摘要
冷轧板带生产中的振动机制复杂多变,工业生产过程采集到的历史数据中蕴含着设备运行状态与产品质量等关联信息,数据驱动的信息挖掘对实现振动状态监控与智能预测具有重要价值。首先针对实际工业生产中采集的历史振动信号进行数据预处理;然后对典型振动钢卷的信号进行经验模态分解,选取相关度较高的主要本征模态分量进行时频分析;随后,基于相关度最高的本征模态分量提取10个时频特征指标作为输入,以振动有效值作为输出标记,构建样本空间;最后,运用随机森林回归算法对多个材质和规格下的钢种振动信号进行识别。结果表明,所提出的经验模态分解与随机森林相结合的方法可以适应样本数量不平衡情况下的振动信号识别问题,能够为冷轧机振动的状态监控与智能预报提供依据。
The vibration mechanism in cold rolled strip production is complex and variable, but the historical data collected in the industrial production process contains information related to the equipment operation status and product quality, and data-driven information mining is of great value to realize vibration monitoring and intelligent forecasting. The historical vibration signals collected in actual industrial production were first preprocessed, and then empirical modal decomposition(EMD) on the signals of typical vibrating steel coils was performed, and the main intrinsic mode function(IMF) with high correlation for time-frequency analysis was selected. Subsequently, 10 time-frequency feature indicators were extracted as inputs based on the highest correlation IMF, and the sample space was constructed with the vibration energy value as the output marker. Finally, the random forest regression algorithm was applied to identify the vibration signals of steel species under multiple materials and specifications. The results show that the proposed method of combining empirical modal decomposition and random forest can be adapted to the problem of identifying vibration signals with unbalanced sample size, and can provide a basis for state monitoring and intelligent prediction of cold mill vibration.
作者
宋寅虎
郜志英
周晓敏
赵潇雅
SONG Yinhu;GAO Zhiying;ZHOU Xiaomin;ZHAO Xiaoya(College of Mechanical Engineering,University of Science and Technology Beijing,Beijing 100083,China)
出处
《钢铁研究学报》
CAS
CSCD
北大核心
2023年第3期303-312,共10页
Journal of Iron and Steel Research
基金
国家自然科学基金资助项目(51775038)。
关键词
冷轧机
振动信号
时频特征
经验模态分解
随机森林
cold rolling mill
vibration signal
time-frequency feature
empirical mode decomposition
random forest
作者简介
宋寅虎(1997-),男,硕士,E-mail:15732575177@163.com;通讯作者:郜志英(1979-),女,博士,教授,E-mail:gaozhiying@me.ustb.edu.cn。
