Integration of in situ imaging and chord length distribution measurements for estimation of particle size and shape

Abstract

Efficient processing of particulate products across various manufacturing steps requires that particles possess desired attributes such as size and shape. Controlling the particle production process to obtain required attributes will be greatly facilitated using robust algorithms providing the size and shape information of the particles from in situ measurements. However, obtaining particle size and shape information in situ during manufacturing has been a big challenge. This is because the problem of estimating particle size and shape (aspect ratio) from signals provided by in-line measuring tools is often ill posed, and therefore it calls for appropriate constraints to be imposed on the problem. One way to constrain uncertainty in estimation of particle size and shape from in-line measurements is to combine data from different measurements such as chord length distribution (CLD) and imaging. This paper presents two different methods for combining imaging and CLD data obtained with in-line tools in order to get reliable estimates of particle size distribution and aspect ratio, where the imaging data is used to constrain the search space for an aspect ratio from the CLD data

Enabling precision manufacturing of active pharmaceutical ingredients: workflow for seeded cooling continuous crystallisations

“…However, processing of the chord length data captured is required to convert from the characteristic chord length to volume density PSD typically needed for PBMs. 78,79 Model building itself is not without its challenges. The more crystallisation phenomena that occur (secondary nucleation, agglomeration, etc.)…”

Determining particle‐size distributions from chord length measurements for different particle morphologies

“…In practice, both continuous distributions in Equation are approximated as piecewise discrete distributions (bins).Then assuming that the kernel is constant in the bin size, Equation 13 in discrete form becomes where q mea is the vector of CLD % in each bin, A is a transition matrix and y id is the vector of bin %s predicted by the idealized geometric model. The solution of Equation to predict the ideal PSD () from a geometric model has been the main approach used in the literature over the last 18 years, including recently proposed methodologies . In this work, we tested the applicability of this approach for the type of crystals and concentrations being studied.…”

Application of inline imaging for monitoring crystallization process in a continuous oscillatory baffled crystallizer

“…Various techniques have been reported in the literature for different operations during image analysis. 17 Ochsenbein et al segregated agglomerates from primary particles using image analysis by utilizing a pregenerated training set to classify the particles as agglomerates or primary particles based on a number of descriptors. The multi segmented Canny method has been shown to result in enhanced segmentation by accounting for the variation of the pixel intensity across the objects in the images.…”

“…16 Recently Agimelen et al used the aspect ratio distribution, sampled over the crystal population from in situ imaging, as a constraint in combination with the CLD data from FBRM to predict accurately the PSD. 17 Ochsenbein et al segregated agglomerates from primary particles using image analysis by utilizing a pregenerated training set to classify the particles as agglomerates or primary particles based on a number of descriptors. 18 Ferreira et al, have also combined image analysis with discreet factorial analysis to quantify the degree of agglomeration and growth rates of sucrose crystals by online monitoring of the crystallization process.…”