To prevent the batch occurrence of surface cracks and achieve anomaly localization, a longitudinal crack prediction model based on an extreme gradient boosting tree(XGBoost) and just-in-time learning was proposed. First, a local sample set was selected, and a support vector machine oversampling algorithm was used to oversample the cracked samples. Then, historical samples were updated according to a dual-weighting strategy of temporal relevance and Chebyshev distance. Finally, a longitudinal crack prediction model based on an extreme gradient boosting tree was constructed, achieving an accuracy of 90%. A root cause diagnosis model for surface longitudinal cracks based on a Bayesian network was proposed. First, key influencing factors were selected to construct an initial directed acyclic graph; each feature was discretized into three categories: low, medium, and high, and secondary factors were eliminated using mutual information. Second, the optimal network structure was determined using a Dirichlet prior scoring search algorithm combined with a hill-climbing algorithm. Finally, parameter learning was performed using maximum likelihood estimation (MLE), and a causal network was obtained through a belief propagation algorithm. The results indicate that silicon content and the temperature difference of crystallizer cooling water outlet are important factors for the occurrence of longitudinal cracks.
ZouLei-lei, HuangJun-xiong, LiQuan-hui, et al. Research progress on the crack and segregation prediction of continuous casting strand [J]. Continuous Casting, 2022, 41(2): 2-9.
[3]
SharafiS M, EsmaeilyH R. Applying data mining methods to predict defects on steel surface [J]. Journal of Theoretical & Applied Information Technology, 2010, 20(2): 87-92.
[4]
WangB R, YangY K, YuZ H, et al. Real-time surface crack defects prediction of 82B steel billet based on multi-model fusion [J]. Metallurgical and Materials Transactions B, 2025, 56(4): 4287-4298.
LiZhong-zheng, WuZhao-xia, WangJin-yang, et al. FeO content prediction model in sinter based on GA-BiLSTM with feature optimization [J]. Journal of Northeastern University(Natural Science), 2025, 46(6): 56-65.
[7]
WangQ C, ZhangL D, WeiZ J, et al. Intelligent optimization method to reduce longitudinal cracks of continuous casting slabs [J]. Metallurgical and Materials Transactions B, 2025, 56(5): 4510-4523.
[8]
NorrenaJ, LouhenkilpiS, VisuriV V, et al. Coupling of solidification and heat transfer simulations with interpretable machine learning algorithms to predict transverse cracks in continuous casting of steel [J]. Steel Research International, 2024, 95(4): 2300529.
[9]
WeiL X, ZhaiB H, SunH, et al. An ensemble JITL method based on multi-weighted similarity measures for cold rolling force prediction [J]. ISA Transactions, 2022, 126: 326-337.
[10]
JinH P, LiJ G, WangM, et al. Ensemble just-in-time learning-based soft sensor for mooney viscosity prediction in an industrial rubber mixing process [J]. Advances in Polymer Technology, 2020, 2020(1): 6575326.
[11]
VenkataV S, MohantaH K, PaniA K. Support vector regression modeling in recursive just-in-time learning framework for adaptive soft sensing of naphtha boiling point in crude distillation unit [J]. Petroleum Science, 2021, 18(4): 1230-1239.
[12]
KitsonN K, ConstantinouA C, GuoZ G, et al. A survey of Bayesian network structure learning [J]. Artificial Intelligence Review, 2023, 56(8): 8721-8814.
[13]
GiudiceE, KuipersJ, MoffaG. The dual PC algorithm and the role of Gaussianity for structure learning of Bayesian networks [J]. International Journal of Approximate Reasoning, 2023, 161: 108975.
[14]
MurphyK, SchölkopfB, ColomboD, et al. Order-independent constraint-based causal structure learning [J]. Journal of Machine Learning Research, 2014, 15(1): 3741-3782.
[15]
LiuJ X, TianZ L. Verification of three-phase dependency analysis Bayesian network learning method for maize carotenoid gene mining [J]. BioMed Research International, 2017, 2017(1): 1813494.
[16]
MasegosaA R, MoralS. An interactive approach for Bayesian network learning using domain/expert knowledge [J]. International Journal of Approximate Reasoning, 2013, 54(8): 1168-1181.
[17]
BehjatiS, BeigyH. Improved K2 algorithm for Bayesian network structure learning [J]. Engineering Applications of Artificial Intelligence, 2020, 91: 103617.
[18]
NeathA A, CavanaughJ E. The Bayesian information criterion: background, derivation, and applications [J]. Wiley Interdisciplinary Reviews, 2012, 4(2): 199-203.