System Thermodynamics Analysis And Model Development For Temperature Prediction Of Molten Salt In Alumina Tube Digestion

Because tube digesting technology adopts the indirect heating mode, the energy dissipation could be reduced in aluminum production process. However, the incrustation scales formation on tube surfaces is obvious in china during the application of digesting system to bauxite. That will result in variation to the state of heat balance. Moreover, the temperature of molten salt should be adjusted continuously. Therefore, awareness to the exergy and heat balance system and establishing prediction model of molten salt temperature are significant for reducing energy consumption to tube digesting system.This paper studied the molten salt heating section according to the tube digesting system of an aluminum plant in China. Based on the heat balance testing, two kinds of thermodynamic models had been used to analysis the energy balance on that system. According to curve fitting of incrustation scales formation, a prediction model of molten salt temperature was built by the mechanisms of heat transfer. In order to improve the prediction precision, principal component analysis (PCA) and multiple neural network (MNN) were set up by neural network toolbox of matlab7.0.1.The main conclusions in this paper are as follows:1. The calculate results indicate that the thermal efficiency (99.4%) and the exergy efficiency (85%) are both high in the molten salt heating section of tube digesting system. Along with circulation of the system, system thermal efficiency is nearly stabilization, whereas exergy efficiency decreases a lot .The exergy loss of heat transfer is primary element of the exergy loss system. So reduction the range of temperature-difference is a main way to save exergy loss.2. Based on heat transfer theory, the thickness of tube incrustation scales formation was calculated according to the test data. The results indicate that the incrustation scales formation in high temperature section is thicker than scale in lower temperature section. In conformity to this result, a heat transfer model to predict the temperature of molten salt was established. Simulation results show that the prediction precision is low.3. Data smoothing, principal component analysis (PCA) and normalization were applied to data preprocessing technology. In order to reduce input variables of network, 13 input variables had be regrouped to 8 principle components through principal component analysis method. That will help establish the neural network (NN).4. Because the heating-up law of molten salt was not the same in each periodic time, a molten salt prediction model of tube digesting system was established with the multiple neural network (MNN) method. Besides, that model includes three subnets and consulted to the initial temperature of molten salt.5. Due to different producing data under different work condition, the molten salt prediction model was tested based on NN. The result shows that the model predicted value whose absolute error was not exceeding±5°C was not lower than 85%. Moreover, it performed better in prediction than the heat transfer model. At the same time, the simulation result indicated that 28.7KJ calorific power could be saved if optimizing control was applied on the temperature of molten salt in that model. So the PCA-MNN model made great contribution to energy saving in industry process.