(6jj) Active Process Control in Pharmaceutical Continuous Manufacturing — the Quality By Control (QbC) Paradigm

Su, Q., Purdue University
Research Interests:

Pharmaceutical manufacturing is traditionally operated in batch mode by processing a certain amount of raw material, defined as a batch or lot, through various steps of unit operations, such as reaction, crystallization, filtration, drying, blending, tableting, etc., to obtain final drug products, for example., oral solid dosages. Quality attributes of in-process material or final product are tested batch wise at the end of each processing step, or known as the Quality by Testing (QbT)1. One example is in the blending step, where powders of active ingredients and excipients are mixed together in a rotating Y- or V-shaped vessel before the mixing uniformity end-point is detected at the batch end 2. Although various process monitoring and control strategies are proposed for batch manufacturing in pharmaceutical or other manufacturing industries to achieve the batch-end product quality, for example, the multiway partial least squares for batch process monitoring3, 4, a shrinking horizon nonlinear model predictive control (NMPC) for batch-end product quality control5, etc., it should be noted that many of them are relied on the known priori optimal or nominal process operation trajectories (or recipes). It is the optimizing and tracking of these operational trajectories during the batch that lead to the batch-end product quality. Rarely, rework or recycle of the material at the batch end, particularly at the secondary manufacturing process, is allowed in the highly regulated pharmaceutical manufacturing industry. Hence, the loss of a whole batch is possible after the end-testing and remedial control actions can only start from the next batch. In other words, there is a large time delay of the entire batch time in controlling the batch-end product qualities by mere end-testing, which is usually known as the batch-to-batch control strategy6. Over the years, the US Food and Drug Administration (FDA) recognized that increased testing does not necessarily improve product quality and quality must be built into the product7.

Quality by Design (QbD) is a concept first developed by the quality pioneer Dr. Joseph M. Juran, who suggests that quality should be designed into a product, and that most quality crises and problems relate to the way in which a product was designed in the first place8. Over the years, pharmaceutical QbD has evolved with the issuance of ICH Q8 (R2) (Pharmaceutical Development), ICH Q9 (Quality Risk Management), ICH Q10 (Pharmaceutical Quality System), and ICH Q11 (Development and Manufacture of Drug Substance). These documents provide high level directions with respect to the scope and definitions of QbD as it applies to the pharmaceutical industry7. QbD elements include the following: (1) a quality target product profile (QTPP) that identifies the critical quality attributes (CQAs) of the drug product; (2) product design and understanding including identification of critical material attributes (CMAs); (3) process design and understanding including identification of critical process parameters (CPPs), linking CMAs and CPPs to CQAs; (4) a control strategy that includes specifications for the drug substance(s), excipient(s), and drug products as well as controls for each step of the manufacturing process; and (5) process capability and continual improvement. QbD tools and studies include priori knowledge, risk assessment, mechanistic models, design of experiments (DoE) and data analysis, and process analytical technology (PAT).

Though the QbD principles are equally implemented in the conventional pharmaceutical batch manufacturing9, at the meantime of the QbD issuance, there has been significant advancements in science and engineering to support the implementation of pharmaceutical continuous manufacturing over the past decade10. The conventional batch manufacturing faces many challenges in scaling-up, manufacturing cost, product quality variance, etc.; while the continuous manufacturing, defined as processing of raw materials without interruption and with continuity of production over a sustained period of time, offers many benefits in minimization of uncertainty in scaling-up, reduced space and capital requirement, and improved quality with consistent operations, etc. 11. A great deal of potentials of improved agility, flexibility, and robustness in the continuous manufacture of pharmaceuticals have also been demonstrated. These investments along with the adoption of the QbD paradigm for pharmaceutical development and the advancement of PAT for designing, analyzing, and controlling manufacturing have progressed the scientific and regulatory readiness for continuous manufacturing12. For example, pharmaceutical continuous manufacturing has already been identified as an emerging technology by FDA and it is now staging from conceptual designs to pilot or production processes. Recently, a few continuous manufacturing facilities and drug have also been approved by FDA, e.g., the Orkambi (lumacaftor/ivacaftor) from Vertex in 2015 and the darunavir from Janssen in 2016 13.

Beside the flexibility and cost-saving features with pharmaceutical continuous manufacturing, a unique advantage of continuous manufacturing compared to the batch manufacturing when both under the QbD guidance is that those identified CQAs and CPPs can be continuously monitored and controlled in real time, which therefore paves the way for a superior real-time product quality assurance, viz., the real-time release (RTR) strategy. For example, in batch blending, CQA variable of final mixing uniformity can only be available at the end of each batch blending step, even though this CQA variable is also continuously measured during the batch. Whereas this final CQA can be continuously measured at the outlet of a continuous blender and sent back to the control system as feedback signals to support a closed-loop control of this CQA variable by adjusting CPP variables, e.g., the blender rotation speed. In such, it is acknowledged that continuous manufacturing, integrated with online/inline-PAT tools and efficient control systems, can accelerate the comprehensive implementation of the QbD principles for the next generation of pharmaceutical products14, 7. It is also worth mentioning that active process control strategies are not new in other continuous manufacturing processes, e.g., in bulk chemicals and petrochemical industries. However, despite the feasibility of implementing active process control, which has been demonstrated for a number of pharmaceutical continuous manufacturing facilities, e.g., continuous crystallization, continuous direct compaction, etc., there remains hesitation in widespread adoption of active process control strategies due to existing investments in established and mature batch technologies15. Besides, most of the pharmaceutical continuous unit operations are still operated locally with very limited flexibility to implement the much more advanced active process control system.

Hence, owing to the substantial change in the control strategy for QbD implementation in pharmaceutical batch and continuous manufacturing on the one hand, and the rare application of active process control system in the pharmaceutical industry on the other hand, a Quality by Control (QbC) paradigm has since been evolving and nurturing the implementation of efficient advanced active process control of the CPPs and CQAs in continuous manufacturing. It has also been shown recently that the QbC is indispensable to the QbD implementation in pharmaceutical continuous manufacturing, and that it ensures more robustness and efficiency than the techniques used in conventional batch manufacturing, providing the key ingredient towards the comprehensive implementation of QbD16.

To this end, my three tentative research plans in pharmaceutical continuous manufacturing and crystallization are listed as follows.

  1. On-site crystallization process monitoring, modeling, and control

Crystallization is an important unit operation for separation and purification of pharmaceutical and fine chemicals. With the advent of in situ real-time process analytical technologies in the past decades, e.g., Raman spectroscopy, Focused Beam Reflectance Measurement (FBRM), or the Attenuated Total Reflection Fourier-transform infrared spectroscopy (FTIR), process knowledge on the crystallization kinetics of nucleation and crystal growth has been advanced unprecedentedly by using mathematical modelling tools of population balance equations, mass balance equations, etc. Specifically, the software package of gCRYSTAL from Process Systems Enterprise Ltd. (PSE), has initiated the development of general and convenient mathematical tools for the crystallization community, compared to the most of in-house developed codes or toolkits in the academic sector.

It should be noted that there is an obvious gap usually observed between the experimental and modelling works in crystallization, viz., the pale description and explanation of experimental observations without insights in crystallization kinetics and dynamics, or the plausible accuracy and reliability of theoretical predictions without strength of proofs in crystallization experiments. One of the examples is in the use of FBRM which measures the chord length distribution (not exactly the crystal size distribution), experimental authors believe the gradual increase in total counts represents the occurrence of secondary nucleation, while the modelling authors may believe the growth of crystal size and morphology changes accountable. One of the root causes for the gap is that mathematical tools are not readily available on site of the experimental set-ups. For example, the gCRYSTAL modelling package is not yet efficiently integrated with those in situ PAT sensors, and those house-made toolkits, e.g., the CryPIN from Loughborough University or CryMOCO from Purdue University, are lack of powerful mathematical solutions.

To equip researchers with both experimental and mathematical expertise, it is proposed here an integrated computational platform for crystallization communication, monitoring, modelling, and control, i.e., communicating with the experimental instruments, statistical monitoring of the measurements, mathematical modelling of crystallization process, and implementing advanced model-based control strategies to the process. The platform would stand out with many innovative features such as using industrial standard OPC servers for instrument communication, on-site modelling, moving-window kinetic parameters and state variables estimation of crystallization process, robust product quality control with nonlinear model-predictive control strategies. The proposed platform would surely support the practise of FDA Quality by Design (QbD) guidance which emphasizes the process knowledge in product quality control.

Dr. Su has solid research background in crystallization process modelling and control with working experience in UK EPSRC Continuous Manufacturing and Crystallization (CMAC) center and under Prof. Zoltan Nagy at Purdue University, who is a distinguished contributor to the crystallization process control. Dr. Su is also a fluent user and developer in gCRYSTAL software, whose work involving gCRYSTAL was awarded the PSE Model-Based Innovation Award in 2017. Dr. Su is also responsible for the plant-wide control strategy development in a research pilot plant for pharmaceutical continuous manufacturing at Purdue University and has the problem solving-skills in Emerson DeltaV OPC server for control strategy development. The proposed platform would combine his unique professional knowledge and supports from industrial and academic partners to promote the development and implement of process system engineering tools in crystallization community, narrowing the gap between experimental and modelling practise, and advancing the research in crystallization towards a sound scientific understanding.

  1. Process-oriented high reliable Process analytical technology (PAT) sensors for pharmaceutical continuous manufacturing

Decade after the US Food and Drug Administration (FDA)’s Process Analytical Technology initiative, its implementation and advantages of monitoring and controlling of critical quality attributes (CQAs) in pharmaceutical manufacturing processes are world-wide recognized in industry and academy. For example, the Near Infrared (NIR) spectroscopy has been extensively reported in monitoring several CQA variables, e.g., active pharmaceutical ingredient (API) mass fraction in formulated powder, ribbon density after roller compaction, or the water content in wet granulation. Worth mentioning is that the pharmaceutical industry is now undergoing a mindset change from batch to continuous manufacturing, which depends more and more on the high reliability and availability of in situ PAT sensors to maintain product quality in real time.

However, the acute operating conditions of phase separation, e.g., crystallization, or the involvement of fine cohesive powders, e.g., API, has led the fouling one of the challenging issues in many PAT sensor applications. Furthermore, besides the chemometric model calibration for most of the PAT sensors based on spectroscopy, common raw spectra data pretreatment methods, e.g., Savitzky-Golay filter, Extended Multiplicative Scatter Correction, are inefficient in handling environmental uncertainties, raw material property variations, or process disturbances, e.g., humidity, powder flowrate, particle size distribution, etc. For example, the NIR spectroscopy collects the complex diffusions that are sensitive to both chemical compositions and physical properties, making a reliable NIR sensor especially challenging. Specifically, though the NIR sensing has been acknowledged as a convenient spectroscopy method without the need of sample pretreatment, except the various in-house designs of NIR boxes or flow chutes in academic research centers, there is a limited number of reports in open literature in terms of reliable application of NIR sensors in the pharmaceutical industry.

Hence, a complete process engineering solution to the secondary development of most commercially available spectroscopic probes are indeed necessary to produce reliable PAT sensors. It is proposed here that systematic design and risk analysis of PAT sensors, including sensor selection, sampling automation, sensor placement, chemometric calibration, model maintenance, sensor redundancy and network design, data reconciliation, are critically important for a high reliable PAT sensor to keep a good track of CQAs in pharmaceutical manufacturing. Unlike the previous works that are interested in the demonstration of sensing mechanisms, the proposed systematic design and risk analysis are process oriented and are focused more on the practical engineering solutions for reliable PAT sensor design, and should have a great impact on the industrial applications of PAT sensors. The proposed work also has the potentials for several patent applications.

Dr. Su has been financially supported by an US FDA funded project (grant number DHHS-FDA U01FD005535-01) on real-time release in continuous solid dose manufacturing: systematic characterization of material properties, and optimal design of sensing and control methods. As a main contributor to the risk-based monitoring and control strategies in this project, Dr. Su has been an extensive user of PAT sensors in the pilot plant at Purdue University and has a profound understanding and personal experience of the challenges in the sensor design.

  1. Intensified process integration for pharmaceutical continuous manufacturing

Advantages of continuous manufacturing in the pharmaceutical industry have been demonstrated through extensive research studies in the last two decades, but mostly at the conceptual or theoretical level. Only few continuous manufacturing processes have been approved by the US Food and Drug Administration thus far, e.g., the Vertex Orkambi and Janssen Darunavir. To move from a conceptual continuous design to a practically operating pilot plant or manufacturing process, the efforts of regulators, academics, and industry are now focused more on process integration. For example, unit operations that are operated in batch mode, e.g., crystallization, filtration, drying, blending, granulation, tableting, etc., are to be integrated together and operated in continuous streamline mode, i.e., the end-to-end manufacturing as showcased in Novartis-MIT center for continuous manufacturing.

Advantages of continuous operation with the above process integration include the increased production capacity by running continuously and the reduced product variation by using steady-state operation. Nonetheless, it should be pointed out that pharmaceutical continuous manufacturing is not limited to this simple process integration, it also opens up another avenue of research towards a much more intensified process integration. For example, the wet granulation can be intensified by using a twin-screw extruder that integrates unit operations of feeding, blending, granulation. Another example is the spherical crystallization by which API molecule either crystallizes on porous excipient polymer materials or simultaneously crystallizes and aggregates to form an agglomerate. These spherical API particles could be directly tableted, shortening almost the whole secondary manufacturing process. Furthermore, other innovative manufacturing technologies are also emerging, e.g., the scaled-down drop-on-demand manufacturing for individualized medicines, which is prototyped and started-up at Purdue University at Professor G.V. Reklaitis’s research group.

Therefore, it is envisioned that an intensified process integration for pharmaceutical continuous manufacturing, specifically a miniaturized and individualized manufacturing, could play an important role in the current trend of Industry 4.0 for smart manufacturing. Dr. Su has research experience covering the whole pharmaceutical secondary manufacturing process from crystallization to the end of tableting and was also involved in several discussions on the Mini Pharm project at Purdue University and is interested in setting up an intensified “smart factory” in the future, specialized in biomolecular or protein drugs.

Teaching Interests:

  1. Undergraduate level

Teaching at undergraduate level would require the experience and capability to interpret the rigid scientific terminologies in understandable language, which is usually a challenge for new lecturers. Being trained in Essential Teaching Skills and as a teaching assistant for two chemical engineering course modules in Loughborough University (7th in 2018 Times Good University Guide) in UK prepared me the first experience of this journey. Also, having laid a solid foundation in Chemical Engineering, especially through the advanced course modules in National University of Singapore (8th in 2017 QS world university ranking for Chemical Engineering), I would feel comfortable teaching across a broad range of subjects at undergraduate level. Furthermore, I strongly believe that mathematical tools are the best scientific language to verify the understanding of complex and dynamic chemical and biochemical processes. My background in Process Systems Engineering (PSE) has helped me gain a wide spectrum of knowledge on process modelling, simulation, control, and optimization. Hence, teaching for undergraduate course modules related to Process Dynamics and Control would be a preference for me, covering general topics of introduction to process control, chemical process modelling, Laplace transform, transfer functions, and control design, etc.

Reference books of “Process Dynamics and Control” by Dale E. Seborg, Duncan A. Mellichamp, and Thomas F. Edgar or “Process Control: Modelling, Design, and Simulation” by B. Wayne Bequette are recommended.

Besides, it is also a good practise to have undergraduate students involved in ongoing research projects as helpers to senior PhD students to deepen their understanding in course modules and to have hands-on experience in research and engineering practise. As a faculty project advisor at Purdue University, I have the experience of supervising two undergraduate students in crystallization process control and NIR sensor design. I feel this experience help me a lot on how to communicate and learn with the new generations.

  1. Postgraduate level

The greatest advantage of being a frontline researcher is our fresh first-hand experience on tackling the technical difficulties in research and better understanding of the expectations from the young Master or PhD students. For example, the experience of developing the state-of-the-art high resolution finite volume method for population balance model, solving complex integral partial derivative and algebraic equations (IPDAEs), multi-objective optimization using Non-Dominated Sorting Genetic Algorithm (NSGA-II), implementing the nonlinear model predictive control (NMPC), etc., would make me confident not only to help postgraduate students understand those methodologies, but also to encourage them to explore many open problems in chemical engineering.

As a lecturer for the course modules in the Doctoral Training Centre (DTC) in UK EPSRC Continuous Manufacturing and Crystallization (CMAC) research center, I have had the chances to experience the teaching of specific advanced topics on population balance modelling for crystallization, modelling and optimization of batch and continuous crystallization using MATLAB and gCRYSTAL software, etc. Hence, depending on the departmental research areas, course modules at postgraduate level may focus on specific topics, e.g., separation and purification using crystallization, crystallization process modelling and control, crystallization process optimization and control, or the advanced process control, etc.

Reference books of “Handbook of Industrial Crystallization” by Allan S. Myerson and “Crystallization: Basic Concepts and Industrial Applications” edited by Wolfgang Beckmann are recommended. Besides, tutorial questions adapted from current research publications would be used to encourage postgraduate students to have comprehensive and critical literature-reading and problem-solving skills.



Hubert C, Lebrun P, Houari S, Ziemons E, Rozet E, Hubert P. Improvement of a stability-indicating method by Quality-by-Design versus Quality-by-Testing: a case of a learning process. J. Pharm. Biomed. Anal. 2014;88:401-409.


Igne B, Juan AD, Jaumot J, et al. Modeling strategies for pharmaceutical blend monitoring and end-point determination by near-infrared spectroscopy. International Journal of Pharmaceutics. 2014;473(1-2):219-231.


Nomikos P, MacGregor JF. Multi-way partial least squares in monitoring batch processes. Chemometr. Intell. Lab. Syst. 1995;30:97-108.


Su QL, Chiu MS. Monitoring pH-shift reactive crystallization of L-Glutamic acid using moving window MPCA. Journal of Chemical Engineering of Japan. 2016;49(7):680-688.


Su Q, Hermanto M, Braatz R, Chiu M. Just-in-Time-Learning based extended prediction self-adaptive control for batch processes. Journal of Process Control. 2016;43:1-9.


Mockus L, Peterson JJ, Lainez JM, Reklaitis GV. Batch-to-Batch variations: a key compoenet for modeling chemical manufacturing processes. Organic Process Research & Design. 2015;19:908-914.


Yu LX, Amidon G, Khan MA, et al. Understanding pharmaceutical quality by design. The AAPS Journal. 2014;16(4):771-783.


Juran JM. Juran on quality by design: the new steps for planning quality into goods and services. New York: The Free Press; 1992.


Potter C. PQLI application of science- and risk-based approaches (ICH Q8, Q9, and Q10) to exisiting products. Journal of Pharmaceutical Innovation. 2009;4(1):4-23.


Ierapetritou M, Muzzio F, Reklaitis G. Perspectives on the continuous manufacturing of powder-based pharmaceutical processes. AIChE Journal. 2016;62(6):1846-1862.


Pernenkil L. Continuous blending of dry pharmaceutical powders. Boston, MA 2008. PhD Thesis.


Lee SL, O'Connor TF, Yang X, et al. Modernizing pharmaceutical manufacturing: from batch to continuous production. Journal of Pharmaceutical Innovation. 2015;10(3):191-199.


Yu L. Continuous Manufacturing Has a Strong Impact on Drug Quality. FDA Voice. April 12, 2016. https://blogs.fda.gov/fdavoice/index.php/2016/04/continuous-manufacturing-has-a-strong-impact-on-drug-quality/. Accessed April 11, 2017.


Singh R, Godfrey A, Gregertsen B, et al. Systematic substrate adoption methodology (SAM) for future flexible, generic pharmaceutical production process. Computer and Chemical Engineering. 2013;58:344-368.


Diab S, Gerogiorgis DI. Process modelling, simulation and technoeconomic evaluation of crystallization antisolvents for the continuous pharmaceutical manufacturing of rufinamide. Computers and Chemical Engineering. 2018;111:102-114.


Koswara A, Nagy ZK. On-Off feedback control of plug-flow crystallization: a case of Quality-by-Control in continuous manufacturing. IEEE Life Sciences Letters. 2017;3(1):1-4.