Modelling Fluidized Wet Granulation Processes

Introduction

Fluidized wet granulation represents a critical manufacturing process across pharmaceutical, chemical, and food industries. This technology transforms fine powders into larger, more manageable granules by spraying a binding liquid in a fluidized bed of particles. The process involves complex multiphase phenomena including particle fluidization, droplet deposition, wetting, nucleation, growth, consolidation, and breakage.

Mathematical modeling of fluidized wet granulation has evolved significantly over recent decades, progressing from empirical correlations to mechanistic models and increasingly sophisticated computational approaches. This evolution has been driven by industrial needs for design tools that can reduce development time, improve product quality, enable robust scale-up, and support regulatory initiatives like Quality by Design (QbD).

This article provides a structured overview of modelling approaches for fluidized wet granulation, examining their theoretical foundations, implementation strategies, advantages, and limitations. It also explores practical applications in process development, optimization, and scale-up, addressing current challenges and future directions.

Fundamentals of Fluidized Wet Granulation

Before discussing modelling approaches, it's essential to understand the fundamental phenomena occurring during fluidized wet granulation. The process involves several simultaneous and often competing mechanisms:

• Fluidization: Air flows upward through a bed of particles, suspending them and creating a fluid-like behavior
• Wetting: Binder solution is atomized and contacts particles, forming liquid bridges
• Nucleation: Formation of initial granules from dust particles wetted by binder droplets
• Coalescence: Two granules combine to form a larger one when their liquid layers overlap
• Layering: Material deposits onto existing granules, increasing their size gradually
• Breakage: Granules may fracture due to collisions or mechanical stress
• Drying: Evaporation of liquid from granules, altering their mechanical properties

The interplay between these mechanisms determines the final granule properties such as size distribution, shape, density, porosity, and moisture content. Effective models must capture these phenomena at appropriate spatial and temporal scales while connecting particle-level interactions to bulk process behavior.

Key Process Parameters

Several critical parameters influence the granulation process and must be incorporated in any model:

Modelling Approaches

Population Balance Models (PBM)

Population Balance Models represent one of the most established approaches for modelling granulation processes. Based on tracking the number of particles in different size classes, PBMs describe the evolution of the particle size distribution through mechanistic rate expressions for aggregation and breakage events.

The general form of a population balance equation for granulation is:

Where n(v,t) is the number density function for particles of volume v at time t, and (u,v) is the aggregation kernel describing the rate of collision between particles of volumes u and v. The equation represents how the number of particles in a given size class changes due to formation from smaller particles and loss to larger particles through aggregation events.

Several aggregation kernels have been developed for granulation modeling:

• Size-independent kernel: (u,v) = constant (simplest form, often unrealistic)
• Size-dependent kernel: (u,v) = ku/ v/ (more flexible, accounts for collision frequency)
• Physically-based kernels: Derived from collision theory, incorporating collision velocity, success probability, and other physical parameters
• Mechanism-specific kernels: Different kernel forms for different regimes (nucleation, steady growth, induction periods)

PBM applications typically involve solving the population balance equation using numerical methods such as Method of Moments, Method of Classes, or Monte Carlo simulation. These approaches can predict granule size distributions under varying operating conditions and inform process optimization.

Despite their widespread use, PBMs have limitations. They often require empirical parameters that need experimental fitting, typically don't capture spatial heterogeneities in the process, and usually focus on size as the sole granule property. Advanced implementations address some of these limitations but at increased computational cost.

Computational Fluid Dynamics (CFD) Models

CFD models solve the governing equations of fluid flow (mass, momentum, energy) to predict the hydrodynamics in a fluidized bed granulator. These models provide detailed information on gas velocity profiles, particle mixing patterns, temperature distribution, and residence time.

For fluidized bed modeling, several CFD approaches are available:

• Two-Fluid Model (Eulerian-Eulerian): Treats both gas and solid phases as interpenetrating continua with their own sets of conservation equations. Suitable for large-scale systems requiring reasonable computational resources, but depends on closure models for particle phase stress.
• Discrete Phase Model (Lagrangian-Eulerian): Tracks individual particles or parcels in a continuous fluid phase. Provides more detailed particle histories but becomes computationally intensive for systems with many particles.
• Discrete Element Method (DEM)-CFD coupling: Combines Lagrangian particle tracking with detailed CFD for the fluid phase, providing the most comprehensive description of particle-particle interactions but requiring substantial computational resources.

Recent advances in CFD modeling of granulation include:

• Heat and mass transfer models to calculate drying rates and binder distribution
• Liquid bridge models to capture wet particle interactions and cohesive forces
• Spray models to predict droplet trajectories, dispersion, and deposition
• Multiphase formulations to track liquid content on particle surfaces
• Advanced turbulence models for more accurate flow prediction
• Adaptive meshing to concentrate computational effort in regions of interest

CFD models provide valuable insights into process scale-up, nozzle positioning, and chamber design. However, standalone CFD models typically don't predict granule properties like size distribution directly. They are often combined with population balance or other modeling approaches to create more comprehensive simulations.

Discrete Element Method (DEM) Models

DEM models track individual particles by solving Newton's equations of motion for each particle, accounting for particle-particle and particle-wall collisions. For wet granulation applications, these models must incorporate additional physics to describe liquid bridges between particles.

Key components of DEM models for wet granulation include:

• Contact mechanics: Normal and tangential forces during collisions, typically modeled using spring-dashpot systems or more sophisticated formulations
• Liquid bridge models: Calculation of cohesive forces between wet particles based on liquid content, viscosity, surface tension, and separation distance
• Surface liquid tracking: Methods to track liquid layers on particle surfaces, including droplet deposition, spreading, and layering
• Coalescence criteria: Rules determining when colliding wet particles will stick together versus bounce apart based on relative velocity, liquid content, and particle size
• Breakage mechanisms: Algorithms to simulate granule fracture under stress based on mechanical strength models

A significant advantage of DEM models is their ability to predict granule properties beyond size, including shape evolution, porosity development, and internal structure formation. They can also capture segregation phenomena and naturally predict the evolution of the granule size distribution based on particle-level interactions.

However, DEM models are computationally intensive, limiting simulations to relatively small systems. Various strategies are employed to make DEM more applicable to real industrial processes:

• Scaling up particle size while maintaining dimensionless parameters
• Using simplified particle shapes instead of actual irregular shapes
• Parallel computing implementation and GPU acceleration
• Hybrid modeling approaches that combine DEM with other techniques
• Coarse-graining methods to reduce the number of tracked particles

Hybrid and Multiscale Models

Recognizing the strengths and limitations of individual modeling approaches, researchers have developed hybrid and multiscale models that combine different techniques to achieve more comprehensive simulations with reasonable computational cost.

Common hybrid modeling approaches for fluidized wet granulation include:

• CFD-PBM Coupling: Combining CFD for hydrodynamics with population balance for granule size evolution. The fluid flow information from CFD informs the rate expressions in the PBM, creating a more physically-based model that accounts for spatial heterogeneity.
• CFD-DEM Coupling: Coupling CFD for gas flow with DEM for particle motion, providing detailed description of fluid-particle and particle-particle interactions. Often extended with population balance methods to track granule size changes.
• PBMs with Internal Coordinates: Extending population balances to track multiple granule properties simultaneously (size, moisture content, porosity, composition) through multi-dimensional distributions.
• Mechanistic-Statistical Hybrids: Combining mechanistic models with statistical techniques like response surface methodology or design of experiments to reduce computational cost while maintaining accuracy.
• Machine Learning-Enhanced Models: Integrating artificial intelligence with mechanistic models to accelerate calculations, optimize parameters, or discover new relationships.

These hybrid approaches represent the state-of-the-art in granulation modeling, offering more complete representations of the process and enabling prediction of multiple granule properties under varying operating conditions.

Industrial Applications

Process Development and Optimization

Granulation models have become invaluable tools in pharmaceutical and chemical process development, enabling:

• Reduced experimental effort through virtual screening of process parameters
• Identification of optimal operating conditions for targeted granule properties
• Prediction of batch-to-batch variability and process sensitivity
• Root cause analysis of product quality issues
• Design of experiments optimization and parameter space exploration

For instance, population balance models can systematically explore the effect of binder spray rate on the final granule size distribution without conducting actual experiments, while CFD models can identify potential dead zones or poor mixing regions in a granulator design before equipment is built.

Scale-up and Technology Transfer

One of the most challenging aspects of granulation process development is scaling from laboratory to production scale. Models assist this transition by:

• Simulating the impact of different equipment geometries and sizes on hydrodynamics
• Identifying dimensionless groups to maintain consistent process behavior across scales
• Predicting how changes in scale affect kinetic rates and mixing patterns
• Suggesting modifications to operating conditions to compensate for scale effects
• Identifying potential scale-up issues such as segregation or poor fluidization quality

CFD models are particularly valuable for scale-up, as they can predict flow patterns, residence times, and mixing behavior across different scales. Meanwhile, PBMs help understand how population dynamics change with scale, especially regarding collision frequencies and kinetic rates.

Quality by Design (QbD) Implementation

In the pharmaceutical industry, regulatory initiatives like Quality by Design emphasize understanding and controlling manufacturing processes. Granulation models support QbD implementation by:

• Identifying critical process parameters (CPPs) that affect critical quality attributes (CQAs)
• Establishing design spaces where process parameters guarantee acceptable product quality
• Enabling risk assessment for process variations and their impacts
• Supporting the development of control strategies including PAT (Process Analytical Technology)

Models for fluidized bed granulation have been used to establish design spaces for granule size distribution, moisture content, bulk density, and compressibility. These design spaces contribute directly to regulatory filings and support more flexible manufacturing approaches within the pharmaceutical industry.

Current Challenges and Future Directions

Despite significant advances, several challenges remain in modelling fluidized wet granulation:

• Model parameterization: Determining model parameters, especially for complex mechanistic models, can be difficult and often requires extensive experimental data. The development of predictive relationships for these parameters remains an active research area.
• Computational limitations: Detailed models like DEM remain computationally expensive for scale-up studies. Even CFD models can be resource-intensive for large systems, limiting their use in virtual experimentation.
• Particle shape and complexity: Most models use simplified particle shapes (spheres) and ignore realistic distributions in size and shape, which can affect predictions, especially for non-spherical or irregular feed materials.
• Multiscale physics: Capturing phenomena ranging from molecular-level liquid bridge formation to system-scale process behavior remains challenging, requiring innovative coupling strategies.
• Model validation: Comprehensive validation of models across different formulations, equipment, and scales is limited by available experimental data, particularly for internal granule structure and composition.
• Regulatory acceptance: While models are increasingly used in development, regulatory acceptance varies across regions and applications, creating uncertainty in their implementation for critical quality decisions.

The field of granulation modelling continues to evolve, with several promising directions:

• Integrated process-product models: Models that link granule properties directly to final product performance (e.g., drug dissolution, tablet hardness, flowability).
• Machine learning enhancements: Using artificial intelligence to accelerate model calculations, optimize parameters, discover new relationships, or develop surrogate models based on limited high-fidelity simulations.
• Real-time digital twins: Combining models with advanced sensors and machine learning for process monitoring, control, and anomaly detection during operation.
• First-principle parameter prediction: Developing methods to predict model parameters from material properties rather than fitting them to data, improving model transferability.
• Standardized validation protocols: Establishing benchmark cases and validation procedures to improve model reliability and enable better comparison between different modeling approaches.
• Cloud-based modeling platforms: Making advanced models more accessible through web-based interfaces and cloud computing, reducing barriers to adoption in industry.

Conclusion

Modelling fluidized wet granulation processes has matured from simple empirical correlations to sophisticated computational approaches that can predict granule properties and process behavior. Population balance models, computational fluid dynamics, discrete element methods, and their various combinations offer different levels of detail and complexity, each with their own strengths and limitations.

These models have proven valuable for process development, optimization, scale-up, and regulatory compliance in pharmaceutical and chemical industries. They enable virtual experimentation, support Quality by Design initiatives, and facilitate technology transfer across different scales and equipment configurations.

Despite progress, challenges remain in model parameterization, computational requirements, validation, and regulatory acceptance. Ongoing research continues to advance the field, with machine learning, integrated process-product models, and real-time digital twin applications representing particularly promising future directions.

As computational power increases and modeling techniques continue to evolve, virtual experimentation and simulation-based process design will become increasingly central to fluidized bed granulation development. This evolution will reduce time-to-market, improve product quality, and enable more flexible manufacturing approaches across many industries that rely on granulation processes.