纺纱过程质量波动预测新方法

Novel method for prediction of quality fluctuation in spinning process

依托纺纱过程所产生的海量数据,对纺纱质量特征值波动的成因、规律,以及影响质量特征值的各类不确定因素的产生机制进行了分析,并对不确定因素与质量特征值之间的相互作用机制进行了研究;利用人机环境系统工程学理论,从纺纱质量特征值修正、波动规律表达、人-机-环境脆性模型构建,以及TARCH(1,1)模型对影响因素异常行为辨识4个方面对纺纱质量特征值波动的内在机制进行了建模与设计,进而提出了基于数据的纺纱质量波动预测"四步法"。实验与对比分析结果表明,"四步法"实现了纺纱质量特征值波动过程的可视化,做到了影响因素异常行为的事前预警以及成纱质量的实时在线检测,为构建基于数据的纺纱质量预测与控制提供了新方法。

On the basis of massive data generated from the spinning process,the fluctuation cause and law of spinning quality characteristic value,and generation mechanism of uncertain factors affecting the quality characteristic value were analyzed,and the interaction mechanism between the uncertainty factors and the quality characteristic value were studied. By using man-machine-environment system engineering theory,the inherent mechanism of the fluctuation of the spinning quality characteristic value was modeled and designed from four aspects,which include the correction of the spinning quality characteristic output value,the expression of fluctuation law,the construction of man-machine-environment brittle model,and identification of abnormal behavior of factors with the TARCH（ 1,1） model. Finally,a four-step spinning quality fluctuation prediction method based on data was proposed. By experimentation,simulation and comparison, the results show that the four-step method proposed can realize the visualization of the fluctuation process of the spinning quality characteristic value,achieves the prewarning of abnormal behavior of uncertainty factors and the real-time online detection of the yarn quality.Furthermore,it will be conducive to provide a novel method for the construction of the prediction and control of the yarn quality based on data.