面向智能烧结的机尾断面烧结矿FeO预测研究
Research on FeO prediction of sintered ore in machine tail section for intelligent sintering
张学锋 1闻亦昕 1熊大林 2龙红明3
作者信息
- 1. 安徽工业大学计算机科学与技术学院,安徽马鞍山 243032
- 2. 首钢自动化信息研究院,河北唐山 063000
- 3. 安徽工业大学冶金工程学院,安徽马鞍山 243032
- 折叠
摘要
针对烧结工艺FeO预测准度问题,提出一种基于机器学习的预测模型.通过对机尾断面温度数据的采集和处理,建立包含多个特征的数据集.采用基于MIV(平均影响值)的特征选择方法,筛选出对预测模型权重占比较高的特征.使用Bi-LSTM(双向长短时记忆神经网络)算法对生产工艺数据进行训练和测试,得到高精度的预测模型.通过实验验证了该模型的预测效果,并与其他神经网络模型方法进行了比较,比较结果表明该模型具有较高的预测精度和实用性.在企业误差允许的范围内,准确率达90.2％,因此可以为智能烧结技术的应用和烧结质量的控制和优化提供重要的参考.
Abstract
Aiming at the prediction accuracy of FeO in sintering process,a prediction model based on machine learning was proposed.By collecting and processing the temperature data of the tail section,a data set containing multiple features was established.The feature selection method based on MIV(Mean Impact Value)was used to screen out the features that account for a higher weight of the prediction model.The Bi-LSTM(bidirectional long-short-term memory neural network)algorithm was used to train and test the production process data to obtain a high-precision prediction model.The prediction effect of the model was verified by experiments,and compared with other neural network model methods,the comparison results show that the model has high prediction accuracy and practicability.Within the allowable range of enterprise error,the accuracy rate reaches 90.2％,so it can provide an important reference for the application of intelligent sintering technology and the control and optimization of sintering quality.
关键词
智能烧结/预测模型/烧结矿FeO含量/大数据/神经网络Key words
intelligent sintering/prediction model/FeO content of sinter/big data/neural network引用本文复制引用
基金项目
安徽省教育厅重点实验室项目(TZJQR007-2023)
安徽省高等学校自然科学研究项目(2022AH050290)
出版年
2024
NETL
NSTL
万方数据