Ziyang Liu

Ziyang Liu

Auto-generated profile, based on publicly available information

Ziyang Liu


Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

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

Representative Publications

View All Publications
13 May 2022 in Aerospace Science and Technology
Verified
Journal Article
Closed access
ZLZiyang Liu
Ziyang Liu
RHRenkun Han
Renkun Han
MZMiao Zhang
Miao Zhang
YZYang Zhang
Yang Zhang
HZHao Zhou
Hao Zhou
GCGang Chen
Gang Chen

27 December 2022 in Aerospace Science and Technology
Verified
Journal Article
Closed access
ZDZhiwen Deng
Zhiwen Deng
HLHongsheng Liu
Hongsheng Liu
BSBeiji Shi
Beiji Shi
ZWZidong Wang
Zidong Wang
FYFan Yu
Fan Yu
GCGang Chen
Gang Chen

1 May 2024 in Physics of Fluids
Verified
Journal Article
Closed access
PDF file
JDJiaoxi Du
Jiaoxi Du
XLXin Li
Xin Li
SDSiwei Dong
Siwei Dong
ZLZiyang Liu
Ziyang Liu
Gang ChenGang Chen
Gang Chen
High Reynolds number turbulent flow of hypersonic vehicles exhibits multi-scale flow structures and non-equilibrium high-frequency characteristics, presenting a significant challenge for accurate prediction. A deep neural network integrated with attention mechanism as a reduced order model for hypersonic turbulent flow is proposed, which is capable of capturing spatiotemporal characteristics from high-dimensional numerical turbulent data directly. The network model leverages encoder–decoder architecture where the encoder captures high-level semantic information of input flow field, Convolutional Long Short-Term Memory network learns low-dimensional characteristic evolution, and the decoder generates pixel-level multi-channel flow field information. Additionally, skip connection structure is introduced at the decoding stage to enhance feature fusion while incorporating Dual-Attention-Block that automatically adjusts weights to capture spatial imbalances in turbulence distribution. Through evaluating the time generalization ability, the neural network effectively learns the evolution of multi-scale high-frequency turbulence characteristics. It enables rapid prediction of high Reynolds number turbulence evolution over time with reasonable accuracy while maintaining excellent computational efficiency.

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.