多源数据驱动的轧机振动预测及可解释性分析

Vibration Prediction and Interpretability Analysis of Rolling Mill Driven by Multi-source Data

为研究轧制过程动态工艺参数对轧机振动的影响规律,改善现有研究中机理模型精度较低且数据模型缺乏可解释性的问题,采用极端梯度提升(Extreme Gradient Boosting,XGBoost)算法建立基于多源数据的轧机振动预测模型,并使用SHapley Additive exPlanations(SHAP)框架对预测模型进行解释。通过与其他预测模型相比,XGBoost预测模型可以利用工艺参数实现对轧机运行状态的高精度预测。基于SHAP框架解释的结果表明,出入口厚度、轧制力、轧制速度对轧机振动影响较大,后张力对轧机振动影响较小。研究为提高轧机设备与工艺参数的匹配度,实现将工业数据应用于轧机振动预测和分析提供理论基础。

In order to study the influence law of dynamic process parameters on rolling mill vibration,and improve the problem of low accuracy of mechanism model and lack of interpretability of data model in existing research,the Extreme Gradient Boosting(XGBoost)algorithm was used to establish the rolling mill vibration prediction model based on multisource data,and the SHapley Additive exPlanations(SHAP)framework was used to explain the prediction model.Compared with other prediction models,the XGBoost prediction model can achieve high-precision prediction of the operating state of the mill by using the process parameters.Based on the SHAP framework interpretation,the results show that the thickness of the inlet and outlet,the rolling force and the rolling speed have great influences on the rolling mill vibration,while the posttension has small influence on the rolling mill vibration.This study provides a theoretical basis for improving the matching degree of rolling mill equipment and process parameters,and realizing the application of industrial data in rolling mill vibration prediction and analysis.