Xin Li

Xin Li

Auto-generated profile, based on publicly available information

Xin Li

State Key Laboratory for Strength and Vibration of Mechanical Structures, School of Aerospace, Xi'an Jiaotong University 1 , Xi’an 710049

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 September 2024 in Computers & Fluids
Verified
Journal Article
Closed access
XLXin Li
Xin Li
ZDZhiwen Deng
Zhiwen Deng
RFRui Feng
Rui Feng
ZLZiyang Liu
Ziyang Liu
RHRenkun Han
Renkun Han
Gang ChenGang Chen
Gang Chen

1 July 2025 in Physics of Fluids
Verified
Journal Article
Closed access
XLXin Li
Xin Li
XNXiaotian Niu
Xiaotian Niu
ZLZiyang Liu
Ziyang Liu
LJLulu Jiang
Lulu Jiang
XJXing Ji
Xing Ji
Gang ChenGang Chen
Gang Chen
Efficient and accurate aerothermal prediction has become a crucial technology for the design of hypersonic vehicles. To enhance the generalization capability for complex geometries and reduce the cost of model training, this paper introduces a novel framework for aerothermal prediction of complex vehicles, based on a hybrid graph neural network model, named hybrid-MGNN. The model consists of three main components: normalized aerodynamic heat distribution prediction, statistical feature prediction, and flow field reconstruction. Initially, the method constructs graph-structured data from discrete mesh information, enabling precise representation of intricate geometries and flow field structures. A graph neural network model based on a message-passing mechanism establishes a nonlinear mapping from incoming flow conditions to the surface heat flux distribution of the vehicle. The model then encodes the flow conditions and global geometric features, which are fed into a feature prediction network to accurately estimate statistical features of the heat flux distribution, such as maximum values. Finally, a de-normalization technique is applied to reconstruct the aerothermal flow field on the vehicle's surface. The performance of the proposed hybrid-MGNN was validated using hypersonic double ellipsoids, blunt cones, double blunt cones, and lifting body shapes. The results show that the predicted aerodynamic heat loads have an average error of less than 6% compared to computational fluid dynamics methods, with a computational efficiency improvement of approximately 400 times. Furthermore, the hybrid-MGNN framework is highly adaptable to arbitrary mesh types and complex geometries, and transfer learning significantly reduces model training costs.

27 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
XNXiaotian Niu
Xiaotian Niu
SDSiwei Dong
Siwei Dong
XYXianxu Yuan
Xianxu Yuan
XLXi Li
Xi Li
GCGang Chen
Gang Chen
The aerodynamic shape design largely determines the performance and flight quality of an aircraft, and the response speed of aerodynamic verification directly dictates the design iteration cycle. However, there remains a lack of rapid and accurate verification methods for high-fidelity three-dimensional flow fields around complex configurations. So this study presents two innovations: 1. A data extraction method named Transfinite Interpolation Multi-Scale Grid Fusion (TFI-MSGF) is proposed. Based on the MSGF method, combined with structured grid generation techniques, multi-block orthogonal grids are mapped onto a streamlined wing-body configuration. Subsequently, using deformation information, the TFI method is applied to perform 2D and 3D TFI deformation on the 3D multi-scale grid, thereby obtaining high-fidelity, body-fitted, multi-shape flow field data suitable for deep learning. The proposed TFI-MSGF method is applicable to most complex 3D configurations, ensures data accuracy, and lays a good data foundation for shape generalization in deep learning prediction. 2. A conditional guidance Res-Unet-3D deep learning model based on Feature-wise Linear Modulation (FiLM-Unet-3D) is proposed. Inspired by CFD solving, the FiLM mechanism is introduced into Res-Unet-3D, fundamentally altering the injection method of condition information. It transforms freestream conditions into dynamic scaling and shifting of feature channels. This enables hierarchical adaptive modulation of flow structures at different scales, effectively enhancing the model's capability to represent complex flows and generalize across variable conditions. The results demonstrate that the deep learning framework satisfies seven challenging indicators: complex configuration, 3D spatial flow field, high accuracy, high time efficiency, shape generalization, condition generalization, and large single-sample data volume. To the best of the authors' knowledge, this is the first time this breakthrough has been achieved. The model requires only 0.006 seconds for high-accuracy prediction of steady flow fields around complex configurations. This enables an optimization framework to directly optimize complex aerodynamic characteristics of aircraft, such as boundary layer structure, shock waves, and boundary layer velocity profiles. This breakthrough marks a paradigm shift from "optimizing only macroscopic aerodynamic coefficients" to "directly optimizing the details of the complete 3D flow field", opening new possibilities for the fine-grained aerodynamic shape design of aircraft.

1 October 2026 in Physics of Fluids
Verified
Journal Article
Closed access
SYShizhe Yu
Shizhe Yu
MLMingyue Lin
Mingyue Lin
SWShengbao Wu
Shengbao Wu
XLXin Li
Xin Li
Gang ChenGang Chen
Gang Chen
Rapid prediction of aerodynamic and aerothermal surface fields is critical for the design and real-time monitoring of cross-domain flight vehicles, yet conventional computational fluid dynamics (CFD) remains prohibitively expensive. Here, a condition-driven globally aware point cloud network (CD-GPN) is presented, achieving millisecond-scale, near-CFD-accuracy predictions of surface pressure and heat flux. Based on PointNeXt, CD-GPN integrates three core components: geometrically enhanced inverted residual modules for local geometric perception; a condition feature modulation mechanism via feature-wise linear modulation to inject flight conditions into deep feature streams; and a global interaction module on condition-modulated latent modes to capture long-range flow-field correlations. Extensively tested across Mach 5–15, altitudes of 20–40 km, and angles of attack of 0°–10°, CD-GPN outperforms mainstream baselines. Specifically, it yields a mean normalized root mean square error of 1.77% for single-configuration pressure prediction, and 1.03% and 2.07% for multi-configuration pressure and heat flux predictions, respectively. Furthermore, it demonstrates good extrapolation robustness to unseen flight conditions and novel geometries, and transfer fine-tuning with only a small number of target-configuration samples can further reduce the cross-topology blind-prediction error. Meanwhile, in terms of efficiency, it reduces the single evaluation time from 45 min (CFD) to 134 ms. These results establish CD-GPN as an efficient, physics-aware surrogate bridging offline simulation and online monitoring for next-generation aerospace vehicle design.