Nhảy đến nội dung

Seminar khoa học của GS.TS. Nguyễn Xuân Hùng (HUTECH), PGS.TS David D.L. Minh (Illinois Institute of Technology), và TS. Nguyễn Thành Nhơn (AIMaS)

Vào ngày 24/2/2023, Viện AIMaS tổ chức buổi trao đổi học thuật với các chuyên gia hàng đầu trong lĩnh vực tính toán tại Phòng họp C.


Đến với buổi trao đổi học thuật, AIMaS vinh dự đón các chuyên gia hàng đầu trong và ngoài nước: GS.TS. Phạm Đức Chính (VAST), GS.TS. Nguyễn Xuân Hùng (HUTECH), PGS.TS David D.L. Minh (Illinois Institute of Technology), TS. Phùng Văn Phúc (HUTECH).

Ngoài ra, Viện cũng vinh dự tiếp đón TS. Nguyễn Trường Huy (Phụ trách khoa Dược) và TS. Nguyễn Ngọc Tuấn (Trợ lý Trưởng khoa Khoa học Ứng dụng) đến và giao lưu với Viện.


Các báo cáo gồm:

“3D printing process optimization based on machine learning” – GS. Nguyễn Xuân Hùng (HUTECH) 
“Asymmetry and ligand binding in the SARS-CoV-2 main protease” – PGS. TS David D. L. Minh (Illinois Institute of Technology)
“Adaptive higher oder phase-field modeling of brittle and ductile fracture” – TS. Nguyễn Thành Nhơn (AIMaS)


Chi tiết hơn:

GS. Hùng trình bày về: 

3D printing technology or additive manufacturing has become highly applicable in engineering, especially in the Era of Digital Tranformation. The industry has brought a facelift to most others. However, this technology still exists some drawbacks, and it therefore has not been generalised to bring the best benefits to users. Recently, we proposed a data-driven machine learning platform [1] for predicting optimised parameters of the 3D printing process from a model design to a complete product. This finding can open up great advances in the current 3D printing technology. Accordingly, the results obtained allow us to predict quickly and accurately some decisive parameters of the traditional 3D printing process such as time, weight and length while the input was fuzzy with a part of the initial information missing. The method does not only need to account for the shape, size and material of the printed object, but also it can perform the process automatically without other extra factors. After completing the model, a configurator is proposed to set the parameters for the respective printer types, which makes the 3D printing process simple and fast. More detail for this work has been recently published in [2,3]. 

[1] Thang Le-Duc, Quoc-Hung Nguyen, Jaehong Lee, H Nguyen-Xuan, Strengthening Gradient Descent by Sequential Motion Optimization for Deep Neural Networks, IEEE Transactions on Evolutionary Computation, in press, 2022.
[2] Phuong Dong Nguyen, Thanh Q Nguyen, QB Tao, Frank Vogel, H Nguyen-Xuan, A data-driven machine learning approach for the 3D printing process optimisation, Virtual and Physical Prototyping, 17, 768-786.
[3]  https://ht3dprint.com/ 



GS. Minh trình bày về:

The coronavirus 3C-like main protease (MPro) is an important target of COVID-19 drugs. Over the last few years, my group has worked with the COVID Moonshot, an effort to produce an open-source antiviral that targets MPro. We have analyzed molecular simulations to better understand the asymmetry of the enzyme and have developed improved statistical analyses of concentration-response curves from MPro enzymatic inhibition assays. Although it is a homodimer with chemically identically subunits, evidence from pH-dependent enzymatic activity, crystal structures, and simulations suggest that at least under certain conditions the enzyme has half-of-the-sites activity due to geometric asymmetry of the subunits. To investigate the transitions between symmetric and asymmetric states, we have analyzed a series of molecular dynamics trajectories with an aggregate time of 2.9 milliseconds from the Folding@Home distributed computing project. Our analysis identified key residues that change conformation during the transitions and interactions that stabilize asymmetric states.
While trying to develop mechanistic models to fit to concentration-response curves for MPro enzyme inhibition assays, we realized that standard statistical analyses of this ubiquitous class of data do not use all available information. National Center for Advancing Translational Sciences (NCATS) guidelines specify that readouts should be normalized by the controls before fitting the curves, but that the control data are not used in fitting. We developed a fitting procedure that includes the control data in the fitting.
Simulations indicate that the proposed procedure provides more precise estimates of parameters compared to previously recommended practices. Analysis of MPro enzymatic inhibition assays demonstrates that the proposed procedure provides more precise estimates of all parameters, including the IC50 and the Hill Slope.


TS. Nguyễn Thành Nhơn trình bày về:

    The phase field method (PFM) is a numerical simulation technique that has received increasing attention in the field of fracture mechanics in recent years. The main idea of the PFM is to represent the crack propagation process in a continuous manner, rather than the traditional discrete representation. This enables the model to capture the complex and multi-scale behavior of crack initiation and propagation.

    One of the key advantages of the PFM is its ability to simulate the crack propagation process in a way that is consistent with the underlying physical laws, such as energy conservation and the equivalence of energetic and dissipated crack tip fields. Therefore, the application of phase-field modeling eliminates the need for additional effort to algorithmically monitor the fracture surfaces. Although techniques such as the extended finite element method, which involve adding discontinuities to the displacement field through the enhancement of basis functions, have proven to be effective in modeling two-dimensional fracture issues, they encounter difficulties when applied to three-dimensional problems. Another benefit of the PFM is its ability to capture the influence of various microstructural and material factors on crack initiation and propagation. For example, the method can account for the effect of microstructure on the crack path, the effect of material inhomogeneities on the crack tip fields, and the effect of stress triaxiality on crack stability.

    Despite these advantages, the PFM also has some limitations, particularly in terms of computational cost. The method requires a large number of iterations and fine mesh, which can be time-consuming and computationally demanding. To enhance the computational efficiency and simplify the process of combining adaptive mesh refinement with the PFM, the isogeometric-meshfree method has been proposed for the phase-field modeling of brittle and ductile fracture in elasto-plastic materials. Moreover, the discrete equations for displacement and phase-field have been generalized to accommodate the calculation of both second- and fourth-order gradients, which are required in resolving phase-field models. The fourth-order phase-field model has the ability to accurately depict crack surfaces with less mesh density compared to the second-order model. In addition, the fourth-order phase-field equation can be resolved directly without the need to divide it into two second-order differential equations, due to the smoothness and higher order continuity of the proposed method.

    In conclusion, the PFM of fracture is a promising numerical simulation technique that has the potential to revolutionize our understanding of fracture mechanics. With continued advances in computational capabilities and the development of more compticated models, the phase field method is likely to become an increasingly important tool for solving complex fracture problems. Moreover, the present approach has the capability to efficiently implement the C1-continuity of a crack phase-field, which is essential for the fourth-order model. The results show that the proposed method is effective in capturing various features of the brittle and ductile fracture phenomenon, including plastic localization, crack propagation, and merging.