Study on Preparation of Crude ZrCl₄ by Industrial Desiliconization Zirconia with Carbochlorination Based on Thermodynamics, Dynamics and Neural Network

To optimize the operating conditions of the industrial fluidized bed reactor with a daily output of 3 tons of crude ZrCl₄, crude ZrCl₄ was prepared by carbochlorination using desiliconization zirconia, C powder, and Cl₂ as raw materials. The thermodynamics and kinetics of carbochlorination reaction were investigated. Furthermore, quality data for the crude ZrCl₄ were modeled using a neural network gate recurrent unit (GRU) model. The experimental findings demonstrated that at elevated temperatures, the dominant formation equation is CO, with the reaction rate primarily governed by the surface reaction. Optimal results in chlorination rate, product quality, and slag discharge were achieved with a ZrO₂: C of 100: 18–19 wt.%, a push speed of 70 kg/h, a material layer height of 1 m, and a chlorine flow rate of 75 kg/h. The loss value of GRU model training prediction is 0.021, and the curve convergence effect is good.