Zihao Wang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Zihao Wang has a total of 60 co-authors. Here are the top co-authors by number of shared publications:
- Guiyong Zhang
State Key Laboratory of Structural Analysis, Optimization and CAE Software for Industrial Equipment, School of Naval Architecture Engineering, Dalian University of Technology 1 , Dalian 116024
26 shared publications - Tiezhi Sun
School of Naval Architecture and Ocean Engineering, Dalian University of Technology 4 , Dalian 116081
12 shared publications Huakun Huang
6 shared publications- Bo Zhou
Dalian University of Technology
3 shared publications - Chongbin Shi
School of Naval Architecture, State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian 116024, China
3 shared publications Yichen Jiang
3 shared publications
Latest Publications
- 19 September 2026
- Published by Elsevier in Ocean Engineering
- Vol. 367
Beaufort-geometry ordinal learning for cross-domain sea-state image classification
- 19 September 2026
- Published by Elsevier in Ocean Engineering
- Vol. 367
Beaufort-geometry ordinal learning for cross-domain sea-state image classification
- 1 August 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (8)
Boundary-layer-informed surrogate modeling of airfoil trailing-edge noise: From wall-pressure excitation to far-field radiation
When turbulent-boundary-layer wall-pressure fluctuations convect past an airfoil trailing edge, they are scattered and radiated to the far field as broadband trailing-edge noise. Predicting this process requires reliable near-trailing-edge boundary-layer quantities, whose repeated evaluation across airfoil geometries and inflow conditions is computationally costly. This study develops AcoustiFoil, a boundary-layer-informed surrogate model that maps class-shape transformation-parameterized airfoil geometry and inflow conditions to aerodynamic coefficients and local boundary-layer quantities required by empirical wall-pressure-spectrum models. Trained on millions of samples generated by XFOIL, a viscous-inviscid airfoil analysis code, AcoustiFoil is coupled with the Lee wall-pressure-spectrum model and Amiet's trailing-edge-noise theory, retaining an explicit sequence from boundary-layer state to far-field radiation without directly regressing acoustic spectra. Validation against National Advisory Committee for Aeronautics 0012 cases from the benchmark problems for airframe noise computations shows that the coupled predictions reproduce the principal trends in the measured one-third-octave spectra. For the 56 m/s case, the integrated sound pressure level over 1–6.3 kHz differs from the Braunschweig aeroacoustic wind tunnel measurement by 0.63 dB, reflecting the complete prediction chain rather than the surrogate alone. The framework is further incorporated into a blade element momentum calculation for the Siemens Wind Turbine-2.3-93 rotor. Although only trailing-edge noise is considered, the calculation captures the dominant broadband trend in the measured turbine spectrum. Within the tested range, AcoustiFoil reduces the need for repeated viscous-flow calculations while retaining the modeled boundary-layer-to-acoustic coupling.
- 1 August 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (8)
Boundary-layer-informed surrogate modeling of airfoil trailing-edge noise: From wall-pressure excitation to far-field radiation
When turbulent-boundary-layer wall-pressure fluctuations convect past an airfoil trailing edge, they are scattered and radiated to the far field as broadband trailing-edge noise. Predicting this process requires reliable near-trailing-edge boundary-layer quantities, whose repeated evaluation across airfoil geometries and inflow conditions is computationally costly. This study develops AcoustiFoil, a boundary-layer-informed surrogate model that maps class-shape transformation-parameterized airfoil geometry and inflow conditions to aerodynamic coefficients and local boundary-layer quantities required by empirical wall-pressure-spectrum models. Trained on millions of samples generated by XFOIL, a viscous-inviscid airfoil analysis code, AcoustiFoil is coupled with the Lee wall-pressure-spectrum model and Amiet's trailing-edge-noise theory, retaining an explicit sequence from boundary-layer state to far-field radiation without directly regressing acoustic spectra. Validation against National Advisory Committee for Aeronautics 0012 cases from the benchmark problems for airframe noise computations shows that the coupled predictions reproduce the principal trends in the measured one-third-octave spectra. For the 56 m/s case, the integrated sound pressure level over 1–6.3 kHz differs from the Braunschweig aeroacoustic wind tunnel measurement by 0.63 dB, reflecting the complete prediction chain rather than the surrogate alone. The framework is further incorporated into a blade element momentum calculation for the Siemens Wind Turbine-2.3-93 rotor. Although only trailing-edge noise is considered, the calculation captures the dominant broadband trend in the measured turbine spectrum. Within the tested range, AcoustiFoil reduces the need for repeated viscous-flow calculations while retaining the modeled boundary-layer-to-acoustic coupling.
- 30 July 2026
- Published by Elsevier in Ocean Engineering
- Vol. 365
A Bayesian event-constrained reduced-order model for trimaran multiphysics slamming transients
- 30 July 2026
- Published by Elsevier in Ocean Engineering
- Vol. 365
A Bayesian event-constrained reduced-order model for trimaran multiphysics slamming transients
- 29 June 2026
- Published by Elsevier in Ocean Engineering
- Vol. 363
Machine learning for recovering three-dimensional turbulent effects in two-dimensional RANS simulations
- 29 June 2026
- Published by Elsevier in Ocean Engineering
- Vol. 363
Machine learning for recovering three-dimensional turbulent effects in two-dimensional RANS simulations
- Open Access
- 30 May 2026
- Published by Elsevier in SSRN Electronic Journal
Numerical Study on the Effects of Wave Phases on Cavity Dynamics and Multiscale Vortex Evolution for Oblique Water Entry of a Cylinder
This study investigates the complex fluid-structure interactions during the oblique water entry of a cylinder in a regular wave environment. A three-dimensional, six-degree-of-freedom (6DOF) numerical wave tank was established by coupling Large Eddy Simulation (LES) with the Volume of Fluid (VOF) method. Validated against experiments, the framework systematically tracks transient loads, rigid-body kinematics, and cavity evolution across four typical wave phases (crest, trough, upstream, and downstream). Furthermore, the Q-criterion and the Vorticity Transport Equation (VTE) were introduced to quantitatively dissect the multi-scale vortex topologies and underlying momentum source terms. Results demonstrate that wave interference mechanisms decouple into surface geometric and deep-water kinematic effects. During initial impact, wave slope governs the relative impact angle: the upstream node triggers extreme flat-bottom slamming with peak pressures, whereas the downstream node induces progressive wetting. In deep water, orbital velocities dominate; reverse flow at the trough provokes severe hydrodynamic squeezing, forcing premature pinch-off. Macroscopically, asymmetric cavity collapse under the downstream node generates a high-velocity re-entrant jet that strikes the cylinder's aft, causing the most violent attitude deflection. Microscopically, VTE analysis confirms that baroclinic torque dictates early interfacial vorticity generation, while vortex stretching drives deep-water wake dissipation.
- 30 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessNumerical Study on the Effects of Wave Phases on Cavity Dynamics and Multiscale Vortex Evolution for Oblique Water Entry of a Cylinder
This study investigates the complex fluid-structure interactions during the oblique water entry of a cylinder in a regular wave environment. A three-dimensional, six-degree-of-freedom (6DOF) numerical wave tank was established by coupling Large Eddy Simulation (LES) with the Volume of Fluid (VOF) method. Validated against experiments, the framework systematically tracks transient loads, rigid-body kinematics, and cavity evolution across four typical wave phases (crest, trough, upstream, and downstream). Furthermore, the Q-criterion and the Vorticity Transport Equation (VTE) were introduced to quantitatively dissect the multi-scale vortex topologies and underlying momentum source terms. Results demonstrate that wave interference mechanisms decouple into surface geometric and deep-water kinematic effects. During initial impact, wave slope governs the relative impact angle: the upstream node triggers extreme flat-bottom slamming with peak pressures, whereas the downstream node induces progressive wetting. In deep water, orbital velocities dominate; reverse flow at the trough provokes severe hydrodynamic squeezing, forcing premature pinch-off. Macroscopically, asymmetric cavity collapse under the downstream node generates a high-velocity re-entrant jet that strikes the cylinder's aft, causing the most violent attitude deflection. Microscopically, VTE analysis confirms that baroclinic torque dictates early interfacial vorticity generation, while vortex stretching drives deep-water wake dissipation.
- Open Access
- 27 March 2026
- Published by Elsevier in SSRN Electronic Journal
Data-driven Machine Learning for Recovering Three-Dimensional Turbulent Effects in Two-Dimensional RANS Simulations
Accurately predicting the mean flow fields with three-dimensional (3D) instability turbulent flow effects using two-dimensional (2D) Reynolds-averaged Navier–Stokes (RANS) simulations is widely deemed problematic, primarily due to the inherent loss of vortex stretching mechanism. In this paper, we propose a machine learning method based on the previously developed data-driven turbulence closure framework (Huang et al., J. Comput. Phys., 2025, 533, 113995) to simulate the effect of 3D instability within a 2D computational domain by learning the additional implicit dissipation laws. These laws are incorporated into the RANS equations represented by four non-dimensional source parameters. In this way, the baseline model is improved by allowing the energy production and dissipation to fit with the high-fidelity energy budget, while maintaining a 2D computational framework. We validate the proposed model by simulating the unseen varied flows, including different geometries and boundary conditions compared with the high-fidelity results. The good performance of the proposed method is observed in varied cases, for which the model demonstrates numerical stability and achieves a computational speedup of at least 49.5 times in 2D simulation compared with 3D high-fidelity simulations in terms of the mean fields in the case of flow past a cylinder of Reynolds number of 3,900.
- 27 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessData-driven Machine Learning for Recovering Three-Dimensional Turbulent Effects in Two-Dimensional RANS Simulations
Accurately predicting the mean flow fields with three-dimensional (3D) instability turbulent flow effects using two-dimensional (2D) Reynolds-averaged Navier–Stokes (RANS) simulations is widely deemed problematic, primarily due to the inherent loss of vortex stretching mechanism. In this paper, we propose a machine learning method based on the previously developed data-driven turbulence closure framework (Huang et al., J. Comput. Phys., 2025, 533, 113995) to simulate the effect of 3D instability within a 2D computational domain by learning the additional implicit dissipation laws. These laws are incorporated into the RANS equations represented by four non-dimensional source parameters. In this way, the baseline model is improved by allowing the energy production and dissipation to fit with the high-fidelity energy budget, while maintaining a 2D computational framework. We validate the proposed model by simulating the unseen varied flows, including different geometries and boundary conditions compared with the high-fidelity results. The good performance of the proposed method is observed in varied cases, for which the model demonstrates numerical stability and achieves a computational speedup of at least 49.5 times in 2D simulation compared with 3D high-fidelity simulations in terms of the mean fields in the case of flow past a cylinder of Reynolds number of 3,900.
- 17 February 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
Integrating Spatial and Temporal Modeling: A Deep Learning Framework for Super-Resolution Reconstruction and Prediction of Fluid Flows
In numerical simulations and experimental observations, flow field data are often limited by insufficient spatial resolution and incomplete time series, which in turn affect the accurate capture and modeling of flow structures and their evolution processes. To address this challenge, this paper proposes an end-to-end high-resolution flow field prediction framework that can rely solely on low-resolution input data to reconstruct the fine spatial structure at the current moment and predict the high-resolution evolution state at future moments. The framework consists of three types of deep neural network modules working together: the U-Net-based super-resolution reconstruction module (U-Net-SR) implements high-resolution spatial reconstruction at the current moment, the long short-term memory module (LSTM) models the temporal evolution trend of low-resolution sequences, and the U-Net-based multi-source fusion prediction module (U-Net-Pred) integrates the information from both to achieve accurate prediction of the high-resolution flow field at future moments. This method is validated in three typical flow field tasks: laminar flow around a cylinder, laminar flow over a controlled pitch airfoil, and experimental turbulent flow field under cross-wind gusts passing through a flat plate airfoil. In multiple typical flow field tasks, the framework demonstrates excellent spatiotemporal modeling capabilities. Overall, the average peak signal-to-noise ratio (PSNR) of the prediction results generally remains in the range of 40–48[ ]dB, and the structural similarity index measure (SSIM) index is mostly stable above 0.95, accurately capturing key structural features and maintaining good temporal evolution consistency. Compared with traditional single-path models that only handle spatial reconstruction or temporal prediction, the innovation of this framework lies in its dual U-Net collaborative mechanism: U-Net-SR provides the high-resolution state at the current moment as a spatial benchmark, LSTM captures the overall evolution trend of the flow field, and U-Net-Pred completes cross-scale fusion and information collaboration in the feature space, thereby achieving better super-resolution prediction.
- 17 February 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Computational Methods
Integrating Spatial and Temporal Modeling: A Deep Learning Framework for Super-Resolution Reconstruction and Prediction of Fluid Flows
In numerical simulations and experimental observations, flow field data are often limited by insufficient spatial resolution and incomplete time series, which in turn affect the accurate capture and modeling of flow structures and their evolution processes. To address this challenge, this paper proposes an end-to-end high-resolution flow field prediction framework that can rely solely on low-resolution input data to reconstruct the fine spatial structure at the current moment and predict the high-resolution evolution state at future moments. The framework consists of three types of deep neural network modules working together: the U-Net-based super-resolution reconstruction module (U-Net-SR) implements high-resolution spatial reconstruction at the current moment, the long short-term memory module (LSTM) models the temporal evolution trend of low-resolution sequences, and the U-Net-based multi-source fusion prediction module (U-Net-Pred) integrates the information from both to achieve accurate prediction of the high-resolution flow field at future moments. This method is validated in three typical flow field tasks: laminar flow around a cylinder, laminar flow over a controlled pitch airfoil, and experimental turbulent flow field under cross-wind gusts passing through a flat plate airfoil. In multiple typical flow field tasks, the framework demonstrates excellent spatiotemporal modeling capabilities. Overall, the average peak signal-to-noise ratio (PSNR) of the prediction results generally remains in the range of 40–48[ ]dB, and the structural similarity index measure (SSIM) index is mostly stable above 0.95, accurately capturing key structural features and maintaining good temporal evolution consistency. Compared with traditional single-path models that only handle spatial reconstruction or temporal prediction, the innovation of this framework lies in its dual U-Net collaborative mechanism: U-Net-SR provides the high-resolution state at the current moment as a spatial benchmark, LSTM captures the overall evolution trend of the flow field, and U-Net-Pred completes cross-scale fusion and information collaboration in the feature space, thereby achieving better super-resolution prediction.
- 26 November 2025
- Published by Elsevier in Ocean Engineering
- Vol. 344
A collision avoidance framework for autonomous surface vehicles considering rule-violating behaviors
- 26 November 2025
- Published by Elsevier in Ocean Engineering
- Vol. 344
A collision avoidance framework for autonomous surface vehicles considering rule-violating behaviors
- 1 November 2025
- Published by AIP Publishing in Physics of Fluids
- Vol. 37 (11)
Spatio-temporal graph neural network for multi-step prediction of cylinder water entry flow fields
This study introduces a Spatio-Temporal Graph Neural Network (ST-GNN) for the multi-step prediction of highly dynamic cylinder water entry flow fields. The proposed model integrates a Graph Neural Network (GNN) to capture complex spatial dependencies on unstructured meshes with a Long Short-Term Memory (LSTM) network to model temporal evolution. We generate a dataset from numerical simulations across different cases, varying both Froude numbers and initial entry angles. Extensive ablation studies are conducted to systematically optimize the model's architecture and hyperparameters, revealing that a single-step prediction strategy with an input sequence of eight timesteps yields the most robust performance, significantly reducing error accumulation in long-term prediction. The optimized model achieves high accuracy (R2>0.97) and successfully predicts key physical phenomena during water entry. Furthermore, generalization tests on unseen cases demonstrate the model's excellent interpolation capability within the training data range and reveal the challenges of extrapolation, particularly for the complex phase interface. Crucially, while the multi-step prediction of the velocity field degrades due to error accumulation, the inferred prediction of the water volume fraction remains remarkably stable. This indicates that the model has learned not just the instantaneous spatial correlations between the fields, but more importantly, has captured the underlying temporal dynamics that govern the evolution of the phase interface.
- 1 November 2025
- Published by AIP Publishing in Physics of Fluids
- Vol. 37 (11)
Spatio-temporal graph neural network for multi-step prediction of cylinder water entry flow fields
This study introduces a Spatio-Temporal Graph Neural Network (ST-GNN) for the multi-step prediction of highly dynamic cylinder water entry flow fields. The proposed model integrates a Graph Neural Network (GNN) to capture complex spatial dependencies on unstructured meshes with a Long Short-Term Memory (LSTM) network to model temporal evolution. We generate a dataset from numerical simulations across different cases, varying both Froude numbers and initial entry angles. Extensive ablation studies are conducted to systematically optimize the model's architecture and hyperparameters, revealing that a single-step prediction strategy with an input sequence of eight timesteps yields the most robust performance, significantly reducing error accumulation in long-term prediction. The optimized model achieves high accuracy (R2>0.97) and successfully predicts key physical phenomena during water entry. Furthermore, generalization tests on unseen cases demonstrate the model's excellent interpolation capability within the training data range and reveal the challenges of extrapolation, particularly for the complex phase interface. Crucially, while the multi-step prediction of the velocity field degrades due to error accumulation, the inferred prediction of the water volume fraction remains remarkably stable. This indicates that the model has learned not just the instantaneous spatial correlations between the fields, but more importantly, has captured the underlying temporal dynamics that govern the evolution of the phase interface.