关键词: 冷轧颤振, 工业数据, 定量递归分析, 核密度估计

Abstract: The cold rolling chatter problem in the production of strip is always the key factor restricting the improvement of production efficiency and the development of new high-strength strip products, and it is difficult to detect and restrain the chatter in time by relying too much on production experience. Therefore, the rapid identification and early warning of chatter state is an important basis for intelligent strip rolling. A novel method of cold rolling chatter identification based on the combination of recurrence quantification analysis and kernel density estimation is proposed, and the historical vibration data of rolling mill accumulated in the actual production process are studied. Firstly, the construction of recurrence plot and recurrence quantification analysis of the acceleration signals under rolling stability state and rolling chatter state are carried out respectively. The difference of topological structure of the recurrence plot and recurrence quantification analysis indexes under two different states are compared. Then, the kernel density estimation method is used to calculate the vibration alarm threshold of different recurrence quantification analysis indexes to distinguish the stable state and chatter state. Finally, the fast and effective identification of chatter state is realized by the combination of recurrence quantification analysis and the alarm threshold. The research results show that, compared with the traditional time domain index of vibration acceleration signals, recurrence quantification analysis can identify and alarm the cold rolling chatter more quickly, improving the identification sensitivity of chatter state and the fast response ability when chatter occurs, and playing an important role in practical production and the intellectualization of strip cold rolling.

Key words: cold rolling chatter, industrial data, recurrence quantification analysis, kernel density estimation