Wang Xuezhong¹，， Zhou Meng¹， Mao Xiyue¹， Liu Yan²， Wu Tao^1*， Zhou Lei¹

1.College of Biomedicine and Health，Beijing Key Laboratory of Enze Biomass Fine Chemicals，

Beijing Institute of Petrochemical Technology， Beijing 102617， China；

2.Jinggema （Qingdao） Intelligent Technology Company Limited， Qingdao 266000， China

Abstract

Significance Fluidized bed granulation （FBG） represents a cornerstone unit operation in solid dosage manufacturing， integrating powder mixing， liquid binder spraying， wet massing， and drying within a single multifunctional vessel. The primary objective of FBG is the efficient production of uniform and homogeneous granules characterized by enhanced flowability， high solubility， and minimal dust generation， which are critical downstream properties for processes such as tableting， compaction， and capsule filling. To facilitate the design of intelligent closed-loop control systems and provide robust theoretical foundations for industrial process scale-up， it is essential to conduct comprehensive research on the online monitoring of critical quality attributes （CQAs） and advanced process simulation. Such research aims to establish a systematic and multiscale understanding of both the macroscopic granulation process parameters and the microscopic particle evolution mechanisms. Deciphering these transient physical changes—ranging from microscale liquid bridge formation to macroscale particle aggregation—enables developers to maintain process stability and prevent batch failures. During scale-up， nonlinear thermodynamic variations and localized fluidization anomalies often occur， making empirical prediction models highly unreliable. Historically， FBG process optimization has relied heavily on empirical trial-and-error methodologies and offline characterization. However， traditional destructive manual sampling and subsequent offline testing introduce significant measurement delays and human errors， making real-time closed-loop adjustments impractical. Furthermore， to meet modern regulatory requirements， such as continuous quality verification， a shift toward science-based process understanding is necessary. To address these limitations， process analytical technology （PAT） and multiscale numerical models have emerged as complementary and non-invasive alternatives. Online PAT sensors provide non-destructive and continuous tracking of physical and chemical attributes during operation， while numerical simulations offer thermodynamic and mechanistic insights into the underlying gas-solid flow dynamics and particle growth mechanisms. The integration of PAT and process simulation represents a viable pathway for building intelligent FBG closed-loop control systems and facilitating efficient industrial scale-up.

Progress Optical and spectroscopic techniques currently represent the predominant approaches for FBG online monitoring. In the domain of PAT， image analysis methods have been successfully implemented for real-time particle size distribution （PSD） measurement， achieving high spatiotemporal resolutions that capture transient growth， aggregation， and breakage phenomena. Meanwhile， near-infrared spectroscopy （NIRS） has been extensively adopted for the simultaneous and non-destructive monitoring of granule moisture content and active pharmaceutical ingredient （API） homogeneity. Despite these advances， mainstream single-point inline sensors—including focused beam reflectance measurement （FBRM） and spatial filter velocimetry （SFV）—frequently face limitations related to spatial localization. When positioned in low-shear or stagnant regions of the fluidized bed， these probes fail to acquire spatially representative data， which can compromise the overall reliability of real-time monitoring results. Parallel to experimental monitoring， FBG numerical modeling has transitioned from macroscopic empirical correlations to multiscale phenomenological simulations. The population balance model （PBM）， computational fluid dynamics （CFD）， and discrete element method （DEM） serve as three core computational frameworks. While early PBMs model macroscopic PSD evolution via specialized growth， aggregation， and breakage kernels， they assume spatial homogeneity and perform poorly in highly dynamic fluidization regimes. To address this spatial limitation， coupled CFD-DEM simulations have been widely applied to analyze particle residence time distributions and resolve detailed spatial flow patterns. Subsequent research has led to integrated CFD-DEM-PBM frameworks that couple microscale liquid bridge physics with macroscale multiphase hydrodynamic behaviors. These models explicitly capture localized binder distribution and interparticle liquid bridge forces to accurately predict granule growth kinetics. Additionally， coarse-grained particle methodologies have been explored to reduce the high computational costs of full-scale CFD-DEM simulations， thereby improving the feasibility of simulating larger industrial-scale equipment. On the application front， PAT-based feedback control systems have demonstrated the ability to adaptively adjust binder spray rates based on inline moisture and PSD measurements， maintaining consistent granule attributes despite process disturbances.

Conclusions and Prospects Over the past decade， substantial progress has been achieved in FBG online monitoring and multiscale simulation， enabling more accurate CQA tracking and detailed modeling of complex gas-solid-liquid interactions. Nonetheless， several critical challenges continue to restrict large-scale industrial deployment. Most existing numerical frameworks rely on idealized spherical particle assumptions and simplified liquid bridge physics， which limit their predictive accuracy when applied to irregular and real-world granules. Furthermore， the high computational overhead associated with coupled multiscale CFD-DEM simulations currently prevents their direct integration into real-time model predictive control （MPC） systems. To overcome these limitations and advance the field， several promising research directions are proposed. First， the development of physics-informed machine learning （PIML） and optimized coarse-grained DEM techniques is necessary to reduce simulation times and accelerate computations without compromising physical fidelity. Second， a deeper and more seamless integration of online PAT tools and real-time numerical models is required to establish robust and adaptive digital twins of industrial FBG systems. By synchronizing virtual models with physical sensors， these digital twins can dynamically predict process drift， facilitate intelligent closed-loop control， and support reliable industrial scale-up. Finally， extending intelligent closed-loop control frameworks from traditional batch operations to continuous granulation lines represents a crucial step toward facilitating the modernization， automation， and digital transformation of pharmaceutical solid dosage manufacturing.

Keywords:  fluidized bed granulation； online detection technology； image analysis； process simulation； machine learning

Get Citation:  Wang Xuezhong， Zhou Meng， Mao Xiyue， et al. Research progress on online monitoring and process simulation of fluidized bed granulation［J］. China Powder Science and Technology， 2027， 33（3）： 1-11.

Received：2026-06-02， Revised： 2026-08-14， Online： 2026-11-04。

Funding：The research was supported by the National Natural Science Foundation of China Young Scientists Fund (Grant No. 52102208), the Science and Technology Plan of Beijing Municipal Commission of Education (Grant No. KM202410017011), and the Beijing Nova Program (Grant No. Z211100002121024).

DOI：10.13732/j.issn.1008-5548.2027.03.013

CLC No.: TQ460.6+3；TQ051.13；TB4

Type Code：A

Serial No.:  1008-5548（2027）03-0001-11