Abstract: To identify abnormal fluctuation of quality characteristic values in spinning process,the fluctuation mechanism of spinning quality characteristic values were theoretically analyzed.On the basis of massive spinning data,a Dk-means clustering algorithm was designed.Meanwhile,the optimal fiber attribute variable (elongation) was abstracted,the spinning quality output characteristic values were corrected,and the linear relationship between variable and characteristic value was completed.An estimating method for abnormal factors of spinning quality based on Dk-means algorithm and gray correlation analysis was proposed,and the abnormal fluctuation problem of spinning quality was solved.Combined with the simulation and experiment,results showed that the proposed estimation method had realized the identification of spinning quality's abnormal factors from the correlation coefficient and correlation degree,and the visualization from the fluctuation causes of quality characteristic values to the relationship expression between the abnormal factors and the spinning quality characteristic values,and then to the behavior identification of abnormal factors.

Key words: abnormal factor, spinning quality, identification, massive data, correlation degree

摘要： 为识别纺纱过程质量特征值的异常波动,对纺纱质量特征值的波动机理进行了理论分析。基于海量纺纱数据设计了一种Dk-means聚类算法,进而提取最优纤维属性变量(断裂伸长),对纺纱质量特征值进行修正,实现变量与特征值之间关系的线性化。提出一种基于Dk-means算法与灰色关联分析的纺纱质量异常因素识别方法,解决了纺纱过程质量异常波动的问题。仿真与实验结果表明,所提出的估计方法从关联系数和关联度的角度实现了纺纱质量异常因素的识别,并实现了质量特征值的波动成因到异常因素与纺纱质量特征值之间的相关关系表达,以及异常因素辨识的可视化。

关键词: 异常因素, 纺纱质量, 识别, 海量数据, 关联度