中铝青海分公司大型铝电解槽效应情况分析与决策系统研究

【作者】 蒋英刚；

【导师】 李劼；

【作者基本信息】 中南大学 ， 有色金属冶金， 2004， 硕士

【摘要】 本文从铝电解生产实际出发，结合智能模糊控制策略，对铝电解槽效应发生情况进行统计分析，设计了基于控制系统参数调整的新型效应分析报表，报表结构新颖，针对性强，具有很好的可操作性，为数据挖掘提供了结构合理的数据源。针对氧化铝浓度降低引起的效应，总结了控制系统的效应预报机制；针对槽况不稳引起的温度效应，对基于人工神经网络的热平衡诊断专家系统进行了优化。 改进的神经网络专家系统网络结构简洁实用，诊断速度和准确度均有所提高，为效应分析系统的构造提供了辅助决策支持。同时为设定电压、出铝量及氟盐添加量提供了简化的线性决策机制。 开发了基于灰关联度算法的数据挖掘型效应分析系统，运用数据挖掘技术对效应数据进行加工，得到一定时期内效应发生的主要影响因素，并首次将数据挖掘技术引入工业控制系统的决策及管理分析中，初步实验表明这种数据分析技术对提高控制系统的决策水平及强化科学化管理具有很强的实用性，值得做进一步的深入研究。 最后，基于数据挖掘和神经网络专家系统的输出结果，对控制系统的控制参数进行调整，达到效应分析及决策的智能化处理。通过对效应受控情况进行动态调整，为电解槽的平稳运行及生产技术指标的进一步提高提供技术保障。更多还原

【Abstract】 Based on the production practice of the aluminum electrolysis and combined the intelligent fuzzy control strategy, the occurrence of the effect in aluminum electrolysis cells were analyzed in statistical way and the newly effect analyses report form, which has novel structure, strong pertinence and perfect maneuverability, was designed founded on the parameter adjustment of control system .This provide the data source with reasonable structure for data mining. To resolve the effect caused by the drop of the alumina concentration, the effect prediction mechanism of control system was summarized; to resolve the temperature effect caused by instability of cell condition, the heat diagnose expert system that based on the manual neural network was optimized.The ameliorated neural network, which is compacted and practicably and has been improved in the speed and veracity of the diagnose, give support to the assistant decision of the effect analyses system conformation. It also provides simplified linear decision mechanism for the voltage setting, the amount of siphon aluminum and the quantum of the fluoride additive.The data mining effect analyses system was developed based on the gray scale association arithmetic. The effect data were processed by the data mining technology and the main affect factor were gained. For the first time, the data mining technology was introduced into the decision and management analyses of industrial control system, and the elementary experiments show that it can be of great practicability for the decision level improvement of control system and the strengthen of scientific management. So, this is deserved to the further research.At last, based on the output of the data mining and neural network expert system, adjust the control parameter of the control system, so as to achieve the effect analyses and the intelligent process of decision. By dynamic regulation of the control of the effect, offer the technological guarantee for the stable operation of the aluminum electrolysis cell and the further improvement of the production index.更多还原