Cong Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Cong Liu has a total of 21 co-authors. Here are the top co-authors by number of shared publications:
- Bin Zhang
School of Nuclear Science and Engineering, North China Electric Power University, Beijing 102206, China
8 shared publications - Liang Zhang
North China Electric Power University, No. 2 Beinong Road, Changping District, Beijing 102206, China
8 shared publications Yixue Chen
North China Electric Power University, School of Nuclear Science and Engineering, Beijing 102206, China
6 shared publicationsBin Zhang
4 shared publicationsShuang Tan
3 shared publicationsLiang Zhang
China Ship Development and Design Center, Wuhan 430064, China
2 shared publications
Latest Publications
- Open Access
- 12 June 2026
- Published by Elsevier in SSRN Electronic Journal
Multi-level parallel SN DFEM algorithm for neutron transport problems on 3-D unstructured grids
Efficient computational methods have long been a major focus and challenge in the field of particle transport, and the unstructured grid discrete ordinates (SN) method is one of the effective numerical approaches for neutron transport simulations. However, the massive computational and storage demands of three-dimensional (3-D) unstructured grid SN calculations constitute a bottleneck constraining the large-scale multi-physics coupling simulations. The serial nature of SN sweeps and the complex data dependencies within 3-D unstructured grids result in limited algorithmic parallelism, complex task scheduling, and high communication latency. These factors pose significant challenges to achieving efficient, scalable parallel computing for large-scale application problems. To address these issues, we develop an SN angle-space dynamic scheduling algorithm and a load-balanced energy group local communication algorithm. Furthermore, we advance a multi-level parallel computations integrating space, angle, and energy groups. Numerical test results demonstrate the high scalability of our multi-level parallel algorithm, and the parallel efficiency at 65536 cores relative to 2048 cores reaches 85%. This supports efficient simulations of neutron transport problems on 3-D unstructured grids, scalable to 100000 cores and over 100 billion degrees of freedom.
- 12 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessMulti-level parallel SN DFEM algorithm for neutron transport problems on 3-D unstructured grids
Efficient computational methods have long been a major focus and challenge in the field of particle transport, and the unstructured grid discrete ordinates (SN) method is one of the effective numerical approaches for neutron transport simulations. However, the massive computational and storage demands of three-dimensional (3-D) unstructured grid SN calculations constitute a bottleneck constraining the large-scale multi-physics coupling simulations. The serial nature of SN sweeps and the complex data dependencies within 3-D unstructured grids result in limited algorithmic parallelism, complex task scheduling, and high communication latency. These factors pose significant challenges to achieving efficient, scalable parallel computing for large-scale application problems. To address these issues, we develop an SN angle-space dynamic scheduling algorithm and a load-balanced energy group local communication algorithm. Furthermore, we advance a multi-level parallel computations integrating space, angle, and energy groups. Numerical test results demonstrate the high scalability of our multi-level parallel algorithm, and the parallel efficiency at 65536 cores relative to 2048 cores reaches 85%. This supports efficient simulations of neutron transport problems on 3-D unstructured grids, scalable to 100000 cores and over 100 billion degrees of freedom.
- Open Access
- 20 October 2022
- Published by Elsevier in Nuclear Engineering and Technology
- Vol. 55 (2) , 756-768
Verification of multilevel octree grid algorithm of S transport calculation with the Balakovo-3 VVER-1000 neutron dosimetry benchmark
- 20 October 2022
- Published by Elsevier in Nuclear Engineering and Technology
- Vol. 55 (2) , 756-768
Open AccessVerification of multilevel octree grid algorithm of S transport calculation with the Balakovo-3 VVER-1000 neutron dosimetry benchmark
- 14 September 2020
- Published by Taylor & Francis in Nuclear Science and Engineering
- Vol. 194 (12) , 1175-1201
Nonmatching Discontinuous Cartesian Grid Algorithm Based on the Multilevel Octree Architecture for Discrete Ordinates Transport Calculation
Obtaining sufficiently accurate geometric descriptions is a crucial prerequisite for reliable particle transport calculations. Conventional transport algorithms on Cartesian grids use a highly efficient sweep technique and numerous mature discretization methods despite their modeling deficiency for complex geometries. To achieve a more accurate geometric description, a cell-based nonmatching Cartesian grid algorithm is proposed on the basis of the multilevel octree architecture. Transport sweep is performed according to the tree branch relationship of nested mesh distributions. The angular flux transmission between discontinuous grids is handled by the flux spatial moment mapping technique, and multiple zero-order mapping schemes are developed, including finite element interpolation, distance interpolation, and exponential fitting methods for treating upwind flux distributions of different relative shapes. The first-order mapping schemes are modified and improved for linear and exponential short characteristic discretization methods. The mapping accuracy is evaluated for polynomial and exponential functions, and a new spatial shape factor is presented for measuring the degree of nonlinearity. The multilevel octree grid (MLTG) algorithm is tested for neutron transport benchmarks, and good agreement with Monte Carlo and standard SN results is achieved. The number of meshes in the VVER shielding model is reduced from 18 million to 2 million using 3-level octree grids with the same geometric description accuracy. Numerical verification of a one-group fixed-source problem shows that 4-level and 5-level MLTG results with proper spatial discretization schemes can achieve relative deviations of less than 3% and 5% for detector region flux, respectively.
- 14 September 2020
- Published by Taylor & Francis in Nuclear Science and Engineering
- Vol. 194 (12) , 1175-1201
Nonmatching Discontinuous Cartesian Grid Algorithm Based on the Multilevel Octree Architecture for Discrete Ordinates Transport Calculation
Obtaining sufficiently accurate geometric descriptions is a crucial prerequisite for reliable particle transport calculations. Conventional transport algorithms on Cartesian grids use a highly efficient sweep technique and numerous mature discretization methods despite their modeling deficiency for complex geometries. To achieve a more accurate geometric description, a cell-based nonmatching Cartesian grid algorithm is proposed on the basis of the multilevel octree architecture. Transport sweep is performed according to the tree branch relationship of nested mesh distributions. The angular flux transmission between discontinuous grids is handled by the flux spatial moment mapping technique, and multiple zero-order mapping schemes are developed, including finite element interpolation, distance interpolation, and exponential fitting methods for treating upwind flux distributions of different relative shapes. The first-order mapping schemes are modified and improved for linear and exponential short characteristic discretization methods. The mapping accuracy is evaluated for polynomial and exponential functions, and a new spatial shape factor is presented for measuring the degree of nonlinearity. The multilevel octree grid (MLTG) algorithm is tested for neutron transport benchmarks, and good agreement with Monte Carlo and standard SN results is achieved. The number of meshes in the VVER shielding model is reduced from 18 million to 2 million using 3-level octree grids with the same geometric description accuracy. Numerical verification of a one-group fixed-source problem shows that 4-level and 5-level MLTG results with proper spatial discretization schemes can achieve relative deviations of less than 3% and 5% for detector region flux, respectively.
- 21 October 2019
- Published by Springer Nature in Nuclear Science and Techniques
- Vol. 30 (11) , 161
Numerical study of scattering Legendre moments and effect of anisotropic scattering on SN shielding calculation
- 21 October 2019
- Published by Springer Nature in Nuclear Science and Techniques
- Vol. 30 (11) , 161
Numerical study of scattering Legendre moments and effect of anisotropic scattering on SN shielding calculation
- 1 January 2019
- Published by Elsevier in Progress in Nuclear Energy
- Vol. 110, 236-244
Numerical analysis of spatial discretization schemes in ARES transport code for shielding calculation
- 1 January 2019
- Published by Elsevier in Progress in Nuclear Energy
- Vol. 110, 236-244
Numerical analysis of spatial discretization schemes in ARES transport code for shielding calculation
- 7 December 2018
- Published by Elsevier in Fusion Engineering and Design
- Vol. 138, 267-271
Angular adaptive discrete ordinates method for deuterium–tritium neutron streaming experiments simulating narrow gaps
- 7 December 2018
- Published by Elsevier in Fusion Engineering and Design
- Vol. 138, 267-271
Angular adaptive discrete ordinates method for deuterium–tritium neutron streaming experiments simulating narrow gaps
- Open Access
- 9 July 2018
- Published by Wiley in Science and Technology of Nuclear Installations
- Vol. 2018, 1-10
Analysis of Spatial Discretization Error Estimators Implemented in ARES Transport Code for EquationsThe discrete ordinates method (SN) is one of the mainstream methods for neutral particle transport calculations. Assessing the quality of the numerical solution and controlling the discrete error are essential parts of large-scale high-fidelity simulations of nuclear systems. Three error estimators, a two-mesh estimator, a residual-based estimator, and a dual-weighted residual estimator, are derived and implemented in the ARES transport code to evaluate the error of zeroth-order spatial discretization for SN equations. The difference in scalar fluxes on coarse and fine meshes is adopted to indicate the error in the two-mesh method. To avoid zero residual in zeroth-order discretization, angular fluxes within one cell are reconstructed by Legendre polynomials. The error is estimated by inverting the discrete transport operator using the estimated directional residual as an anisotropic source. The inner product of the forward directional residual and the adjoint angular flux is employed to quantify the error in quantities of interest which can be denoted by a linear functional of forward angular flux. Method of Manufactured Solutions (MMS) is adopted to generate analytical solutions for SN equation with scattering and the determined true error is used to evaluate the effectivity of these estimators. Promising results are obtained in the numerical results for both homogeneous and heterogeneous cases. The larger error region is well captured and the average effectivity index for the local error estimation is less than unity. For the series test problems, the estimated goal quantity error can be contained within an order of magnitude around the exact error.
- 9 July 2018
- Published by Wiley in Science and Technology of Nuclear Installations
- Vol. 2018, 1-10
Open AccessAnalysis of Spatial Discretization Error Estimators Implemented in ARES Transport Code for EquationsThe discrete ordinates method (SN) is one of the mainstream methods for neutral particle transport calculations. Assessing the quality of the numerical solution and controlling the discrete error are essential parts of large-scale high-fidelity simulations of nuclear systems. Three error estimators, a two-mesh estimator, a residual-based estimator, and a dual-weighted residual estimator, are derived and implemented in the ARES transport code to evaluate the error of zeroth-order spatial discretization for SN equations. The difference in scalar fluxes on coarse and fine meshes is adopted to indicate the error in the two-mesh method. To avoid zero residual in zeroth-order discretization, angular fluxes within one cell are reconstructed by Legendre polynomials. The error is estimated by inverting the discrete transport operator using the estimated directional residual as an anisotropic source. The inner product of the forward directional residual and the adjoint angular flux is employed to quantify the error in quantities of interest which can be denoted by a linear functional of forward angular flux. Method of Manufactured Solutions (MMS) is adopted to generate analytical solutions for SN equation with scattering and the determined true error is used to evaluate the effectivity of these estimators. Promising results are obtained in the numerical results for both homogeneous and heterogeneous cases. The larger error region is well captured and the average effectivity index for the local error estimation is less than unity. For the series test problems, the estimated goal quantity error can be contained within an order of magnitude around the exact error.
- 6 November 2017
- Published by Taylor & Francis in Nuclear Science and Engineering
- Vol. 189 (2) , 120-134
Goal-Oriented Regional Angular Adaptive Algorithm for the SN Equations
Angular discretization errors inherent in the discrete ordinates method are a major problem, especially for localized source problems and problems with strongly absorbing media or large-volume void regions, where angular discretization errors would be totally unacceptable. This paper proposes a regional angular adaptive algorithm together with a goal-oriented error estimate to solve the SN equations. Standard angular adaptive refinement techniques are based on estimated local errors. We compare an interpolated angular flux value against a calculated value to generate local errors. The adaptive quadrature sets can be created by subdividing a spherical quadrilateral into four spherical subquadrilaterals that have positive weights and can be locally refined. Techniques for mapping angular fluxes from one quadrature set to another are developed to transfer angular fluxes on the interfaces of different spatial regions. To provide a better detector response, local errors are weighted by the importance of a given angular region toward the computational goal, providing an appropriate goal-oriented angular adaptivity. First collision source methods are employed to improve adjoint flux calculation accuracy. We tested the performance and accuracy of the proposed goal-oriented regional angular adaptive algorithm within the ARES code for a number of benchmark problems and present the results of a one-region test model and the Kobayashi benchmark problems. The reduction of angular number is at least one order of magnitude for adaptive refinement. The benchmarks demonstrate that the proposed goal-oriented adaptive refinement can achieve the same level of accuracy as the SN method, which has significantly higher computation cost. Thus, adaptive refinement is a viable approach for investigating difficult particle radiation transport problems.
- 6 November 2017
- Published by Taylor & Francis in Nuclear Science and Engineering
- Vol. 189 (2) , 120-134
Goal-Oriented Regional Angular Adaptive Algorithm for the SN Equations
Angular discretization errors inherent in the discrete ordinates method are a major problem, especially for localized source problems and problems with strongly absorbing media or large-volume void regions, where angular discretization errors would be totally unacceptable. This paper proposes a regional angular adaptive algorithm together with a goal-oriented error estimate to solve the SN equations. Standard angular adaptive refinement techniques are based on estimated local errors. We compare an interpolated angular flux value against a calculated value to generate local errors. The adaptive quadrature sets can be created by subdividing a spherical quadrilateral into four spherical subquadrilaterals that have positive weights and can be locally refined. Techniques for mapping angular fluxes from one quadrature set to another are developed to transfer angular fluxes on the interfaces of different spatial regions. To provide a better detector response, local errors are weighted by the importance of a given angular region toward the computational goal, providing an appropriate goal-oriented angular adaptivity. First collision source methods are employed to improve adjoint flux calculation accuracy. We tested the performance and accuracy of the proposed goal-oriented regional angular adaptive algorithm within the ARES code for a number of benchmark problems and present the results of a one-region test model and the Kobayashi benchmark problems. The reduction of angular number is at least one order of magnitude for adaptive refinement. The benchmarks demonstrate that the proposed goal-oriented adaptive refinement can achieve the same level of accuracy as the SN method, which has significantly higher computation cost. Thus, adaptive refinement is a viable approach for investigating difficult particle radiation transport problems.