Predicting complex fluid dynamics to scale up bioreactors for production

About the customer

Bristol Myers Squibb is a global biopharmaceutical company headquartered in New York, New York, United States (U.S.) that manufactures innovative pharmaceuticals and biologics in several therapeutic areas, including oncology, cardiovascular, immunoscience and fibrosis.

The following case study explores research conducted jointly with Bristol Myers Squibb, published in Chemical Engineering Science by authors John A. Thomas, Xiaoming Liu, Brian DeVincentis, Helen Hua, Grace Yao, Michael C. Borys, Kathryn Aron and Girish Pendse.

Their challenge

For pharmaceutical companies like Bristol Myers Squibb, the biomanufacturing process of biologic drugs – produced by living organisms within stirred-tank bioreactors – presents significant challenges to process scale-up and intensification. Because bioreactors are designed for specific scales, the team needed a way to predict how a process optimized at a small, tabletop scale would translate to full-scale production while still maintaining optimal bioreactor operation.

Traditional predictive modeling in these environments is often slow and difficult due to complex mechanics. For turbulent bioprocess simulations, time-averaged flow fields derived from Reynolds-averaged Navier-Stokes (RANS) equations provide limited value. Furthermore, dissolved gas concentrations – a critical environmental parameter for living organisms – are driven by complex dynamics that make single-phase fluid models inapplicable.

To accurately capture the complexity of cell culture, Bristol Myers Squibb required a two-phase fluid model capable of handling simultaneous agitation and gassing while also supporting species transport across the bubble and liquid interface.

Our solution

To successfully scale up production and bring compounds to market faster, Bristol Myers Squibb required a way to predict complex, multi-fluid mixing processes. The team turned to Simcenter™ M-Star™ software, part of the Siemens Xcelerator business platform of software, hardware and services, which utilizes graphics processing unit (GPU)-native lattice-Boltzmann method (LBM) algorithms to deliver high-fidelity, fully transient simulations.

By leveraging this modern computational fluid dynamics (CFD) paradigm, Bristol Myers Squibb successfully developed a framework for building time-dependent, bubble-resolved, two-phase models. This allowed them to investigate real-time blending and mass transfer dynamics within stirred-tank bioreactors with unprecedented efficiency.

With the support of Simcenter M-Star, including in-depth training and specialized consulting in fluid mechanics, Bristol Myers Squibb was able to model, solve and validate a robust mechanistic approach for predicting critical process parameters.

Results

Simcenter M-Star delivered high-fidelity results that consistently agreed with measured conditions across a wide range of operating environments. This predictive accuracy arms Bristol Myers Squibb with the critical data needed to understand exactly how a process will translate from small-scale development to full production, ensuring a seamless technology transfer.

The transition to the GPU-native architecture of Simcenter M-Star provided multiple order-of-magnitude improvements in computational speed compared to traditional central processing unit (CPU) clusters. These accelerated simulations maintained strong agreement with experimental data while requiring no model re-parameterization between scales or operating conditions – a significant technical advantage that streamlined the modeling workflow.

Ultimately, this integrated approach empowered Bristol Myers Squibb to significantly compress their production timelines and reduce the overall cost of biologics manufacturing. By leveraging these high-speed simulation results, the team was able to accelerate the delivery of life-saving compounds to market.