摘要
人工智能技术在数据预测方面具有显著的优势,利用人工智能技术对焦炭质量数据进行预测,对于高炉冶炼产品品质预测具有关键影响。首先阐述了焦炭质量预测的影响因素,以及现有的人工智能技术在冶炼行业的研究。然后提出了一种基于人工智能的高炉冶炼焦炭质量预测的架构。接着论述了人工智能模型中用于生产预测的随机向量函数链接网络(RVFL),同时对该网络进行了神经元权重和网络结构的改进,以进一步提高模型预测的准确率。再者依据生产实际,选取了配合煤和生产工艺的各项指标作为特征向量,并以焦炭成品的质量作为标签,构建并训练了基于人工智能的焦炭质量预测模型。最后通过对某钢铁厂实际生产中的焦炭质量预测,对不同RVFL网络的性能进行对比和讨论,证明了改进算法后的RVFL的预测误差低于未改进的网络模型,在预测的准确率方面具有优势,能够满足冶炼生产的需要。
Artificial intelligence technology offers significant advantages in data prediction,with the prediction of coke quality data using artificial intelligence technology having a critical impact on product quality prediction in blast furnace smelting.Factors that influence coke quality prediction and the existing research on AI technology in the smelting industry is described.Subsequently,the random vector function link network(RVFL)for production prediction in the AI model is discussed,with the neuron weights and network structure of this network being improved to further enhance the accuracy of the model prediction.After that,based on actual production data,the indexes of matching coal and production process are selected as feature vectors,with the quality of the finished coke product being used as the label.The data is then used to construct and train a coke quality prediction model based on artificial intelligence.Finally,the performance of different RVFL networks in predicting coke quality in an iron and steel plant is evaluated.The results demonstrate that the improved algorithm of RVFL outperforms the unimproved network model in terms of prediction accuracy,and it is capable of meeting the needs of smelting production.
作者
朱兆松
周云磊
张胜伟
Zhu Zhaosong;Zhou Yunlei;Zhang Shengwei(School of Computer Science,Xi’an University of Posts and Telecommunications,Xi’an 710061,China;School of Mathematics and Computer Science,Nanjing Normal University of Special Education,Nanjing 210038,China)
出处
《机电工程技术》
2024年第9期47-50,75,共5页
Mechanical & Electrical Engineering Technology
基金
陕西省自然科学基金资助项目(2022JQ-376)。
关键词
焦炭质量
人工智能
预测模型
RVFL模型
coke quality
artificial intelligence
predictive modelling
RVFL model