基于PCA-RF的热轧带钢板凸度预测  

Strip Crown Prediction of Hot Rolled Based on PCA-RF

作　　者：赵志挺 ZHAO Zhiting(School of Mechanical and Power Engineering,Shenyang University of Chemical Technology,Shenyang110142,China)

机构地区：[1]沈阳化工大学机械与动力工程学院,沈阳110142

出　　处：《机械工程师》2023年第10期42-46,共5页Mechanical Engineer

摘　　要：为了满足日益增长的带钢板凸度预测精度和速度要求,建立了一种基于降维的主成分分析(Principal Component Analysis,PCA)协同随机森林(Random Forest,RF)的板凸度预测模型。首先,应用Pauta准则去除异常值,用五点三次平滑公式进行降噪处理;其次采用主成分分析法对数据进行降维,利用载荷矩阵选取关键控制变量;最后利用关键控制变量建立基于随机森林的板凸度预测模型,并与支持向量机回归(Support Vector Regression,SVR)、最近邻(K Nearest Neighbor,KNN)、轻量梯度提升机(light Gradient Boosting Machine,LightGBM)、极端梯度增强(Extreme Gradient Boosting,XGBoost)和梯度提升决策树(Gradient Boosting Decision Tree,GBDT)模型进行比较。结果表明,PCA-RF模型将参数由93维降低到15维,极大地减少了建模时间,且PCA-RF对测试集预测的决定系数(Coefficient of Determination,R^(2))、平均绝对误差(Mean Absolute Error,MAE)和均方根误差(Root Mean Squared Error,RMSE)分别为0.9820、1.4852μm和2.2603μm,均优于其他预测模型,且98%以上样本点的预测误差为-3~3μm,满足板凸度预测的精度要求。从而证明该模型能够通过降维减少建模时间,同时实现了带钢板RF凸度的高精度预测,为热轧带钢板凸度的研究提供了一定的参考。In order to meet the increasing accuracy and speed requirements of strip crown prediction,a crown prediction model based on principal component analysis(PCA)and random forest(RF)is established.Firstly,the Pauta criterion is used to remove outliers,and the five-point cubic smoothing formula is used for noise reduction.Secondly,the principal component analysis is used to reduce the dimension of data,and the load matrix is used to select the key control variables.Finally,a strip crown prediction model based on random forest is established by using key control variables.It is compared with the support vector regression(SVR),K-Nearest Neighbor(KNN),light gradient boosting machine(LightGBM),extreme gradient boosting(XGBoost)and gradient boosting decision trees(GBDT)models.The results show that the PCA-RF model reduces the parameters from LightGBM dimensions to 15 dimensions,which greatly reduces the modeling time.The coefficient of determination(R^(2)),mean absolute error(MAE)and root mean squared error(RMSE)of PCA-RF based on the test set are 0.9820,1.4852μm and 2.2603μm,respectively.The prediction error of more than 98%sample points is-3—3μm,which meets the accuracy requirements of strip crown prediction.It is concluded that the PCA-RF model can reduce the modeling time by dimensionality reduction while achieving the high-precision prediction of strip crown,which provides a certain reference for the study of strip crown.

关 键 词：热轧带钢 板凸度预测 主成分分析 随机森林 

分 类 号：TG335.11[金属学及工艺—金属压力加工]