中文摘要：

针对无缝钢管连轧生产中衡量荒管质量好坏的横向和纵向壁厚不均检验滞后和难以建立其机理模型的问题，提出了基于步进均值子时段MPLS方法的荒管质量预报模型．介绍了步进均值子时段MPLS方法中过程数据时段分解、均值求取、回归模型建立和模型在线预报等关键内容．将建立的预报模型用于荒管质量预报中，为提高连轧生产的无缝钢管质量奠定了良好的基础．其实时性良好，可靠性和精度高，可用于荒管质量的在线预报和优化。

英文摘要：

The shell is produced by the semi-floating mandrel mill, whose shell longitudinal and transversal wall thickness unevenness is often checked laggingly. The quality prediction of the shell model based on step mean value staged MPLS （multiway partial least square） method is proposed to overcome the disadvantage. The staged decomposition of the productive data, calculation of the mean value in the mean value MPLS method, modeling, and on-lined prediction are introduced. The model used in the shell quality prediction can improve seamless tubes quality of mandrel mill made effectively. Its obvious benefits are good real time function, high reliability precision, and can be used on-line for the prediction and optimization on the quality of the shell.