建筑结构钢板热轧轧机DBN-PSO振动预报及应用
Prediction and Application of DBN-PSO Vibration in Hot Rolling Mill of Structural Steel Plate
王莹 1马晓力 2王强3
作者信息
- 1. 河南建筑职业技术学院,河南 郑州 450000
- 2. 郑州大学,河南 郑州 450000
- 3. 河南飞工重型机械制造有限公司,河南 郑州 450064
- 折叠
摘要
利用实时监测数据(Real-Time Monitoring Data,RMD)参数分析轧机振动状态,综合运用深度置信网络(Deep Be-lief Networks,DBN)与粒子群(Particle Swarm Optimization,PSO)算法构建轧机振动仿真模型,实现RMD参数的深度挖掘,并达到轧机振动的预报效果.通过融合处理能够获得非常接近实际振动过程的预测数据,具备优异预测能力.结合现场测试的初始数据预测误差在3.5%范围内,跟轧机振动情况相符.当轧制速率变慢后,振动加速度出现了降低结果;入口张力对轧机的振动加速度具有反向作用;轧机振动加速度相对出口张力表现为正相关特点;以不同宽度的轧件进行测试发现轧机振动加速度保持基本恒定的状态.该研究对提高热轧轧机运行稳定性,对保证建筑结构钢板成形精度具有很好的指导意义,可以拓宽到其它的成形设备优化领域.
Abstract
Real-Time Monitoring Data(RMD)was used to analyze mill vibration state.The Deep Belief Networks(DBN)and Particle Swarm Optimization(PSO)algorithms are used to construct a simulation model of rolling mill vibration to realize the Deep mining of RMD parameters and achieve the prediction effect of rolling mill vibration.Through fusion processing,the predic-tion data which is very close to the actual vibration process can be obtained,and it has excellent prediction ability.The prediction error of the initial data combined with the field test is within 3.5%,which is consistent with the rolling mill vibration.When the rolling rate slows down,the vibration acceleration decreases.The inlet tension has a reverse effect on the vibration acceleration of the rolling mill.The vibration acceleration of the rolling mill is positively correlated with the outlet tension.It is found that the vi-bration acceleration of the rolling mill is basically constant when the rolling parts of different widths are tested.The research has a good guiding significance for improving the running stability of hot rolling mill and ensuring the forming accuracy of building structural steel plate,and can be extended to other forming equipment optimization fields.
关键词
热轧/钢板/轧机振动/振动预报/DBN算法/PSO算法Key words
Hot-Rolling/Steel Plate/Mill Vibration/Vibration Prediction/DBN Algorithm/PSO Algorithm引用本文复制引用
基金项目
河南省高等学校重点科研项目(21B57001)
出版年
2024
NETL
NSTL
万方数据