An automated and extensible workflow for the design and evaluation of an axial blood pump, based on parametric CAD and CFD analysis
CFD Simulation Report
The purpose of this case study is to start a new long-term research initiative focused on biomedical flows, with a particular emphasis on the simulation and optimization of axial blood pumps and other medical devices. This ambitious project is a collaboration between:
- CFD SUPPORT, Prague
- Faculty of Mechanical Engineering, Czech Technical University in Prague
- Institute of Mathematics, Czech Academy of Sciences
- others …
The primary objective is to develop a robust, automated, and extensible simulation framework for accurately modeling complex biomedical phenomena such as:
- Blood flow dynamics (including non-Newtonian behavior)
- Hemolysis and shear-induced damage
- Coagulation modeling
- Elastic deformation of device components (FSI)
At the core of the framework lies a fully parametric CAD model of an axial blood pump, developed in Rhinoceros (Rhino) with Grasshopper for flexible design control. This model is tightly coupled with advanced simulation workflows built using TCAE, integrating:
- Automated meshing and solver setup
- Parametric simulation and optimization loops
- Real-time post-processing
- Automated report generation
- Validation support based on experimental and published data
As part of this initiative, we are excited to release this case study as a public benchmark within the project. It demonstrates the current capabilities of TCAE in biomedical applications and provides a fully reproducible simulation workflow, including geometry, mesh, setup, and results.
By extending TCAE into the biomedical domain, this initiative aims to deliver a validated, modular, and scalable simulation concept that meets the rigorous requirements of medical engineering — including traceability, repeatability, and physical accuracy.
The workflow for axial blood pump development is designed to enable systematic performance assessment and shape refinement, while remaining flexible and extensible. The long-term goal is to establish a reproducible and shareable benchmark that can support future extensions, ranging from advanced shape optimization to the integration of biomedical metrics, such as blood damage prediction, shear stress evaluation, or thrombus formation modeling (Blood Coagulation).
TCAE Workflow For Axial Blood Pump
TCAE is an engineering simulation framework based on a set of modular tools built around reliable open-source software. Its structure allows users to adapt the environment to specific tasks, making it suitable for both standard workflows and more advanced, customized simulations.
The core of TCAE consists of five main modules: TMESH (meshing), TCFD (flow simulation), TFEA (structural analysis), TOPT (optimization), and TCAA (acoustic analysis). These modules can be used individually or in combination, depending on the user’s goals, experience, and available resources.
In the current workflow for axial blood pump, the main components in use are TMESH, TCFD , and TOPT—providing a robust setup for CFD simulations and design optimization. There are also considerations for extending the workflow in the future with TBASE, a database module currently in development, and potentially TFEA, to incorporate structural simulations.
TCAE is designed for integration. It works well alongside other commercial, in-house, or open-source tools, allowing users to build flexible workflows around their specific needs. Whether operated as a simple, user-friendly “black box” or as a fully customizable simulation pipeline, TCAE supports a wide range of user expertise and applications.
Automatic reporting is included for both individual simulations and optimization tasks. The planned TBASE module will further enhance this by organizing simulation data into a structured database. This will enable the use of surrogate modeling and machine learning to accelerate design iterations without relying solely on repeated full-scale simulations.
Overall, TCAE offers a practical and flexible simulation environment that can grow with the user’s needs—making it a solid foundation.
Geometrical Model & Parametric Design of Axial Blood Pump
The geometry of the axial blood pump was constructed as a fully parametric CAD model, defined entirely by a set of user-controlled input parameters. This approach provides a structured and repeatable framework for modifying the design, which is essential for both exploratory case studies and automated optimization tasks.
By simply adjusting input parameters, new axial blood pump design variants can be generated automatically, without any manual intervention. The CAD model is created through a script-driven process, ensuring consistency between iterations and enabling seamless integration into simulation and optimization workflows. This methodology combines flexibility and precision, allowing for systematic exploration of the design space.
Such a parametric framework offers several key advantages:
- Fully automated geometry generation
- Once input values are defined, the entire CAD model is built without manual editing.
- Flexibility in design modification
- The model can be easily adapted to explore various design concepts by adjusting input parameters.
- 30 geometry parameters.
- Efficient exploration of the design space
- New geometries are generated automatically through scripted construction, allowing systematic exploration of the design space and direct integration with optimization workflows.
- Consistency
- Script-driven modelling ensures that all design variants follow the same construction logic, reducing human error and increasing reliability.
- Reproducibility and traceability
- Every geometry is fully defined by a parameter set, making design steps transparent, documented, and repeatable.
The 3D surface geometry is constructed using an internally developed parametric model in Rhinoceros and Grasshopper. However, end users do not interact directly with these software tools. Instead, the geometry is automatically generated based on a set of input parameters, or alternatively imported from a predefined geometry file.
To facilitate meshing and solver setup, the geometry is further subdivided into distinct stl files, representing both physical boundaries and artificial internal interfaces as artificial inlet and outlet boundaries. The computational domain is extended enough to allow the flow to develop naturally before and after passing through the pump. However, in principle, this could be avoided — for example by applying appropriate boundary conditions.
TMESH – CFD Meshing
In this particular study of the axial blood pump, the model is divided into three components to support flexible and modular mesh generation. For each component, the computational mesh can be either generated in automated software module TMESH, using the snappyHexMesh open-source mesh generation utility or imported from an external source. Both approaches can be combined within the same simulation setup, allowing maximum freedom in how the mesh is constructed.
Although snappyHexMesh was used in this case, it is not required by the TCAE environment—any external mesh may be loaded directly in supported formats such as MSH, CGNS, or OpenFOAM format.
For each model component, the initial background mesh can be based on a Cartesian block structure or a cylindrical topology and further refined along with the simulated object. This meshing approach is fully integrated into the automated simulation workflow, enabling repeatable and consistent mesh generation across different design variants.
Basic mesh cell size is defined with the keyword “background mesh size”. The internal point must be selected manually. To better support workflows involving geometry variation and optimization, automatic placement is being developed and is planned for inclusion in a future release.
CFD Simulation Setup – MRF
The CFD simulation is performed using the TCFD module, part of the TCAE simulation framework. The entire CFD simulation setup and execution are carried out through the TCFD GUI integrated within ParaView. TCFD uses OpenFOAM open-source application.
For the computational purposes the blood pump model is subdivided into two static and one rotating part.
The static straightener and diffuser blocks are defined in a fixed reference frame, while a rotating frame of reference is considered for the impeller block. This approach known as the Multiple Reference Frame (MRF) technique allows for simple and efficient approximation of flow problems with moving (often rotating) parts. Instead of solving fully unsteady flow problem with moving and sliding boundaries, a steady flow simulation is performed with “frozen”-like rotor and all the effects of rotation are included as extra volume forces acting on the fluid in the local reference frame.
The resulting velocity field needs thus be interpreted as being relative to the given static/rotating reference frame. In this case, the impeller operates at a constant speed of –8000 RPM, rotating around the body’s axis (specifically, the x-axis).
- TCAE Simulation type: Pump
- Time management: steady-state
- Physical model: Incompressible
- Number of components: 3 [-]
- Wall roughness: none
- Speed: -8000 [RPM]
- Inlet: 2.5 [l/min]
- Outlet: Static Pressure 0 [m2/s2]
- Shear Thinnig model: Newtonian
- Turbulence: k-ω-SST
- Turbulence intensity: 5%
- Wall treatment: Wall functions
- Speedlines: 1 [-]
- Simulation points: 1 [-]
- Fluid: Blood
- Reference pressure: 1 [atm]
- Dynamic viscosity: 5 [cP]
- Air density: 1060 [kg/m3]
TCAE Simulation & Results Evaluation
The TCAE simulation run for the axial blood pump is completely automated. The whole workflow can be run by a single click in the GUI, or the whole process can be run in batch mode on a background.
The TCFD module includes a built-in post-processing module that automatically evaluates all the required quantities, such as efficiency, forces, flow rates, pressure, velocity, and much more. These quantities are continuously monitored and evaluated throughout the simulation, and the key results are summarized in an HTML report. This report is generated automatically for each simulation run and can be updated at any time during the process.
All relevant integral quantities and simulation statistics are computed automatically for every simulation. Each simulation run in TCAE is accompanied by its own unique report, ensuring clarity and traceability of results.
The evaluated integral quantities are sorted, stored in corresponding CSV files, and made readily available for further analysis or post-processing.
While the integral results are exported in CSV format, the volume fields are post-processed in the open-source visualization tool ParaView, which offers a wide range of tools and methods for postprocessing and results evaluation.
Automated Optimization with TOPT
All components — Rhino/Grasshopper for geometry, TMESH, TCFD and TOPT are connected into a unified, fully automated loop.
Within the TCAE environment, the optimisation is handled by the TOPT module, which supports two modes:
Design of Experiments (DOE)
In the DOE phase, parameter values can be defined as explicit lists or as linear or geometric sequences. A Cartesian product of these sets forms a grid of points, where one simulation (computational case) is run for each point.
Optimisation
The Optimisation mode in TOPT aims to identify parameter combinations that minimise or maximise a defined objective function. Two built-in algorithms are available: Golden Section Search and DIRECT. In both cases, TOPT first performs the DOE runs, identifies the best-performing design, and narrows the optimisation domain to its neighboring DOE points. \footnote{The DIRECT algorithm does not require a DOE initialization. The number of initial samples can be set to zero.}
Before launching a full-scale optimisation, it is advisable to verify that the chosen mesh resolution is suitable for the range of geometries and operating conditions expected during chosen study.
A mesh that performs well for one configuration may prove insufficient — or overly detailed — for others. A coarse mesh may fail to capture critical flow phenomena, while an unnecessarily fine mesh increases computational costs without a proportional gain in accuracy.
In this context, TOPT offers a convenient way to estimate an appropriate mesh resolution for the intended optimisation task. By running selected design points and flow conditions across multiple mesh levels, users can evaluate the sensitivity of results to mesh resolution and identify a suitable compromise between accuracy and efficiency.
For our purposes, we tested various flow rates ranging from 0.5 to 10.0 l/min across different mesh sizes. The simulations were run until a convergence criterion of efficiency absolute tolerance 0.001 was met.
At this level of accuracy, the results suggest that mesh size has only a minor influence on efficiency for flow rates up to 7.5 l/min — efficiency values remain relatively stable across the tested meshes. However, at the highest flow rate, a noticeable dependence on mesh resolution emerges: efficiency declines significantly on coarser meshes and stabilises only as the mesh is refined, indicating that key flow features may be under-resolved at high throughput.
More broadly, the required mesh resolution depends not only on the operating conditions but also on which physical quantities are of primary interest and the accuracy with which they need to be predicted. When efficiency is a target metric — as in the present optimisation task — the mesh must be fine enough to resolve the flow features that significantly impact it.
Within this approach, TOPT can be used also as a pre-optimisation tool for mesh resolution assessment and workflow verification, helping ensure that the results are consistent and reliable throughout the design space.
Conclusion
The TCAE simulation run is completely automated. The whole workflow can be run by a single click in the GUI, or the whole process can be run in batch mode on a background.
To enable performance improvements and systematic evaluation of the axial blood pumps, we developed a parametric and automation-ready simulation framework. This framework allows controlled variation of geometric parameters and evaluation through CFD simulations. While the current study does not yet include blood related phenomena, such as hemolysis or shear induced damage, it provides the necessary technical foundation for such investigations in the future. By linking parametric geometry, automated mesh generation, simulation workflows and optimization, the framework enables efficient exploration of design variants with minimal manual input.
The goal is to support future research that may extend the framework not only towards bio-related analyses, such as blood damage prediction, shear stress evaluation, or thrombosis risk assessment, but also towards application-specific optimizations and interdisciplinary investigations, including fluid–structure interaction or integration with physiological models.
While the presented framework provides a solid foundation, it remains an ongoing project under active development.
Researchers or groups interested in contributing models, extending the workflow, or collaborating on related topics — such as blood damage modelling, multiphysics coupling, or patient-specific applications — are warmly invited to contact us. We welcome interdisciplinary input and aim to make the framework adaptable to a wide range of research needs.
