【作者】 胡友明；

【导师】 姜涛； 范晓慧；

【作者基本信息】 中南大学 ， 钢铁冶金， 2011， 博士

【摘要】 随着现代钢铁工业的不断发展,铁矿石资源日趋紧张,使得国内烧结厂原料结构波动大,生产极不稳定。如何在有限的资源条件下,快速准确地获得产质量指标满足要求、经济性好的配矿方案及相应的工艺参数,在原料条件发生变化时能快速稳定烧结生产过程已成为烧结乃至炼铁工作者的研究热点和重点。本文以原料主要靠外购的涟钢烧结厂作为主要研究和成果应用对象,在系统研究了原料物化性能、制粒性能、成矿性能和烧结性能及其关系的基础上,将理论计算模型、BP神经网络模型、遗传优化模型和专家系统相结合,构建了烧结优化配矿专家系统,并在涟钢进行了工业应用,取得了良好的效果。采用Access2003数据库技术构建了系统数据库,包括原料基础性能(物化性能、制粒性能和成矿性能)数据库、混合料烧结性能(单矿及配矿烧结)数据库、烧结生产信息库(配矿方案、工艺参数、产质量指标)、配矿专家知识库、模型库(模型参数及BP网络模型)等；并采用VC++软件开发工具,搭建了数据管理、模型建立和知识管理平台,实现了文字、数据和图片等多种信息的综合管理。数据管理系统具有添加、删除、修改、查找、加载、保存文件等功能。根据烧结杯试验研究结果,结合生产经验,以混合料的化学成分、矿物组成、碱度、不同粒径颗粒的含量等作为输入参数,以烧结速度(利用系数)、转鼓强度和固体燃耗作为模型的输出参数；采用三层BP神经网络模型对烧结矿的产质量指标进行了预测,模型的预报命中率达85%以上；系统还提供了建模平台,用户可根据预报效果要求修正模型的输入、输出参数,模型学习参数等,模型的适应性强。根据原料物化性能、制粒性能、成矿性能和烧结性能的研究结果,采用线性回归分析的方法,建立了烧结适宜混合料水分和燃料配比计算模型,模型准确率达到了93%以上。系统获得烧结优化配矿方案的过程如下：在已知原料条件(原料物化性能、供应条件)和烧结矿产质量指标要求的条件下,先采用线性规划法获得化学成分满足要求的配料方案组；采用产质量指标预测模型和工艺参数优化模型,预测各方案的产质量指标和适宜的工艺参数,根据配矿经济模型评价各方案的经济性；在成分满足要求的配矿方案组内,以产质量指标作为约束函数,以经济性作为评价函数,采用遗传算法获得产质量指标满足要求、经济性好的配矿方案,并给出配矿方案相应的工艺参数、产质量指标；若用户对系统提供的方案不满意,可以采用配矿调整专家系统,根据专家经验对方案进行调整,直到用户满意为止。更多还原

【Abstract】 With the development of modern steel industry, iron ore resource shortage becomes more and more severe, which causes major fluctuations of domestic raw material structure, and the production is very unstable. Obtaining qualified and economical ore proportioning scheme and corresponding technological parameters under the condition of limited resources to stabilize sinter production under fluctuant raw materials condition becomes a hotspot and key point of sintering and even iron making groups.In this paper, sinter plant of Liangang which is outsourcing materials relied is the main object of research and application. Expert system of ore proportioning optimization of sintering is constructed combining computation model, BP neural network model, genetic optimization model and expert system, based on study of physical and chemical properties of raw materials, granulation properties, metallogenic properties, sintering properties and their relationships. This system was put into application and good results were achieved.Database was built up by using Access2000database technology, including basic properties database of raw materials (physical and chemical properties, granulation properties and metallogenic properties), sintering properties database (data of single ore and ore proportioning sinter pot test), sintering information database (ore proportioning schemes, technological parameters and indicators of yield and quality), knowledge base of ore proportioning, model library (model parameters and BP network model), etc., Data management, modeling and knowledge management platform was built using VC++. Integrated management of text, data and images information, as well as files adding, deleting, modifying, locating, loading and saving were achieved.Based on results of sinter pot tests, combined with production experience, sinter quality and productivity indicators of sinter were predicted using improved three-layer BP neural network model, with chemical composition, mineral composition, basicity, particle content as input parameters and sintering speed (productivity), tumbler strength and solid fuel consumption as outputs. The model accuracy was over85%. Modeling platform was provided, input parameters, output parameters, learning parameters could be modified based on forecast results, and good adaptability was achieved.Calculation model of suitable moisture and fuel proportions was built based on results of physical and chemical properties of raw materials, granulation properties, metallogenic properties and sintering properties, using linear regression analysis method. The model accuracy was over93%.Procedures of obtaining optimal ore proportioning scheme are as follow:under the given conditions of raw materials (physical and chemical properties, supply conditions) and required indicators of sinter yield and quality, ore proportioning scheme groups which meet the requirements of sinter chemical compositions are obtained by linear programming method; predict indicators of sinter yield and quality and proper technological parameters of each scheme using prediction model of sinter yield and quality and optimization model of technological parameters, evaluate the economical efficiency with economic model of ore proportion; within the ore proportioning scheme group which can meet the requirements of chemical compositions, sinter yield and quality indicators are selected as constraint function and economical efficiency as evaluation function, ore proportioning scheme is obtained with satisfying sinter yield and quality, as well as good economical efficiency using genetic algorithm, and corresponding technological parameters and indicators of sinter yield and quality are provided; the scheme can be modified based on expertise using expert system of ore proportioning modification, until the users are satisfied.更多还原