Search

CN-121278795-B - Neutron transport simulation method, related device and computer program product

CN121278795BCN 121278795 BCN121278795 BCN 121278795BCN-121278795-B

Abstract

The present disclosure provides a neutron transport simulation method, a related device and a computer program product, and relates to the technical field of computer technology and nuclear physics. The method comprises the specific implementation modes of carrying out CSG modeling on a reactor core to construct a CSG tree, generating N projection surfaces around the reactor core, generating an incident point array on all the projection surfaces, constructing a blank Segment list, carrying out recursive characteristic line tracking on all the incident points from a root node of the CSG tree, sequentially filling Segment data into the blank Segment list, carrying out load balancing on a GPU layer and a process layer in the GPU, transmitting characteristic lines, FSR and material data to each GPU, carrying out transport scanning calculation on each GPU, calculating global FSR angle torque conversion flux, calculating cracking and fission sources of each GPU, calculating K effective values and verifying whether convergence is achieved. According to the technical scheme, the limit of CSG geometric construction in the existing neutron transport simulation method can be broken, unnecessary characteristic line generation is reduced to save calculation resources, and the iteration solving speed is improved to accelerate simulation.

Inventors

  • WU ZHENGHONG
  • WANG JUE
  • WANG ZONGGUO
  • WANG YANGANG

Assignees

  • 中国科学院计算机网络信息中心

Dates

Publication Date
20260512
Application Date
20251029

Claims (10)

  1. 1. A neutron transport simulation method, comprising: Performing CSG modeling on a reactor core to construct a CSG tree, wherein the reactor core can be any geometric reactor core, the CSG modeling supports the intersecting operation of father-child nodes of the CSG tree and the intersecting and supplementing operation between the maximum four curved surfaces, the CSG tree nodes are segmented into FSRs and are endowed with first FSR numbers of all the FSRs, the first FSR numbers are random numbers, the nature of the CSG tree nodes is an intersecting point calculation sequence of a characteristic line and an FSR boundary, and the intersecting and supplementing operation maintains the intersecting point calculation sequence as a tree relation; Generating N projection surfaces around the reactor core, and generating an incident point array on all the projection surfaces, wherein the projection surfaces are arranged on the same plane as the core All the incident points generate a characteristic line perpendicular to the projection plane where the incident points are located; Constructing a blank Segment list, starting from the CSG tree root node, carrying out recursive feature line tracking on all the incidence points, and sequentially filling Segment data into the blank Segment list; the characteristic line tracking is carried out, and according to the intersection point calculation sequence, a polynomial of the highest quadric surface and a characteristic line polynomial are combined, and the intersection compensation operation is converted into an intersection point sequence solving operation; The method comprises the steps of carrying out load balancing on a GPU layer and a thread layer in the GPU, carrying out first characteristic line grouping on the characteristic lines according to the directions of the characteristic lines, distributing continuous serial numbers to the characteristic lines in the first characteristic line grouping to obtain first characteristic line numbers, distributing characteristic line polling to the GPU according to the first characteristic line grouping and the first characteristic line numbers, dividing the characteristic lines with the same first FSR numbers penetrating through a first FSR in the GPU into a group to obtain second characteristic line grouping, distributing continuous serial numbers to the characteristic lines in the second characteristic line grouping to obtain second characteristic line numbers, distributing the characteristic lines to adjacent GPU threads in a greedy grouping mode according to the second characteristic line numbers, and enabling the number of segments in the GPU threads to be not smaller than a preset upper limit according to the ending condition of the greedy grouping; Transmitting the characteristic line data, FSR data and material data to each GPU; Each GPU performs transport scanning calculation, wherein the transport scanning calculation means that each GPU traverses the characteristic line and updates the local FSR angular flux; calculating global FSR angular flux, wherein the global FSR angular flux is obtained through full-specification operation of an inter-GPU communication library NVSHMEM; Each GPU calculates a cracking torque and a cracking source; transmitting the total fissile source value to the CPU; calculating a K effective value and verifying whether convergence is achieved, wherein if the verification is negative, the transportation scanning calculation, the global FSR angular flux calculation, the split moment and split source calculation, the total split source value transmission, the K effective value calculation and the verification are continued, and if the verification is positive, the simulation is ended.
  2. 2. The method of claim 1, wherein generating an array of incidence points on the projection surface comprises: generating an incident point array on all the projection surfaces at a preset density; Counting the total FSR nFSR _base crossed by the characteristic lines generated at the preset density; Encrypting all the incident point arrays according to a preset rule; counting the total FSR nFSR _now which is traversed by all the characteristic lines at present; delta _ FSRs = nFSR _ now-nFSR _ base is calculated, then nFSR. Mu. base = nFSR _now; And judging whether delta_ FSRs is smaller than a preset delta_ FSRs threshold value, if yes, ending encryption, if no, setting delta_ FSRs to 0, continuing encrypting all the incident point arrays according to the preset rule, recalculating delta_ FSRs, and judging again, wherein the preset delta_ FSRs threshold value is larger than 0.
  3. 3. The method according to claim 2, characterized by comprising: the preset density is that the incident point row interval and the incident point column interval of the incident point array are both 1; The preset rule is that the incident point row interval and the incident point column interval are halved.
  4. 4. The method according to claim 1 or 2, comprising: Distributing continuous serial numbers for FSRs in the CSG leaf child nodes; Reconnecting FSRs in the leaf nodes according to the subsequent traversal sequence of the CSG tree to obtain a second FSR number; Sequentially searching the longest characteristic line passing through each FSR according to the second FSR numbering sequence, and obtaining an FSR numbering sequence passing through the longest characteristic line, wherein the longest characteristic line is the characteristic line with the largest Segment number; maintaining a global variable G and initializing to 0, and assigning a third FSR number to the FSR; Maintaining a boolean array having a total length of the FSR number, for recording whether the FSR is assigned the third FSR number; And according to the second FSR number and the FSR number sequence, G values are sequentially given to the FSRs to obtain the third FSR number, wherein the G values are increased by 1 during each assignment.
  5. 5. A neutron transport simulation device, comprising: The system comprises a CSG modeling unit, a CSG tree node, a characteristic line and a FSR boundary, wherein the CSG modeling unit is used for carrying out CSG modeling on a reactor core to construct a CSG tree, the reactor core can be any geometric reactor core, the CSG modeling supports the intersecting operation of the father-child nodes of the CSG tree and the intersecting and supplementing operation between the highest four curved surfaces, the CSG tree node is segmented into FSRs and is endowed with first FSR numbers of all the FSRs, the first FSR numbers are random numbers, the nature of the CSG tree node is an intersecting point calculation sequence of the characteristic line and the FSR boundary, and the intersecting and supplementing operation maintains the intersecting point calculation sequence as a tree relation; an incident point array generation unit configured to generate N projection planes around the core and generate an incident point array on all the projection planes, wherein the core is configured to generate a plurality of light sources All the incident points generate a characteristic line perpendicular to the projection plane where the incident points are located; The recursive feature line tracking unit is configured to construct a blank Segment list, start from the CSG tree root node, carry out recursive feature line tracking on all the incidence points, and fill Segment data into the blank Segment list in sequence; the characteristic line tracking is carried out, and according to the intersection point calculation sequence, a polynomial of the highest quadric surface and a characteristic line polynomial are combined, and the intersection compensation operation is converted into an intersection point sequence solving operation; The system comprises a load balancing unit, a GPU layer load balancing unit, a greedy grouping mode, a first characteristic line number distribution unit, a greedy grouping mode and a second characteristic line grouping unit, wherein the load balancing unit is configured to perform GPU layer load balancing and GPU inner thread layer load balancing, the GPU layer load balancing is used for performing first characteristic line grouping on the characteristic lines according to the directions of the characteristic lines, distributing continuous sequence numbers to the characteristic lines in the first characteristic line grouping to obtain first characteristic line numbers, distributing the characteristic line polling to a GPU according to the first characteristic line grouping and the first characteristic line numbers, grouping the characteristic lines which pass through a first FSR in the GPU and are the same in the GPU into a group to obtain second characteristic line grouping, distributing continuous sequence numbers to the characteristic lines in the second characteristic line grouping to adjacent GPU inner threads according to the second characteristic line numbers, and the greedy grouping mode is used for ending the greedy grouping, wherein the number of segments in the GPU inner threads is not smaller than a preset upper limit; a first data transmission unit configured to transmit the feature line data, the FSR data, and the material data to each GPU; the transport scanning calculation unit is configured to carry out transport scanning calculation on each GPU, wherein the transport scanning calculation means that each GPU traverses the characteristic line and updates the local angular flux of the FSR; a global FSR angle flux calculation unit configured to calculate a global FSR angle flux, wherein the global FSR angle flux is obtained through a full-specification operation of the inter-GPU communication library NVSHMEM; A split torque and fission source calculation unit configured to calculate split torque and fission sources for each GPU; a second data transmission unit configured to transmit the total fissile source value to the CPU; and the convergence verification unit is configured to calculate a K effective value and verify whether convergence exists, wherein if the verification is negative, the transportation scanning calculation, the global FSR angular flux calculation, the split moment and fission source calculation, the total fission source value transmission, the K effective value calculation and the verification are continued, and if the verification is positive, the simulation is ended.
  6. 6. The apparatus of claim 5, wherein the incident point array generation unit comprises an incident point array generation subunit configured to step up an incident point density, the incident point array generation subunit further configured to: generating an incident point array on all the projection surfaces at a preset density; Counting the total FSR nFSR _base crossed by the characteristic lines generated at the preset density; Encrypting all the incident point arrays according to a preset rule; counting the total FSR nFSR _now which is traversed by all the characteristic lines at present; delta _ FSRs = nFSR _ now-nFSR _ base is calculated, then nFSR. Mu. base = nFSR _now; And judging whether delta_ FSRs is smaller than a preset delta_ FSRs threshold value, if yes, ending encryption, if no, setting delta_ FSRs to 0, continuing encrypting all the incident point arrays according to the preset rule, recalculating delta_ FSRs, and judging again, wherein the preset delta_ FSRs threshold value is larger than 0.
  7. 7. The apparatus of claim 5 or 6, further comprising a two-stage FSR permute unit configured to: Distributing continuous serial numbers for FSRs in the CSG leaf child nodes; Reconnecting FSRs in the leaf nodes according to the subsequent traversal sequence of the CSG tree to obtain a second FSR number; Sequentially searching the longest characteristic line passing through each FSR according to the second FSR numbering sequence, and obtaining an FSR numbering sequence passing through the longest characteristic line, wherein the longest characteristic line is the characteristic line with the largest Segment number; maintaining a global variable G and initializing to 0, and assigning a third FSR number to the FSR; Maintaining a boolean array having a total length of the FSR number, for recording whether the FSR is assigned the third FSR number; And according to the second FSR number and the FSR number sequence, G values are sequentially given to the FSRs to obtain the third FSR number, wherein the G values are increased by 1 during each assignment.
  8. 8. An electronic device, comprising: at least one processor, and A memory communicatively coupled to the at least one processor, wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to implement the neutron transport simulation method of any one of claims 1-4.
  9. 9. A non-transitory computer readable storage medium storing computer program instructions for causing the computer to perform the neutron transport simulation method of any one of claims 1-4.
  10. 10. A computer program product comprising a computer program which, when executed by a processor, implements the neutron transport simulation method according to any one of claims 1-4.

Description

Neutron transport simulation method, related device and computer program product Technical Field The present disclosure relates to the field of computer technology and nuclear physics technology, and in particular, to a neutron transport simulation method, a related apparatus, and a computer program product. Background A Virtual Reactor (Virtual Reactor) is a comprehensive software system for performing highly realistic reproduction of the physical process, engineering properties and operating states of a nuclear Reactor by means of computer modeling and simulation techniques. The method can realize the omnibearing study on the design, operation and safety analysis of the reactor under the condition of not depending on a real reactor core experiment. The virtual reactor is not only beneficial to reducing research and development cost and experimental risk, but also can support the optimal design of advanced reactor type, the prediction and verification of accident conditions and the education and training of nuclear engineering, and is regarded as an important tool and platform for promoting the development of nuclear energy science. Neutron transport equation (Neutron Transport Equation) is the core of the virtual reactor simulation. The method is used for describing the distribution and evolution of neutrons in space and energy, and can describe the change of neutrons in a nuclear reactor core caused by scattering, absorption, fission and other processes. Currently, one of the mainstream numerical solutions of neutron transport equations is the eigenvector method (Method of Characteristics, MOC). The method has high fidelity and excellent geometric adaptability, so that the method is widely applied to the full-reactor simulation of various reactor cores. The main flow of MOC solving is divided into two parts, namely, feature line generation and iterative solving. In the feature line generation, the core needs to be geometrically modeled first, and a common method is to construct a solid geometry (Constructive Solid Structure, CSG). The method can accurately represent the geometric structure and construct a modeling tree by performing intersection and interpolation calculation on the space plane. Node types on the modeling tree can be divided into Cell, universe, lattice, and Material. Cell is a geometric space surrounded by one or more surfaces, wherein the geometric space is filled by Universe or Material, universe is a three-dimensional space range formed by a series of cells, lattice can rapidly represent a plurality of universes with similar structures, and Material is one of Cell nodes, is a leaf node of a CSG tree, and is filled by a single substance only. From the node type dimension, the CSG tree structure may be as shown in fig. 1. These nodes may correspond to fuel rods and moderator components in the core. Each node on the modeling tree is then further segmented into a plurality of base space units, named uniform source regions (Flat Source Region, FSR). Then, a plurality of differently angled projection surfaces are generated around the core with an array of start points of the feature lines thereon. Each starting point generates a ray perpendicular to the plane, i.e. a feature line. The signature line, during its passage through the core, will be cut into segments by the FSR that has been built. The process of determining which FSRs a feature line will traverse, and the actual length of each Segment on the feature line, is referred to as feature line tracking. After feature line tracking is performed on all incidence points on the projection surface, the feature line generation stage is finished. In the iterative solution, the calculation of each step of iteration can be divided into two parts. First, transport scan calculations will traverse segments on each feature line in sequence, forward and reverse, updating the angular flux delta for each FSR traversed. And secondly, numerical conversion, converting the angular flux increment of each FSR into a split torque distribution, an effective value K and a global fission source. The fission moment represents the power distribution inside the core, and the K-space value represents the neutron valueincrement coefficient. After the above calculation is completed, the algorithm will determine whether the program is terminated based on the K-valid value differences of the previous and current steps and whether the global fission source difference is below a preset threshold. The existing neutron transport simulation method has some technical problems, such as limited CSG modeling geometric construction, calculation resource waste caused by unnecessary feature line generation, low iteration solving efficiency and the like. Therefore, there is a need in the art for an efficient neutron transport simulation method that is adaptable to any geometric core. Disclosure of Invention The present disclosure provides a neutron transport simulation method, appa