Reliability analysis method for large-scale nuclear power systems based on parallel computing

Abstract

This application discloses a large-scale nuclear power system reliability analysis method based on parallel computing, belonging to the field of nuclear power system reliability analysis. The method comprises the following steps: constructing a nuclear power system fault tree model T based on the functional principle, failure mode, and reliability logic relationship of the nuclear power system; using a fault tree block method to identify independently analyzable fault tree subtrees in the nuclear power system fault tree model T, and constructing a vertex directed graph of the independently analyzable fault tree subtree vertices based on the connection relationship of the fault tree subtrees; using the vertex directed graph and a topological sorting method to identify the analysis order of the fault tree subtrees, and then analyzing the nuclear power system fault tree model T in parallel according to this order to obtain the probability of occurrence of the top event of the nuclear power system fault tree model T. The present invention can be applied to the analysis of large-scale fault tree models in nuclear power systems.

Description

Technical Field

[0001] The present application relates to a large-scale nuclear power system reliability analysis method based on parallel computing, which belongs to the field of nuclear power system reliability analysis. Background Art

[0002] The development of nuclear power technology and the increasing proportion of nuclear power plants supplying electricity are having a significant impact on national industries and people's lives. Due to the high risk of nuclear leaks, the reliability and safety requirements of nuclear power are extremely high. The complex structure and diverse operating conditions of nuclear power systems lead to large-scale and complex fault trees. Therefore, accurately analyzing nuclear power reliability has become a major challenge in nuclear power system reliability design and analysis.

[0003] Existing analysis techniques typically employ precise modeling and approximate analysis methods, only partially analyzing the causes of critical and important events while ignoring events with lower probability of occurrence. While this improves analysis efficiency, it cannot fully meet the requirements for analyzing system reliability. However, with the development of parallel computing technology and several parallel frameworks and tools, it has become possible to analyze the reliability of large-scale nuclear power systems using parallel computing. Summary of the Invention

[0004] According to one aspect of the present application, a large-scale nuclear power system reliability analysis method based on parallel computing is provided, which can be applied to the analysis of large-scale fault tree models in nuclear power systems.

[0005] A large-scale nuclear power system reliability analysis method based on parallel computing is characterized by comprising the following steps:

[0006] S1: Construct a nuclear power system fault tree model T based on the functional principle, failure mode and reliability logic relationship of the nuclear power system;

[0007] S2: Using the fault tree block method, identify the fault tree subtrees that can be analyzed independently in the nuclear power system fault tree model T, and construct a vertex directed graph of the fault tree subtree vertices that can be analyzed independently based on the connection relationship of the fault tree subtrees ;

[0008] S3: Using the vertex directed graph , and based on the topological sorting method, identify the analysis order of the fault tree subtrees, and then analyze the nuclear power system fault tree model T in parallel according to the order to obtain the probability of occurrence of the top event of the nuclear power system fault tree model T.

[0009] Furthermore, the S2 includes:

[0010] S21: Using the linear block method, identify the fault tree subtree vertices that can be analyzed independently in the fault tree model. The set of the fault tree subtree vertices is recorded as , where n represents the number of vertices of the fault tree subtree;

[0011] S22: Based on the connection relationship of the fault tree subtree and the fault tree subtree vertex, construct a vertex directed graph that forms a mapping with the fault tree subtree vertex ,in, is the set of fault tree subtree vertices, and E is the edge formed by these vertices.

[0012] Furthermore, the S3 includes:

[0013] S31: Using the vertex directed graph ,identify parallelizable fault tree subtrees based on the topological sorting method;

[0014] S32: According to the independent parallel and analysis dependency relationships of the fault tree subtrees, the fault tree subtrees are converted into corresponding binary decision diagrams in sequence to obtain analysis results of the fault tree subtrees, and the fault tree subtrees are replaced with newly created nodes until the probability of occurrence of the top event of the nuclear power system fault tree model T is obtained.

[0015] Furthermore, in S31, identifying parallelizable fault tree subtrees based on a topological sorting method includes:

[0016] 1) Create an empty queue Q, which is used to store a set of nodes with an in-degree of 0 , initialize the topological sorting result queue R, so that the topological sorting result queue R contains a node set ;

[0017] 2) Traverse each node in the vertex directed graph G and calculate its in-degree;

[0018] 3) Add all nodes with in-degree 0 to the empty queue Q;

[0019] 4) Take values ​​from queue Q in sequence until queue Q is empty, and then take out the node set from queue Q , create a new empty node set ,from Take out the nodes one by one , traverse the nodes All adjacent nodes of , the node The in-degree of the node is reduced by 1; if If the in-degree of is reduced to 0, the node Joining a node collection ; Traverse the nodes After connecting all adjacent nodes, the node Delete from the vertex directed graph G; node set After the node traversal is completed, the node set Join queue Q and topological sort result queue R;

[0020] The obtained topological sorting result queue R is a queue containing multiple node sets that can be processed in parallel.

[0021] Furthermore, the S32 includes:

[0022] a) Traverse the topological sorting result queue R, extract the fault tree subtree vertices that can be analyzed in parallel, and convert the fault tree subtrees into binary decision diagrams for parallel analysis;

[0023] b) After the parallel analysis is completed, a new node is created and the analysis results are assigned to the new node, and then the new node replaces the original fault tree subtree;

[0024] c) Repeat the above steps until the topological sorting result queue R is empty, and the analysis results of the large-scale nuclear power system fault tree model T are obtained.

[0025] Furthermore, the implementation of converting the fault tree subtrees into binary decision diagrams in parallel includes:

[0026] ;

[0027] ;

[0028] ;

[0029] ;

[0030] ;

[0031] ;

[0032] in, represents the left branch of the binary decision diagram with event x as the vertex; represents the right branch of the binary decision diagram with event x as the vertex; Represents the left branch of the binary decision diagram with event y as the vertex; Represents the right branch of the binary decision diagram with event y as the vertex;

[0033] and They represent the binary decision diagrams corresponding to the fault tree subtree containing event x and the fault tree containing event y respectively;

[0034] “ " indicates logical AND operation, " ” represents a logical OR operation;

[0035] represents the result of a logical AND operation on two binary decision diagrams with the same vertices;

[0036] represents the result of a logical OR operation on two binary decision diagrams with the same vertices;

[0037] represents the result of a logical AND operation on two binary decision diagrams with different vertices and x < y;

[0038] represents the result of a logical OR operation on two binary decision diagrams with different vertices and x < y.

[0039] The beneficial effects that this application can produce include:

[0040] The method for reliability analysis of a large-scale nuclear power system based on parallel computing provided by this application can decompose the large-scale fault tree model of the nuclear power system that could not be processed originally through parallel computing, analyze the decomposed model in parallel, and finally merge the analysis nodes of each module to obtain the analysis result of the large-scale fault tree model in the nuclear power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is the flow chart of the steps of the method for reliability analysis of a large-scale nuclear power system based on parallel computing in an embodiment of this application;

[0042] Figure 2 is the flow chart of the method for reliability analysis of a large-scale nuclear power system based on parallel computing in an embodiment of this application;

[0043] Figure 3 is the fault tree model of the opening failure of the shutdown circuit breaker of the IP channel in the emergency shutdown system of the nuclear reactor in an embodiment of this application;

[0044] Figure 4 is the directed graph formed by mapping the vertices of the sub-tree module of the fault tree in an embodiment of this application;

[0045] Figure 5 is the fault tree sub-tree in an embodiment of this application after being converted into a binary decision diagram;

[0046] Figure 6 is the updated fault tree model after replacing the sub-tree in an embodiment of this application;

[0047] Figure 7 is the fault tree sub-tree in an embodiment of this application The converted binary decision diagram;

[0048] Figure 8 In one embodiment of the present application, a subtree is replaced The updated fault tree model. DETAILED DESCRIPTION

[0049] The present application is described in detail below with reference to embodiments, but the present application is not limited to these embodiments.

[0050] like Figure 1-8 As shown, the large-scale nuclear power system reliability analysis method based on parallel computing is characterized by comprising the following steps:

[0051] S1: Construct a nuclear power system fault tree model T based on the functional principle, failure mode and reliability logic relationship of the nuclear power system;

[0052] S2: using a fault tree block method, identifying independently analyzable fault tree subtrees in the nuclear power system fault tree model T, and constructing a vertex directed graph of the independently analyzable fault tree subtree vertices based on the connection relationship of the fault tree subtrees;

[0053] S3: Using the vertex directed graph and based on a topological sorting method, identify the analysis order of the fault tree subtrees, and then analyze the nuclear power system fault tree model T in parallel according to the order to obtain the probability of occurrence of the top event of the nuclear power system fault tree model T.

[0054] Specifically, based on the design principles and operating mechanisms of the nuclear power system, the functions and interrelationships of each component are clarified. Various possible failure modes within the nuclear power system are identified, including hardware failures, software errors, and human errors. Logical relationships between faults, such as "AND" and "OR," are determined to construct a fault tree model. Combining this information, a fault tree model is constructed that reflects the reliability characteristics of the nuclear power system. Independently analyzed subtrees within the fault tree model are identified. These subtrees are logically independent and can be processed in parallel. The fault tree model is divided into multiple subtrees, each representing an independent fault analysis unit. Based on the connectivity between the fault tree subtrees, a vertex-directed graph is constructed. Vertices in the graph represent fault tree subtrees, and edges represent the logical dependencies between subtrees. Using this vertex-directed graph, a topological sort is used to determine the analysis order of the fault tree subtrees. Topological sorting ensures that when analyzing a subtree, all its dependent subtrees have already been analyzed. Each fault tree subtree is analyzed in parallel, following the topological sorting order. Because the subtrees are logically independent, they can be analyzed simultaneously, significantly improving analysis efficiency. After the parallel analysis is completed, the analysis results of each subtree are integrated to calculate the probability of occurrence of the top event of the nuclear power system fault tree model T.

[0055] Therefore, parallel computing technology can significantly shorten the time required for large-scale nuclear power system reliability analysis. During the parallel analysis process, topological sorting ensures the sequential nature of subtree analysis, thereby guaranteeing the accuracy of the analysis results. This method is applicable to reliability analyses of nuclear power systems of various sizes and complexities.

[0056] In a specific implementation, taking a nuclear reactor emergency shutdown system as an example, if the IP channel trip circuit breaker fails to open or the trip signal fails, either event will result in an output event, namely, the IP channel trip circuit breaker failure to open. Therefore, an OR gate is used to represent this static failure behavior. During a manual shutdown, operator error or manual switch failure will also result in the manual trip signal failure. Therefore, an OR gate is used to represent this static failure behavior. Within the same protection channel, a logical voting failure of the "local trip" signal or a failure of the Accelerated Processing Unit (APU) module will both result in the automatic shutdown signal failure. Therefore, the logical relationship between these two can be described using an OR gate. Causes of APU module failure include backplane bus lockup, processor failure, power supply failure, and rack failure. Any of these possible causes will cause the APU module to fail. Therefore, an OR gate is used to represent this static failure behavior. The "local trip" signal undergoes a two-out-of-four logical voting to generate the emergency shutdown signal. For this two-out-of-four voting, if three inputs fail, the output fails. Therefore, a three-out-of-four voting gate is used to represent this static characteristic. The failure of the input modules of the IP protection channel, IIP protection channel, IIIP protection channel, and IVP protection channel is affected by the sensors, signal preprocessing units, and signal acquisition and processing units of each protection channel. The failure of these components will lead to the failure of the input of the corresponding protection channel. Therefore, an OR gate is used to characterize this static characteristic.

[0057] If the automatic trip signal fails, the operator can use the manual trip switch to perform an emergency shutdown of the reactor. When both the automatic and manual trip signals fail, the output event is the failure of the trip signal. Therefore, an AND gate is introduced to describe this characteristic. The fault logic of other module characteristics can also be derived from the above analysis.

[0058] The fault tree model of the IP channel trip circuit breaker opening failure is obtained through analysis as follows: Figure 3 shown.

[0059] The S2 includes:

[0060] S21: Using the linear block method, identify the fault tree subtree vertices that can be analyzed independently in the fault tree model. The set of fault tree subtree vertices is recorded as , where n represents the number of vertices of the fault tree subtree;

[0061] S22: Based on the connection relationship of the fault tree subtree and the fault tree subtree vertex, construct a vertex directed graph that forms a mapping with the fault tree subtree vertex ,in, is the set of fault tree subtree vertices, and E is the edge formed by these vertices.

[0062] Specifically, linear block partitioning is a method that decomposes a complex system into multiple relatively independent parts, which is suitable for block processing of fault tree models. In the fault tree model, the linear block partitioning method is used to identify the fault tree subtree vertices that can be analyzed independently. These subtree vertices are relatively independent in logic and can be analyzed separately without being affected by other subtrees. After linear block partitioning, a set of fault tree subtree vertices is obtained. Each vertex in this set represents a fault tree subtree that can be analyzed independently. The set of fault tree subtree vertices can be expressed as ,in, Indicates the nth fault tree subtree vertex.

[0063] Furthermore, the connection relationship between fault tree subtrees refers to the logical dependency between different subtrees. For example, the output of one subtree may be the input of another subtree. By analyzing these connection relationships, the logical order and dependency between subtrees can be determined. The vertices in the vertex directed graph correspond to the elements in the fault tree subtree vertex set V, that is, each fault tree subtree vertex is mapped to a vertex in the directed graph. The edge E in the directed graph represents the connection relationship between the fault tree subtrees. If subtree v i Depends on subtree v j , then add a line from v to the directed graph j Point to v i The vertex directed graph can be represented as , where V is the vertex set and E is the edge set.

[0064] It's worth noting that the vertex-directed graph clearly represents the logical order and dependencies between fault tree subtrees, providing a foundation for subsequent topological sorting and parallel analysis. This vertex-directed graph allows us to identify which subtrees can be analyzed in parallel and which need to wait until the analysis of their dependent subtrees is complete.

[0065] Among them, using the linear block method, the vertices of the fault tree subtrees that can be analyzed independently in the fault tree model can be obtained as shown in Table 1.

[0066] Table 1

[0067]

[0068] Based on these vertices and their connection relationships in the original fault tree model, we can get the following: Figure 4The vertex directed graph G is shown.

[0069] The S3 includes:

[0070] S31: Using the vertex directed graph ,identify parallelizable fault tree subtrees based on the topological sorting method;

[0071] S32: According to the independent parallel and analysis dependency relationships of the fault tree subtrees, the fault tree subtrees are converted into corresponding binary decision diagrams in sequence to obtain analysis results of the fault tree subtrees, and the fault tree subtrees are replaced with newly created nodes until the probability of occurrence of the top event of the nuclear power system fault tree model T is obtained.

[0072] Furthermore, in S31, identifying parallelizable fault tree subtrees based on a topological sorting method includes:

[0073] 1) Create an empty queue Q, which is used to store a set of nodes with an in-degree of 0 , initialize the topological sorting result queue R, so that the topological sorting result queue R contains a node set ;

[0074] 2) Traverse each node in the vertex directed graph G and calculate its in-degree;

[0075] 3) Add all nodes with in-degree 0 to the empty queue Q;

[0076] 4) Take values ​​from queue Q in sequence until queue Q is empty, and then take out the node set from queue Q , create a new empty node set ,from Take out the nodes one by one , traverse the nodes All adjacent nodes of , the node The in-degree of the node is reduced by 1; if If the in-degree of is reduced to 0, the node Joining a node collection ; Traverse the nodes After connecting all adjacent nodes, the node Delete from the vertex directed graph G; node set After the node traversal is completed, the node set Join queue Q and topological sort result queue R;

[0077] The obtained topological sorting result queue R is a queue containing multiple node sets that can be processed in parallel. At the same time, the topological sorting result queue R has a first-come-first-out feature, which ensures the dependency relationship of node analysis at different levels.

[0078] Specifically, topological sorting is a method for linearly ordering the vertices of a directed acyclic graph (DAG) such that for each directed edge (u, v) in the graph, vertex u always comes before vertex v in the sorting. In fault tree analysis, topological sorting can determine the dependencies between subtrees, thereby identifying subtrees that can be processed in parallel.

[0079] The empty queue Q is used to store nodes with an in-degree of 0. Initially, queue Q is empty. The topological sort result queue R is used to store the final topological sort result and is initially empty. Traverse the vertex directed graph G and, for each node in the graph, calculate its in-degree (i.e., the number of edges pointing to it). Add all nodes with an in-degree of 0 to queue Q: these nodes have no dependencies and can be analyzed or processed immediately. Sequentially remove values ​​from queue Q until queue Q is empty. Create a new empty node set to temporarily store the currently processed node set. Remove the node set from queue Q and process it one by one. Traverse all adjacent nodes of a node and decrement the node's in-degree by 1 to indicate that one of its dependencies has been processed. If the node's in-degree is reduced to 0, add the node to the empty node set, indicating that it can now be analyzed or processed. After processing all of the node's adjacent nodes, remove the node from the graph. Adding the node set to queue Q and the topological sort result queue R indicates that processing of the node set is complete and the next node set can be processed in parallel. The topological sorting result queue R contains multiple node sets that can be processed in parallel. The nodes in each node set have no direct dependencies and can therefore be processed in parallel. By using the topological sorting result queue R, the order in which the fault tree subtrees are analyzed can be determined, allowing for efficient parallel analysis.

[0080] It's worth noting that topological sorting clarifies the dependencies between fault tree subtrees, thereby determining the analysis order. It also identifies sets of nodes that can be processed in parallel, improving analysis efficiency. By ensuring that all dependent nodes are analyzed before analyzing a node, deadlock issues during analysis are avoided.

[0081] Furthermore, the S32 includes:

[0082] a) Traverse the topological sorting result queue R, extract the fault tree subtree vertices that can be analyzed in parallel, and convert the fault tree subtrees into binary decision diagrams for parallel analysis;

[0083] b) After the parallel analysis is completed, a new node is created and the analysis results are assigned to the new node, and then the new node replaces the original fault tree subtree;

[0084] c) Repeat the above steps until the topological sorting result queue R is empty, and the analysis results of the large-scale nuclear power system fault tree model T are obtained.

[0085] Specifically, a set of fault tree subtree vertices that can be analyzed in parallel is extracted from a queue R and converted into a binary decision diagram (BDD). A BDD is a data structure used to represent Boolean functions and can efficiently handle logical operations and probability calculations. Each fault tree subtree is converted in parallel into a BDD to facilitate efficient logical operations and probability analysis. Using parallel computing resources, each BDD is analyzed to calculate its corresponding failure probability or other reliability metric. After the parallel analysis is complete, a new node is created for each analysis result. The failure probability or other reliability metric obtained from the parallel analysis is assigned to the new node. The original fault tree subtree is replaced with the new node, and the entire fault tree model is updated. Steps (a) and (b) are repeated until the topological sorting result queue R is empty. When queue R is empty, all fault tree subtrees have been analyzed, resulting in the final analysis result of the large-scale nuclear power system fault tree model T.

[0086] Therefore, through parallel conversion to binary decision diagrams and parallel analysis, computing resources are fully utilized, significantly improving analysis efficiency. Rigorous parallel analysis and result integration ensure the accuracy and reliability of the final analysis results. This approach is suitable for reliability analysis of large-scale and complex nuclear power systems and can efficiently process complex fault tree models.

[0087] At the same time, the implementation of parallel conversion of fault tree subtrees into binary decision diagrams includes:

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] ;

[0094] in represents the left branch of the binary decision diagram with event x as the vertex; represents the right branch of the binary decision diagram with event x as the vertex; Represents the left branch of the binary decision diagram with event y as the vertex; Represents the right branch of the binary decision diagram with event y as the vertex;

[0095] and They represent the binary decision diagrams corresponding to the fault tree subtree containing event x and the fault tree containing event y respectively;

[0096] “ ” represents the logical AND operation, and “ ” represents the logical OR operation;

[0097] represents the result of the logical AND operation of two binary decision diagrams with the same vertices;

[0098] represents the result of the logical OR operation of two binary decision diagrams with the same vertices;

[0099] represents the result of the logical AND operation of two binary decision diagrams with different vertices and x < y;

[0100] represents the result of the logical OR operation of two binary decision diagrams with different vertices and x < y.

[0101] Specifically, the set of nodes with an in-degree of 0 , add it to the queue Q, and also add it to the topological sorting result queue R, obtaining , …;

[0102] Take out from the queue Q, create a new empty node set , take out the node M6 from , visit the adjacent node M4 of M6, reduce the in-degree of M4 by 1, and update its in-degree to 3; then take out the node M7, visit the adjacent node M4 of M7, reduce the in-degree of M4 by 1, and update its in-degree to 2; then take out the node M8, visit the adjacent node M4 of M8, reduce the in-degree of M4 by 1, and update its in-degree to 1; then take out the node M9, visit the adjacent node M4 of M9, reduce the in-degree of M4 by 1, and update its in-degree to 0. At this time, add the node M4 to the set . Then take out the node M10, visit the adjacent node M5 of M10, reduce the in-degree of M5 by 1, and update its in-degree to 1; then take out the node M11, visit the adjacent node M5 of M11, reduce the in-degree of M5 by 1, and update its in-degree to 0. At this time, add the node M5 to the set , and at this time ; [[ID=,47]]

[0103] Add the set [[ID=5l]] to the queue Q and the queue R. At this time , ;

[0104] Take out the node set in the queue Q again, repeat the above process, and obtain the final topological sorting result array:

[0105] ;

[0106] in, , , , .

[0107] According to the above results, we can parallel The fault tree subtrees with the nodes as vertices are converted into binary decision diagrams in parallel. The binary decision diagrams after these nodes are converted are shown in the attached figure. Figure 5 shown.

[0108] After obtaining the result of the fault tree subtree, replace the original fault tree subtree with a separate node. The updated fault tree model is shown in the attached figure. Figure 6 shown.

[0109] Then, the fault tree subtrees M4 and M5 are converted and analyzed in parallel, and the binary decision diagram obtained is shown in the attached figure. Figure 7 The original subtree is replaced by a single node, and the updated fault tree model is shown in the attached Figure 8 As shown, the fault tree subtree M7 is transformed in sequence, and finally the fault tree T1 is processed to obtain the final analysis result.

[0110] It is worth noting that the present invention provides a method path for analyzing the fault tree model of a large-scale nuclear power system, which can solve the large-scale fault tree model and provide strong support for analyzing the reliability of the large-scale nuclear power system.

[0111] The above descriptions are merely a few embodiments of the present application and do not constitute any form of limitation to the present application. Although the present application discloses the preferred embodiments as above, they are not intended to limit the present application. Any technical personnel familiar with the present profession, without departing from the scope of the technical solution of the present application, using the technical content disclosed above to make slight changes or modifications are equivalent to equivalent implementation cases and fall within the scope of the technical solution.

Claims

1. A large-scale nuclear power system reliability analysis method based on parallel computing, characterized by: The steps include: S1: Construct a nuclear power system fault tree model T based on the functional principle, failure mode and reliability logic relationship of the nuclear power system; S2: using a fault tree block method, identifying independently analyzable fault tree subtrees in the nuclear power system fault tree model T, and constructing a vertex directed graph G of independently analyzable fault tree subtree vertices based on the connection relationship of the fault tree subtrees; S3: using the vertex directed graph G and based on a topological sorting method, identifying the analysis order of the fault tree subtrees, and then analyzing the nuclear power system fault tree model T in parallel according to the order to obtain the probability of occurrence of the top event of the nuclear power system fault tree model T; The S2 includes: S21: Using the linear block method, identify the fault tree subtree vertices that can be analyzed independently in the fault tree model, and the set of the fault tree subtree vertices is recorded as V = (v1, v2, ..., v n ), where n represents the number of vertices of the fault tree subtree; S22: Based on the connection relationship of the fault tree subtrees and the fault tree subtree vertices, construct a vertex directed graph G=(V, E) that forms a mapping with the fault tree subtree vertices, where V is a set of fault tree subtree vertices and E is an edge formed by these vertices; The S3 includes: S31: using the vertex directed graph G, identifying parallelizable fault tree subtrees based on a topological sorting method; S32: According to the independent parallel and analysis dependency relationships of the fault tree subtrees, the fault tree subtrees are sequentially converted into corresponding binary decision diagrams to obtain analysis results of the fault tree subtrees, and the fault tree subtrees are replaced with newly created nodes until the probability of occurrence of the top event of the nuclear power system fault tree model T is obtained; In S31, identifying parallelizable fault tree subtrees based on a topological sorting method includes: 1) Create an empty queue Q, the empty queue Q is used to store the node set V0 with an in-degree of 0, initialize the topological sorting result queue R, and make the topological sorting result queue R contain the node set V0; 2) Traverse each node in the vertex directed graph G and calculate its in-degree; 3) Add all nodes with in-degree 0 to the empty queue Q; 4) Take values ​​from queue Q in sequence until queue Q is empty, and then take out node set V from queue Q i , create a new empty node set V i , from V i Take out the node u in turn j , traverse node u j All adjacent nodes v kj , change node v kj The in-degree of node v is reduced by 1; if the node v kj If the in-degree of node v is reduced to 0, kj Join node set V i ; Traverse the node u j After connecting all adjacent nodes, node u j Delete from the vertex directed graph G; node set V i After the node traversal is completed, the node set V i Join queue Q and topological sort result queue R; The obtained topological sorting result queue R is a queue containing multiple node sets that can be processed in parallel.

2. The large-scale nuclear power system reliability analysis method based on parallel computing according to claim 1 is characterized in that: The S32 includes: a) Traverse the topological sorting result queue R, extract the fault tree subtree vertices that can be analyzed in parallel, and convert the fault tree subtrees into binary decision diagrams for parallel analysis; b) After the parallel analysis is completed, a new node is created and the analysis results are assigned to the new node, and then the new node replaces the original fault tree subtree; c) Repeat the above steps until the topological sorting result queue R is empty, and obtain the analysis results of the large-scale nuclear power system fault tree model T.

3. The large-scale nuclear power system reliability analysis method based on parallel computing according to claim 2 is characterized in that: The implementation of parallel conversion of fault tree subtrees into binary decision diagrams includes: g=(x,G1,G2); h=(y,H1,H2); ite(x,G1,G2)·ite(x,H1,H2)=ite(x,G1·H1,G2·H2); ite(x,G1,G2)+ite(x,H1,H2)=ite(x,G1+H1,G2+H2); ite(x,G1,G2)·ite(y,H1,H2)=ite(x,G1·h,G2·h); ite(x,G1,G2)+ite(y,H1,H2)=ite(x,G1+h,G2+h); Among them, G1 represents the left branch of the binary decision diagram with event x as the vertex; G2 represents the right branch of the binary decision diagram with event x as the vertex; H1 represents the left branch of the binary decision diagram with event y as the vertex; H2 represents the right branch of the binary decision diagram with event y as the vertex; g = (x, G1, G2) and h = (y, H1, H2) respectively represent the subtree of the fault tree containing event x and the binary decision diagram corresponding to the fault tree containing event y; "·" represents logical AND operation, and "+" represents logical OR operation; ite(x, G1, G2)·ite(x, H1, H2) = ite(x, G1·H1, G2·H2) represents the result of the logical AND operation of two binary decision diagrams with the same vertex; ite(x, G1, G2)+ite(x, H1, H2) = ite(x, G1 + H1, G2 + H2) represents the result of the logical OR operation of two binary decision diagrams with the same vertex; ite(x, G1, G2)·ite(y, H1, H2) = ite(x, G1·h, G2·h) represents the result of the logical AND operation of two binary decision diagrams with different vertices and x < y; ite(x, G1, G2)+ite(y, H1, H2) = ite(x, G1 + h, G2 + h) represents the result of the logical OR operation of two binary decision diagrams with different vertices and x < y.

Images