文档介绍：基于主元分析和模糊聚类的浮选过程的数据预处理
摘要
随着信息时代的来临,人类在各种领域中面临着越来越多的数据信息,与此同时,这些数据的规模还在以惊人的速度不断增长。鉴于主元分析法的降维特性和模糊C-均值聚类算法良好的分类性能,本文针对反浮选过程的被控对象复杂、数学模型不确定以及控制要求高等特点,提出一种基于主元分析和模糊聚类的数据预处理算法。采用模糊C-均值聚类算法得到聚类中心,然后进行线形回归从而对过程变量数据进行了预处理。主元分析法则用来进行辅助变量的选取和输入高维向量降维简化。在保留原有信息的基础上,去除了冗余数据,加快了聚类速度,在实现对模型的输入简化以及输入数据的故障诊断,为过程建模、先进控制和优化控制等打好基础。然后针对主元变量采用径向基函数网络建立了系统经济技术指标的预测模型。根据工业实际生产数据进行的模型校验和误差分析表明,能够满足浮选过程控制的精度要求。
关键词:数据预处理;模糊C均值聚类;主元分析;浮选过程
Data Pretreatment of Flotation Process Based on ponent Analysis and Fuzzy C-means Clustering
Abstract
With ing of information age, human are confronted with increasing data and information in different fields. At the same time, these data are developing in surprisingly speed. A data pretreatment algorithm based on ponent analysis and fuzzy c-means clustering for flotation process is proposed in this paper. Linear regression of clustering centers gained by fuzzy c-means clustering algorithm is introduced to carry through data pretreatment. The paper adopts ponent analysis to select the primary variables and reduce dimensions of input vectors. By dong so, original information is kept down and redundant information is removed, which builds up the foundation for process modeling, advanced control technology and optimized control, and so on. Then the paper uses radial basis work to set up the prediction model of economy and technology index in flotation process aiming at ponent variables. Model verification presented by using real operating data from industrial experiments indicates that the model’s precision is good enough to satisfy the request of floatation process control.
Keywords: Data pretreatment; Fuzzy C-means clustering (FCM); ponent analysis (PCA); Flotation process
目录
摘要 I
Abstract II
1 绪论 1
研究背景 1
数据预处理技术的研究 2
2 模糊C-均值聚类算法 4
模糊C-均值简介及算法分析 4
模糊C-均值聚类算法的实现原理 5
FCM聚类算法的一般步骤 6
数据预处理结果 6
3 基于主元分析的数据预处理 9
引言 9
主元分析基本思路 9
基于主元分析的数据降维 11
基于PCA-RBF的浮选过程软测量模型 14
RBF神经网络的学习算法 14
RB