Zhen Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zhen Zhang has a total of 16 co-authors. Here are the top co-authors by number of shared publications:
- Qingkuan Liu
Ate Key Laboratory of Mechanical Behavior and System Safety of Traffic Engineering Structures, Shijiazhuang Tiedao University, Shijiazhuang 050043, China
4 shared publications - Jing Huo
School of Civil Engineering, Shijiazhuang Tiedao University, Shijiazhuang 050043, P. R. China
2 shared publications - Yuexin Cao
School of Civil Engineering, Shijiazhuang Tiedao University 2 , Shijiazhuang 050043
2 shared publications Haohan Li
School of Civil Engineering, Shijiazhuang Tiedao University 2 , Shijiazhuang 050043
1 shared publication- Jingzhu Wang
Chinese Academy of Sciences
1 shared publication - Renfang Huang
Key Laboratory for Mechanics in Fluid Solid Coupling Systems, Institute of Mechanics, Chinese Academy of Sciences, Beijing, China
1 shared publication
Latest Publications
- 1 September 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (9)
Physics-informed super-resolution of dual-cylinder wake flow fields from coarse-grid data
High-resolution flow fields are essential for resolving wake interaction and pressure-coupled unsteady features in bluff-body flows, yet their acquisition from experiments or high-fidelity simulations remains expensive. In dual-cylinder configurations, the interaction between the cylinders can substantially alter the near-wake topology, pressure evolution, and dominant vortex-shedding characteristics, making physically reliable reconstruction from coarse data particularly challenging. In this study, physics-informed neural networks (PINNs) are employed for super-resolution reconstruction of dual-cylinder flow fields at Reynolds number 200 using direct numerical simulation data. To address the non-uniqueness of pressure reconstruction in standard PINNs, a pressure-constrained PINNs (PC-PINNs) framework is proposed by introducing sparse pressure supervision at specific locations to anchor the absolute pressure field. Furthermore, a transfer-learning-enhanced framework (TL-PINNs) is developed to accelerate convergence across related cylinder arrangements. Rather than focusing only on reconstruction accuracy, the present study examines whether the reconstructed fields preserve physically meaningful observables, including near-wake structures, pressure distributions, pressure-fluctuation histories, and dominant spectral signatures. The results show that standard PINNs can recover the major velocity structures but suffer from pressure drift, whereas PC-PINNs markedly improve pressure reconstruction and enable more reliable recovery of pressure-related wake features. TL-PINNs further reduce the training cost while maintaining predictive fidelity and physical interpretability. These findings demonstrate the potential of pressure-constrained physics-informed reconstruction for mechanism-oriented analysis of interference-induced bluff-body flows from coarse-resolution data.
- 1 September 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (9)
Physics-informed super-resolution of dual-cylinder wake flow fields from coarse-grid data
High-resolution flow fields are essential for resolving wake interaction and pressure-coupled unsteady features in bluff-body flows, yet their acquisition from experiments or high-fidelity simulations remains expensive. In dual-cylinder configurations, the interaction between the cylinders can substantially alter the near-wake topology, pressure evolution, and dominant vortex-shedding characteristics, making physically reliable reconstruction from coarse data particularly challenging. In this study, physics-informed neural networks (PINNs) are employed for super-resolution reconstruction of dual-cylinder flow fields at Reynolds number 200 using direct numerical simulation data. To address the non-uniqueness of pressure reconstruction in standard PINNs, a pressure-constrained PINNs (PC-PINNs) framework is proposed by introducing sparse pressure supervision at specific locations to anchor the absolute pressure field. Furthermore, a transfer-learning-enhanced framework (TL-PINNs) is developed to accelerate convergence across related cylinder arrangements. Rather than focusing only on reconstruction accuracy, the present study examines whether the reconstructed fields preserve physically meaningful observables, including near-wake structures, pressure distributions, pressure-fluctuation histories, and dominant spectral signatures. The results show that standard PINNs can recover the major velocity structures but suffer from pressure drift, whereas PC-PINNs markedly improve pressure reconstruction and enable more reliable recovery of pressure-related wake features. TL-PINNs further reduce the training cost while maintaining predictive fidelity and physical interpretability. These findings demonstrate the potential of pressure-constrained physics-informed reconstruction for mechanism-oriented analysis of interference-induced bluff-body flows from coarse-resolution data.
- 1 November 2024
- Published by AIP Publishing in Physics of Fluids
- Vol. 36 (11)
Physics-informed dynamic mode decomposition for reconstruction and prediction of dense particulate pipe flows
Dynamic mode decomposition (DMD) effectively captures the growth and frequency characteristics of individual modes, enabling the construction of reduced-order models for flow evolution, thereby facilitating the prediction of fluid dynamic behavior. However, DMD's predictive accuracy is inherently constrained by its inability to inherently incorporate physical principles. Therefore, for dense particulate pipe flows with complex flow mechanisms, we introduce a physics-informed dynamic mode decomposition (PIDMD) approach, which augments the purely data-driven DMD framework by incorporating the conservation of mass as a constraint. This ensures that the extracted dynamic modes adhere to known physical principles. Initially, we apply the DMD to reconstruct and predict the velocity field, comparing the results against benchmark computational fluid dynamics-discrete element method (CFD-DEM) simulations. Findings indicate that while DMD can reconstruct the flow field simulated by CFD-DEM and provide predictions of future flow states, its predictive accuracy gradually deteriorates over time. Next, we utilize both PIDMD and DMD to reconstruct and predict particle volume fraction, evaluating both models based on CFD-DEM outcomes. The results indicate that both PIDMD and DMD can predict particle aggregation toward the center, but PIDMD provides more accurate predictions regarding the size of particle aggregations and their distribution near the tube wall. Furthermore, the average prediction error for particle volume fraction using PIDMD is 6.54%, which is lower than the error of 13.49% obtained by DMD. Both qualitative and quantitative comparisons highlight the superior predictive capability of PIDMD. The methodology developed in this study provides valuable insights for high-precision predictions of particulate flows.
- 1 November 2024
- Published by AIP Publishing in Physics of Fluids
- Vol. 36 (11)
Physics-informed dynamic mode decomposition for reconstruction and prediction of dense particulate pipe flows
Dynamic mode decomposition (DMD) effectively captures the growth and frequency characteristics of individual modes, enabling the construction of reduced-order models for flow evolution, thereby facilitating the prediction of fluid dynamic behavior. However, DMD's predictive accuracy is inherently constrained by its inability to inherently incorporate physical principles. Therefore, for dense particulate pipe flows with complex flow mechanisms, we introduce a physics-informed dynamic mode decomposition (PIDMD) approach, which augments the purely data-driven DMD framework by incorporating the conservation of mass as a constraint. This ensures that the extracted dynamic modes adhere to known physical principles. Initially, we apply the DMD to reconstruct and predict the velocity field, comparing the results against benchmark computational fluid dynamics-discrete element method (CFD-DEM) simulations. Findings indicate that while DMD can reconstruct the flow field simulated by CFD-DEM and provide predictions of future flow states, its predictive accuracy gradually deteriorates over time. Next, we utilize both PIDMD and DMD to reconstruct and predict particle volume fraction, evaluating both models based on CFD-DEM outcomes. The results indicate that both PIDMD and DMD can predict particle aggregation toward the center, but PIDMD provides more accurate predictions regarding the size of particle aggregations and their distribution near the tube wall. Furthermore, the average prediction error for particle volume fraction using PIDMD is 6.54%, which is lower than the error of 13.49% obtained by DMD. Both qualitative and quantitative comparisons highlight the superior predictive capability of PIDMD. The methodology developed in this study provides valuable insights for high-precision predictions of particulate flows.
- 1 January 2023
- Published by AIP Publishing in Physics of Fluids
- Vol. 35 (1)
Data-driven turbulence model for unsteady cavitating flow
Unsteady Reynolds-averaged Navier–Stokes (URANS) equations have been widely used in engineering fields to investigate cavitating flow owing to their low computational cost and excellent robustness. However, it is challenging to accurately obtain the unsteady characteristics of flow owing to cavitation-induced phase transitions. In this study, we propose an implicit data-driven URANS (DD-URANS) framework to analyze the unsteady characteristics of cavitating flow. In the DD-URANS framework, a basic computational model is developed by introducing a cavitation-induced phase transition into the equations of Reynolds stress. To improve the computational accuracy and generalization performance of the basic model, the linear and nonlinear parts of the anisotropic Reynolds stress are predicted through implicit and explicit methods, respectively. A data fusion approach, allowing the input and output of characterized parameters at multiple time points, is presented to obtain the unsteady characteristics of the cavitating flow. The DD-URANS model is trained using the numerical results obtained via large-eddy simulation. The training data consist of two parts: (i) the results obtained at cavitation numbers of 2.0, 2.2, and 2.7 for a Venturi flow, and (ii) those obtained at cavitation numbers of 0.8 and 1.5 for a National Advisory Committee for Aeronautics (NACA) 66 hydrofoil. The DD-URANS model is used to predict the cavitating flow at cavitation numbers of 2.5 for a Venturi flow and 0.8 for a Clark-Y hydrofoil. It is found that the DD-URANS model is superior to the baseline URANS model in predicting the instantaneous periodic shedding of a cavity and the mean flow fields.
- 1 January 2023
- Published by AIP Publishing in Physics of Fluids
- Vol. 35 (1)
Data-driven turbulence model for unsteady cavitating flow
Unsteady Reynolds-averaged Navier–Stokes (URANS) equations have been widely used in engineering fields to investigate cavitating flow owing to their low computational cost and excellent robustness. However, it is challenging to accurately obtain the unsteady characteristics of flow owing to cavitation-induced phase transitions. In this study, we propose an implicit data-driven URANS (DD-URANS) framework to analyze the unsteady characteristics of cavitating flow. In the DD-URANS framework, a basic computational model is developed by introducing a cavitation-induced phase transition into the equations of Reynolds stress. To improve the computational accuracy and generalization performance of the basic model, the linear and nonlinear parts of the anisotropic Reynolds stress are predicted through implicit and explicit methods, respectively. A data fusion approach, allowing the input and output of characterized parameters at multiple time points, is presented to obtain the unsteady characteristics of the cavitating flow. The DD-URANS model is trained using the numerical results obtained via large-eddy simulation. The training data consist of two parts: (i) the results obtained at cavitation numbers of 2.0, 2.2, and 2.7 for a Venturi flow, and (ii) those obtained at cavitation numbers of 0.8 and 1.5 for a National Advisory Committee for Aeronautics (NACA) 66 hydrofoil. The DD-URANS model is used to predict the cavitating flow at cavitation numbers of 2.5 for a Venturi flow and 0.8 for a Clark-Y hydrofoil. It is found that the DD-URANS model is superior to the baseline URANS model in predicting the instantaneous periodic shedding of a cavity and the mean flow fields.