【作者】 吕学伟；

【导师】 白晨光；

【作者基本信息】 重庆大学 ， 冶金工程， 2010， 博士

【摘要】 目前,国内多数的钢铁企业大量依赖铁矿石的进口,个别企业的进口矿比例甚至达到了80%以上。从2004年开始,进口矿的价格开始飙升;2009年,达到了惊人的150美元/吨。5年的时间,价格增长了5倍多。2004年后,每次铁矿石价格谈判过程都十分艰难。由于我国迫于铁矿石的需求和维持钢铁厂正常生产经营的压力,谈判往往处于被动。即使价格攀升,也只能无奈接受。昂贵的进口矿使得国内的钢铁企业积极的利用国内便宜的低品位矿和含有害元素多的矿石,一些含铁的工业物料和废料也被广泛的利用。混合料的化学成分波动频繁,原料的不稳定性给制粒和烧结带来很大困难。在这种情况下,开发一个全面的、准确的铁矿石评价体系,对矿石在全炼铁流程中表现进行评价,并提出优化配料方案,无疑是每个企业所亟需的。本文围绕铁矿石评价体系这一研究目标,按照工艺流程,分别把铁矿石在混合料制粒、烧结、炼铁中的表现进行了理论计算和实验研究。针对混合料制粒过程中加水量的优化问题,提出了湿容量的概念,并开发出了测试设备和测量方法。对数种铁矿粉湿容量的测量和分析表明:随着矿物粒度的减小,湿容量逐渐增大。基于影响矿物湿容量的因素,建立了表征矿物湿容量的数学模型。湿容量的无孔模型合理地解释了矿物的湿容量随颗粒尺寸减小而增加的现象。湿容量的有孔模型考虑了矿石颗粒表面孔隙对于矿物吸水能力的影响,并理论上推导了开孔和闭孔的差别,理论计算表明闭孔对于湿容量的影响很小,可以忽略。通过实验数据的回归,得到了以矿石的比表面积、孔容、堆密度和真密度之差为自变量的表达式。基于矿物的吸水特性曲线,建立了宏观和微观动力学模型。宏观动力学模型研究表明铁矿石的吸水过程符合Lagergren一阶动力学方程,并得到了水在不同矿物中的传质系数。研究还发现大颗粒矿物的吸水速率比小颗粒矿物大。矿物的微观动力学模型基于水在颗粒间传输时的受力分析。颗粒间的空隙尺寸和颗粒表面的闭孔体积是影响矿物吸水动力学的主要因素,其中空隙尺寸是主要因素。混料实验表明:湿容量(x)和铁矿粉最佳配水量(y)具有很好的线性关系。对于本研究所涉及的制粒系统而言,这个关系为y = 6 .94+0.12x。说明湿容量越大,料层得到最佳透气性时所需要的最佳配水量也应该越大。通过人工神经网络研究了制粒效果评价指标及其影响因素,采用三层BP神经网络结构,确定了模型各层节点数、激励函数、训练函数、训练次数等网络参数,最终建立了基于矿物湿容量和实际加水量的多输入单输出的粒度和透气性预测模型,预测效果在样本趋势上取得较好吻合,精度基本在可接受范围内,可以指导制粒实验及实际生产。采用FACTSage软件对铁矿粉烧结过程中的物相变化、液相量、热效应与温度的关系进行了理论计算,并利用多种实验方法对计算结果进行了验证。结果表明:FACTSage计算得到的矿物的物相变化规律、液相量与温度的关系与实验基本吻合,可以通过理论计算对铁矿粉的烧结进行优化。由于对矿物的原始物相缺乏准确的表征,FACTSage计算得到的理论热效应和实际差热分析得到的数据数量级一致,但次序吻合较差。采用正交实验的方法对影响烧结矿各项指标的因素进行了考察,对实验数据的极差分析表明,配碳量对烧结矿的物理性能和技术指标影响最大;随着配碳量的增加,烧结速度、利用系数以及烧结强度均有不同程度改善;碱度升高对改善烧结强度也有促进作用。以正交试验的结果为基础,采用BP神经网络建立了烧结矿性能的预测模型。并对各烧结矿性能预测子模型的结构及参数进行了优化。经过检验,在误差范围内,还原度和利用系数的预测命中率可以达到75%以上,落下强度、转鼓强度、烧结速度的预测均命中率则达到87.5%以上,且预测趋势吻合,模型能够指导烧结实验及生产。针对烧结过程的配料优化而言,对于关系简单、规模较小的模型,线性规划方便易用,求解效率高。对于大规模复杂问题,当约束条件的重要程度不同时,遗传算法能灵活有效地解决问题。随着变量和约束条件继续增多,模型规模和复杂度的增大,遗传算法能够满足高性能求解优化模型的要求,并且其独有的惩罚函数可以灵活地处理各种约束条件,通过控制惩罚度的大小对约束条件划分优先顺序,使配料过程的优化模型求解更符合烧结操作者的意愿,实现配料的人工智能。为实现对烧结矿矿相的准确表征,采用图像处理技术对烧结矿的灰度值计算、灰度直方图的分布特征、矿相的纹理特征等内容进行了研究。其中,矿物的反射率计算模型合理、准确。矿相的灰度正态分布模型与实际矿物的灰度分布特征吻合,利用该模型统计得到了常见矿物的正态分布参数;并结合遗传算法,实现了矿物含量的智能计算。基于灰度共生矩阵的图像特征提取方法,研究了灰度共生矩阵的参数如灰度级数、图像窗口尺寸、共生距离和共生角度等与矿物纹理结构的关联性。并利用该方法和兰氏空间距离实现了对矿相的识别。在上述模型的基础上开发出了智能矿相识别处理软件。综上所述,本文对铁矿石在每个工艺环节中的行为都进行了定量的评价,并且采用计算机编程语言和数据库技术开发出了炼铁体系铁矿石评价软件。该软件已经在工业过程中得到应用,效果良好。更多还原

【Abstract】 There are many Iron and Steel companies in China which depend on the iron ores imported greatly. In some companies, the iron ore imported from abroad takes up even more than 80%. The price of the iron ores imported increased quickly because of the huge amount of demand in China, especially after 2004. The price increased from less than 30 to about 150 dollar per ton, meaning about 5 times as that 5 years ago. The Chinese association for steel negotiated about the price of iron each year with the iron ore providers from abroad since 1981. The negotiation is very hard after 2004. In every negotiation, the Chinese iron and steel companies are in the dock, because they need a huge of iron ores and cannot shut down due to the society pressure. Always, they cannot help but accept the high price. The high price of the iron ore made the iron ore incorporations use some low grade ores from domestic and some waste materials containing Fe to cut down the cost of raw materials. The chemical composition always vary a lot in the industry production. Therefore, the control of the process during granulation, sintering, and the blast furnace become difficult. Under this conditions, the system which can evaluate the iron ores in the whole process is very necessary for the iron and steel company in China.In present study, the iron making process is divided into granulation, sintering, and blast furnace process, and the behavior of the iron ore in each of the procedure were studied through the theoretical and experimental methods.The conception of moisture capacity, equipments, and measurement method was suggested for the optimization in the granulation process. The measurements indicated that the moisture capacity increases with the decrease of particle size. The mathematical models were developed for the prediction of the moisture capacity. The no-pore model give a good explanation of the phenomenon that the moisture capacity increase with decreasing the particle size. The pore model consider the effect of the pore size on the ability of water absorption, and calculate the difference between the open pore and closed pore. The calculations indicated that the closed pore has little effect on the moisture capacity. The equation was got by fitting the measurement that the moisture capacity is expressed by surface area per unit mass, pore volume, bulk density and the real density. The macro and micro kinetic models were developed based on the water absorption curves. The macro kinetic model indicated that the water absorption into the iron ore particles agree with the first order Lagergren kinetic equation, and got the mass transfer coefficients of the water in the iron ores. The micro kinetic model was based on the force analysis of the water in the particles. The calculations indicated that the size of space between the particles and the closed pore volume in the particle surface are the main factors for the kinetic. The granulation experiments with the rotating cylinder indicated that the optimal water content for the granulation increases with increasing the moisture capacity, and they obey a good linear relationship. For the system used in this study, the equation is y = 6 .94+0.12x.The artificial neural network was used to build the prediction model of the granulation results. The model used three layer BP structure. The number of nodes in the three layer, activation function, training function, training times were optimized. Finally built the multi-input and one output model with moisture capacity and water content added. The tendency predicted agrees well with the measurements, and the prediction accuracy is accepted. The models can be used to improve the industrial production.On the physical and chemical behaviors of the ore in the sintering, the phase transformation, mass of liquid, thermal effect with the temperature were calculated with FACTSage, and the results were validated by various experimental data. It was indicated that the calculation by FACTSage agreed well with the measurements. This calculation can be used to optimize the sintering parameters under various raw materials conditions. The thermal effect calculated have the same scale, but had a great deviation with the measurements. The probable reason is lack of right characterization of the chemical composition.As for the influence of burden on the sintering, the orthogonal method was used to check the effect of various factors and levels. The results showed that the coal dosage the first factor influencing the sinter properties and technical index of the sintering. The sintering velocity, utilization coefficient, strength of the sinter were all improved with increasing the coal dosage. The increase of basicity will lead the improvement. The prediction model of the sintering was built with the artificial neural network and the measurements. The parameters in the model were all optimized. The validations showed that the hit rate can reach >75% for the reduction ratio and utilization coefficient in the error limit, and that of tumbler index, sintering velocity reach 87.5%. The tendency predicted agree well with the measurements, and the models can be used in industry.The optimization model for the burden based on the physical and chemistry behavior in the sintering process was developed. However, the model is complex and huge, therefore, the linear program can not solve the model. The comparison among the many methods showed that the genetic algorithm can satisfied the demand of solution. On the importance of the restrictions, the punishments function in the algorithm can deal with this importance by adjust the punishments degree. This method can get the solution to agree with the operator’s goal, realizing the intelligent burden.The mineralogy recognition and analysis were also studied. The reflective power model, mineralogy qualification model, and the texture features extraction mode were all developed based on the digital image processing techniques. The reflective power model is simple but reasonable, and the results are accuracy. The Gauss model of the gray distribution for the mineralogy agree with the feature of sinter. The distribution parameters were calculated by the model. The intelligent qualification of mineralogy was realized by combing the Gauss model and genetic algorithm. The texture feature extractive method based on the gray level co-existence matrix can get the features of various mineralogy exactly. The machine recognition of the mineralogy can be realized with canberra space distance. Finally the intelligent software was developed with the models above.In summary, the behaviors of the iron ore in the iron making were qualified in the sub-section, and finally developed the evaluation system. The software was developed with C# program language and SQL database. The software has been used in the plant, and got a good application.更多还原