( 1. 中南大学 冶金科学与工程学院，长沙 410083；
2. 中南大学 信息科学与工程学院， 长沙 410083)
摘 要: 针对烧结法氧化铝生产过程中生料浆配料工艺的特点， 根据物料平衡的原理建立机理模型， 作为生料浆质量预测的主规律模型； 针对碱液成分波动大且难以实时检测的问题， 对碱液成分含量建立了神经网络预测模型， 并和机理模型进行嵌套集成； 利用灰色理论对机理模型的偏差数据进行信息挖掘， 建立了GM(1, 1)补偿模型， 并与机理模型进行并联集成， 获得生料浆质量预测模型。 验证结果表明， 该质量预测模型能获得较理想的生料浆质量预测精度， 其应用可使生料浆质量得到显著的提高。
关键字: 生料浆； 神经网络； 机理模型； 预测模型； 灰色理论
（ 1. School of Metallurgical Science and Engineering, Central South University, Changsha 410083, China;
2. School of Information Science and Engineering, Central South University, Changsha 410083, China）
Abstract: Based on the analysis of the characteristics of the raw mix slurry preparing process in alumina sintering production process, firstly, a mechanism model based on material balance principle was established as the master-rule model for the quality prediction; secondly, considering the problem that the alkali liquor composition was unstable and its real-time measurement was difficult, a NN (neural networks) prediction model for the prediction of the alkali liquor composition was set up and nesting-integrated with the mechanism model; finally, using the gray theory for the information mining from the errors of the mechanism model, a GM(1, 1) compensation model was put forward and parallel-connection-integrated with the mechanism model, achieving a raw mix slurry quality prediction model. Verification results show that the quality prediction model is with satisfactory prediction accuracy, and its industrial application will benefit to improving the quality of raw mix slurry.
Key words: raw mix slurry; neural network; mechanism model; prediction model; gray theory
