基于AOD炉口火焰图像特征的喷溅预测方法  

Splash prediction method based on AOD furnace flame image feature

作　　者：毛家怡 尤文[1] MAO Jia-yi;YOU Wen(College of Electrical and Electronic Engineering,Changchun University of Technology,Changchun 130012,China)

机构地区：[1]长春工业大学电气与电子工程学院,吉林长春130012

出　　处：《冶金自动化》2020年第3期36-41,79,共7页Metallurgical Industry Automation

基　　金：国家自然科学基金资助项目(51374040);吉林省科技发展计划资助项目(20180201019GX)。

摘　　要：氩氧精炼低碳铬铁合金过程中,炉口火焰能够表征冶炼阶段和冶炼的状态,喷溅是一种不正常的工作状态,喷溅前的火焰与正常工作的状态有所区别。为了降低喷溅率,提高喷溅的预测准确率,对炉口火焰的图像进行分析,利用灰度直方图统计特征进行特征提取。确定了最大灰度值、平均灰度值、火焰丰度、能量、均方差、信息熵6个特征量,分别计算正常工作和3种喷溅强度下的6种特征向量的类间距离,并以此进行评价,采用决策层融合方法,对特征向量进行赋予权值,试验比较决策层融合方法与其他特征值两两作为识别特征的结果。经过在线的检测试验验证,采用决策层融合方法有较好的识别率。In the process of argon-oxygen refining of low-carbon ferrochrome,the furnace flame can characterize the stage and state of smelting. Splash is an abnormal working state,and the flame before splashing is different from the normal working state. In order to reduce the splash rate and improve the prediction accuracy of the splash,the image of the flame of the mouth was analyzed,and the feature extraction was performed by statistical features of the gray histogram. The six characteristic quantities including maximum gray value,average gray value,flame abundance,energy,mean square error and information entropy were determined,and the inter-class distances of the six eigenvectors of normal working and three kinds of splashing strength were calculated respectively. This evaluation was carried out by using the decision-making layer fusion method to assign weights to the feature vectors,and comparing the results of the decision-making layer fusion method with other feature values as the identification features. After online testing experiments,the decision-making layer fusion method had a good recognition rate.

关 键 词：氩氧精炼 炉口火焰 喷溅 特征提取 类间距离 决策层融合 

分 类 号：TF641[冶金工程—钢铁冶金] TP391.41[自动化与计算机技术—计算机应用技术]