摘要
传统转炉炼铜工艺非常依赖个人经验,一般通过人工观察炉口火焰判断炉体内温度、造渣和造铜终点,该过程存在重大的安全与环保隐患,同时粗铜的品位得不到保障且易损坏炉体。随着环保与本质化安全指标要求的提高,各家企业开始转向闭窗吹炼。为应对绿色化与安全性要求,本文基于转炉炉口火焰分析结果并综合吹炼工艺中涉及的多种因素,设计了一种转炉炉口图像智能监测系统,实现了转炉炉口火焰的智能化监测。在此基础上,通过对不同阶段炉口火焰图像的分析,设计了基于自适应曝光阈值的颜色特征计算方法,能够对图像进行预处理并提取到可预测终点时间的关键特征,解决了目前传感器曝光度参数影响图像特征的关键问题。最后设计了基于深度神经网络的终点时间预测模型,实验结果表明造渣一期、造渣二期和造铜期终点的预测误差分别为0.74,0.83和1.4 min,显示了设计系统的有效性。
The traditional copper smelting process of converter depends very much on personal experience.By manually observing the flame of the furnace to judge the temperature in the furnace body and the end point of slag making and copper making,there are major safety and environmental protection problems.At the same time,the grade of copper in the corresponding stage is not guaranteed and the furnace body is easily damaged.With the improvement of environmental protection and intrinsic safety indicators,companies began to turn to closed window blowing.In order to meet the requirements of green and safety,based on the results of flame analysis and various factors involved in the blowing process,an intelligent monitoring system for the image of the converter mouth was designed to realize the intelligent monitoring of the converter mouth flame.On this basis,through the analysis of different stages of the furnace flame image,the color feature calculation method based on adaptive exposure threshold was designed,which could preprocess the image and extract the key features of the predictable end time,so as to solve the key problem that the sensor exposure parameters affect the image features.Finally,a prediction model of the end point time based on deep neural network was designed.The experimental results showed that the prediction errors of the end point in the first slag making stage,the second slag making stage and the copper making stage were 0.74,0.83 and 1.4 min,respectively,and the success rate was higher than 91.4%,which showed the effectiveness of the system designed in this paper.
作者
张红哲
徐子昂
刘光远
郭廷谦
张冰洁
ZHANG Hongzhe;XU Ziang;LIU Guangyuan;GUO Tingqian;ZHANG Bingjie(China Enfei Engineering Technology Co.,Ltd.,Beijing 100038,China;School of Navigation,Northwestern Polytechnical University,Xi’an 710072,China)
出处
《铜业工程》
CAS
2024年第2期1-7,共7页
Copper Engineering
基金
北京市自然科学基金项目(8234060)资助。
关键词
转炉吹炼
炉口火焰图像
辅助炼铜
深度学习
终点预测
bessemer blowing
furnace flame image
auxiliary copper smelting
deep learning
endpoint prediction
