A5035394736- Nuclear reactor physics and engineering
- Nuclear Materials and Properties
- Fusion materials and technologies
- Parallel Computing and Optimization Techniques
- Matrix Theory and Algorithms
- Hydrology and Watershed Management Studies
- Nuclear Engineering Thermal-Hydraulics
- Computational Physics and Python Applications
- Advanced Data Storage Technologies
- Heat transfer and supercritical fluids
- CO2 Sequestration and Geologic Interactions
- Ferroelectric and Negative Capacitance Devices
- Advanced Numerical Methods in Computational Mathematics
- Machine Learning in Materials Science
- Computer Graphics and Visualization Techniques
- Flood Risk Assessment and Management
- Hydrological Forecasting Using AI
- Graphite, nuclear technology, radiation studies
- Distributed and Parallel Computing Systems
University of Science and Technology Beijing
2019-2023
The reactor is an extreme environment of high temperature, pressure, and radiation dose. Damage accumulates in structural materials over service time. It will eventually lead to material failure, such as hardening, embrittlement, swelling. damage accumulation inherent, complex, multiscale process, which has been studied extensively with modeling simulation methods. rapid development high-performance computing makes it possible accurately operate for the microstructure evolution irradiated...
Sparse matrix-vector multiplication (SpMV) is a fundamental building block for various numerical computing applications. However, most existing GPU-SpMV approaches may suffer from either long preprocessing overhead, load imbalance, format conversion, bad memory access patterns. In this paper, we proposed two new SpMV algorithms:flat andline-enhance, as well their implementations, GPU systems to overcome the above shortcomings. Our algorithms work directly on CSR sparse matrix format. To...
ABSTRACT Parameter calibration is an important part of hydrological simulation and affects the final results. In this paper, we introduce heuristic optimization algorithms, genetic algorithm (GA) to cope with complexity parameter problem, use particle swarm (PSO) as a comparison. For large-scale simulations, multilevel parallel framework make full processor resources, accelerate process solving high-dimensional calibration. Further, test apply experiments on domestic supercomputers. The...
High fidelity simulation of fission gas behavior is able to help us understand and predict the performance nuclear fuel under different irradiation conditions. Cluster dynamics (CD) a mesoscale method which rapidly developed in research area recent years, it can effectively describe microdynamic fuel; however, due huge cost computation needed for CD model solution, application scenario has been limited. Thus, how design acceleration algorithm given computing resources improve efficiency...
Summary The performance of sparse stiffness matrix‐vector multiplication is essential for large‐scale structural mechanics numerical simulation. Compressed row (CSR) the most common format storing matrices. However, high sparsity matrix makes number nonzero elements per very small. Therefore, CSR‐scalar algorithm, light and HOLA algorithm in calculation will cause some threads GPU to be idle state, which not only affect computing but also waste resources. In this article, a new CSR‐vector...
Molecular dynamics is a significant method in computational material and many other fields, computationally intensive. In this paper, we present new molecular software, MISA-MD, which mainly designed for high performance on GPU architecture. The main contributions include redesign of the atom data structure, communication optimizations, parallel strategies series optimizations memory accessing. Compared to CPU version, presented version MISA-MD can achieve more than 130 speedup when running...
Radiation damage to the steel material of reactor pressure vessels is a major threat nuclear safety. It caused by metal atom cascade collision, initialized when atoms are struck high-energy neutron. The paper presents MISA-MD, new implementation molecular dynamics, simulate such collision with EAM potential. MISA-MD realizes (1) hash-based data structure efficiently store an and find its neighbors, (2) several acceleration optimization strategies based on SW26010 processor Sunway Taihulight...