李小龙¹， 孙继鹏²， 吴峰¹

1. 西北大学 化工学院，陕西 西安 710069；2.洛阳涧光特种装备股份有限公司，河南 洛阳 471003

引用格式：

李小龙， 孙继鹏， 吴峰. 基于CFD模拟的硫磺湿法成型装备智能优化［J］. 中国粉体技术， 2026， 32（6）： 1-13.

Citation:Li Xiaolong， Sun Jipeng， Wu Feng. Intelligent optimization of sulfur wet granulation equipment based on CFD simulation［J］. China Powder Science and Technology， 2026， 32（6）： 1-13.

DOI：10.13732/j.issn.1008-5548.2026.06.001

收稿日期： 2026-03-16， 修回日期： 2026-06-22，上线日期： 2026-08-03。

基金项目： 国家自然科学基金项目，编号：22478317。

第一作者： 李小龙（2001—），男，硕士生，研究方向为计算流体力学。 E-mail： 18582598906@163.com。

通信作者： 吴峰（1978—），男，教授，博士，博士生导师，研究方向为反应器过程强化。 E-mail： wufeng@nwu.edu.cn。

摘要： 【目的】 为了提升成品中目标粒径颗粒的产率，实现硫磺湿法成型过程的高效控制，对关键工况参数开展系统优化研究。【方法】 基于计算流体力学（computational fluid dynamics，CFD）数值模拟方法对硫磺湿法成型过程进行仿真分析，引入反向传播神经网络模型，对成型工况参数进行优化求解。【结果】 获得最优操作条件为：液硫压力为700 kPa，液硫热力学温度为403.15 K，冷却水热力学温度为333.15 K。【结论】 明确液硫压力是影响硫磺颗粒形态特征的主导参数；通过对照模拟实验验证，该参数组合表现出良好的可靠性与稳定性。

关键词： 计算流体力学； 硫磺湿法成型； 人工神经网络； 装备优化

Abstract

Objective Duringthe wet granulation process of sulfur， there is a significant nonlinear correlation between the degree of particle formation， the quality of the particle size distribution， and the key operating condition parameters. Factors such as melt temperature， cooling medium properties， and flow conditions interact in a complex way， directly influencing granule morphology and size uniformity. To increase the proportion of particles within the desired size range in the final product， it is essential to systematically optimize these operating parameters. Such optimization not only improves product quality and consistency but also enhances process efficiency and stability， thereby providing a more reliable basis for industrial-scale production.

Methods The research employed computational fluid dynamics （CFD） to simulate and analyze the wet granulation process of sulfur， enabling a systematic and in-depth investigation of key physical phenomena such as droplet formation， breakup， cooling， and solidification under a wide range of operating conditions. By establishing a comprehensive multiphase flow and heat transfer model， the study accurately captured the complex interactions between molten sulfur and the cooling medium， including interfacial dynamics， momentum exchange， and phase change behavior. These simulations provided valuable insights into the underlying mechanisms governing particle size distribution， morphology evolution， and granulation efficiency， thereby laying a solid foundation for subsequent data-driven modeling. Building upon the high-quality simulation dataset， a backpropagation （BP） neural network model was introduced to quantitatively characterize the nonlinear relationships between critical process parameters—including liquid sulfur pressure， temperature， and cooling conditions—and key granulation performance indicators. The neural network was systematically trained， tested， and validated using representative data samples to ensure high prediction accuracy， robustness， and generalization capability across diverse operating scenarios. Once trained， the model rapidly predicted granulation outcomes over a broad parameter space and efficiently identified optimal parameter combinations， significantly reducing the need for time-consuming and computationally expensive numerical simulations. Furthermore， the integration of CFD with machine learning established a hybrid modeling framework that not only enhanced computational efficiency but also improved the reliability and applicability of process optimization. This approach effectively overcame the limitations of traditional trial-and-error methods and provided a powerful， scalable tool for guiding process design， optimization， and control. Ultimately， it offered strong support for industrial application and had broad potential for extension to other complex multiphase systems， contributing to the intelligent development and optimization of advanced manufacturing processes.

Results and Discussion The pressure of liquid sulfur played a dominant role in regulating its kinetic energy during the wet granulation process， directly affecting droplet breakup， impact intensity， and the resulting particle size distribution and morphology. Increasing pressure enhanced fragmentation and dispersion， promoting finer and more uniform particles.However， excessive pressure could lead to over-fragmentation and irregular shapes， while insufficient pressure resulted in poor breakup and larger granules. In contrast， the temperature of liquid sulfur and the cooling water temperature influenced particle formation through secondary mechanisms. The former altered viscosity and surface tension， thereby affecting droplet formation and deformation， while the latter governed heat transfer and solidification behavior， influencing crystallization， internal structure， and surface characteristics. Based on systematic analysis and multi-parameter optimization， the optimal operating conditions were determined to be a liquid sulfur pressure of 700 kPa， a liquid sulfur thermodynamic temperature of 403.15 K， and a cooling water thermodynamic temperature of 333.15 K. Simulation and validation results confirmed that this parameter combination achieved a favorable balance between particle size uniformity and structural stability， demonstrating strong reliability and reproducibility， and providing a solid theoretical basis for improving product quality and advancing sulfur wet granulation technology.

Conclusion Among the three key factors—liquid sulfur pressure， liquid sulfur temperature， and cooling water temperature—liquid sulfur pressure emerges as the core dominant parameter due to its significant regulatory effect on the kinetic energy of the molten sulfur. By directly influencing the velocity， impact intensity， and breakup behavior of sulfur droplets， pressure plays a decisive role in determining particle morphology， size distribution， and overall granulation quality. In contrast， liquid sulfur temperature and cooling water temperature primarily affect the process through secondary mechanisms， such as modifying thermophysical properties and controlling heat transfer and solidification rates. Building on this understanding， the BP neural network-based process optimization method developed in this study demonstrates strong application potential. By establishing a data-driven mapping between operating parameters and granulation performance， this approach effectively reduces the uncertainty and inefficiency associated with the traditional trial-and-error method， while also overcoming the computational burden of direct optimization based on high-cost numerical simulations. Furthermore， the proposed methodology provides a scalable and efficient framework for parameter optimization， offering valuable guidance for process control and design. It is expected to be extended to a wide range of material processing and advanced manufacturing fields， thereby promoting the intelligent upgrading and digital transformation of related industrial processes.

Keywords： computational fluid dynamics； sulfur wet granulation； artificial neural network； equipment optimization

参考文献（References）

［1］Zhou Ling，Lyu Wanning，Bai Ling，et al.CFD-DEM study of gas-solid flow characteristics in a fluidized bed with different diameter of coarse particles［J］. Energy Reports， 2022， 8： 2376-2388.

［2］Hong S M， Kim O Y， Hwang S H. Chemistry of polythiols and their industrial applications［J］. Materials， 2024， 17（6）： 1343.

［3］Litster J， Ennis B. The science and engineering of granulation processes［M］. Dordrecht： Springer Science & Business Media， 2004.

［4］Abdoli Rad M， Shahsavand A. Modeling and simulation of heat transfer phenomenon in steel belt conveyer sulfur granulatingprocess［J］. Iranian Journal of Chemistry and Chemical Engineering， 2013， 32（4）： 93-104.

［5］Peeters M， Jiménez A A B， Matsunami K， et al. Analysis of the effect of formulation properties and process parameters on granule formation in twin-screw wet granulation［J］. International Journal of Pharmaceutics， 2024， 650： 123671.

［6］孙继鹏， 刘志平， 张占营， 等. 硫磺湿法造粒成型盘工艺参数模拟优化［J］. 化工与医药工程， 2024， 45（1）： 8-15.

Sun Jipeng， Liu Zhiping， Zhang Zhanying， et al. Simulation and optimization of process parameters for wet sulfur granulation and forming disk technology［J］. Chemical Engineering and Pharmaceutical Engineering， 2024， 45（1）： 8-15.

［7］Kittikunakorn N， Liu T， Zhang F. Twin-screw melt granulation： current progress and challenges［J］. International Journal of Pharmaceutics， 2020， 588： 119670.

［8］Poozesh S， Karam M， Akafuah N， et al. Integrating a model predictive control into a spray dryer simulator for a closed-loop control strategy［J］. International Journal of Heat and Mass Transfer， 2021， 170： 121010.

［9］Alexandrov N M， Lewis R M. Analytical and computational aspects of collaborative optimization formultidisciplinary design［J］. AIAA Journal， 2002， 40（2）： 301-309.

［10］Rahmanian N， Homayoonfard M， Alamdari A. Simulation of urea prilling process： an industrial case study［J］. Chemical Engineering Communications， 2013， 200（6）： 764-782.

［11］Sun Li， Liang Fei， Cui Wutai. Artificial neural network and its application research progress in chemical process ［J］. Asian Journal of Research in Computer Science， 2021，12（4）：177-185.

［12］Zhang Huishu， Zhan Dongping， Jiang Zhouhua. Application of improved BP neural network to final sulfur content prediction of hot metal pre-desulfurization［J］. Journal of Northeastern University （Natural Science）， 2007， 28（8）： 1140.

［13］Queipo N V， Haftka R T， Shyy W， et al. Surrogate-based analysis and optimization［J］. Progress in Aerospace Sciences， 2005， 41（1）： 1-28.

［14］Yondo R， Bobrowski K， Andrés E， et al. A review of surrogate modeling techniques for aerodynamic analysis and optimization： current limitations and future challenges in industry［M］.Cham： Springer International Publishing， 2019.

［15］Zhang Xijun， Yong Si. Optimization control of wastewater treatment based on neural network and multi-objective optimization algorithm［J］. Desalination and Water Treatment， 2024， 320： 100736.

［16］Ghafariasl P， Mahmoudan A， Mohammadi M， et al. Neural network-based surrogate modeling and optimization of a multigeneration system［J］. Applied Energy， 2024， 364： 123130.

［17］Baranya S， Olsen N R B， Stoesser T， et al. Three-dimensional RANS modeling of flow around circular piers using nested grids［J］. Engineering Applications of Computational Fluid Mechanics， 2012， 6（4）： 648-662.

［18］Golovina E， Ivanova I， Kargashilov D， et al. Engineering environmental protection at an industrial facility［J］. E3S Web of Conferences. EDP Sciences， 2023， 381： 01081.

［19］De Michele C， Coppola G. Numerical treatment of the energy equation in compressible flows simulations［J］. Computers & Fluids， 2023， 250： 105709.

［20］Zhang Lijuan， Gao Qiang， Wu Feng， et al. Numerical and experimental study on the suction process of residual kerosene of rocket engines［J］. International Communications in Heat and Mass Transfer， 2022， 139： 106504.

［21］Che Xinin， Wu Feng， Wang Junwu. Multiple field synergy mechanism of the desulfurization process in the intensified spouted beds［J］. Chemical Engineering Journal， 2023， 467： 143521.

［22］Rumelhart D E， Hinton G E， Williams R J. Learning representations by back-propagating errors［J］.Nature， 1986， 323（6088）： 533-536.

［23］Huang Lei， Qin Jie， Zhou Yi， et al. Normalization techniques in training dnns： methodology， analysis and application［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2023， 45（8）： 10173-10196.

［24］Paturi U M R， Cheruku S， Geereddy S R. Process modeling and parameter optimization of surface coatings using artificial neural networks （ANNs）： state-of-the-art review［J］. Materials Today： Proceedings， 2021， 38： 2764-2774.

［25］Pilania G. Machine learning in materials science： from explainable predictions to autonomous design［J］. Computational Materials Science， 2021， 193： 110360.

［26］Liu Lei， Kono Y， Kenney-Benson C， et al. Chain breakage in liquid sulfur at high pressures and high temperatures［J］. Physical Review B， 2014， 89（17）： 174201.

［27］Chamnanthongpaivanh B， Chatchawalsaisin J， Kittithreerapronchai O. Artificial neural network and support vector regression modeling for prediction of mixing time in wet granulation［J］. Journal of Pharmaceutical Innovation， 2022， 17（4）： 1235-1246.

［28］Younis Y M， Kayi H. Artificial neural network prediction of sulfur content of diesel fuel from its physical properties［C］//IOP Conference Series： Materials Science and Engineering. IOP Publishing， 2019， 518（6）： 062008.