Abstract: Our team has developed an efficient visualization system for billion-grid reservoir models, integrating large-scale data loading, parallel rendering optimization, dynamic profile generation, and attribute animation display. The system uses the OSG 3D engine and D2D 2D graphics acceleration, supporting real-time interaction and multi-mode display, significantly improving the visualization efficiency and engineering application value of reservoir simulation results.

Speaker Bio: Professor and doctoral supervisor at Xi'an Shiyou University. Holds a PhD in Mechanics from Peking University and completed postdoctoral research at the China Petroleum Exploration and Development Research Institute. Has long been committed to numerical simulation of complex oil and gas reservoirs and the development of integrated geological engineering methods and software. Since joining Xi'an Shiyou University in 2021, has accumulated extensive experience in developing and industrializing large software systems in teaching and research. Published over 30 SCI/EI papers, authorized 20 invention patents, authored 2 academic monographs, won three provincial and ministerial first prizes for scientific and technological progress, and led 2 national-level projects and over 10 other research projects.

Abstract: This talk focuses on industrial scenarios, exploring the applications of visualization and visual analytics technologies for flow fields and fluorine materials. Flow field visualization technology clearly presents the dynamics of industrial fluid motion, while fluorine material visualization intuitively displays its distribution and state. Through practical cases, the talk elaborates on their specific applications in industrial production monitoring, fault diagnosis, and process optimization, analyzing the benefits and value they bring, providing references for technology upgrades and applications in the industrial field.

Speaker Bio: Daoming Lü, Associate Professor at Sichuan University of Light and Chemical Industry, Deputy Director of Sichuan Province Big Data Visual Analytics Technology Engineering Laboratory. His research focuses on artificial intelligence and visualization, with work published in internationally renowned journals and conferences such as IEEE T-NNLS, IEEE T-ETCI, iScience, AAAI, IJCAI, NeurIPS, and AAMAS. He has served as a reviewer for conferences and journals including AAAI, IJCAI, NeurIPS, ICML, ICLR, UAI, Smart Health, Machine Learning Journal, IEEE-TNNLS, and ACM TIST. He received the Best Paper Award at the AAMAS-2022 OptLearnMAS workshop and the Best Reviewer Award at UAI-2022.

Abstract: With the rapid development of supercomputers, the scale of numerical simulations performed by domain experts is increasing, making visual analytics more challenging. To address this, we summarized the challenges of visual analytics for ultra-large-scale numerical simulations and explored possible solutions. First, to alleviate the storage and transmission pressure of large-scale numerical simulation results, we conducted research on data reduction. Then, to avoid the loss of important features in reduced data, we designed a data reconstruction scheme to provide domain experts with high-precision numerical simulation details. Our data reduction and reconstruction scheme alleviates storage and transmission pressure while ensuring the accuracy of reconstructed scientific data.

Speaker Bio: Zhongke Bi, Professor at Tianjin University's School of Intelligence and Computing, doctoral supervisor, and Chief Scientist of the National Key R&D Program. He received his PhD in Science from the University of Tokyo in 2012 and worked as a researcher at Japan's RIKEN from 2012 to 2016 before joining Tianjin University. His main research areas are high-performance computing, visualization, and artificial intelligence. He serves as a standing committee member of CSIG Visualization and Visual Analytics, a member of CCF High-Performance Computing, and a member of CCF Computer-Aided Design and Graphics. Over the past five years, he has led one National Key R&D Program, one 173 Project, two National Natural Science Foundation projects, two pre-research projects, and two National Numerical Wind Tunnel key projects. He has participated in three National Natural Science Foundation projects (including one key project), four ministerial projects, and one Tianjin key project. He has published over 100 papers in high-performance computing and visualization, with applications in nuclear power, environmental protection, and numerical wind tunnels.

Abstract: High-fidelity visual analytics is a core technology for deconstructing knowledge and patterns in complex flow field data. However, traditional methods are forced to adopt spatiotemporal sparse sampling due to storage limitations, leading to the loss of dynamic features. Existing data-driven models lack deep embedding of physical knowledge and uncertainty quantification, making them difficult to meet engineering credibility requirements. To address these issues, this talk proposes an innovative framework combining 'data-driven + physics-guided + uncertainty quantification' and conducts the following research: (1) Based on the self-developed NNW series CFD software, a high-fidelity dataset HFR-Beach balancing grid resolution and numerical accuracy is constructed to provide high-quality data support for intelligent models; (2) From data representation, model architecture, and training strategy, physics-guided intelligent model design methods (PCSAGAN, PgTransGAN, PEINR) are proposed to enhance model interpretability and generalization; (3) Bayesian methods are introduced to establish a credibility evaluation system integrating uncertainty quantification, ensuring the scientific and reliability of intelligent model decisions. The research results form high-precision, interpretable high-fidelity visual analytics tools, promoting the paradigm shift of intelligent flow field analysis from 'black-box fitting' to 'white-box decision-making'.

Speaker Bio: Dr. Liang Deng, Senior Engineer at the China Aerodynamics Research and Development Center, has long been engaged in high-performance CFD computation and intelligent analysis of fluid simulation big data. He was selected for the 'Wind Control Aerospace' Youth Talent Support Program of the China Aerodynamics Research and Development Center. He has led five projects, including national defense technology fund projects, Sichuan Province key R&D projects, and equipment pre-research projects, and participated in six major projects, including the National Numerical Wind Tunnel Project, National Key R&D Program, and National Natural Science Foundation. He has published over 30 papers in important domestic and international journals and conferences such as ICML, PPoPP, SC, DAC, TACO, TNNLS, and JPDC, authored one academic monograph, and holds over 10 national invention patents and software copyrights.