Chong Pan
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
Chong Pan has a total of 141 co-authors. Here are the top co-authors by number of shared publications:
- Jinjun Wang
Beijing University of Aeronautics and Astronautics
18 shared publications Jinjun Wang
Key Laboratory of Fluid Mechanics of Ministry of Education, Beihang University, Beijing, China
9 shared publications- Shuai Qiao
Beihang University
8 shared publications - Chujiang Cai
Beihang Univ, Key Lab Fluid Mech, Minist Educ, Beijing 100191, Peoples R China
7 shared publications - Qingfu Zhang
Key Laboratory of Fluid Mechanics of Ministry of Education, Beihang University, Beijing, China
4 shared publications - Yanpeng Liu
Institute of Fluid Mechanics, Beihang University, Beijing 100191, China
4 shared publications
Latest Publications
- 23 October 2026
- Published by Springer Nature in Experiments in Fluids
- Vol. 67 (11) , 179
Particle-based profile velocimetry for flow-field velocity measurement
- 23 October 2026
- Published by Springer Nature in Experiments in Fluids
- Vol. 67 (11) , 179
Particle-based profile velocimetry for flow-field velocity measurement
- Open Access
- 24 September 2026
- Published by Elsevier in SSRN Electronic Journal
A High-Performance Tomographic Background-Oriented Schlieren System for Time-Resolved Three-Dimensional Flow Measurements
Time-resolved measurements of three-dimensional variable-density flows are essential for understanding the evolving structures of compressible jets and combustion flows. Tomographic background-oriented schlieren (TBOS) enables nonintrusive measurements of these structures by reconstructing three-dimensional refractive-index fields from multiview background images. Incoherently illuminated TBOS systems are widely used owing to their performance and affordability, but simultaneously achieving short exposures, high acquisition rates, and large numbers of pixels per view remains challenging. This study develops a TBOS instrument that combines overdriven pulsed light-emitting diodes (LEDs), retroreflective backgrounds, and industrial cameras with real-time image output. The combined LED and retroreflective-background gains are estimated to increase received brightness by more than two orders of magnitude relative to the classical continuous LED illumination of a diffuse background. Compared with state-of-the-art incoherently illuminated TBOS systems, the instrument reduces effective exposure by approximately tenfold. It also records approximately ten times as many pixels per second across all views. These gains are achieved with a small aperture and competitive values for the number of views, pixels per view, and acquisition rate. Measurements of an unsteady premixed flame recover consecutive three-dimensional refractive-index fields, resolving the fine-scale flame structure and bulk unsteady dynamics. The results demonstrate a practical TBOS instrument for time-resolved three-dimensional refractive-index measurements in unsteady flows.
- 24 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA High-Performance Tomographic Background-Oriented Schlieren System for Time-Resolved Three-Dimensional Flow Measurements
Time-resolved measurements of three-dimensional variable-density flows are essential for understanding the evolving structures of compressible jets and combustion flows. Tomographic background-oriented schlieren (TBOS) enables nonintrusive measurements of these structures by reconstructing three-dimensional refractive-index fields from multiview background images. Incoherently illuminated TBOS systems are widely used owing to their performance and affordability, but simultaneously achieving short exposures, high acquisition rates, and large numbers of pixels per view remains challenging. This study develops a TBOS instrument that combines overdriven pulsed light-emitting diodes (LEDs), retroreflective backgrounds, and industrial cameras with real-time image output. The combined LED and retroreflective-background gains are estimated to increase received brightness by more than two orders of magnitude relative to the classical continuous LED illumination of a diffuse background. Compared with state-of-the-art incoherently illuminated TBOS systems, the instrument reduces effective exposure by approximately tenfold. It also records approximately ten times as many pixels per second across all views. These gains are achieved with a small aperture and competitive values for the number of views, pixels per view, and acquisition rate. Measurements of an unsteady premixed flame recover consecutive three-dimensional refractive-index fields, resolving the fine-scale flame structure and bulk unsteady dynamics. The results demonstrate a practical TBOS instrument for time-resolved three-dimensional refractive-index measurements in unsteady flows.
- 19 August 2026
- Published by American Chemical Society (ACS) in Langmuir
- Vol. 42 (34) , 25217-25227
Inclination-Induced Interfacial Asymmetry and Tangential Force Balance of Bubbles Pinned at Wetting-Step Boundaries
Bubble growth on inclined surfaces is governed by the coupling among buoyancy, interfacial deformation, and three-phase contact-line constraint. Here, we experimentally investigate quasi-static bubble growth and detachment on superhydrophobic substrates patterned with a circular wetting-step line, which pins the three-phase contact line throughout the prenecking stage. Increasing substrate inclination shortens the detachment period while maintaining cycle-to-cycle coefficients of variation below 2%. Because lateral contact-line motion is suppressed, the substrate-parallel component of buoyancy produces both inclination-dependent global shape deformation and progressively stronger lateral asymmetry, characterized by a higher aspect ratio at a given normalized growth stage and a larger lateral offset of the projected bubble centroid, respectively. The apparent contact-angle distribution also becomes asymmetric: the upslope advancing-side angle decreases during growth, whereas the downslope receding-side angle remains comparatively high, causing the apparent contact-angle difference to increase throughout the prenecking stage. A quasi-static tangential force balance, formulated using the resultant capillary traction along the pinned circular boundary, quantitatively relates the dimensionless tangential load to the measured contact-angle asymmetry. A zero-intercept regression yields an effective retention coefficient of 0.8683 with R2 = 0.9989, while the coefficients evaluated separately at each nonzero inclination differ from the global regression value by less than 4% and exhibit no monotonic dependence on inclination. These results establish a reduced interfacial-mechanics description of the critical prenecking state under geometric contact-line constraint and clarify how inclination-induced tangential buoyancy is accommodated through coupled interfacial and apparent-contact-angle asymmetry, providing a mechanistic basis for regulating bubble growth and release on wettability-patterned surfaces.
- 19 August 2026
- Published by American Chemical Society (ACS) in Langmuir
- Vol. 42 (34) , 25217-25227
Inclination-Induced Interfacial Asymmetry and Tangential Force Balance of Bubbles Pinned at Wetting-Step Boundaries
Bubble growth on inclined surfaces is governed by the coupling among buoyancy, interfacial deformation, and three-phase contact-line constraint. Here, we experimentally investigate quasi-static bubble growth and detachment on superhydrophobic substrates patterned with a circular wetting-step line, which pins the three-phase contact line throughout the prenecking stage. Increasing substrate inclination shortens the detachment period while maintaining cycle-to-cycle coefficients of variation below 2%. Because lateral contact-line motion is suppressed, the substrate-parallel component of buoyancy produces both inclination-dependent global shape deformation and progressively stronger lateral asymmetry, characterized by a higher aspect ratio at a given normalized growth stage and a larger lateral offset of the projected bubble centroid, respectively. The apparent contact-angle distribution also becomes asymmetric: the upslope advancing-side angle decreases during growth, whereas the downslope receding-side angle remains comparatively high, causing the apparent contact-angle difference to increase throughout the prenecking stage. A quasi-static tangential force balance, formulated using the resultant capillary traction along the pinned circular boundary, quantitatively relates the dimensionless tangential load to the measured contact-angle asymmetry. A zero-intercept regression yields an effective retention coefficient of 0.8683 with R2 = 0.9989, while the coefficients evaluated separately at each nonzero inclination differ from the global regression value by less than 4% and exhibit no monotonic dependence on inclination. These results establish a reduced interfacial-mechanics description of the critical prenecking state under geometric contact-line constraint and clarify how inclination-induced tangential buoyancy is accommodated through coupled interfacial and apparent-contact-angle asymmetry, providing a mechanistic basis for regulating bubble growth and release on wettability-patterned surfaces.
- 1 August 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (8)
Effect of wetting-step line geometry on bubble growth and detachment
Bubble growth and detachment under circular and regular polygonal wetting-step line (WSL) constraints, where a WSL denotes the boundary between superhydrophilic and superhydrophobic regions, were investigated experimentally at identical contact-line perimeters and circularities from 0.60 to 1.00. Synchronized dual-view imaging showed that decreasing circularity progressively increased the discrepancy between the two characteristic bubble profiles and strengthened the spatial anisotropy of the apparent contact angle. During quasi-static growth, non-circular WSLs produced a faster reduction in the Laplace pressure ratio between the bubble base and apex, leading to earlier necking and detachment. The plateau-averaged apparent contact-angle difference increased approximately linearly as circularity decreased. An effective capillary-retention coefficient calculated from the measured apparent contact angles increased from approximately 0.86 for the triangular WSL to 0.94 for the circular WSL, and was empirically correlated with circularity. A quasi-static force-balance model linked this coefficient to the normalized detachment volume. The measured normalized detachment volume increased from approximately 0.62 at a circularity of 0.60 to approximately 0.90 at a circularity of 1.00, with a linear regression coefficient of determination of 0.991. The circular result was close to previously reported axisymmetric WSL data. These results show that reduced circularity redistributes interfacial curvature and apparent contact angle, reduces the net vertical capillary-retention force, and causes bubble detachment at a smaller volume under equal-perimeter constraints.
- 1 August 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (8)
Effect of wetting-step line geometry on bubble growth and detachment
Bubble growth and detachment under circular and regular polygonal wetting-step line (WSL) constraints, where a WSL denotes the boundary between superhydrophilic and superhydrophobic regions, were investigated experimentally at identical contact-line perimeters and circularities from 0.60 to 1.00. Synchronized dual-view imaging showed that decreasing circularity progressively increased the discrepancy between the two characteristic bubble profiles and strengthened the spatial anisotropy of the apparent contact angle. During quasi-static growth, non-circular WSLs produced a faster reduction in the Laplace pressure ratio between the bubble base and apex, leading to earlier necking and detachment. The plateau-averaged apparent contact-angle difference increased approximately linearly as circularity decreased. An effective capillary-retention coefficient calculated from the measured apparent contact angles increased from approximately 0.86 for the triangular WSL to 0.94 for the circular WSL, and was empirically correlated with circularity. A quasi-static force-balance model linked this coefficient to the normalized detachment volume. The measured normalized detachment volume increased from approximately 0.62 at a circularity of 0.60 to approximately 0.90 at a circularity of 1.00, with a linear regression coefficient of determination of 0.991. The circular result was close to previously reported axisymmetric WSL data. These results show that reduced circularity redistributes interfacial curvature and apparent contact angle, reduces the net vertical capillary-retention force, and causes bubble detachment at a smaller volume under equal-perimeter constraints.
- Open Access
- 22 July 2026
- Published by Springer Nature in Lasers in Medical Science
- Vol. 41 (1) , 157
Influence of temporal delay in SWEEPS dual-pulse Er: YAG laser-activated irrigation on fluid dynamics in confined domain environment: an in vitro study
The shock wave-enhanced emission photoacoustic streaming (SWEEPS) technique delivers synchronized dual-pulse Er: YAG laser energy to enhance fluid dynamics during laser-activated irrigation. The temporal delay (Tp) between consecutive pulses is a key parameter, but its influence on fluid dynamics within a confined root canal environment remains unclear. This in vitro study investigated the effects of different Tp settings on bubble morphology and transient apical peak pressure (TAPP) in a simulated root canal model. A transparent single-rooted resin anterior tooth model (apical diameter: 0.30 mm, taper: 6%, root canal length: 12 mm) was fabricated. The root canal was filled with deionized water. SWEEPS dual-pulse Er: YAG laser irrigation (20 mJ/pulse, 20 Hz) was performed at five Tp settings: 120, 210, 300, 400, and 500 µs. Bubble dynamics within the root canal were recorded using a high-speed camera system (200,000 fps). The grayscale integral of bubble images was calculated using MATLAB as a surrogate for relative bubble volume. Simultaneously, TAPP at the root apex was measured using a high-frequency pressure sensor (50 kHz). Four repeated measurements were performed per Tp group. Data were analyzed using one-way ANOVA followed by Tukey’s HSD post-hoc test (α = 0.05). Both bubble morphology and TAPP varied significantly with Tp. At Tp = 120 µs, the bubbles generated by the first and second pulses coalesced into a single larger bubble, whereas at Tp = 300–500 µs, two independent bubbles formed sequentially. The grayscale integral (secondary cavitation volume surrogate) was significantly higher at Tp = 120 µs ((0.074 ± 0.010) ×10 6 pixel) than at 210 µs ((0.052 ± 0.012) ×10 6 pixel), P < 0.05), and significantly higher at 500 µs ((2.767 ± 0.195) ×10 6 pixel) than at 300 µs ((1.926 ± 0.06) ×10 6 pixel) and 400 µs ((1.986 ± 0.135) ×10 6 pixel, both P < 0.05). TAPP followed a similar pattern: 120 µs produced significantly higher apical pressure than 210 µs (3.520 ± 0.291 vs. 2.835 ± 0.047 kPa, P < 0.05), and 500 µs produced significantly higher pressure than 300 and 400 µs (3.398 ± 0.068 vs. 2.807 ± 0.099 and 3.128 ± 0.182 kPa, both P < 0.05). No significant differences in TAPP were observed between 120 and 500 µs (P > 0.05). In this in vitro resin root canal model with deionized water, SWEEPS dual-pulse Er: YAG laser-activated irrigation with Tp = 120 or 500 µs produced greater secondary cavitation volume and higher apical pressure than intermediate Tp values (210–400 µs), indicating that Tp significantly influences fluid dynamics under the tested experimental conditions.
- 22 July 2026
- Published by Springer Nature in Lasers in Medical Science
- Vol. 41 (1) , 157
Open AccessInfluence of temporal delay in SWEEPS dual-pulse Er: YAG laser-activated irrigation on fluid dynamics in confined domain environment: an in vitro study
The shock wave-enhanced emission photoacoustic streaming (SWEEPS) technique delivers synchronized dual-pulse Er: YAG laser energy to enhance fluid dynamics during laser-activated irrigation. The temporal delay (Tp) between consecutive pulses is a key parameter, but its influence on fluid dynamics within a confined root canal environment remains unclear. This in vitro study investigated the effects of different Tp settings on bubble morphology and transient apical peak pressure (TAPP) in a simulated root canal model. A transparent single-rooted resin anterior tooth model (apical diameter: 0.30 mm, taper: 6%, root canal length: 12 mm) was fabricated. The root canal was filled with deionized water. SWEEPS dual-pulse Er: YAG laser irrigation (20 mJ/pulse, 20 Hz) was performed at five Tp settings: 120, 210, 300, 400, and 500 µs. Bubble dynamics within the root canal were recorded using a high-speed camera system (200,000 fps). The grayscale integral of bubble images was calculated using MATLAB as a surrogate for relative bubble volume. Simultaneously, TAPP at the root apex was measured using a high-frequency pressure sensor (50 kHz). Four repeated measurements were performed per Tp group. Data were analyzed using one-way ANOVA followed by Tukey’s HSD post-hoc test (α = 0.05). Both bubble morphology and TAPP varied significantly with Tp. At Tp = 120 µs, the bubbles generated by the first and second pulses coalesced into a single larger bubble, whereas at Tp = 300–500 µs, two independent bubbles formed sequentially. The grayscale integral (secondary cavitation volume surrogate) was significantly higher at Tp = 120 µs ((0.074 ± 0.010) ×10 6 pixel) than at 210 µs ((0.052 ± 0.012) ×10 6 pixel), P < 0.05), and significantly higher at 500 µs ((2.767 ± 0.195) ×10 6 pixel) than at 300 µs ((1.926 ± 0.06) ×10 6 pixel) and 400 µs ((1.986 ± 0.135) ×10 6 pixel, both P < 0.05). TAPP followed a similar pattern: 120 µs produced significantly higher apical pressure than 210 µs (3.520 ± 0.291 vs. 2.835 ± 0.047 kPa, P < 0.05), and 500 µs produced significantly higher pressure than 300 and 400 µs (3.398 ± 0.068 vs. 2.807 ± 0.099 and 3.128 ± 0.182 kPa, both P < 0.05). No significant differences in TAPP were observed between 120 and 500 µs (P > 0.05). In this in vitro resin root canal model with deionized water, SWEEPS dual-pulse Er: YAG laser-activated irrigation with Tp = 120 or 500 µs produced greater secondary cavitation volume and higher apical pressure than intermediate Tp values (210–400 µs), indicating that Tp significantly influences fluid dynamics under the tested experimental conditions.
- Open Access
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Research on Low-drag Low-boom Supersonic Transport Configuration using an MDO framework and Deep Learning Methods
The low-drag, low-boom design is a critical enabling technology for the practical deployment of future supersonic civil aircraft. Achieving a balance between aerodynamic efficiency improvement and sonic boom mitigation requires a design method capable of handling strongly coupled, highly nonlinear interactions across multiple disciplines. In this study, two representative design strategies for low-drag, low-boom supersonic aircraft were investigated and compared: a traditional multidisciplinary design optimization (MDO) framework based on computational fluid dynamics (CFD) and a data-driven, deep learning-based method for prediction and inverse design. The MDO framework integrates aerodynamic performance and sonic boom objectives at multiple azimuths by combining a medium-fidelity Euler solver with boundary-layer correction and the augmented Burgers equation. In contrast, the deep learning method employs a multilayer perceptron (MLP) neural network, which is trained on a geometry–near-field–far-field mapping dataset collected by an automatic sampling platform to achieve near-real-time forward prediction and waveform-driven inverse design of low-boom configurations. Both methods were applied to a representative supersonic aircraft configuration, and the results show that the MDO method can achieve robust and quantitatively consistent improvements in aerodynamic efficiency and ground perceived loudness, while the deep learning-based method can drastically reduce computational cost, with the inverse-designed configurations providing physically meaningful reference solutions. The comparative study further demonstrates that the two approaches play complementary roles at different stages of the design process, with deep learning enabling rapid conceptual exploration and MDO ensuring high-fidelity, constraint-consistent optimization. This work provides a comparison and integrated perspective on physics-based and data-driven approaches for low-boom design, demonstrating strong potential for future intelligent aerodynamic–acoustic design systems of supersonic transport.
- 12 May 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessResearch on Low-drag Low-boom Supersonic Transport Configuration using an MDO framework and Deep Learning Methods
The low-drag, low-boom design is a critical enabling technology for the practical deployment of future supersonic civil aircraft. Achieving a balance between aerodynamic efficiency improvement and sonic boom mitigation requires a design method capable of handling strongly coupled, highly nonlinear interactions across multiple disciplines. In this study, two representative design strategies for low-drag, low-boom supersonic aircraft were investigated and compared: a traditional multidisciplinary design optimization (MDO) framework based on computational fluid dynamics (CFD) and a data-driven, deep learning-based method for prediction and inverse design. The MDO framework integrates aerodynamic performance and sonic boom objectives at multiple azimuths by combining a medium-fidelity Euler solver with boundary-layer correction and the augmented Burgers equation. In contrast, the deep learning method employs a multilayer perceptron (MLP) neural network, which is trained on a geometry–near-field–far-field mapping dataset collected by an automatic sampling platform to achieve near-real-time forward prediction and waveform-driven inverse design of low-boom configurations. Both methods were applied to a representative supersonic aircraft configuration, and the results show that the MDO method can achieve robust and quantitatively consistent improvements in aerodynamic efficiency and ground perceived loudness, while the deep learning-based method can drastically reduce computational cost, with the inverse-designed configurations providing physically meaningful reference solutions. The comparative study further demonstrates that the two approaches play complementary roles at different stages of the design process, with deep learning enabling rapid conceptual exploration and MDO ensuring high-fidelity, constraint-consistent optimization. This work provides a comparison and integrated perspective on physics-based and data-driven approaches for low-boom design, demonstrating strong potential for future intelligent aerodynamic–acoustic design systems of supersonic transport.
- Open Access
- 9 April 2026
- Published by Elsevier in SSRN Electronic Journal
A 3D DNS solver using the vorticity-vector potential-scalar potential formulation for incompressible turbulent flow with periodic boundary conditions
In this study, a three-dimensional (3D) Direct Numerical Simulation (DNS) solver for incompressible turbulent flows is developed based on the vorticity-vector potential-scalar potential ($\bm{\omega}-\boldsymbol{\psi}-\Phi$) formulation. By employing periodic boundary conditions, the solver inherently avoids the complex consistency issues between vorticity and vector potential often encountered at physical boundaries. This simplification allows the numerical framework to focus on mitigating aliasing errors induced by strong nonlinear convection and strictly enforcing the solenoidal constraints on both velocity and vorticity fields. By incorporating a scalar potential, the method projects the velocity field onto a divergence-free manifold, effectively neutralizing the accumulation of numerical errors. Regarding the numerical implementation, the framework integrates high-order hybrid finite difference schemes for the vorticity transport equation with an efficient FFT-based spectral solver for the Poisson equations, achieving a balance between accuracy and computational efficiency. Comprehensive validation against the 2D and 3D Taylor-Green vortex flows verifies second-order spatial convergence and robust stability under relaxed CFL constraints. Furthermore, simulations of forced Homogeneous Isotropic Turbulence (HIT) at Taylor-scale Reynolds numbers up to $\mathrm{Re}_\lambda = 201$ demonstrate that the method resolves the turbulent energy spectrum with spectral-like fidelity. In summary, this vorticity-based DNS solver not only obviates the intricacies of pressure-velocity coupling inherent in traditional primitive-variable methods but also overcomes the aliasing error limitations of pseudo-spectral methods. Consequently, this study provides a high-fidelity and robust tool for turbulence research, demonstrating strong potential for broader applications, including compressible flows and multiphase flows.- 9 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessA 3D DNS solver using the vorticity-vector potential-scalar potential formulation for incompressible turbulent flow with periodic boundary conditions
In this study, a three-dimensional (3D) Direct Numerical Simulation (DNS) solver for incompressible turbulent flows is developed based on the vorticity-vector potential-scalar potential ($\bm{\omega}-\boldsymbol{\psi}-\Phi$) formulation. By employing periodic boundary conditions, the solver inherently avoids the complex consistency issues between vorticity and vector potential often encountered at physical boundaries. This simplification allows the numerical framework to focus on mitigating aliasing errors induced by strong nonlinear convection and strictly enforcing the solenoidal constraints on both velocity and vorticity fields. By incorporating a scalar potential, the method projects the velocity field onto a divergence-free manifold, effectively neutralizing the accumulation of numerical errors. Regarding the numerical implementation, the framework integrates high-order hybrid finite difference schemes for the vorticity transport equation with an efficient FFT-based spectral solver for the Poisson equations, achieving a balance between accuracy and computational efficiency. Comprehensive validation against the 2D and 3D Taylor-Green vortex flows verifies second-order spatial convergence and robust stability under relaxed CFL constraints. Furthermore, simulations of forced Homogeneous Isotropic Turbulence (HIT) at Taylor-scale Reynolds numbers up to $\mathrm{Re}_\lambda = 201$ demonstrate that the method resolves the turbulent energy spectrum with spectral-like fidelity. In summary, this vorticity-based DNS solver not only obviates the intricacies of pressure-velocity coupling inherent in traditional primitive-variable methods but also overcomes the aliasing error limitations of pseudo-spectral methods. Consequently, this study provides a high-fidelity and robust tool for turbulence research, demonstrating strong potential for broader applications, including compressible flows and multiphase flows. - Top Cited Papers
- 20 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
Optical-Flow-Guided Recursive Prediction-Refinement Neural Network for Particle Image Velocimetry
Accurate in-plane displacement field estimation is a key procedure in planar two-dimensional Particle Image Velocimetry (PIV). In this paper, we propose an end-to-end deep learning-based model, termed Recursive Prediction-Refinement Neural Network (RPR-NN), for PIV post-processing. The proposed model estimates dense displacement fields from a single pair of PIV images using a lightweight recurrent unit that incorporates key principles of conventional optical flow solvers, including coarse-to-fine multi-resolution pyramids, iterative warping-based refinement, and gradient-based optical flow constraints derived from the brightness constancy assumption. Furthermore, a spatial attention mechanism is introduced to emphasize high-confidence features, effectively addressing the inherent sparsity of PIV images. RPR-NN is trained on the publicly available PIVDataset and achieves superior accuracy and computational running-time efficiency compared to representative PIV post-processing methods. Quantitative assessments on synthetic datasets show that the proposed model exhibits robust generalization across diverse out-of-distribution scenarios that span a wide spectrum of particle seeding density, diameter, image noise levels, and particle displacement. Its effectiveness is further validated through challenging experimental PIV measurements across a wide range of flow regimes, including low-speed and hypersonic flows, where non-uniform particle seeding and low signal-to-noise ratios create substantial difficulties for conventional PIV post-processing algorithms.
- 20 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Instrumentation and Measurement
- Vol. 75 (00189456) , 1-16
Top Cited PapersOptical-Flow-Guided Recursive Prediction-Refinement Neural Network for Particle Image Velocimetry
Accurate in-plane displacement field estimation is a key procedure in planar two-dimensional Particle Image Velocimetry (PIV). In this paper, we propose an end-to-end deep learning-based model, termed Recursive Prediction-Refinement Neural Network (RPR-NN), for PIV post-processing. The proposed model estimates dense displacement fields from a single pair of PIV images using a lightweight recurrent unit that incorporates key principles of conventional optical flow solvers, including coarse-to-fine multi-resolution pyramids, iterative warping-based refinement, and gradient-based optical flow constraints derived from the brightness constancy assumption. Furthermore, a spatial attention mechanism is introduced to emphasize high-confidence features, effectively addressing the inherent sparsity of PIV images. RPR-NN is trained on the publicly available PIVDataset and achieves superior accuracy and computational running-time efficiency compared to representative PIV post-processing methods. Quantitative assessments on synthetic datasets show that the proposed model exhibits robust generalization across diverse out-of-distribution scenarios that span a wide spectrum of particle seeding density, diameter, image noise levels, and particle displacement. Its effectiveness is further validated through challenging experimental PIV measurements across a wide range of flow regimes, including low-speed and hypersonic flows, where non-uniform particle seeding and low signal-to-noise ratios create substantial difficulties for conventional PIV post-processing algorithms.
- 1 March 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (3)
Optimal growth in a high-speed Görtler vortex flow
Optimal growth in Görtler vortex flow over a concave wall at Mach 6 is investigated via well-established parabolized governing equations (PSE3D-APSE3D). The influence of parameters, i.e., the initial/final positions and the frequency, is studied, and the physical mechanism of optimal growth is explored. The optimal growth consists of an algebraic growth stage near the initial location and a subsequent quasi-exponential growth stage. The algebraic growth is dominated by a combination of the lift-up effect and the Orr mechanism. High-frequency disturbances initially tilt against the flow direction and amplify via the Orr mechanism during downstream rotation. The quasi-exponential growth of optimal disturbances is governed by the modal instability mechanisms. Energy analysis results indicate that both the spanwise and wall-normal components of the optimal antisymmetric disturbance are relatively important downstream. For the optimal antisymmetric disturbance, the wall-normal component consistently dominates throughout the growth evolution. Optimal growth gains energy from the strong wall-normal and spanwise shear.
- 1 March 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (3)
Optimal growth in a high-speed Görtler vortex flow
Optimal growth in Görtler vortex flow over a concave wall at Mach 6 is investigated via well-established parabolized governing equations (PSE3D-APSE3D). The influence of parameters, i.e., the initial/final positions and the frequency, is studied, and the physical mechanism of optimal growth is explored. The optimal growth consists of an algebraic growth stage near the initial location and a subsequent quasi-exponential growth stage. The algebraic growth is dominated by a combination of the lift-up effect and the Orr mechanism. High-frequency disturbances initially tilt against the flow direction and amplify via the Orr mechanism during downstream rotation. The quasi-exponential growth of optimal disturbances is governed by the modal instability mechanisms. Energy analysis results indicate that both the spanwise and wall-normal components of the optimal antisymmetric disturbance are relatively important downstream. For the optimal antisymmetric disturbance, the wall-normal component consistently dominates throughout the growth evolution. Optimal growth gains energy from the strong wall-normal and spanwise shear.
- 1 February 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (2)
Vibration onset of a flexible membrane wing in an accelerating free stream
This study focuses on the vibration behavior of a flexible membrane wing encountering an accelerating free stream, aiming to unravel its transient response. Wind-tunnel experiments are carried out to synchronously measure the time-resolved membrane deformations and surrounding flow fields. Based on the time-frequency spectrum obtained by continuous wavelet transform of the vibration signal, the time interval from the non-vibration state to the establishment of the dominant second mode state is defined as the vibration onset stage. During this stage, a chordwise first vibration mode (approximately 20 Hz) initially emerges and subsequently transitions to a second mode (approximately 40 Hz). The vibration frequency of each mode increases over time due to the increasing camber and membrane tension. The nondimensional frequencies of membrane vibration generally match the Strouhal number range reported in the literature. Time-frequency analysis reveals a coupling between the membrane vibration and the velocity fluctuations within the leading-edge shear layer. As the accelerating flow impacts, the membrane gradually approaches the shear layer, perceives the disturbances therein, and destabilizes to vibrate. Notably, the generation and transition of vibration modes stem from the coupling of Kelvin–Helmholtz instability and membrane's natural frequencies, which is due to an internal resonance within the coupled fluid–structure system. The wake retains its inherent bluff-body flow features at the beginning of the vibration onset stage, after which it becomes coupled with the membrane vibration.
- 1 February 2026
- Published by AIP Publishing in Physics of Fluids
- Vol. 38 (2)
Vibration onset of a flexible membrane wing in an accelerating free stream
This study focuses on the vibration behavior of a flexible membrane wing encountering an accelerating free stream, aiming to unravel its transient response. Wind-tunnel experiments are carried out to synchronously measure the time-resolved membrane deformations and surrounding flow fields. Based on the time-frequency spectrum obtained by continuous wavelet transform of the vibration signal, the time interval from the non-vibration state to the establishment of the dominant second mode state is defined as the vibration onset stage. During this stage, a chordwise first vibration mode (approximately 20 Hz) initially emerges and subsequently transitions to a second mode (approximately 40 Hz). The vibration frequency of each mode increases over time due to the increasing camber and membrane tension. The nondimensional frequencies of membrane vibration generally match the Strouhal number range reported in the literature. Time-frequency analysis reveals a coupling between the membrane vibration and the velocity fluctuations within the leading-edge shear layer. As the accelerating flow impacts, the membrane gradually approaches the shear layer, perceives the disturbances therein, and destabilizes to vibrate. Notably, the generation and transition of vibration modes stem from the coupling of Kelvin–Helmholtz instability and membrane's natural frequencies, which is due to an internal resonance within the coupled fluid–structure system. The wake retains its inherent bluff-body flow features at the beginning of the vibration onset stage, after which it becomes coupled with the membrane vibration.