SavedAdd your journal

Journals / Gazi Mühendislik Bilimleri Dergisi / 2026 / Volume: 12 Issue: 2

CFD-Driven Machine Learning Surrogate for Predicting Transient Outlet Velocity Profiles in Aneurysmal Aortic Geometries

Hakan Bayrakcı(kahramanmaraş sütçü imam ünivierstesi)0009-0001-0885-1972
Arif Çutay(KAHRAMANMARAS SUTCU IMAM UNIVERSITY)0000-0002-0057-9417
Özdeş Çermik(KAHRAMANMARAS SUTCU IMAM UNIVERSITY)0000-0001-9308-4589
Pages
—
Publication Date
2026-08-31
🏛️Bibliographic metadata and full text access for this article are provided via DergiPark.
DergiPark Record ↗
✨

AI Summary & Key Takeaways

AI Insight
⚡ TL;DR(Too Long; Didn't Read)

Results show that the ML framework accurately reproduces peak systolic velocity, waveform morphology, and phase alignment across all outlets, with RMSE values below 0.026 m/s and mean relative errors generally under 7%.

📌 3 Key Takeaways

🎯
Objective & Hypothesis:Transient hemodynamic analysis using Computational Fluid Dynamics (CFD) provides detailed insight into blood flow behavior in aneurysmal aortic geometries; however, repeated simulations for multiple configurations are computationally expensive.
📊
Key Findings & Data:In this study, a hybrid CFD–Machine Learning (ML) surrogate framework was developed to predict time-dependent outlet velocity profiles for unseen vascular geometries characterized by different aneurysm diameters (45 mm and 50 mm).
💡
Conclusion & Impact:Results show that the ML framework accurately reproduces peak systolic velocity, waveform morphology, and phase alignment across all outlets, with RMSE values below 0.026 m/s and mean relative errors generally under 7%.
🔬Design:Modeling & Simulation
📍Scope:Comprehensive Sample
⚙️Variables:CFD (Computational Fluid Dynamics) · Aortic Aneursym · Machine Learning

Abstract

Transient hemodynamic analysis using Computational Fluid Dynamics (CFD) provides detailed insight into blood flow behavior in aneurysmal aortic geometries; however, repeated simulations for multiple configurations are computationally expensive. In this study, a hybrid CFD–Machine Learning (ML) surrogate framework was developed to predict time-dependent outlet velocity profiles for unseen vascular geometries characterized by different aneurysm diameters (45 mm and 50 mm). High-fidelity transient CFD simulations were conducted under pulsatile flow conditions, and the resulting outlet velocity data were used to train a data-driven interpolation-based surrogate model. The trained model was evaluated on geometries not included in the training dataset. Results show that the ML framework accurately reproduces peak systolic velocity, waveform morphology, and phase alignment across all outlets, with RMSE values below 0.026 m/s and mean relative errors generally under 7%. Under a fixed inlet flow rate of 8.75 L/min, global mass conservation was maintained, while minor local flow redistribution differences were observed. The proposed approach significantly reduces computational cost and offers a rapid, reliable alternative for parametric hemodynamic analysis.

🔖

Cite This Article

Bayrakcı, H., Çutay, A., Çermik, Ö. (2026). CFD-Driven Machine Learning Surrogate for Predicting Transient Outlet Velocity Profiles in Aneurysmal Aortic Geometries. Gazi Mühendislik Bilimleri Dergisi, Volume: 12 Issue: 2. https://doi.org/10.30855/gmbd.070526A07

📄 Full Text PDF Viewer

📄

PDF görüntüleyici hazırlanıyor...

References (18)

  1. E. Pajaziti, J. Montalt-Tordera, C. Capelli, R. Sivera, E. Sauvage, M. Quail, S. Schievano, and V. Muthurangu, “Shape-driven deep neural networks for fast acquisition of aortic 3D pressure and velocity flow fields,” PLOS Computational Biology, vol. 19, no. 4, Art. no. e1011055, 2023. doi: 10.1371/journal.pcbi.1011055
  2. D. R. Rutkowski, A. Roldán-Alzate, and K. M. Johnson, “Enhancement of cerebrovascular 4D flow MRI velocity fields using machine learning and computational fluid dynamics simulation data,” Scientific Reports, vol. 11, no. 1, Art. no. 10240, 2021. doi: 10.1038/s41598-021-89636-z
  3. S. Padhee, M. Johnson, H. Yi, T. Banerjee, and Z. Yang, “Machine learning for aiding blood flow velocity estimation based on angiography,” Bioengineering, vol. 9, no. 11, Art. no. 622, 2022. doi: 10.3390/bioengineering9110622
  4. G. Li, H. Wang, M. Zhang, S. Tupin, A. Qiao, Y. Liu, M. Ohta, and H. Anzai, “Prediction of 3D cardiovascular hemodynamics before and after coronary artery bypass surgery via deep learning,” Communications Biology, vol. 4, no. 1, Art. no. 99, 2021. doi: 10.1038/s42003-020-01638-1
  5. S. Wang, D. Wu, G. Li, Z. Zhang, W. Xiao, R. Li, A. Qiao, L. Jin, and H. Liu, “Deep learning-based hemodynamic prediction of carotid artery stenosis before and after surgical treatments,” Frontiers in Physiology, vol. 13, Art. no. 1094743, 2023. doi: 10.3389/fphys.2022.1094743
  6. A. Arzani, J.-X. Wang, and R. M. D’Souza, “Uncovering near-wall blood flow from sparse data with physics-informed neural networks,” Physics of Fluids, vol. 33, no. 7, Art. no. 071905, 2021. doi: 10.1063/5.0055600
  7. L. Sun, H. Gao, S. Pan, and J.-X. Wang, “Surrogate modeling for fluid flows based on physics-constrained deep learning without simulation data,” Computer Methods in Applied Mechanics and Engineering, vol. 361, Art. no. 112732, 2020. doi: 10.1016/j.cma.2019.112732
  8. X. Zhang, B. Mao, Y. Che, J. Kang, M. Luo, A. Qiao, Y. Liu, H. Anzai, M. Ohta, Y. Guo, and G. Li, “Physics-informed neural networks (PINNs) for 4D hemodynamics prediction: An investigation of optimal framework based on vascular morphology,” Computers in Biology and Medicine, vol. 164, Art. no. 107287, 2023. doi: 10.1016/j.compbiomed.2023.107287
  9. P. Du, X. Zhu, and J.-X. Wang, “Deep learning-based surrogate model for three-dimensional patient-specific computational fluid dynamics,” Physics of Fluids, vol. 34, no. 8, Art. no. 081906, 2022. doi: 10.1063/5.0101128
  10. D. Lin and S. Kenjereš, “Towards fast and reliable estimations of 3D pressure, velocity and wall shear stress in aortic blood flow: CFD-based machine learning approach,” Computers in Biology and Medicine, vol. 191, Art. no. 110137, 2025. doi: 10.1016/j.compbiomed.2025.110137.
  11. J. Suk, D. Alblas, B. A. Hutten, A. Wiegman, C. Brune, P. van Ooij, and J. M. Wolterink, “Physics-informed graph neural networks for flow field estimation in carotid arteries,” Medical Image Analysis, vol. 110, Art. no. 103974, 2026. doi: 10.1016/j.media.2026.103974 [12] G. Nannini, J. Suk, P. Rygiel, S. Saitta, L. Mariani, R. Maranga, A. Baggiano, G. Pontone, J. M. Wolterink, and A. Redaelli, “Learning hemodynamic scalar fields on coronary artery meshes: A benchmark of geometric deep learning models,” Computers in Biology and Medicine, vol. 195, Art. no. 110477, 2025. doi: 10.1016/j.compbiomed.2025.110477
  12. B. Morgan, A. R. Murali, G. Preston, Y. A. Sima, L. A. Marcelo Chamorro, C. Bourantas, R. Torii, A. Mathur, A. Baumbach, M. C. Jacob, S. Karabasov, and R. Krams, “A physics-based machine learning technique rapidly reconstructs the wall-shear stress and pressure fields in coronary arteries,” Frontiers in Cardiovascular Medicine, vol. 10, Art. no. 1221541, 2023. doi: 10.3389/fcvm.2023.1221541
  13. W. Rajhi, Z. Ahmed, A. B. M. Ali, A. Alizadeh, Z. A. Hussein, N. S. S. Singh, B. Louhichi, and W. Aich, “Use of proper orthogonal decomposition and machine learning for efficient blood flow prediction in cerebral saccular aneurysms,” Scientific Reports, vol. 15, no. 1, Art. no. 32335, 2025. doi: 10.1038/s41598-025-17823-3
  14. N. Maul, K. Zinn, F. Wagner, M. Thies, M. Rohleder, L. Pfaff, M. Kowarschik, A. Birkhold, and A. K. Maier, “Transient hemodynamics prediction using an efficient octree-based deep learning model,” in Information Processing in Medical Imaging: IPMI 2023, A. Frangi, M. de Bruijne, D. Wassermann, and N. Navab, Eds., Lecture Notes in Computer Science, vol. 13939. Springer, 2023, pp. 183–194. doi: 10.1007/978-3-031-34048-2_15
  15. E. S. Kökten and M. Tekir, “Energy-efficient optimization of radiant floor heating systems via random forest-based surrogate models and evolutionary algorithms,” Case Studies in Thermal Engineering, vol. 80, Art. no. 107931, 2026. doi: 10.1016/j.csite.2026.107931
  16. E. Taşkesen, M. Dirik, M. Tekir, and H. K. Pazarlıoğlu, “Predicting heat transfer performance of Fe₃O₄–Cu/water hybrid nanofluid under constant magnetic field using ANN,” Journal of Thermal Engineering, vol. 9, no. 3, pp. 811–822, 2023. doi: 10.18186/thermal.1300854
  17. M. J. F. G. Ramaekers, I. B. van der Vlugt, J. J. M. Westenberg, R. Perinajová, H. J. Lamb, J. E. Wildberger, S. Kenjereš, and S. Schalla, “Flow patterns in ascending aortic aneurysms: Determining the role of hypertension using phase contrast magnetic resonance and computational fluid dynamics,” Computers in Biology and Medicine, vol. 172, Art. no. 108310, 2024. doi: 10.1016/j.compbiomed.2024.108310
  18. J. Park, J. Kim, S. Hyun, and J. Lee, “Hemodynamics in a three-dimensional printed aortic model: A comparison of four-dimensional phase-contrast magnetic resonance and image-based computational fluid dynamics,” Magnetic Resonance Materials in Physics, Biology and Medicine, vol. 35, no. 5, pp. 719–732, 2022. doi: 10.1007/s10334-021-00984-3