基于火焰图像特征与BT-SVM的窑况识别方法  

Status Recognition Research for Rotary Kiln Based on Flame Image Features and BT-SVM

作　　者：孙鹏[1] 周晓杰[2] 柴天佑[1,2] 

机构地区：[1]东北大学流程工业综合自动化教育部重点实验室,沈阳110004 [2]东北大学自动化研究中心,沈阳110004

出　　处：《系统仿真学报》2009年第13期4019-4022,4027,共5页Journal of System Simulation

基　　金：国家自然科学基金重点项目(60534010);国家创新研究群体科学基金项目(60521003);长江学者和创新团队发展计划资助(IRT0421);国家863高技术计划重点项目(2007AA041404)

摘　　要：针对氧化铝回转窑烧成带工况变化复杂难以实现连续在线检测,长期依赖人工看火操作的难题,提出了利用计算机图像处理技术模拟传统的人工看火过程进行窑况识别研究的方法,方法包括两个部分:提取烧成带火焰图像特征,融合关键工艺过程数据组成混合特征;建立具有准正态二叉树结构的支持向量机窑况识别模型对混合特征数据进行分类识别。最后,应用该方法对采集得到的火焰图像数据与过程数据进行仿真实验研究,获得了满意的效果。Considering the complexity, importance of the condition variation of the alumina rotary kiln burning zone and the deficiency of process detection method, a new method was put forward, in order to simulate traditional man-watch operation with computer image processing techniques. At first, flame image features were extracted, hybrid features were composed with important process data, then status recognition model was constructed based on semi-symmetrical binary tree structure and SVM theory, and hybrid features were used as model input, status recognition results as model ou^out, status recognition research was carried out for burning zone of alumina rotary Mln. At last, this method was applied to the application research based on the flame image and process data from practical industrial process, then a satisfied result was got.

关 键 词：回转窑 窑况识别 火焰图像处理 支持向量机 准正态二叉树 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]