氧化铝生产四效逆流降膜式蒸发过程出口浓度预测模型研究

【作者】 毕庆华；

【导师】 唐朝晖；

【作者基本信息】 中南大学 ， 控制科学与工程， 2008， 硕士

【摘要】 生产实践表明,在氧化铝生产蒸发过程中,料液出口浓度是重要的蒸发质量指标,也是对蒸发操作参数(料液入口流量、生蒸汽流量和压力)进行调整的主要依据。由于料液出口浓度的人工检测滞后数小时以上,难以及时起到反馈修正操作参数的作用;并且由于蒸汽流量等操作参数是人工调节的,为控制出口浓度而进行的操作参数的调整,也是建立在操作人员经验的基础上,系统调整时间长。为了解决蒸发过程出口浓度预测的问题,需要建立出口浓度变化与各操作参数之间的关系,论文以某氧化铝厂四效逆流降膜式蒸发过程为研究对象,建立蒸发过程出口浓度预测模型。在分析蒸发过程的工艺机理分析的基础上,建立了基于物料平衡、热量平衡和相平衡的机理模型;同时,基于大量实际生产运行数据,建立了以直接影响出口浓度的当前状态参数为输入,出口浓度为输出的BP神经网络预测模型。两种模型仿真研究表明:BP神经网络模型总体的拟合性好,但在工况不稳定时由于数据的不完备使预测精度降低;机理模型误差相对较大,而在工况不稳定时,比BP神经网络的预测精度高。根据二者的特点,建立了由这两个模型结合的智能集成预测模型。实际生产应用验证了所建立的智能集成模型在工况稳定和不稳定时都能以较高的精度实现氧化铝生产蒸发过程出口浓度的预测,预测精度满足工业生产要求,对蒸发过程的优化操作具有参考价值。更多还原

【Abstract】 The practices of production indicated that, in the evaporation process of alumina production, the feed outlet concentration is an important indicator; it is also the primary base for regulating the operational parameters (flow of imported feed, the fresh steam flow and pressure). Because the manual measurement of the feed outlet concentration lag behind production process several hours later, it’s difficult for us to modify the model by feedback timely. Further, owing to the artificial regulation to steam flow and other operating parameters, the adjustment of operating parameters for the control of outlet concentration is also built on the experience of the operators and the adjustment time of system is too long.In order to solve the problem of the outlet concentration forecast in the evaporation process and establish the relationship between outlet concentration and the operating parameters, the forecast model of the outlet concentration of the evaporation process is established by studying the four-effect back-feed falling film evaporation system of a certain alumina plant. This paper establishes a mechanism model based on materiel balance, heat balance and phase balance by analyzing the mechanism. Further, based on industrial running data, a BP neural network model for the forecast of the outlet concentration is established, whose inputs are the direct influencing factors of the outlet concentration and the output is the outlet concentration, too.The characters of the two models are analyzed respectively, and some conclusions can be drawn as follows: the BP neural network model’s fitting performance is fine by and large, but it can not predict exactly when the work conditions are unstable because of the shortage of data; the mechanism model’s precision is lower than the former generally, but its effect is better than it when the work conditions change abruptly. So an intelligent integrated forecast model is established by combination the two models above. Actual production application indicate that the intelligent integrated model could do estimation well in any condition and could provide reference to the optimization of the industrial operation in the evaporation process of alumina production.更多还原