【作者】 鲁春燕；

【导师】 李炜； 王君；

【作者基本信息】 兰州理工大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 在铝电解生产过程中,氧化铝物料输送是一个极为重要的环节,其输送技术会直接影响铝电解生产的稳定和氧化铝单耗。随着超浓相输送技术的广泛应用和自动化水平的日益提高,通过优化控制确保产品质量,进而实现降耗增效已成为企业追求的目标。为了达到这一目标需对更多的工艺参数与变量进行在线精确测量,而铝电解中氧化铝粉流量即为其中主要参数之一。软测量作为一种解决产品质量等关键性生产参数在线测量问题的新技术,近年来已成为有效提高生产效益、保证产品质量的有力手段。本文以电解铝厂氧化铝输送过程为背景,针对氧化铝粉流量或因进口仪表价格昂贵、或因现存国产仪表精度不高难以在线有效测量这一问题,提出了将软测量技术应用于氧化铝粉流量的在线估计。首先在对氧化铝输送过程工艺深入分析的基础上,结合自动化系统可提供的测量信息,确定了供料离心风机风流量和风压作为辅助变量;其次利用实际工业现场数据,分别以PLS、RBF和LS-SVM等单一模型,对氧化铝粉流量的软测量进行了仿真研究;进而考虑到模型参数选择对LS-SVM模型性能的影响,利用粒子群优化算法确定其参数,得到了精度更高、泛化能力更强的预测模型;再者针对单一模型难以全面描述复杂系统全局特性的问题,采用一种基于模糊C均值聚类的双层多模型软测量方法,对氧化铝粉流量的预测估计进行了仿真研究,该方法以模糊C均值聚类作为分类基础层,分别以多PLS、多RBF、多LS-SVM以及PLS+RBF+LSSVM作为各分类数据模型层,在线运行时将各模型的预测输出通过实际数据在各模型的隶属度上进行加权求和,从而获得被估计参数的软测量值。研究结果表明单一模型和多模型软测量建模方法,对于氧化铝粉流量的估计均具有一定的泛化能力和较好的预测精度,且异类多模型更优,LS-SVM更适宜模型的在线修正,从而验证了软测量技术用于氧化铝粉流量的在线估计是可行有效的,同时也为软测量技术在氧化铝粉流量预测的工业实现提供了依据。更多还原

【Abstract】 Alumina conveying is one of the most important parts in the process of electrolytic aluminum plant for the production of alumina.The transportation technique is straight influence the stabilization of the production of alumina and the consumption per unit of alumina.Along with widely application of hyper dense phase conveying system and the improvement of automatic level day by day,through optimize and control to ensure the quality of product,and then realizing consumption reduction and Enlarging has already become the goal that enterprises have pursued.For the goal more craft parameters and variables are needed to be measured on-line,and alumina powder flow in electrolytic aluminum plant is one of the main parameters.The soft sensor technology have been presented itself as a new technology that settles the main parameter measured on-line, improving the productivity benefit and guarantee the quality of production in recent years.On the background of the aluminium transport course in electrolytic aluminum plant,the soft sensor technology have been presented to measure the flow of alumina powder in electrolytic aluminum plant because of the difficult problem that alumina powder flow can not be precise measured on-line for import instruments cost expensive and the precision of domestic instrument is relatively low.At first on the basis of further investigate the craft of Alumina conveying and the measured information which the automation system can be offered the wind pressure and the wind flux of the centripetal fan are chosen as assistant variable to estimate the flow of alumina powder.Secondly on the basis of the actual industry’s on-the-spot data PLS,RBF and LS-SVM simulation research has been carried out with the single model soft sensor of alumina powder flow. And then considering the impact of model parameter chosen on model performance Pso algorithm has been utilized to confirm its parameters and achieved higher precision and better generalization ability forecast model.It is difficult for single model to describe global properties of complex system,and multi-point of complex systems in work is taken into account,so a multi-modeling soft sensor based on double layers intelligent structure is proposed in this paper,in which fuzzy c-means clustering is classification layer,and many PLS,may RBF neural network,many LS-SVM and PLS+RBF+LS-SVM are modeling layers.The degrees of membership are used for combining the output of sub-models to obtain the finial result which is the measured value of estimated parameters,and then the model of alumina powder flow is founded. The result of simulation research proves that the single model and multi-modeling modeling technique all have better generalization ability and higher precision accuracy and different multi-models are superior.The soft sensor technique of LS-SVM is more suitable model that could be revision on-line and proved that this technique is feasible and effective,at the same time the soft sensor technology have offered the basis for the real-time monitoring of alumina powder flow in the field of the industry of aluminium.更多还原