Researchers have gone on to develop a physics-enhanced autocorrelation-based estimator (PEACE) machine learning algorithm that can go ahead with the extraction of the particle size distribution (PSD) of a pharmaceutical powder surface from its laser spot. The process provides a non-invasive, real-time, as well as a far-field optical probe so as to monitor particle size distribution in a quantitative way.
Forecasting the appearance of the large particles by way of consistent monitoring happens to be pivotal for process control, as per the research that got published in Nature Communications.
Electronic Speckle Pattern Interferometry (ESPI), notably, can go on to measure the motion distribution of the surface even at scales as tiny as a nanometre and is used to denote the roughness of the surface.
That said, extracting quantitative information when it comes to highly scattering surfaces from an imaging system is quite an uphill task, as per Zhang et al. This is due to the fact that the scattered light phase goes through multiple folds upon proliferation, thereby resulting in speckle patterns that happen to be complex.
Nevertheless, laser speckle pattern mostly functions only when the surface height fluctuation happens to be smaller than or is comparable to the light wavelength. This, therefore, curtails its application to the surfaces that are encountered in pharmaceutical manufacturing. Recent developments in machine learning have gone on to show imaging success via scattering media as well as suppression of the speckle. However, the speckle pattern is regarded as a disturbance that is unwanted.
According to Zhang et al., there happen to be no real-time monitoring approaches that can spot such particle size changes in the early stages and thereby safeguard such changes, which are abnormal for wet powder drying.
It is well to be noted that quantitative granularity characterization is indeed desirable when it comes to the powder drying process. Although these parameters happen to be well controlled, the progression of the particle sizes during the process of agitation isn’t entirely predictable. Hence, it is of utmost essence to monitor particle sizes in real time, and that too quantitatively and correctly when it comes to abnormal size changes by way of feedback control in terms of process parameters like agitation, speed, and temperature.
Apparently, the free space multiplication equations helped the scientists relate the function pertaining to ensemble-averaged spatial integration autocorrelation to that of powder surface statistics like the particle size distribution.
One of the major advantages of this strategy happens to be its understandability. It is important to note that the inferences drawn from the activation map went on to be a complete match with the forward model that happens to be used by the researchers. The method goes on to solve both forward and inverse challenges all at once, as per the authors.
Especially in scenarios pertaining to densely concentrated wet powders, this method happens to be the first in-line measurement and is, as a matter of fact, easily deployable when it comes to the industrial instrument, making it apt for the processes related to blending as well as milling in the pharmaceutical manufacturing gamut.