A UV-Vis spectroscopic method for monitoring of additive particle properties during polymer compounding

Abstract

Determination of the dispersion characteristics/or morphology of additives in polymer melts by fast, reliable and accurate on-line methods is highly desired in the polymer industry. An ultraviolet-visible (UV-Vis) spectroscopic methodology is described which meets these demands. It is demonstrated that the applied methodology may be developed on a cheap, packaging grade of Polylactic Acid (PLLA), an important bioresorbable polymer for the medical device industry, and still be accurate when implemented on a production line using a more expensive (medical) grade of the polymer compound. Simple chemometric algorithms are applied allowing the data processing step to be carried out in near real time, thus providing vital information to process operators which allows any out of control process to be identified and rectified without product loss.

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

“…As described above, a few studies have been reported in which the transferability of machine learning models was studied in HME processes. Whitaker et al [ 31 ] studied the transferability of PLS and LDA calibration models for monitoring filler particle size in different grades of PLA (packaging and medical grade). They achieved good model transferability, attributed to the similar optical properties of the melt in each case.…”

“…Whitaker et al [ 31 ] used PLS and linear discriminant analysis (LDA) (a classification algorithm) with UV–Vis spectroscopy to monitor the particle size (D50) of β-TCP (beta-tricalcium phosphate (Ca 3 (PO 4 ) 2 ) extruded in a packaging grade PLA. Two different particle sizes of β-TCP (5 and 30 µm) were used and processed separately.…”

“…The main aim is to improve the understanding of the mechanism of the manufacturing process, enhance process monitoring, and to reduce the processing time. In the literature, spectroscopic techniques, including Raman [ 2 ], NIR [ 30 ], and UV–Vis spectroscopy [ 31 ], have been widely implemented as PAT tools for in/on-line monitoring of the HME process, and an in-line slit die rheometer has also been implemented as a PAT tool in some studies [ 32 , 33 ]. Machine learning (ML) algorithms are generally used to infer the required information from in/on-line collected spectra.…”

“…Since 2004, many research studies have been reported, in which different machine learning algorithms have been applied to in-process data to analyse product and process parameters in real-time. The applications include the monitoring of product critical quality attributes (CQAs), including the following: the solid state of the polymer/drug [ 2 , 3 ]; API/additive concentration [ 34 , 35 ]; degradation of the polymer [ 36 , 37 ]; the particle size of additive/s [ 31 ]; and mechanical properties [ 33 ]. Other works have examined the monitoring of critical process properties (CPPs), including melt temperature [ 38 ], pressure [ 39 ], and viscosity [ 40 ], and for process fault detection [ 41 ].…”

In‐line measurement and characterization in glass bead‐filled polypropylene extrusion based on machine vision

“…[ 13 ] Whitaker et al used UV–VIS spectroscopy in combination with a chemometric algorithm to measure the particle size of additive particles in polymer extrusion. [ 14 ] Becker et al used a spectroscopic sensor and turbidimetric measurement to determine the size distribution of clay particles in polypropylene/clay blends, and the width and median particle radius of the particle size distribution function agreed with the result measured by SEM. [ 15 ] Barnes et al used FT‐NIR to determine the inorganic filler concentration in polymer melts during extrusion with the error of 0.27 wt%.…”