Massachusetts Institute of Technology researchers have developed an algorithm that extracts powder particle size distribution from laser speckle in real-time.
Researchers have developed a physics-enhanced autocorrelation-based estimator (PEACE) machine learning algorithm that can extract the particle size distribution (PSD) of a pharmaceutical powder surface from its laser speckle.
The method provides a real-time, non-invasive, far-field optical probe to monitor particle size distributions quantitatively. Estimating the appearance of large particles from the continuous monitoring is important for process control, according to the research published in Nature Communications.
Electronic Speckle Pattern Interferometry (ESPI) can measure the surface motion distribution even at nanometre scales and is used to characterise surface roughness. Yet extracting quantitative information about highly scattering surfaces from an imaging system is challenging, according to Zhang et al. This is because the phase of the scattered light undergoes multiple folds upon propagation, resulting in complex speckle patterns.
However, the laser speckle pattern usually works only when the surface height fluctuation is smaller than or comparable to the light wavelength. This limits its application to surfaces encountered in pharmaceuticals manufacturing.
Recent advances in machine learning have showed success in imaging through scattering media and speckle suppression, yet speckle pattern is treated as an unwanted disturbance, the authors noted.
Real-time online monitoring methods for wet powder drying
Zhang et al. stated that to their knowledge, there are no real-time online monitoring methods that can detect early on and prevent such abnormal particle size changes for wet powder drying.
Quantitative granularity characterisation is desirable in the powder drying process. While these parameters are generally well-controlled, the evolution of the particle sizes during agitation is not fully predictable. Therefore, it is crucial to monitor particle sizes quantitatively in real-time and correct for abnormal size changes through feedback control on process parameters, such as temperature, agitation and speed.
How was the particle size distribution monitoring algorithm effective?
Free-space propagation equations enabled the researchers to relate the ensemble-averaged spatial-integral autocorrelation function to the statistics of powder surface, eg, the PSD.
Credit: Massachusetts Institute of Technology
Advantages and applications
One key advantage of the strategy is interpretability. Results from the activation map matched the forward model used by the researchers. The method solves both the forward and inverse problems together, the authors stated.
Zhang et al. concluded: “Especially for densely concentrated wet powders, this method is the first in-line measurement, and it is easily deployable in the industrial instrument” and is therefore suitable for processes such as blending and milling in pharmaceutical manufacturing.
Millennium Pharmaceuticals, Inc., a subsidiary of Takeda Pharmaceuticals, supported the research.
Source link
#MIT #unlocks #realtime #powder #particle #size #distribution #monitoring
