氧化铝生产过程苛性比值与溶出率智能集成预测模型研究

【作者】 应苏如；

【导师】 桂卫华；

【作者基本信息】 中南大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 苛性比值与溶出率是氧化铝高压溶出过程中两个重要的经济技术指标。它们不仅决定了氧化铝溶出的效果及碱耗，而且对氧化铝的后续生产具有极大的影响。然而，目前苛性比值与溶出率的检测严重滞后、波动范围大且难以及时调整，从而导致整个生产流程的实时控制陷入被动。为此，研究如何运用智能集成建模方法建立苛性比值与溶出率的预测模型从而实现苛性比值与溶出率的在线检测，对实现氧化铝生产过程的稳产高产、提高企业竞争力都具有重要意义。 本文以中国铝业公司河南分公司氧化铝生产过程为背景，着重研究了苛性比值与溶出率的智能集成预测模型的建立和应用。首先，在分析氧化铝高压溶出过程机理的基础上，确定了影响苛性比值与溶出率的主要因素，提出了苛性比值与溶出率的机理模型；然后，提出了基于主元分析的多神经网络模型，用来补偿机理模型的偏差，从而建立机理模型与神经网络的集成模型；接着又提出了具有自校正样本库的基于聚类分析的匹配模型，用来预测苛性比值与溶出率；最后在深入分析机理模型与神经网络集成模型、匹配模型两者的优点与不足的基础上，提出了采用基于专家知识和统计学知识的智能协调器对两者的输出进行协调，从而建立苛性比值与溶出率的智能集成预测模型。 智能集成预测模型的现场运行结果表明该模型具有较高精度，能很好地实现苛性比值与溶出率的在线预测。更多还原

【Abstract】 Ratio of Soda to Aluminate (RSA) and Leaching Rate(LR) are two very important economical and technical indices in the process of High-Pressure Digestion (HPD) of alumina. Not only they affect output of alumina and alkali consumption, but also they make great influence on successive production of alumina. However, at present it is difficult to measure RSA and LR timely, which makes it hard to achieve real-time control of the whole process of production. Therefore, How to implement on-line measurement of RSA and LR by establishing prediction model using intelligent integrated modeling, is significant for realizing steady and high production of alumina and enhancing the competitive power of enterprises.On the background of production process of alumina in Henan Brach China Aluminium, this paper mainly does research in the establishment and application of intelligent integrated prediction model of RSA and LR. Firstly, the key factors that influence RSA and LR are acquired by analyzing the mechanism of process of HPD, then mechanism model of RSA and LR is established. Secondly, in order to compensate errors of mechanism model, multiple neural networks based on principle component analysis is put forward, thereby integrated model of mechanism model and neural network is proposed. Thirdly, matching model with self-tuning sample library based on cluster analysis is put forward. Finally, on the base of analyzing merits and flaws of integrated model of mechanism model and neural network and matching model, intelligent coordinator based on expert knowledge and statistics is designed to harmonize outputs of these two models, therefore intelligent integrated prediction model of RSA and LR is established.An real-world application in Henan Brach China Aluminium shows that intelligent integrated model is effective and it can predict RSA and LR on line very precisely.更多还原