Interpretable Machine Learning Methods for Monitoring Polymer Degradation in Extrusion of Polylactic Acid.

This work investigates real-time monitoring of extrusion-induced degradation in different grades of PLA across a range of process conditions and machine set-ups. Data on machine settings together with in-process sensor data, including temperature, pressure, and near-infrared (NIR) spectra, are used as inputs to predict the molecular weight and mechanical properties of the product. Many soft sensor approaches based on complex spectral data are essentially 'black-box' in nature, which can limit industrial acceptability. Hence, the focus here is on identifying an optimal approach to developing interpretable models while achieving high predictive accuracy and robustness across different process settings. The performance of a Recursive Feature Elimination (RFE) approach was compared to more common dimension reduction and regression approaches including Partial Least Squares (PLS), iterative PLS (i-PLS), Principal Component Regression (PCR), ridge regression, Least Absolute Shrinkage and Selection Operator (LASSO), and Random Forest (RF). It is shown that for medical-grade PLA processed under moisture-controlled conditions, accurate prediction of molecular weight is possible over a wide range of process conditions and different machine settings (different nozzle types for downstream fibre spinning) with an RFE-RF algorithm. Similarly, for the prediction of yield stress, RFE-RF achieved excellent predictive performance, outperforming the other approaches in terms of simplicity, interpretability, and accuracy. The features selected by the RFE model provide important insights to the process. It was found that change in molecular weight was not an important factor affecting the mechanical properties of the PLA, which is primarily related to the pressure and temperature at the latter stages of the extrusion process. The temperature at the extruder exit was also the most important predictor of degradation of the polymer molecular weight, highlighting the importance of accurate melt temperature control in the process. RFE not only outperforms more established methods as a soft sensor method, but also has significant advantages in terms of computational efficiency, simplicity, and interpretability. RFE-based soft sensors are promising for better quality control in processing thermally sensitive polymers such as PLA, in particular demonstrating for the first time the ability to monitor molecular weight degradation during processing across various machine settings.

Machine Learning for Process Monitoring and Control of Hot-Melt Extrusion: Current State of the Art and Future Directions.

In the last few decades, hot-melt extrusion (HME) has emerged as a rapidly growing technology in the pharmaceutical industry, due to its various advantages over other fabrication routes for drug delivery systems. After the introduction of the 'quality by design' (QbD) approach by the Food and Drug Administration (FDA), many research studies have focused on implementing process analytical technology (PAT), including near-infrared (NIR), Raman, and UV-Vis, coupled with various machine learning algorithms, to monitor and control the HME process in real time. This review gives a comprehensive overview of the application of machine learning algorithms for HME processes, with a focus on pharmaceutical HME applications. The main current challenges in the application of machine learning algorithms for pharmaceutical processes are discussed, with potential future directions for the industry.

Development of a Surface-Enhanced Raman Spectroscopic Methodology to Detect Immobilized Organic Materials in Biogeological Contexts.

The likelihood of finding intact cellular structures on the surface or in the near subsurface of the martian regolith is slim, due in part to the intense bombardment of the surface by ionizing radiation from outer space. Given that this radiation is predicted to be so intense that it would render a living cell inactive within minutes, it is logical to search for evidence of microbial life by looking for molecules produced by the breakdown of cellular matter. This "pool" of molecules, known as biomarkers, consists of a range of species with various functionalities that make them likely to interact with minerals in the martian regolith. Raman spectroscopy, a molecularly specific analysis method utilized for detecting organic biomarkers among inorganic geomaterials, suffers from low signal intensity when the concentration of organics is as low as it appears to be on the martian surface. This article describes the utility of a surface-enhanced Raman spectroscopy (SERS) method used to detect extremely low levels of biomarkers that were passively adhered to mineral surfaces in a method that represents how this interaction would take place in a natural environment on Mars. The methodology showed promise for the detection of multiple classes of biomarkers.

Compartmental approach for modelling twin-screw granulation using population balances.

In this study, a compartmental population balance model (CPBM) is developed as a predictive tool of particle size distribution (PSD) for wet granulation in co-rotating twin-screw granulator (TSG). This model is derived in terms of liquid to solid ratio (L/S) and screw speed representing the main process parameters of the TSG. The mathematical model accounts for aggregation and breakage of the particles occurring in five compartments of the TSG with inhomogeneous screw configurations (3 conveying zones and 2 kneading zones). Kapur's aggregation kernel is implemented in granulation and finite volume numerical method is adapted for solving the mathematical model. The results show a dramatic improvement in solution accuracy compared to the cell average numerical method. Moreover, Kriging interpolation is used to interpolate for new values of empirical parameters at different L/S and screw speeds. Finally, the CPBM model is calibrated and validated using the experimental data.

Investigation of polymorphic transitions of piracetam induced during wet granulation.

Piracetam was investigated as a model API which is known to exhibit a number of different polymorphic forms. It is freely soluble in water so the possibility exists for polymorphic transformations to occur during wet granulation. Analysis of the polymorphic form present during lab-scale wet granulation, using water as a granulation liquid, was studied with powder X-ray diffraction and Raman spectroscopy as off-line and inline analysis tools respectively. Different excipients with a range of hydrophilicities, aqueous solubilities and molecular weights were investigated to examine their influence on these solution-mediated polymorphic transitions and experimental results were rationalised using molecular modelling. Our results indicated that as an increasing amount of water was added to the as-received piracetam FIII, a greater amount of the API dissolved which recrystallised upon drying to the metastable FII(6.403) via a monohydrate intermediary. Molecular level analysis revealed that the observed preferential transformation of monohydrate to FII is linked with a greater structural similarity between the monohydrate and FII polymorph in comparison to FIII. The application of Raman spectroscopy as a process analytical technology (PAT) tool to monitor the granulation process for the production of the monohydrate intermediate as a precursor to the undesirable metastable form was demonstrated.