基于LSTM算法的冷轧机架振动动态预警分析
Dynamic early-warning analysis of cold tandem rolling stand vibration based on LSTM algorithm
马志刚1
作者信息
- 1. 新乡职业技术学院 智能制造学院,河南 新乡 453000
- 折叠
摘要
在实际生产阶段冷轧机具有多态性与时变性,需要对轧机振动动态预警进行转换形成包含多变量的时间序列预警.建立了一种基于LSTM算法的冷轧机振动预警模型.研究结果表明:提高步长后模型预警性能获得明显提升,随着步长到达 5 后,模型表现也逐渐变差,步长为 4时,获得了最优预警效果.结合实际振动报警阈值,在预警振动能量值升高至阈值 75%时激发形成振动预报,第一卷与第二卷分别提前预报 1.6s与 3.2s.该研究对控制板材的精度具有很好的指导意义.
Abstract
In the actual production stage,the cold rolling mill has polymorphism and time variation,so it is necessary to convert the dynamic warning of mill vibration to form a multi-variable time series warning.A vibration warning model of cold rolling mill based on LSTM algorithm was established.The results show that the early-warning performance of the model is obviously improved after the step size is increased,and the performance of the model gradually deteriorates when the step size reaches 5,and the optimal early-warning effect is obtained when the step size is 4.Combined with the actual vibration alarm threshold,vibration pre-diction is generated when the early-warning vibration energy value rises to 75%of the threshold value.The first and second volumes are predicted in advance for 1.6s and 3.2s,respectively.The research has a good guiding significance for controlling the precision of sheet metal.
关键词
轧机振动/长短时记忆循环神经网络/预报/模型Key words
Rolling mill vibration/Cyclic neural network of short and long term memory/Forecast/Model引用本文复制引用
出版年
2024
NETL
NSTL
万方数据