现代生产中的大量生产数据蕴藏着丰富的生产过程和质量信息,通过聚类分析可以了解生产状态,进行生产故障诊断或有针对性的质量检测,而经常使用的相似性的度量欧式距离只能反映数据空间分布为球形或超球形的结构特性.难以刻画复杂数据分布特性,将流形距离引入到生产过程状态的聚类分析中,利用标准数据、田纳西——伊斯曼过程和热轧带钢实际生产过程数据对方法的有效性进行验证,进而可以更加有效地了解生产过程的状态.%More and more data are collected in model manufacturing process. There are rich information of the production state and quality among the data. The clustering method with process data is used to acquire the production status, thus for process diagnosis and enhancing the focal points of the quality inspect. The Euclidean distance as the common similarity measure, can only extract the features of the spherically distribution data and can not express the complex distribution data. This paper introduced the manifold distance to do the production state clustering. It used the benchmark data, Tennessee-Eastman process data and hot steel rolling process data for model validation. As a result the proposed method has better performance on clustering, compared with the Euclidean distance.
展开▼
