【作者】 李军；

【导师】 庄剑鸣；

【作者基本信息】 中南大学 ， 矿物加工工程， 2004， 硕士

【摘要】 配矿是目前烧结研究的重点，配矿专家系统是烧结配矿研究的一个新领域，本论文结合湘钢烧结配矿试验，对配矿烧结专家系统进行了研究。 配矿专家系统由三部分组成：配矿方案的产生模块、配矿调整指导系统和产、质量指标预测模型。配矿方案的产生模块主要是产生满足化学成分要求的配矿方案；产、质量指标预测模型预测中和矿的烧结性能；配矿调整指导系统通过预测指标和配矿方案的状态判断对配矿方案进行调整。 1．配矿方案的产生模块 配矿方案采用用户输入的方式产生。当参加配矿的各种矿石用量之和是为100％时，系统对配矿方案进行了化学成分检验。主要检验中和矿的SiO₂、TFe、CaO、MgO和Al₂O₃的含量是否在各自适宜的范围。 2．产、质量指标预测模型 烧结产、质量指标预测模型采用BP神经网建立。模型的输入参数为参加配矿的精矿的SiO₂、FeO、CaO、MgO、Al₂O₃和+0.25mm粒级的含量，粉矿的SiO₂、FeO、CaO、MgO、Al₂O₃和-0.25mm粒级及1～3mm的含量；模型的输出为烧结速度和烧结矿的转鼓强度。 BP网采用单隐层，隐单元的个数为27个；采用固定应学习率，初始值为0.9，动量项初始值为0.7，初始权值为0.5，初始阈值为0.2；激励函数分别为：f（x）=1／（1+exp（-x））和f（x）=-（1／2）+1／（1+exp（-x）） 3．配矿调整指导系统 配矿调整指导系统的知识来自烧结试验的一般性结论，调整的知识为启发性知识，知识采用产生式规则的表示方式。调整的策略是：以SiO₂的含量为调整的目的，根据矿石的用量确定调整对象，再根据预测结果和调整对象的状态，采用相应的调整措施。更多还原

【Abstract】 Nowadays iron ore matching has been a focus in sintering study, and matching expert system (MES) has been a new field in study on iron ore matching. In this paper, study of MES bases on the sintering test of sinterplant, Xiangtan I&S co.MES is consists of three parts: producing part of Matching scheme, forecast model of sinter quality and output, and the redressal system. The aim of producing part is to produce a matching scheme in which the chemical composition is in a proper range. The forecast model forecasts the sinter property of matching scheme in a certain condition and the redressal system adjusts the matching scheme through the state judgement of the forecast and matching scheme.1. Producing part of matching schemeMatching scheme is produced by user’s input. The system examine chemical composition of the scheme when the sum of all of the ores which are participate in the matching is 100%, chemical composition examine is mainly examine the proper range of SiO2,TFe,CaO,MgO and A12O3 in the mixed ores.2. Forecast model of sinter quality and outputThe forecast model is based on the BP Neural Networks. The input parameter of model is consist of SiO2, TFe, CaO, MgO, A12O3 and +0.25mm size ores of participated concentrates and SiO2, TFe, CaO, MgO, A12O3, -0.25mm size ores and 1 ~ 3mm size ores of powder ores, the output parameter is the sintering speed and strength of sinter.The number of latent layer in BP Neural Networks is only one, and which member is 27, the velocity of study is aptotic and which initialization is 0.9, initialization of right and threshold is 0.5 and 0.2, power function is3. The redressal systemThe knowledge of the redressal system comes from the general conclusion of sintering test, and the adjusting knowledge is the heuristic knowledge that expressed by produced rule. The aim of adjusting is the content of SiO2 in the mixed ores, based on the ore’s content to determine the adjusting target;adjusting step is confirmed based on the forecast and the state of the adjustingtarget.更多还原