Understanding powder blending mechanics for applications in pharmaceutical continuous direct compression

Jones-Salkey, Owen ORCID: 0009-0002-7931-5955 (2025). Understanding powder blending mechanics for applications in pharmaceutical continuous direct compression. University of Birmingham. Eng.D.

Jones-Salkey2025EngD.pdf Text - Accepted Version Available under License All rights reserved. Download (23MB)

Abstract

This thesis presents a holistic example of accelerating pharmaceutical process development by integrating experimental platforms, user-centric process design, and data-driven modelling. At the heart of this approach lies the modular process optimisation suite, the MACH-1, a Good Laboratory Standard (GLS) equipment suite operating in parallel with a fully fledged Good Manufacturing Practice (GMP) development facility. This dual setup not only expands the experimental capacity but also allows project-specific work packages to seamlessly integrate into AstraZeneca’s established workflows for oral product development. Beyond its on-site utility, the GLS suite has supported off-site mechanistic studies, highlighting its flexibility for delivering diverse R&D needs.

Accordingly, Positron Emission Particle Tracking (PEPT), a unique noninvasive measurement technique, was leveraged to characterise steady-state powder behaviours in con�tinuous blending. Through the tracking of discrete particle spatial-temporal data, the study revealed how adjusting incline linear blender parameters, such as rotational speed and blade configuration, can enhance content uniformity and even induce a ’folding region’, where powder is transported away from and returned to a bed of bulk mass residing at the base of the inclined blender, in a convection-like cycle. These insights also provide the means for calibrating and validating future Discrete Element Method (DEM) simulations.

Building on these experimental insights, the relationship between product formulation and process conditions was modelled using three distinct AI/ML approaches: Random Forest Regression, Symbolic Regression, and Artificial Neural Network (ANN). When predicting fill levels in inclined linear blenders, the ANN outperformed other methods, achieving an r2 of 0.97 and generalising effectively to multicomponent commercial formulations. By reducing experimental trials by up to 60–70%, these models substantially cut material use and development time.

An additional innovation is the digital layer integrated into the MACH-1 workflow: a democratised ANN for real-time assessment of blending quality. This user-friendly platform, accessible via a hyperlink on the intranet, empowers scientists and engineers to refine process parameters independently, minimising the reliance on specialised expertise. Consequently, the interface increases both operational agility and the broader adoption of advanced analytics.

This holistic approach ultimately yielded the following benefits per experimental campaign:

• Reduced experimentation: Up to a 60–70% reduction in the number of experiments, significantly conserving high-cost APIs.

• Cost efficiency: Decreasing the need for expensive materials translates into a projected savings of £112k.

• Environmental sustainability: Cutting down on experimental runs reduces carbon emissions by almost 6 tonnes of CO2.

• Accelerated timelines: Replacing multi-day in-vitro trials with minutes long in silico studies, facilitated by real-time, democratised access to advanced modelling tools.

In summary, this thesis illustrates the core objective of an industrially based Engineering Doctoral programme– conducting research that not only advances the broader scientific field, but also directly serves the needs of AstraZeneca. Through the fusion of modular process design and digital transformation, it advances process understanding, sustainability, innovation, and cost efficacy. In so doing, it demonstrates that, despite decades of brilliant work, there is a growing market for opportunities to make drug product development more environmentally and economically responsible

Type of Work:	Thesis (Doctorates > Eng.D.)
Award Type:	Doctorates > Eng.D.
Supervisor(s):	Supervisor(s) Email ORCID Windows-Yule, Christopher UNSPECIFIED orcid.org/0000-0003-1305-537X Ingram, Andrew UNSPECIFIED UNSPECIFIED Clifford, Sean UNSPECIFIED UNSPECIFIED Reynolds, Gavin UNSPECIFIED UNSPECIFIED
Licence:	All rights reserved
College/Faculty:	Colleges > College of Engineering & Physical Sciences
School or Department:	School of Chemical Engineering
Funders:	Engineering and Physical Sciences Research Council
Subjects:	T Technology > TA Engineering (General). Civil engineering (General)
URI:	http://etheses.bham.ac.uk/id/eprint/16951

Actions

	Request a Correction
	View Item

Downloads