A new pharmaceutical manufacturing study is highlighting a familiar challenge in Operational Research: how to connect sophisticated models with live operational data quickly enough to support real-time decision-making.
The research focuses on continuous filtration, washing, and drying in pharmaceutical production, a downstream stage that plays a major role in determining whether crystalline active ingredients leave manufacturing with the correct moisture content, purity, and handling characteristics. While pharmaceutical continuous manufacturing has advanced significantly in recent years, researchers argue that downstream separation and isolation processes remain difficult to control consistently under changing operational conditions.
The broader issue is one of system integration. In continuous pharmaceutical manufacturing, performance depends not just on optimising individual process stages independently, but on understanding how upstream crystallisation behaviour affects downstream filtration and drying performance across the entire production chain.
Previous work in this area has already demonstrated the value of mechanistic simulation models. Researchers behind the ContCarSim benchmark simulator showed how intensified filtration-drying systems could be tested virtually before deployment, allowing both design and control strategies to be evaluated in advance. Separate modelling work using the PharmaPy framework has also demonstrated how integrated crystallisation, filtration, washing, and drying models can help define viable operating spaces for continuous production systems.
What distinguishes the latest study is its attempt to bridge the gap between simulation and live plant operation through the use of a digital twin connected directly to real-time sensor data.
Built using PharmaPy and linked to industrial communication systems, the digital twin combines first-principles modelling with in-line mass spectrometry measurements. The system continuously updates its internal parameters using moving horizon estimation, allowing the virtual model to remain aligned with the physical process as operating conditions change.
In trials involving an aspirin continuous filtration-drying carousel, the approach reportedly tracked residual moisture with high accuracy and used that information to improve control of drying time through a PID control loop. According to the researchers, this reduced the risk of over-drying or under-drying the final crystalline product while improving operational stability.
For Operational Research practitioners, the study reflects a growing trend toward integrated operational decision systems where simulation, optimisation, sensing, and control are increasingly combined into a single adaptive framework.
The work also reinforces a broader shift taking place across manufacturing industries. Rather than relying solely on static process models or offline optimisation, organisations are increasingly exploring digital twins, embedded analytics, and closed-loop control systems capable of supporting operational decisions in real time.
Industry groups involved in continuous pharmaceutical manufacturing have increasingly promoted hybrid digital-twin approaches as a way to improve process robustness, accelerate technology transfer, and reduce waste. What makes this latest work particularly notable is its focus on operational deployment rather than theoretical modelling alone.
If the approach proves scalable, it could help continuous manufacturing systems operate less conservatively while improving efficiency, consistency, and responsiveness across pharmaceutical production environments.
For the OR community, the study offers another example of how systems thinking, dynamic modelling, and real-time decision support are becoming central to the management of increasingly complex industrial operations.
References
https://aiche.onlinelibrary.wiley.com/doi/10.1002/aic.70291?af=R
https://www.sciencedirect.com/science/article/abs/pii/S0098135422001478