模糊控制器的输出和自适应PID控制器的输出进行加权集成,使得控制器在误差较大时,主要由模糊控制器起作用,具有较快的响应能力;而在误差较小时主要由自适应PID控制器起作用,具有较高的控制精度,实现了模糊控制器和自适应PID控制器输出的连续平滑切换。更多还原

【Abstract】 The imperial updrafted-sintering is an important loop of the Lead-Zinc Imperial Smelting Process. As the first working procedure of the Lead-Zinc sintering process, the blending process plays an important role, which exerts an influence on the sintering production cost, quantityt-quality and energy sources waste. As the chief tache for stabilization and optimization production, the role of Lead-Zinc sintering blending process has not been played at present and there are problems of low economics and accuracy in the sintering blending process. At the same time, high cost and low quantity-quality have resulted from the low level of the Lead-Zinc sintering blending process control, which result in energy sources squander and environment pollution. Amied at the above problems, the intelligent integrated modeling and optimization control strategy in the Lead-Zinc sintering blending process have been mainly studied and the main achievements in this dissertation include several aspects as follows:(1) Intelligent integrated prediction model for component of Pb-Zn agglomerateTo deal with the problem of the component prediction for Pb-Zn agglomerate, an intelligent integrated prediction model based on process neural network (PNN) and grey system theory (GST) was presented. First, the component of agglomerate was predicted by PNN and GST models. Then, from the viewpoint of the information theory, a kind of entropy method was proposed and the deviation of forecasting error series was defined renewal, then the optimal weightling coefficient of each prediction model was obtained. The exact prediction results for component of Pb-Zn agglomerate were acquired by integrating the reaults of two prediction models. The results show that the integrated model has high prediction precision, it predicts the component of agglomerate effectively, and it meets the requirements of the data completeness for proportioning computation.(2) Intelligent integrated prediction model for quantity of sintering return powdersBased on the fact that the complexity of return powders is affected by various factors and it is hard to predict accurately with a single prediction model, an intelligent integrated prediction model based on improved grey system (IGS) and support vector machine (SVM) was proposed. Firstly, the quantity of sintering return powders was respectively predicted by using IGS prediction model and SVM prediction model. Then an intelligent integrated prediction model based on two precision indicators of mean and deviate, was introduced to predict the quantity of sintering return powders by calculating optimal weighting coefficient. The prediction results show that the prediction precision of the integrated model is higher than that of single prediction model and it can predict the quantity of sintering return powders effectively. The prediction results can offer data support to the determination of blending ratio of return powders.(3) Qualitative and quantiative synthetic optimization algorithm for the primary blendingTo deal with the problem of high cost and low accuracy existed in conventional methods for the primary blending process in lead-zinc sintering, a kind of qualitative and quantiative synthetic optimization algorithm was proposed. First, based on the analysis for the characteristic and economy of the sintering material and establishment for the intelligent integrated prediction model for the agglomerate component, a blending optimization model was established for the purpose of minimizing the costs. The mixture ratios were optimized respectively by using the expert reasoning strategies and the immune genetic algorithm. Then, from the viewpoint of the system theory, the mixture ratios were optimized ulteriorly by using the qualitative and quantiative synthetic methodology integrated the process neural network technology, grey system theory, expert reasoning strategy and improved immune genetic algorithm. The blending accuracy was enhanced, the sintering cost was reduced and the considerable economic benefit was procured.(4) Intelligent optimization strategy for the secondary blending process based on comprehensive evaluation for sintering statusBased on the analysis of the relationship between parameters in the whole sintering process, the intelligent optimization strategy for the secondary blending process based on comprehensive evaluation for sintering status was proposed. The comprehensive evaluation model for sintering production status was established and the optimization algorithm for the operation parameters matching based on clustering analysis was presented. Firstly, based on the establishment of the models for sintering return powder, Pb, Zn and S component prediction, the outputs of these models were considered as factors for sintering status evaluation. The comprehensive evaluation for sintering status was implemented by using the fuzzy evaluation method. Secondly, according to the results of comprehensive status evaluation model, the optimum operation parameters were acquired by using matching optimization algorithm based on weighting fuzzy C-means clustering. The results show that the fluctuation of status can be meliorated efficiently and the quantity and quality of agglomerate were improved.(5) Intelligent integrated control strategy for sintering blending processThe material flow in the sintering blending process was impacted by many uncertain factors and fluctuates greatly with characters of high nonlinear and large lag. It is hard to solve the sintering blending control problem by conventional control theory and methods since the mathematic model is difficult to be established. In order to improve the accuracy and stability of the sintering blending, a fuzzy adaptive PID intelligent integrated control strategy for sintering blending based on weighing coefficients was proposed by combining with the features of the fuzzy control and PID control. The fuzzy controller and the adaptive PID controller were designed respectively. The outputs of the fuzzy controller and the adaptive PID controller were integrated by using the weighting coefficients. When the error of the material flow is greater, the fuzzy controller mostly determines the output so that fast response ability can be shown, and when the error is smaller, the adaptive PID controller mostly determines the output, so that a higher control precision can be obtained. The switch of the output between fuzzy controller and adaptive PID controller is continuous.更多还原