烧结终点(burning through point,BTP)位置是烧结过程中重要的参数,直接影响烧结机效率.由于烧结生产过程具有多工况、时变等特性,使得全局模型预测性能不足,为此提出了一种在即时学习框架中使用极端梯度提升(extreme gradient boosting,XGBoost)作为局部模型的烧结终点预测模型,即JITL(just-in-time learning)-XGBoost.首先采用KL散度(Kullback-Leibler divergence)相似性度量方法提取待测样本的特性,选出与待测样本最相关的数据集.然后将该数据集作为XGBoost模型的输入来预测烧结终点的位置.此外,考虑了相关数据集数量对模型预测精度和计算时间的影响.最后与其他模型对比,结果表明,所建模型在合理的时间内具有最佳预测精度,为提高烧结机效率提供新的指导方向.
Abstract
The burning through point (BTP) is an important parameter in the sintering process, which directly affects the efficiency of the sintering machine. Due to the multi-working conditions and time-varying characteristics of the sintering production process, the prediction performance of the global model is insufficient. Therefore, a burning through point prediction model using XGBoost as a local model in the just-in-time learning framework is proposed, namely JITL-XGBoost. Firstly, the KL divergence similarity measurement method is used to extract the characteristics of the sample to be tested, and the most relevant data set of the sample to be tested is selected. Secondly, this dataset is used as input to the XGBoost model to predict the location of the burning through point. In addition, the impact of related dataset numbers on model prediction accuracy and model computation time is considered. Finally, by comparing with other models, the results show that the model built has the best prediction accuracy within a reasonable time, providing new guidance for improving the efficiency of sintering machines.
针对传统模型存在的局限性,即时学习建模框架采用“边建模、边预测”的方式,通过相似度计算选出与待测样本最相关的样本集构建局部模型,实现短时间内高精度预测[9].当前,即时学习算法已在部分领域用于构建预测模型.任明仑等[10]将即时学习算法引入灰铸铁抗拉强度预测,并与传统全局模型对比验证即时学习算法的有效性.王通等[11]利用即时学习算法来提高油田液面测量的精度.针对烧结过程中时变导致预测性能不足的问题,本文首先对烧结数据进行预处理,得到可以直接建模的有效数据.其次,通过最大互信息系数(maximal information coefficient,MIC)得到26个与BTP最相关的变量.然后,利用KL散度相似性度量方法计算出查询样本和历史数据库的距离,进而选取相关数据集.最后将选取的相关数据集作为XGBoost模型的输入,得到查询样本的最终BTP预测值,并将BTP数据库及时更新.烧结生产过程中的实际数据验证了该模型的优越性和可靠性,对BTP的及时、准确预测具有重要意义.
DuS, WuM, ChenL F, et al. A fuzzy control strategy of burn-through point based on the feature extraction of time-series trend for iron ore sintering process[J]. IEEE Transactions on Industrial Informatics, 2020, 16(4): 2357-2368.
[2]
PatissonF, BellotJ P, AblitzerD, et al. Mathematical modelling of iron ore sintering process[J]. Ironmaking & Steelmaking, 1991, 18(2): 89-95.
[3]
CummingM J, ThurlbyJ A. Developments in modelling and simulation of iron ore sintering[J]. Ironmaking & Steelmaking, 1990, 17(4): 245-254.
[4]
ChenX X, ShiX H, TongC D. Multi-time-scale TFe prediction for iron ore sintering process with complex time delay[J]. Control Engineering Practice, 2019, 89: 84-93.
[5]
DuS, WuM, ChenL F, et al. Prediction model of burn-through point with fuzzy time series for iron ore sintering process[J]. Engineering Applications of Artificial Intelligence, 2021, 102: 104259.
[6]
ChenX, HuJ, WuM, et al. T–S fuzzy logic based modeling and robust control for burning-through point in sintering process[J]. IEEE Transactions on Industrial Electronics, 2017, 64(12): 9378-9388.
[7]
LiuS, LyuQ, LiuX J, et al. A prediction system of burn through point based on gradient boosting decision tree and decision rules[J]. ISIJ International, 2019, 59(12): 2156-2164.
[8]
WangD D, YangK, HeZ J, et al. Application research based on GA-FWA in prediction of sintering burning through point[C]//Proceedings of 2018 International Conference on Computer, Communications and Mechatronics Engineering. Shanghai,2018: 378-385.
Zhao Ji-min, HeYang, Liu Jian-hua, et al. Predictive method of casting slab quality based on just-in-time learning algorithm[J]. Continuous Casting, 2022, 47(6): 54-60, 67.
Ren Ming-lun, Song Yue-li, ChuWei. Locally weighted linear regression modeling for tensile strength prediction of gray cast iron[J]. Journal of Electronic Measurement and Instrumentation, 2019, 33(3): 65-71.
WangTong, Gao Xian-wen, Liu Wen-fang. Soft sensor for determination of dynamic fluid levels based on enhanced just-in-time learning algorithm[J]. Journal of Northeastern University (Natural Science), 2015,36(7): 918-922.
[15]
XiaG L, WuZ X, LiuM Y, et al. Prediction interval estimation of sinter drum index based on light gradient boosting machine and kernel density estimation[J]. Ironmaking & Steelmaking, 2023, 50(8): 909-920.
[16]
LiD C, HuangW T, ChenC C, et al. Employing box plots to build high-dimensional manufacturing models for new products in TFT-LCD plants[J]. Neurocomputing, 2014, 142: 73-85.
[17]
SimonN, TibshiraniR. Comment on “detecting novel associations in large data sets” by reshef et Al, science dec 16, 2011[EB/OL].(2014-06-14)[2023-05-15]. 2014: 1401.7645.
[18]
LiG Y, LiW, TianX L, et al. Short-term electricity load forecasting based on the XGBoost algorithm[J]. Smart Grid, 2017, 7(4): 274-285.