【作者】 张勇；

【导师】 王伟；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2006， 博士

【摘要】 浮选过程机理模型由于其自身复杂性和其假设条件在工业生产中很难得到满足,因而其应用受到一定的限制。对此问题,本文将神经网络和粗糙集理论引入浮选建模过程中,将数据预处理和软测量模型与浮选生产工艺有机结合,探索浮选过程的建模与智能优化方法。本文主要工作如下所述: (1)介绍了鞍钢集团弓长岭矿山公司选矿厂阳离子反浮选过程工艺流程,并对浮选过程进行了详细系统分析,论述了浮选过程自动控制研究现状,综述了粗糙集理论、神经网络和智能系统的研究概况。 (2)研究了智能系统建模前数据预处理技术。采用模糊聚类-线性回归方法获得采集数据置信区间,去除数据中“坏样”样本。采用控制图法对浮选过程实时数据进行监测,为浮选过程优化控制提供良好的输入数据。 (3)详细研究了浮选工艺流程,了解操作条件对浮选技术指标的影响,为浮选过程经济技术指标(精矿品位和浮选回收率)软测量模型选择合适辅助变量。采用主元分析法和径向基神经网络技术建立浮选技术指标预测模型。主元分析法用来对神经网络模型输入进行降维处理,简化模型复杂度;RBF神经网络采用最近邻聚类学习算法进行训练。 (4)结合粗糙集理论和神经网络的各自特点,提出了一种基于粗糙集一神经网络的浮选过程药剂用量数学模型,并且与基于粗糙集控制思想的浮选过程药剂添加模型进行了比较。将浮选过程经济技术指标软测量模型和浮选过程药剂添加模型的浮选过程控制用于弓长岭矿山公司选矿厂实际生产,取得了很好的应用效果。更多还原

【Abstract】 Mechanism model of flotation process based on some hypotheses is limited in the application for flotation process, since the flotation process is complex and dose not satisfy these hypotheses condition in industrial condition. Focusing on this problem, this paper combines of artificial neural networks (ANN) and rough set (RS) theory to model the flotation industry process by integrating data pretreatment method with soft sensors model to flotation technology. The paper explores the modeling and intelligent optimization of flotation process. The main contents of the paper are as follows:(1) The paper first Introduce the technical flow of cation anti-flotation process of mill factory in GongChangLing mining company of AnGang group and makes a detailed systematic analysis to flotation process. The paper discusses the actuality of automatic control of flotation process and summarizes the research on theory and application of RS theory, ANN and intelligent system.(2) The paper studies the data pretreatment technique of the intelligent system before modeling. Believing region of the sample data set is got by using fuzzy C-means clustering algorithm and linear regression method in order to eliminate bad sample data. Control chart method is used to supervise real-time data from flotation process and provides good input data for optimal control of the flotation process.(3) Based on detailed research on flotation process and influence of operation conditions on flotation technology, the paper chooses proper assistant variables for economy-technology index (extractive ore grade and flotation callback ratio) soft sensors model of flotation process. The paper adopts principal component analysis (PCA) method and radial basis function (RBF) neural network technology to build the soft sensors model. The PCA algorithm is used to deduce the input dimension of RBF neural networks and predigest model complexity. The RBF neural network is trained by the nearest neighbor-clustering algorithm.(4) A medicament dosage model of flotation process based on rough set theory and neural networks is proposed in this paper. It is compared with the medicament dosage model based on rough control theory. The flotation process intelligent control system based on the economy-technology index soft sensors model and the medicament addition model is used in the practice production of mill factory of GongChangLing mining company and A better application result is obtained.更多还原