【作者】 安剑奇；

【导师】 吴敏；

【作者基本信息】 中南大学 ， 控制科学与工程， 2011， 博士

【摘要】 钢铁产业是我国国民经济的重要支柱产业,涉及面广,产业关联度高,在经济建设和社会发展等方面发挥着重要作用。高炉是钢铁生产流程中的最关键设备,也是能源、资源消耗和环境污染的大户。高炉的顺行、稳定、高效生产,关系到整个钢铁公司乃至整个社会的经济效益。高炉炉喉煤气流分布是影响高炉生产状况的主要的因素之一,对于保持高炉的稳定运行和指导高炉优化操作起着十分重要的作用。高炉料面温度场能够准确地反映炉喉煤气流分布状况,本文针对单一检测信息不能准确反映高炉料面温度场的问题,运用信息融合技术、图像处理技术、人工智能技术、软测量技术等进行“基于多源信息融合的高炉料面温度场在线检测系统”的研究,在某钢铁公司2200m3高炉上实现工业应用。应用表明,系统能准确反映高炉料面温度场分布,为高炉布料操作和炉况诊断等提供了实时准确的指导,保证高炉稳顺优化运行,取得了显著的效益。本文的主要研究工作和创新点包括以下4个方面：(1)基于异类信息提取和多源信息融合的料面温度场检测方案本文在充分研究高炉生产工艺,煤气流形成与分布原理基础上,通过建立高炉红外图像、十字测温、矿焦比以及料线等多源检测信息与料面温度场之间的灰色关联度模型,确定影响高炉料面温度场的过程检测参数和特征量。采用高炉生产过程多源信息采集、多尺度异类信息特征提取和多源信息融合三个层次,确定高炉料面温度场在线检测方案。该方案能够充分利用高炉现有检测设备和仪器,有效计算和提取异类多尺度信息特征,合理融合各可信特征信息,为高炉生产过程状态检测提供了新的有效手段。(2)高炉料面温度不同尺度异类检测信息的特征提取技术首先针对多源检测信息在空间和时间尺度上的差异,采用基于空间几何关系的空间配准法和基于最小二乘的时间配准方法,将多尺度信息统一在同一时间和空间维度上。其次,提取了不同尺度异类检测信息的特征,包括在对红外摄像机光学定标的基础上,提取了像素点与料面的几何对应关系,采用遗传最大模糊熵算法提取了图像二值特征,采用专家规则提取了图像炉心位置和图像等灰度线；采用空间插值方法,提取了十字测温平面温度分布；通过高炉布料模型的计算,提取了与料面温度分布有关联的高炉轴向矿焦比。(3)基于可信度理论和卡尔曼滤波方法的高炉料面温度场多源信息融合技术在提取多源异类信息特征的基础上,首先分别通过各种单一异类信息估计料面温度,包括采用热力学原理,通过十字测温平面温度估计料面温度；采用多点动态定标方法,通过红外图像估计料面温度；采用模糊推理方法,通过矿焦比估计料面温度；采用热传导原理,通过炉墙热电偶估计料面边缘温度。其次分别采用可信度理论和卡尔曼滤波方法融合以上估计值,计算出料面中部区域和边缘区域温度,建立完整准确的高炉料面温度场。(4)建立高炉料面温度场在线检测系统,在典型高炉上应用利用红外摄像仪、图像采集卡、工业以太网通信设备和工业控制计算机等硬件设备,采用VC++6.0进行应用软件开发,实现了多源信息的采集、特征提取和信息融合,开发了具有友好人机接口和互联性的高炉料面温度场在线检测系统。针对某钢铁公司2200m3高炉,应用基于多源信息融合的高炉料面温度场在线检测系统,实现了对高炉料面温度场的低成本、高准确度在线检测。更多还原

【Abstract】 As the foundation of our national economy, steel industry plays an important role in the economic and social development respecting its extensive connection with other industries. Blast furnace (BF), the most critical equipment in steel industry and the major energy-consumer as well as the great producer of environmental pollution, has a great influence on the economic benefits of the steel company and even the whole country.The gas flow distribution in the throat is one of the major influences on BF production situation, which keeps it running stablely and provides a direction for its optimization operation. While the gas flow distribution in BF throat could be truly reflected by BF burden surface temperature field. In view of the deficiency of singular detecting information in reflecting BF burden surface temperature field, this dissertation develops an "on-line detecting system for BF burden surface temperature field based on multi-source information fusion" by employing the technique of information fusion, image processing, artificial intelligence and soft sensor. The industrial application to a 2200mj BF of some steel company shows the efficiency of the system in truly reflecting the distribution status of BF burden surface temperature field and in providing a real-time and accurate guidance for the operation of BF burden distribution and the diagnose of BF status, which ensure BF a smooth running and the company a great benefit as well.The main focuses of study and innovations are included in the following four aspects:1. The detecting plan for burden surface temperature field based on heterogeneous information extraction and multi-source information fusion.Based on the comprehensive investigation on BF production process and the principle of gas flow formation and distribution, this dissertation builds a grey correlation model for relating burden surface temperature field with multi-source detecting information, such as infrared image, cross temperature, ratio of ore to coke (OCR), stock line etc. so as to obtained the parameters of process detection and characteristic quantity which are closely related to the burden surface temperature field. And then, the on-line detecting plan for BF burden surface temperature field is settled with the courses of multi-source information collection in the process of BF production, feature extraction of multi-scale heterogeneous information and multi-source information fusion. This plan can make full use of all the detecting devices and facilities on hand, calculate and extract multi-scale heterogeneous information efficiently and generalize all the reliable feature information reasonably, which offers a new idea for the state detection in BF production.2. The technology of feature extraction for BF burden surface temperature field based on multi-scale heterogeneous information.First, to solve the disagreement of multi-source detecting information in spatial dimension and in temporal dimension, spatial registration method based on spatial geometric relation and temporal registration method based on least square are employed for the purpose of conforming the multi-scale information spatially and temporally. Then, the feature of multi-scale heterogeneous information is extracted, which includes a) the geometrical relation between pixel points and burden surface coming from the optical calibration of infrared camera, the binary feature of images based on genetic algorithm and maximum fuzzy-entropy, furnace heart calibration and gray contour in the image based on the expert rules; b) the temperature distribution of gas flow around cross-thermocouple concerning the method of spatial interpolation; c) the OCR related to the distribution of burden surface temperature field based on the model of BF burden distribution.3. The technology of multi-source information fusion for BF burden surface temperature field based on reliability theory and Kalman filter.On the basis of the feature extraction of multi-source heterogeneous information, firstly, the burden surface temperature is estimated respectively according to all kinds of singular heterogeneous information, including:cross-thermocouple based on thermodynamic principle, the infrared images based on multipoint dynamic calibration, OCR based on fuzzy reasoning and thermocouple in furnace wall based on heat conduction. Secondly, the center part of the temperature field and its edge part are respectively calculated with reliability theory and Kalman filter.4. Online detecting system of BF burden surface temperature field and its application to a typical BF.The software of VC++6.0 and the hardware including infrared photography instrument, image acquisition card, industrial Ethernet communication equipment and industrial control computer are employed to realize the collection of multi-source information, feature extraction, information fusion and to build the online detecting system of BF burden surface temperature field featuring the friendly man-machine interface and interconnection. The application of this system to a 2200m3 BF in some steel company achieves the goal of low-cost and high-precision in the detection of BF burden surface temperature field.更多还原