The integration of predictive modeling into bioprocess automation and control systems is transforming the biopharmaceutical industry, enabling enhanced performance and precision. By combining mechanistic models — such as hydrodynamic and cell kinetic equations — with data-driven insights and predictive algorithms, digital twins provide real-time simulations of cell growth, nutrient consumption, hydrodynamic behavior, and other critical variables. Central to this innovation is computational fluid dynamics (CFD), which serves as a foundation for developing effective digital twins. Samsung Biologics is at the forefront of this technological evolution, leveraging CFD and digital twin technologies to achieve unparalleled quality management. Harnessing digital twin technology, Samsung Biologics optimizes process development, streamlines technology transfers, and evaluates process changes and scale-up manufacturing operations.
Driving Precision in Biopharmaceutical Manufacturing
Biopharmaceutical production relies on complex living cell factories, which is why rigorous process control is crucial in ensuring consistent performance. The automation of process monitoring and control systems plays a pivotal role in maintaining optimal conditions during cell growth and product expression. By minimizing manual interventions, automation not only enhances operational consistency but also mitigates the risks associated with human error, leading to higher yields and improved product quality.
The integration of predictive modeling into these automated systems takes efficiency to the next level. Bioprocess digital twins (BPDTs) — virtual replicas of biopharmaceutical processes—empower manufacturers to analyze both historical and real-time data and enable proactive optimization by leveraging their predictive capability. These advanced models offer actionable insights that help maintain optimal parameters during production, pinpoint trends that require attention, identify opportunities for process optimization, and establish ideal conditions for production.
A fundamental aspect of creating effective digital twins is understanding how process mixtures behave within a bioreactor during cell culture. CFD provides critical insights into this behavior, modeling variables such as mixing efficiency, oxygen transfer, and nutrient distribution. By enabling the precise simulation of these parameters, CFD significantly accelerates facility-fit analysis and process development timelines. Consequently, careful planning and execution of CFD modeling are essential for achieving the full potential of digital transformation in bioprocessing operations.
Transforming Biomanufacturing With Digital Twin Technology
A BPDT is an advanced virtual representation of a physical bioprocess — be it a single bioreactor or an entire production line. Continuously synchronized with real-time data from sensors and control systems, the BPDT mirrors actual conditions and captures dynamic changes. By incorporating mechanistic models (such as hydrodynamic and cell kinetic equations), data-driven models, and predictive algorithms, digital twins simulate critical variables, including cell growth, nutrient consumption, metabolite secretion, and hydrodynamic performance.
The benefits of BPDTs are profound in biologic drug development and manufacturing. For instance, in silico development enables manufacturers to simulate, test, and optimize processes virtually before performing any physical experiments. This capability reduces the time, resources, and costs required during process development and optimization. At the production level, real-time data integration with process analytical technologies allows BPDTs to diagnose process deviations immediately, enabling faster troubleshooting and minimizing disruptions.
Leveraging BPDTs to forecast cell growth, key process attributes, and critical quality attributes proactively optimizes process parameters by guiding process manipulation to return to its normal state via predictive simulation. This reduces batch variability, safeguards compliance with quality standards, and boosts overall product reliability.
Samsung Biologics’ Vision for Digital Twin Technology
Samsung Biologics is pioneering the development of BPDTs to revolutionize daily plant operations and mitigate the inherent risks of bioprocessing. By integrating advanced sensors, connected devices, and a hybrid modeling approach, Samsung Biologics is creating a powerful predictive tool. This hybrid model combines state-of-the-art CFD with mechanistic models such as cell kinetics alongside advanced technologies such as artificial intelligence, machine learning, and other deep learning systems. The result is a sophisticated platform designed to enhance bioprocess monitoring and control to deliver highly accurate predictions.
Unlike traditional digital twins, which rely primarily on data-driven models for rapid prediction, Samsung Biologics’ approach goes further. By incorporating mechanistic modeling, such as hydrodynamic equations and cell-specific growth kinetics, into their BPDT, the company improves predictive accuracy and ensures tailored insights for individual customer products. This mechanistic model-based approach enables Samsung Biologics to build bespoke digital twins for each client, reflecting the unique growth and performance characteristics of their bioprocesses (Figure 1).
Figure 1. Schematic of Digital Twin Technology Used by Samsung Biologics
Decoding Fluid Dynamics for Bioprocess Excellence
Mechanistic modeling using CFD is a cornerstone of effective BPDT development. The behaviors of liquid and gas flows within bioreactors directly impact both process performance and product quality. CFD provides the capability to analyze intricate fluid dynamics and mass/heat transfer phenomena by accounting for parameters like temperature, pressure, velocity, and density. This is a complex task given the simultaneous, nonlinear interactions among cells, cell-derived materials, fluid motion, and bioreactor geometry.
CFD software is grounded in primary governing equations (e.g., the Navier–Stokes equation) that define interactions among various forces on the fluid based on the conservation of mass, momentum, and energy. For instance, equations that describe changes in fluid velocity incorporate factors such as pressure, density, viscosity, and external forces that act on the fluid. By applying known initial and boundary conditions, CFD can precisely calculate fluid velocity and other vital aspects of fluid behaviors.
In the bioprocessing industry, CFD-based mechanistic models offer unparalleled accuracy in predicting process performance, eliminating the need for extensive physical experimentation. These models are versatile, finding applications in a wide range of facilities and equipment that involve fluid dynamics, including bioreactors, process vessels, chromatography and ultrafiltration/diafiltration systems, and piping networks (Figure 2). Additionally, CFD can function as a virtual sensor, providing insights into the volumetric oxygen transfer coefficient (kʟa), mixing times, flow rates, gaseous concentrations, and more, thereby enabling enhanced control and optimizing bioprocesses (Figure 3).
Figure 2. Facility and Equipment Implementing Computational Fluid Dynamics
Figure 3. Timeline for Computational Fluid Dynamics Utilization at Each Facility
The CFD Advantage at Samsung Biologics
Samsung Biologics employs CFD across a wide range of applications, from analyzing fluid dynamic properties of cell culture media to advancing digital twin development. CFD plays a critical role in evaluating the potential impacts of client-requested process changes, enabling robust risk assessments and the identification of unexpected issues through detailed what-if scenario analysis. These applications include addressing process volume increases, modifying mixing times, integrating new processes via tech transfer, and optimizing scale-up strategies. Moreover, CFD is pivotal in diagnosing and resolving pipe design issues. Insights from these analyses are systematically captured, guiding the redesign of equipment and facilities to prevent recurrence and bolster operational efficiency.
The CFD workflow begins with the construction of an accurate digital representation of the system geometry. This can be achieved through 3D scanning of physical assets or computer-aided design to digitally replicate the structure. The system is then discretized into a set of control volumes through meshing, where finer mesh resolutions improve accuracy but require greater computational power. Next, a selection of partial differential equations is applied to represent the physical conditions of the system. These equations may include the Navier–Stokes equation-derived turbulence models, multiphase flow equations, and mass transfer equations. The system’s behavior is simulated, generating detailed insights into fluid interactions (Figure 4).
Figure 4. Computational Fluid Dynamics Model Workflow
The precision of CFD modeling hinges on the accuracy of the geometry and the complexity of the equations employed. Post-processing visualization is a key component of this analysis, offering valuable insight into the fluid environment and highlighting opportunities for process optimization. CFD outputs can include fluid flow patterns, shear stress, turbulence indices, mixing times, heat transfer, cell distribution, kʟa, and power consumption, among others.
Among these parameters, kʟa is particularly influential. Statistical models based on probe measurements often fail to capture the complete distribution of gas bubbles following process changes, as probe readings only reflect local conditions. This limitation can mask gradients, hypoxia risks, and areas of excessive shear stress. CFD provides a more comprehensive analysis, enabling precise predictions of oxygen transfer dynamics within bioreactors.
For instance, CFD contour graphs reveal that high local kʟa values occur near the impeller and air inlet due to smaller bubbles (as indicated by the Sauter mean diameter, d32) in these regions. Turbulence indices, such as turbulent kinetic energy and eddy dissipation rates, alongside shear strain rates, also significantly affect oxygen gradients within production bioreactors.
Scale-up processes also benefit from the precision of CFD. Traditional scaling criteria, such as power consumption ratios, often fail to predict bubble distribution, particularly for sensitive cells. By leveraging CFD-based kʟa values, Samsung Biologics is establishing refined scale-up criteria that improve performance consistency and robustness across bioprocesses. Mapping the fluid dynamic characteristics of all bioreactors at Samsung Biologics based on specific parameters, such as kʟa, further enhances the utility of CFD. This enables detailed facility-fit analyses during tech transfers, particularly in terms of fluid dynamics.
Enhancing Digital Twin Development With CFD
To develop effective digital twins, CFD data must accurately replicate the hydrodynamic characteristics and behavior of actual bioprocess. Achieving this level of precision often requires leveraging a high number of models and incorporating extensive data to maximize explanatory power.
Samsung Biologics employs a data-driven approach to digital twin development by utilizing a hybrid modeling strategy that integrates CFD-based mechanistic hydrodynamic models. These models rely on partial differential equations related to fluid dynamics to accurately describe fluid behavior.
The CFD model is designed as a reduced-order model, enabling real-time process data integration (e.g., agitation and aeration rate, volume) to generate hydrodynamic virtual sensing for kʟa, shear stress, and other fluid behaviors in real-time. These feed into a hybrid digital twin model combining fluid dynamics and process variables to deliver highly accurate predictions of cell growth and product quality. By more accurately predicting product quality and titer at harvest, Samsung Biologics can proactively identify and address potential deviations before they occur.
Moreover, BPDTs excel in mitigating the risks associated with scaling up operations or transferring production sites. By simulating variables such as mixing efficiency and oxygen transfer rates, BPDTs deliver accurate, physics-based predictions. These simulations minimize uncertainty and ensure smooth transitions across varying operational scales and locations.
Navigating Regulatory Compliance in CFD and Digital Twin Development
Currently, no formal regulatory standards exist for CFD models or digital twins in the biopharmaceutical industry. However, the U.S. Food and Drug Administration (FDA) encourages the use of predictive models within a risk-informed framework to assess and mitigate associated risks. The agency also regularly engages with industry experts to promote the adoption of advanced technologies in pharmaceutical manufacturing.
The FDA has taken steps to guide the broader application of CFD. General recommendations for CFD modeling were published in a 2017 report, and in 2023 the FDA issued specific guidelines for CFD applications in healthcare devices. Additionally, the American Society of Mechanical Engineers (ASME) Standards Committee on Verification, Validation, and Uncertainty Quantification in Computational Modeling and Simulation has developed standards for evaluating the uncertainty of CFD models. The committee is also exploring the regulatory implications of integrating machine learning into computational frameworks.
To ensure regulatory standards in the development of CFD models and digital twins, Samsung Biologics adheres to the recommendations for uncertainty evaluation outlined by both the FDA and ASME. By following these frameworks, Samsung Biologics ensures that its models meet stringent reliability and validation criteria while supporting the safe and effective integration of cutting-edge technologies into bioprocessing operations.
Delivering Value for Clients Through Innovation
As a leading contract development and manufacturing organization, Samsung Biologics prioritizes delivering high-quality, reliable, and tailored services to its clients. The adoption of CFD models and digital twins exemplifies the company’s commitment to innovation, ensuring clients can bring their products to patients as quickly and effectively as possible. These advanced tools provide significant benefits across productivity, product quality, and scalability.
CFD modeling offers powerful visualization capabilities, allowing inefficiencies and bottlenecks in hydrodynamic parameters to be identified and addressed. Optimizing these parameters enhances overall process performance and streamlines the scaling of processes from lab to production levels. By reducing the hydrodynamic risks associated with scale-up, CFD ensures key parameters remain stable, which safeguards process reliability. When integrated into hybrid digital twins, CFD enables precise control over critical variables, minimizing variability and delivering more consistent, reliable outcomes.
Digital twins further elevate these capabilities by anticipating deviations in cell growth patterns. Early warnings allow for timely adjustments to maintain optimal conditions, enhancing control over key variables. With real-time monitoring and predictive control of the process variables, digital twins ensure consistent product quality. They also provide early notification of potential deviations, enabling faster problem resolution and reducing downtime. These capabilities translate to improved productivity and efficiency, ensuring the uninterrupted delivery of high-quality products.
In addition, the actionable insights offered by digital twins foster deeper collaboration between Samsung Biologics and its clients. This enhanced partnership enables faster, more informed decision-making, accelerating timelines and reinforcing confidence in the manufacturing process.
Driving Efficiency in Process Changes With CFD
A key future application for CFD at Samsung Biologics lies in the in-house evaluation of customer-requested process changes. By leveraging bioreactor hydrodynamic simulations, CFD enables precise predictions of fluid flow and mass transfer. These data are used to conduct gap assessments for requested changes in process operations, ensuring thorough evaluations and risk mitigation.
Currently, statistical analyses are employed to assess changes such as culture volume adjustments or stirring speed modifications, estimating kʟa values. However, statistical models are inherently limited by their dependence on localized sensor data, which often fail to capture fluctuations across the system. As a result, these models may yield less reliable predictions for conditions outside the specific experimental parameters used in their development. Despite this limitation, combining statistical models with historical data, operational expertise, and customer collaboration often allows for risk assessments without requiring test runs.
In the future, Samsung Biologics aims to move beyond basic volume and mixing time analyses. CFD modeling offers the capability to predict more complex process parameters, including pH and dissolved oxygen levels, as functions of time. By incorporating variables such as geometry and advanced equations for turbulence, mass transfer, heat exchange, and fluid-gas interactions, CFD provides more robust and reliable insights than arithmetic-based estimations. This transition will be fully realized once Samsung Biologics’ CFD models meet the stringent FDA guidelines for model credibility.
The reliability of Samsung Biologics’ CFD models is already evident. For example, the production bioreactor model in the fourth plant demonstrated a 95% equivalency with actual process data (RMSE ±0.29), which was statistically significant. This level of precision allows CFD to replace real-world experimentation, significantly reducing costs. Additionally, the enhanced process understanding gained through CFD modeling will streamline operations across Samsung Biologics’ global production network, enabling increased efficiency in the optimization of bioprocesses.
Advancing the Future of CFD and Digital Twins at Samsung Biologics
Samsung Biologics is making continuous improvements to its CFD modeling capabilities, with a focus on enhancing both prediction accuracy and speed. Recognizing the importance of computational power, the company has made significant infrastructure investments. For example, in December 2024, computing capacity was expanded with further annual upgrades planned to support increasingly sophisticated analyses.
These investments not only accelerate computational speed but also enable a broader range of equations to be integrated into CFD models, facilitating advanced and diverse analyses. This capability will allow Samsung Biologics to conduct more complex studies without extending timelines, enhancing the company’s ability to implement digital twins and driving its digital transformation initiatives.
By addressing these challenges head-on, Samsung Biologics is paving the way for more efficient, accurate, and scalable bioprocess solutions, ensuring the company remains at the forefront of biomanufacturing innovation.
