Dang Thai Son Nguyen

Dang Thai Son Nguyen

Auto-generated profile, based on publicly available information

Dang Thai Son Nguyen

School of Mechanical Engineering, Hanoi University of Science and Technology 1 , No. 01, Dai Co Viet, Hai Ba Trung, Hanoi

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
1 July 2024 in Physics of Fluids
Verified
Journal Article
Closed access
PDF file
DNDinh Thang Nguyen
Dinh Thang Nguyen
TPThi Dieu Thuy Phan
Thi Dieu Thuy Phan
VNVan Luc Nguyen
Van Luc Nguyen
TTTien-Anh Tran
Tien-Anh Tran
DNDang Thai Son Nguyen
Dang Thai Son Nguyen
VDViet Dung Duong
Viet Dung Duong
Wake transitions of flow past two tandem semi-circular cylinders near a moving wall were numerically investigated using the lattice Boltzmann method at the Reynolds number of 150 with various gap ratios (G/D, where G and D are the spacing between the cylinder and wall and the cylinder diameter, respectively) and spacing ratios (L/D, where L is the distance between the cylinder centers). The analysis aims to clarify the effects of L/D and G/D on wake structures, hydrodynamic forces, Strouhal number, and spectral energy of the flow exerted on both cylinders. This study reveals five distinct flow regimes in the L/D-G/D space, such as overshoot, continuous reattachment, pair-wise, quasi-coshedding, and coshedding. These regimes are identified using plots of vorticity contour, time history of drag and lift coefficients (CD and CL), power spectral density, and proper orthogonal decomposition of the vorticity fluctuation into deterministic spatial structures. Furthermore, the flow regime maps and diagrams of time-averaged pressure coefficients on the surface of the cylinders are given to analyze the influence of the moving wall. A significant change in CD and CL is observed for both cylinders, depending on the L/D and G/D ratios. The time-averaged drag coefficient of the downstream cylinder is remarkably lower than that of the upstream cylinder. A significant increase in the time-averaged lift coefficient of the upstream cylinder is observed when the G/D is small due to near-moving wall effects. The root-mean-squared value of the lift and drag coefficients of the downstream cylinder is lower than that of the upstream cylinder as a result of the proximity effect. Meanwhile, the Strouhal number for both cylinders remains mostly the same.

1 September 2025 in Physics of Fluids
Verified
Journal Article
Closed access
DNDang Thai Son Nguyen
Dang Thai Son Nguyen
PPPramudita Satria Palar
Pramudita Satria Palar
LZLavi Rizki Zuhal
Lavi Rizki Zuhal
Duc Nguyen DinhDuc Nguyen Dinh
Duc Nguyen Dinh
VDViet Dung Duong
Viet Dung Duong
Reconstruction of flow over porous media from sparse and noisy measurements is a critical challenge in many engineering applications, yet it is hindered by the limitations of well-known experimental and computational methods. While pioneering studies with physics-informed neural networks (PINNs) have focused on pore-scale models or idealized data, a critical gap remains: the applicability and robustness of PINNs for macroscopic, volume-averaged models have not been systematically investigated, particularly for inferring effective material properties like the Darcy number from sparse and noisy experimental data. To address this gap, this study proposes a PINN framework to solve the inverse problem by integrating the Darcy–Brinkman–Forchheimer equations (DBF-PINN). The model is designed to infer the unknown Darcy number (Da) and reconstruct the complete pressure and velocity fields, both inside and outside the porous body, using only velocity data from the external fluid region. The framework's performance is systematically validated against high-fidelity lattice Boltzmann method simulation data for flows at a Reynolds number of 100 across a Darcy number range of 10−2 to 10−4. We demonstrate that an optimized architecture incorporating a modified multi-layer perceptron, random weight factorization, and a random Fourier feature layer, combined with a novel bounded logarithmic transformation for the Darcy parameter, significantly enhances prediction accuracy and training stability. Results show that the DBF-PINN can accurately reconstruct the flow fields and infer the correct order of magnitude for the Darcy number, exhibiting remarkable robustness against sparse and noisy data (up to 20%) in high-permeability regimes (Da=10−2,10−3) and even in low-permeability scenarios Da=10−4. However, the study also identifies a potential limitation: while the PINN remains accurate at Da=10−4, primarily in terms of capturing the correct order of magnitude, its stability becomes more sensitive to noise, highlighting the challenges of applying such models to stiff problems with imperfect data.

1 August 2025 in Physics of Fluids
Verified
Journal Article
Closed access
DNDang Thai Son Nguyen
Dang Thai Son Nguyen
PPPramudita Satria Palar
Pramudita Satria Palar
INIch Long Ngo
Ich Long Ngo
Yong CaoYong Cao
Yong Cao
VDViet Dung Duong
Viet Dung Duong
This comprehensive numerical study examines dynamic flow past a permeable D-shaped cylinder near a moving wall in a wide parameter space of porosity (10−4−0.8) and gap ratio (0.6−5) at Reynolds number of 150. Wake dynamics is classified into four regions: overshoot (OS), overshoot with wavy vortex street (PW), primary and meandering vortex street (PM), and co-shedding vortex shedding street (OP). Wake transitions of these regimes are highly switchable in parameter space, where gap ratio variation controls vortex shedding in intermediate and far wake. For small gap ratios, increasing porosity shifts the wake from PW to OS, enhancing shear layer stability. At moderate gap ratios, PM and OP occur, with OP requiring higher porosity due to flow penetration. Pressure coefficient distributions approach a rigid cylinder's as porosity decreases. Stagnation points shift toward the moving wall at small gap ratios but align with an isolated circular cylinder at large gap ratios. Higher porosity shifts stagnation points away from the wall at small gap ratios. Decreasing porosity reduces the time-averaged drag coefficient due to increased permeability. PM and OP regions show higher drag coefficient than OS and PW, exceeding those of single circular and square cylinders due to the moving wall. Lift coefficients approach a single circular cylinder's, except at small gap ratios, where the moving wall effect dominates. At large gap ratios, porosity has a minor effect on the fluctuation of lift coefficient, but a small gap ratio increases the maximum lift coefficient compared to a single circular cylinder.

28 May 2026
Verified
Book Chapter
Closed access
NNNhi Nguyen Ngoc
Nhi Nguyen Ngoc
DMDo Tran Tuan Minh
Do Tran Tuan Minh
TTTran Thi Phuong Thao
Tran Thi Phuong Thao
NHNguyen Huy Hoang
Nguyen Huy Hoang
STSon Nguyen Dang Thai
Son Nguyen Dang Thai
DVDung Duong Viet
Dung Duong Viet

1 February 2026 in Physics of Fluids
Verified
Journal Article
Closed access
KZKemas Zakaria
Kemas Zakaria
BGBryan Goh
Bryan Goh
DNDang Thai Son Nguyen
Dang Thai Son Nguyen
VDViet Dung Duong
Viet Dung Duong
JMJoseph Morlier
Joseph Morlier
PPPramudita Satria Palar
Pramudita Satria Palar
With the growing use of machine learning in fluid dynamics, autoencoder-based models have become a popular tool for learning low-dimensional representations of high-dimensional flow fields, where improving reconstruction accuracy remains a central focus. To that end, we propose the Fourier neural autoencoder: an autoencoder architecture enhanced with Fourier-based layers inspired by the Fourier neural operator, aimed at improving reconstruction accuracy. To further boost its capability, we develop a hyperbolic feature mapping technique that improves feature extraction. To explore our models in greater depth, we apply them to both non-variational and variational autoencoder variants. Since the Fourier neural autoencoder and hyperbolic feature mapping are independent of each other, we obtain four variants of our proposed approach: the Fourier neural autoencoder with and without hyperbolic feature mapping, each in both non-variational and variational forms. As a benchmark case, we apply our Fourier neural autoencoder to the flow past a single cylinder at a Reynolds number of 70. The results show that the model achieves higher reconstruction accuracy while having fewer parameters compared to a standard convolutional autoencoder. We further evaluate the model on more complex flow scenarios, including the unsteady Reynolds-averaged Navier–Stokes simulation of a NACA0012 airfoil at a Reynolds number of 1000 and the flow over two side-by-side circular cylinders at a Reynolds number of 100 with varying spacing. Our results suggest that both the Fourier neural autoencoder and hyperbolic feature mapping improve accuracy while reducing parameter counts compared to the standard autoencoder.