基于数据挖掘的宽厚板板凸度控制

Wide and heavy plate crown control based on data mining

针对某宽厚板CVC plus大型骨干工业轧机存在板凸度偏大且难以控制的问题,对现场轧制过程工艺及板凸度质量数据进行跟踪采集和分析,通过数据转换、提炼和集成建立标准的数据挖掘数据集,基于历史生产数据建立宽厚板板凸度随机森林预测模型,应用主成分分析以及减法聚类离散化方法进行数据的预处理,通过关联规则挖掘和控制变量影响力评估实现不同产品质量状态下关键控制变量的快速定位,并将其应用于典型规格宽厚板关键工艺参数调整策略;建立三维有限元耦合模型用于调整策略的仿真分析。研究结果表明:3种典型规格宽厚板板凸度分别下降30.9%,14.7%和23.9%;基于数据挖掘提出的调整策略可以有效改善板凸度控制情况,可为宽厚板板形质量控制研究提供参考。

Considering that the crown of a wide and heavy plate CVC plus large mainstay industrial rolling mill was too large and difficult to control,rolling process and crown quality data were collected and analyzed.Standard data mining data sets were established through data conversion,extraction and integration.Based on historical production data,wide and heavy plate crown random forest prediction model was established.Principal component analysis and subtractive clustering discretization methods were used to preprocess the data.The rapid positioning of key control variables under different product quality conditions was realized by mining association rules and evaluating the influence of control variables.Then,it was applied to the adjustment strategy of key control process parameters for typical wide and heavy plates.A three-dimensional finite element coupling model was established for simulation and analysis of adjustment strategy.The results show that the crown of three typical wide and thick plates are reduced by 30.9%,14.7%and 24.0%,respectively.The proposed adjustment strategy based on data mining can effectively improve the control of plate crown,and provides reference for the profile and flatness control of wide and heavy plate.