Real-time characterization of particle size distribution evolution in top-spray fluidized bed granulation via passive acoustic emission

Abstract

Real-time monitoring of particle size distribution (PSD) in fluidized bed granulation remains a significant challenge due to the complex coupling between particle growth and multiphase flow dynamics, alongside a lack of robust non-invasive techniques. In this work, a physics-informed acoustic inversion framework is developed to enable real-time PSD characterization based on passive acoustic emission (AE) signals. Wavelet packet transform extracts frequency-domain energy features, which are linked to particle size through collision-induced acoustic responses. A constrained least-squares approach is then implemented to decouple multi-component contributions and reconstruct PSD evolution. To overcome signal instability caused by wall-sticking-a key limitation in practical systems-an intermittent spray strategy is introduced to stabilize acoustic signals and improve reproducibility. The proposed approach achieves reliable PSD prediction during the drying stage, with an average relative error of 22.13%. The remaining errors are mainly associated with trace component identification (mass fraction <5%) and severe deposition-induced signal distortion. Importantly, this method captures the attenuation of fluidization dynamics during particle growth, providing real-time quantitative feedback for process control. This work establishes a non-invasive and mechanism-informed monitoring strategy, offering a practical pathway toward intelligent control of fluidized bed granulation processes.

Introduction

Particle agglomeration is a fundamental process in modern chemical manufacturing, enabling the transformation of fine powders into structured aggregates with tailored flowability, porosity, and functional properties for downstream applications such as tableting and capsule filling [1], [2], [3], [4]. Controlled agglomeration is widely employed in processes including coating, fluidized bed drying, heterogeneous catalysis, and advanced granulation, where it plays a key role in enhancing process efficiency and product quality [5]. Among these, top-spray fluidized bed granulation is particularly attractive due to its operational simplicity, efficiency, and controllability [6]. In this process, particles are well dispersed in a fluidized state, while a two-fluid nozzle atomizes binder solution into fine droplets. Concurrently, heated air promotes rapid solvent evaporation, leading to solute precipitation at particle contact points and the formation of solid bridges, thereby consolidating particles into stable agglomerates.

Fluidized bed granulation is a complex, multi-parameter process requiring precise control of operating variables, including temperature, humidity, gas velocity, binder properties, spray rate, and droplet size, to achieve desired product quality [7]. Its operation is challenged by stochastic particle motion, non-uniform liquid distribution, and the risk of defluidization when excessive moisture accumulates [8], [9]. Agglomerate quality is traditionally assessed via off-line sampling during operation [10].

However, this approach suffers from inherent limitations, including delayed feedback, poor spatial representativeness of particle size distribution, and disturbance of bed hydrodynamics and thermal stability due to sampling-induced air intrusion. These issues are exacerbated under high-temperature or high-pressure conditions, where invasive sampling poses additional safety risks. Therefore, the development of non-invasive, real-time monitoring techniques is essential to enable dynamic process control, ensuring stable operation and consistent product quality.

Several process analytical technologies have been developed for monitoring fluidized bed granulation and related particulate processes. Optical and image-based techniques can directly provide information on particle size, morphology, and growth dynamics. Recent studies on continuous spray fluidized bed agglomeration indicate that operating parameters, such as gas inlet temperature and binder content, significantly influence both agglomerate growth and morphology. Furthermore, two-dimensional roundness obtained from imaging analysis, when calibrated with X-ray microcomputed tomography, can be used to estimate the three-dimensional fractal dimension of agglomerates for online morphological characterization [11], [12]. Probe-based techniques, including near-infrared spectroscopy, focused beam reflectance measurement, and particle vision and measurement, have been widely applied to monitor moisture content, particle size distribution, bulk density, coating thickness, and morphological evolution in fluidized bed processes [13]. In addition, Raman spectroscopy, pressure fluctuation analysis, and electrical capacitance tomography (ECT) provide complementary insights into particle composition and bed hydrodynamics. Although these techniques have significantly advanced process monitoring from different perspectives, their application in wet and dense fluidized beds may be limited by probe fouling, optical obstruction, local sampling bias, limited spatial representativeness, and the requirement for direct in-bed access.

Among these non-invasive monitoring strategies, acoustic emission (AE) is attractive because it can be implemented through externally mounted sensors and does not require optical access or direct insertion of probes into the bed. AE refers to transient elastic waves generated by the rapid release of localized energy within materials, typically arising from fracture, collision, or mechanical vibration [14]. These signals can be captured and processed using AE sensors coupled with amplification and data acquisition systems. In fluidized beds, AE signals primarily originate from particle–particle and particle–wall interactions [15].

As a non-invasive, real-time monitoring technique, AE has been widely applied in chemical engineering, pharmaceuticals, and structural diagnostics to probe complex physico-chemical processes [16], [17], [18]. In top-spray fluidized bed granulation, AE offers distinct advantages over conventional pressure-based methods by providing not only macroscopic information on fluidization behavior but also real-time characterization of PSD, while enabling early detection of fluidization instability [19], [20], [21].

Fu et al. [22] combined near-infrared (NIR) spectroscopy with AE for simultaneous monitoring of moisture content and particle size, demonstrating higher accuracy of NIR for moisture tracking and superior sensitivity of AE to particle size evolution. In spouted bed granulation, Liu et al. [23] developed a multi-modal monitoring system integrating acoustic, pressure, and electrical capacitance tomography (ECT), where acoustic recurrence analysis and ECT visualization enabled early detection of defluidization. AE has also shown strong potential in industrial process monitoring. Sheahan et al. [24] applied statistical analysis of acoustic signals to qualitatively assess coating quality, while Vervloet et al. [25] used AE to track drying kinetics of pharmaceutical granules. Beyond granulation, acoustic diagnostics have been widely employed to characterize flow dynamics: Briongos et al. [26] analyzed acoustic and pressure signals to identify fluidization regimes, particularly slugging onset, and Zhou et al. [13] correlated acoustic frequency-energy features with flow regime transitions using discrete wavelet transform.

Despite these advances, the application of AE for real-time PSD prediction in fluidized beds remains limited by several challenges. First, real-time analysis imposes a substantial computational burden due to high sampling frequencies (up to 300 kHz) [27], [28], requiring efficient signal-processing algorithms. Second, the strong coupling between non-linear agglomeration dynamics and fluidization behavior complicates the decoupling of particle composition and size information. Third, signal stability is highly sensitive to operational disturbances, particularly wall-sticking, which undermines reproducibility compared with conventional pressure-based methods.

To overcome these limitations, this study extends the PSD prediction framework of Ren et al. [29] toward real-time application in top-spray fluidized bed granulation. An intermittent spray strategy is proposed to suppress wall-sticking interference and enhance signal stability. More importantly, a hybrid analytical approach is developed by coupling Wavelet Packet Transform (WPT) with a physics-constrained least-squares method (LSM), enabling effective decoupling of multi-component mass fractions from complex acoustic signals. This integrated strategy allows reliable, real-time tracking of particle size evolution throughout the granulation process, providing a robust quantitative basis for monitoring fluidization dynamics and enabling feedback-driven process control.

Continue reading here

Yi Wang, Jiliang Ma, Jialiang Cai, Han Pu, Daoyin Liu, Xiaoping Chen, Real-time characterization of particle size distribution evolution in top-spray fluidized bed granulation via passive acoustic emission, Chemical Engineering Journal, Volume 544, 2026, 179214, ISSN 1385-8947, https://doi.org/10.1016/j.cej.2026.179214.

Read more interesting articles on spray fluidized bed granulation here:

• Comprehensive Review of Modern Techniques of Granulation in Pharmaceutical Solid Dosage Forms
• A Comparative Study on the Coating Performance of Wurster, Bottom and Top Spray Fluidized Bed Coaters
• The Functional Attributes of Co-Processed Excipients in Direct Compression