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

• Published in: JOM (1989) Vol. 76; no. 9; pp. 5589 - 5600
• Main Authors: Yang, Zhanxin, Mu, Minxuan, Wang, Junbo, Chen, Jian, Li, Qingchun, Yin, Guili, Qi, Guochao, Wu, Qiong
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.09.2024; Springer Nature B.V
• Subjects: Carbon; Chemical reactions; Chemistry/Food Science; Chlorine; Design; Desiliconizing; Earth Sciences; Engineering; Environment; Fluidized bed reactors; Fluidized beds; Gases; High temperature; Neural networks; Optimization; Particle size; Physics; Raw materials; Surface reactions; Technical Article; Temperature; Thermodynamics; Zirconium; Zirconium dioxide
• ISSN: 1047-4838; 1543-1851
• DOI: 10.1007/s11837-024-06756-3

To optimize the operating conditions of the industrial fluidized bed reactor with a daily output of 3 tons of crude ZrCl 4 , crude ZrCl 4 was prepared by carbochlorination using desiliconization zirconia, C powder, and Cl 2 as raw materials. The thermodynamics and kinetics of carbochlorination reaction were investigated. Furthermore, quality data for the crude ZrCl 4 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 ZrO 2 : 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.