高炉
预测建模
焦炭
一般化
支持向量机
生产(经济)
质量(理念)
数据挖掘
回归分析
过程(计算)
计算机科学
回归
工艺工程
钥匙(锁)
人工神经网络
样品(材料)
机器学习
工程类
长期预测
数据建模
线性回归
贝叶斯网络
煤
基础(线性代数)
贝叶斯概率
人工智能
性能预测
数据驱动
粒子群优化
数据质量
模式识别(心理学)
作者
Haojun Li,Xiaoxia Han,Fengyi Liu,Chaojie Guo,Chenglin Li
出处
期刊:
日期:2025-09-12
卷期号:: 804-809
标识
DOI:10.1109/iotaai66837.2025.11213021
摘要
Accurate prediction of coke quality indicators is crucial for the blast furnace ironmaking process in the steel industry. However, factors such as insufficient samples of actual production data, imbalanced data quality, and difficulties in tuning model parameters have limited the improvement of prediction accuracy. To address these issues, this study proposes a coke quality prediction model that integrates a Generative Adversarial Network (GAN) with Bayesian-Optimized Support Vector Regression (Bayes-Svr). First, GAN is employed to generate and augment historical production data, thereby expanding the sample size and improving the uniformity of the data distribution. Second, the Bayesian optimization algorithm is used to adaptively search for the optimal values of the key parameters c and γ in Support Vector Regression (SVR) to enhance the model’s predictive performance. Finally, using coke quality indicators M40and M10as the prediction targets, experiments are conducted based on actual production data from a coking plant. The results demonstrate that the proposed GAN-Bayes-Svr hybrid model significantly outperforms traditional prediction models in both prediction accuracy and generalization capability, providing an effective theoretical basis for intelligent coal blending in steel enterprises.
科研通智能强力驱动
Strongly Powered by AbleSci AI