XMXuerui Mao

Xuerui Mao

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Xuerui Mao has a total of 108 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Linear perturbation analysis of flow over a flat plate with an elliptic leading edge

    by Htet Myet Ne Ye, Yiping Yao, Dandan Xiao, Xuerui Mao

    Linear perturbation analysis is used to investigate the impact of leading-edge geometry on disturbance amplification, and a flat plate is adopted to eliminate other geometric factors, e.g., thickness, camber, etc. The two-dimensional (2D) base flow is first examined across a range of Reynolds numbers (Re) and leading-edge aspect ratios (AR). The threshold for the formation of a laminar separation bubble (LSB) on the AR-Re plane is identified, with the critical AR increasing monotonically with Re. At Re=300, the optimal energy gain exhibits two competing branches: a three-dimensional (3D) steady branch at inflow frequency ω=0 and finite spanwise wavenumber β, and a 2D unsteady branch at β=0 and finite ω. For higher AR with attached base flow, the dominant response is the 3D steady branch, and its optimal perturbation structure and downstream evolution are consistent with the lift-up mechanism, while finite-frequency responses are weaker. For lower AR with an LSB, the dominant response is the 2D unsteady branch, which exhibits strong finite-frequency amplification over a broad frequency range and downstream wave-packet structures concentrated along the separated shear layer, consistent with Kelvin–Helmholtz instability. A transitional regime is identified near AR≈0.6, where the 3D steady and 2D unsteady branches attain comparable gains, and a 3D unsteady mode with mixed features is observed to be the most energetic one. As Re increases, the crossover between the 2D unsteady and 3D steady branches shifts to lower AR. These results demonstrate that the leading-edge geometry dominates the amplification of free-stream disturbance through its influence on leading-edge separation.

  • Noise-robust temporal super-resolution of three-dimensional turbulent flow using an attention-enhanced convolutional LSTM network

    by Lei Dong, Dandan Xiao, Jie Yao, Xuerui Mao

    Existing temporal super-resolution methods struggle to reconstruct three-dimensional turbulent flows across large time gaps, often losing coherent vortical structures and giving flow-specific performance. We introduce ResSE-LSTM, an attention-enhanced convolutional LSTM (long short-term memory) framework that recovers intermediate 3D velocity fields from two sparse snapshots and remains robust to noisy inputs. A Kolmogorov-scale temporal parameter, \ensuremath{\Pi}\ensuremath{\mathit{t}}, unifies performance across Reynolds numbers. The model recovers Reynolds stresses and coherent structures up to \ensuremath{\Pi}\ensuremath{\mathit{t}} \ensuremath{\le} 4, versus \ensuremath{\Pi}\ensuremath{\mathit{t}} \ensuremath{\le} 0.8 for a CNN baseline, enabling reliable reconstruction from sparse simulations and measurements.

  • Effects of the Atwood number on three-dimensional azimuthal instability of a shock-accelerated spherical bubble

    by Yang Wang, Yongkai Chen, Gang Dong, Xianxu Yuan, Xuerui Mao

    The shock-driven evolution of a three-dimensional (3D) spherical gas bubble occurs frequently in natural phenomena and engineering applications. In the present study, the interaction between a planar shock wave and a spherical interface is investigated numerically by solving the 3D, compressible Euler equations using a fifth-order weighted essentially non-oscillatory scheme. A series of numerical cases spanning a wide range of Atwood (At) numbers is performed to examine the shock–bubble interaction (SBI), with emphasis on azimuthal instability. The results indicate that a primary vortex ring is consistently formed following shock acceleration of the spherical interface, whereas a leading vortex ring appears only in the case of heavy bubbles (At>0). As the distorted interface evolves, a pronounced At-dependent azimuthal instability emerges, exhibiting a clear modal distribution dominated by the m=4 mode: for heavy bubbles, the growth of azimuthal instability increases with increasing At number. Larger At numbers lead to stronger vortex ring formation; the accompanying shock wave reflections further reinforce the azimuthal disturbances, resulting in the growth of azimuthal modes over a broad range of scales. As the At number decreases, the azimuthal instability shifts toward the gradual amplification of small-scale perturbations. In contrast, for light bubbles (At<0), strong convective transport promotes a more pronounced development of azimuthal modes. These findings highlight the critical role of azimuthal instability in SBI beyond traditional streamwise analyses, with the dominant modal behavior consistently observed but its quantitative characteristics dependent on the adopted computational configuration.

  • Optimal free-stream perturbation to an airfoil flow

    by Yiping Yao, Dandan Xiao, Xuerui Mao

    Free-stream perturbations induce large coherent structures, crucial to the laminar–turbulent transition on flight vehicles and turbine blades, significantly affecting drag, heat transfer, noise, etc. This paper computes the optimal free-stream perturbations that undergo the strongest amplification using an adjoint method, specifically for the incompressible flow around a NACA0012 airfoil, in contrast to sharp (or no) leading-edge configurations. These perturbations are examined at different angles of attack (0 Superscript ring ∘ $^{\circ }$ – 6 Superscript ring 6 ∘ $6^{\circ }$ ) and Reynolds numbers ( italic Re equals 1000 Re = 1000 ${\textit{Re}} = 1000$ – 8000 8000 $8000$ ). At low angles of attack, two extreme points of energy growth emerge in the frequency–spanwise wavenumber plane, both on the axes. One has a spanwise wavenumber of zero, corresponding to two-dimensional unsteady modes in the wake, and the other has a frequency of zero, representing three-dimensional steady boundary-layer modes. The former is driven by the inflection point of the base velocity profile in the wake and is stronger, whereas the latter is related to the lift-up non-modal stability mechanism and directly influences the flow on the airfoil surfaces. The distribution of the three-dimensional steady optimal perturbations, similarly to the thickness of the base flow boundary layer, follows a italic Re Superscript negative 0.5 Re − 0.5 ${\textit{Re}}^{-0.5}$ scaling law. This scaling enables the projection of optimal profiles obtained at low italic Re Re ${\textit{Re}}$ to high ones for engineering applications, where the direct calculation is prohibitively expensive. Subsequently, the linear optimal perturbation at italic Re equals 100 000 Re = 100 000 ${\textit{Re}} = 100\,000$ is obtained, and the generation, nonlinear deformation and breakdown of streaks are examined by direct numerical simulation to illustrate the impact of the inflow perturbations on transition.


  • Exploiting temporal coherence for self-supervised denoising of turbulent flows

    by Lei Dong, Dandan Xiao, Jie Yao, Xuerui Mao

    Denoising plays a crucial role in experimental fluid mechanics, where measurement noise can significantly contaminate flow structures and distort statistical quantities. A self-supervised denoising framework, incorporating a squeeze-and-excitation (SE) attention module and a convolutional long short-term memory (ConvLSTM) unit, referred to as SE–ConvLSTM, is developed by exploiting the temporal coherence of flows. The model reconstructs the target flow field from a sequence of preceding noisy snapshots, with performance primarily governed by the temporal correlation of the input data, determined by the input time span and the temporal interval between snapshots. The denoising capability of SE–ConvLSTM is systematically evaluated for two-dimensional flow fields under a wide range of noise intensities, noise types, and flow configurations, with comparisons against both classical and deep-learning methods. The results demonstrate that SE–ConvLSTM achieves superior performance for temporally uncorrelated noise, enabling more accurate noise removal while preserving fine-scale flow structures, spatial continuity, and turbulent statistics, particularly under severe noise contamination. The denoising performance is closely related to the ratio of the integral time scales of the noisy and reference flows, with improved accuracy obtained when temporal coherence is better preserved in the noisy field. Extension to three-dimensional flow fields further confirms the robustness of the framework in reconstructing volumetric structures. Overall, the SE–ConvLSTM provides an effective and generalizable approach for denoising noisy flow measurements.

  • Prediction of bypass transition from wall measurements

    by Dandan Xiao, Jan Oscar Pralits, Xuerui Mao

    Prediction of bypass transition from wall measurements in a zero-pressure-gradient boundary-layer flow over a flat plate with a leading edge is investigated using adjoint-variational optimisation techniques and Hessian analyses. The reference wall data are obtained from separate direct numerical simulations subjected to inflows with increasing levels of complexity. The prediction is achieved by reconstructing an inflow profile that generates wall data matching the reference. Initially, the algorithm is validated through steady inflows constructed from a single vortical mode. Positive streaks are reconstructed accurately, while negative streak predictions worsen with increasing turbulence intensity due to their reduced wall impact. Hessian analysis reveals that, under low turbulence intensity, the most observable inflow structures deviating from the optimal profile align with classical streak-generating modes, whereas at higher turbulence levels, the dominant observable directions shift toward spanwise uniform perturbations in high-shear regions, highlighting a reduced estimation robustness. Secondly, an inflow constructed from multiple vortical modes effectively demonstrates the prediction of streak meandering motions, with the optimised inflow profile differing significantly from the reference one. Finally, a more elaborate inflow in the form of free-stream turbulence is used to trigger bypass transition, and the results indicate that primary flow structures can be reconstructed, while the accuracy of the details improves when more downstream measurement data are included. These findings highlight the potential for accurate reconstruction of boundary-layer transition from wall measurement data.

  • AeroTinyNet: an attention-enhanced framework for tiny object detection in remote sensing imagery

    by Xiaozheng Jiang, Wei Zhang, Xuerui Mao, Haifeng Tan

    Detecting tiny and small objects in remote sensing (RS) imagery has long been a challenging task due to their minimal spatial information, weak feature representations, and dense distributions across complex backgrounds. Despite numerous efforts devoted, mainstream detectors still underperform in such scenarios. To bridge this gap, we introduce AeroTinyNet, a detection framework for tiny and small object detection in various RS scenarios. The core design of AeroTinyNet is a spatial-channel hybrid attention (SCHA) mechanism. It models attention along the channel and spatial dimensions, where both dimensions integrate global and local contextual features to enhance the tiny object representation and suppress background interference. Moreover, SCHA uses lightweight pooling and convolution operations, allowing seamless integration into different stages of the detection pipeline for feature refinement. Comprehensive experiments conducted on the public RS dataset AI-TOD demonstrate that AeroTinyNet improves AP and AP 75 by 4.0% and 6.5% over the best compared method, respectively, and VisDrone shows that our model achieves a 1.7% improvement in mAP 50 over recent high-performing detectors. Evaluations on DIOR and RS-STOD benchmark dataset further validate its favorable detection performance in diverse RS scenarios. These results indicate that the proposed network provides a competitive and effective approach for tiny object detection on the evaluated RS benchmarks.

  • Data-based transient growth analysis

    by Yin Wang, Xuerui Mao

    Transient growth analysis has been extensively studied in asymptotically stable flows to identify their short-term amplification of perturbations. Generally, in global transient growth analyses, matrix-free methods are adopted, requiring the construction of adjoint equations, either in the discrete or continuous form. This paper introduces a data-driven algorithm that circumvents the adjoint equations by extracting the optimal initial perturbation and its energy growth over a specified time horizon from transient snapshots of perturbations. This method is validated using data from the linearised complex Ginzburg–Landau equation, backward-facing step flow and the Batchelor vortex. Unlike model-based methods, which require $S$ sets of integrations of the linearised governing equation and its adjoint for $S$ time horizons, the proposed approach collects the snapshots of $S$ time horizons in one integration of the linearised equation. Furthermore, this study provides a robust framework for utilising proper orthogonal decomposition modes to synthesise optimal modes. The developed capacity to conduct transient growth analyses without solving the adjoint equations is expected to significantly reduce the barriers to transient dynamics research.

  • Open Access

    RShDet: An Adaptive Spectral-Aware Network for Remote Sensing Object Detection Under Haze Corruption

    by Wei Zhang, Yuantao Wang, Haowei Yang, Xuerui Mao

    Remote sensing (RS) object detection faces intrinsic challenges arising from the overhead imaging paradigm and the diversity of climatic conditions. In particular, atmospheric phenomena such as clouds and haze cause severe visual degradation, making reliable object detection difficult. However, most existing detectors are developed under clear-weather conditions, which limits their generalization capability in realistic haze-degraded RS scenarios. To alleviate this issue, an adaptive spectral-aware network for RS object detection under haze interference is proposed, termed RShDet, which is designed to handle both high-altitude RS imagery and low-altitude Unmanned Aerial Vehicle (UAV) scenarios. Firstly, the Object-Centered Dynamic Enhancement (OCDE) module dynamically adjusts the spatial positions of key-value pairs through query-agnostic offsets, enabling the network to emphasize object-relevant regions while suppressing haze-induced background interference. Secondly, the Dynamic Multi-Spectral Perception and Filtering (DSPF) module introduces a multi-spectral attention mechanism that adaptively selects informative frequency components, thereby enhancing discriminative feature representations in hazy environments. Thirdly, the Frequency-Domain Multi-Feature Fusion (FDMF) module employs learnable weights to complementarily integrate amplitude and phase information in the frequency domain, enabling effective cross-task feature interaction between the enhancement and detection branches. Extensive experiments demonstrate that RShDet consistently achieves superior detection performance under hazy conditions across both synthetic and real-world benchmarks. Specifically, it achieves improvements of 2.4% mAP50 on Hazy-DOTA, 1.9% mAP on HazyDet, and 2.33% mAP on the real-world foggy dataset RTTS, surpassing existing state-of-the-art methods.