面向酿酒过程多模型软测量建模的多视角大规模模糊聚类方法研究

项目名称： 面向酿酒过程多模型软测量建模的多视角大规模模糊聚类方法研究

项目编号： No.61300151

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 自动化技术、计算机技术

项目作者： 王骏

作者单位： 江南大学

项目金额： 25万元

中文摘要： 由于酿酒发酵过程数据采集的设备呈现多组化，多视角化以及智能化的特点，导致采集的数据具有数据量大、多视角和演化等重要特性，为模糊聚类技术在此领域的应用提出了重要挑战。本项目拟针对酿酒过程多模型软测量建模的实际需要，探索面向酿酒过程多模型软测量建模的多视角大规模模糊聚类方法。研究内容包括：（1）结合酿酒发酵过程的机理模型，研究面向多视角数据的自适应模糊聚类新方法MV-AFC；（2）研究基于二阶差分的蒙特卡罗自适应采样新方法，以渐进采样为策略，以实现大规模数据环境下的快速模糊聚类；（3）开发领域自适应多视角模糊聚类学习算法DA-MV-AFC，以实现演进渐变式学习。课题组前期工作为本课题的展开提供了充分的技术准备。

中文关键词： 模糊聚类；多视角聚类；无监督距离度量学习；深度特征表示；

英文摘要： The data collected from the brewing process are becoming larger and larger and even contain multi-view information, which poses a great challenge to current studies on fuzzy clustering. Motivated by such a challenge, this project aims to develop large-scaled multi-view fuzzy clustering algorithms for the multiple soft sensing models in practical brewing process. Firstly, a novel multi-view based adaptive fuzzy clustering algorithm named MV-AFC will be developed by integrating with the mechanical model of fermentation for the brewing process. Secondly, a novel progressive sampling strategy using a novel second order differential Monte Carlo sampling will be developed and accordingly a fast multi-view fuzzy clustering algorithm fast-MV-AFC will be proposed for clustering on large-scaled data. Thirdly, a novel multiple view adaptive fuzzy clustering algorithm will be developed for domain adaptation. Our work in the past years has provided solid foundation for this project.

英文关键词： fuzzy clustering；multi-view learning；unsupervised distance metric learning；deep feature representation；