Method and system for analyzing device fault propagation based on unstable point detection

Abstract

The application discloses a kind of equipment failure propagation analysis method and system at unstable point detection, equipment failure propagation analysis method based on unstable point detection includes the following steps: according to the fault propagation relationship between industrial complex system, construct fault propagation graph;Using BRISM method under different rules to detect unstable point, while using PageRank method to evaluate the importance of original fault propagation graph, and update the weight of unstable point;The weight of edge in hierarchical result is calculated, and the key fault propagation path between different systems is identified according to the weight of edge.The application effectively reduces the influence of unstable point on key propagation path, further improves the accuracy of fault propagation path identification, and finally effectively identifies the fault propagation path of different fault components.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of failure propagation of key components in rotating machinery devices, and discloses a device failure propagation analysis method and system based on unstable point detection. BACKGROUND

[0002] Industrial failures can have serious consequences, including economic losses, environmental damage, and even death. For industries, particularly disastrous is failure propagation, which occurs when a system failure serially affects other systems.

[0003] Nowadays, with more and more complex systems emerging, the danger of failure propagation is also rising. For example, if a power failure occurs in a large factory, it can trigger a chain reaction, causing other systems and machines to fail. Another problem is that a failure of one system can cause failures of other systems, such as systems sharing components or similar operations. Therefore, it is of great significance to determine the key propagation path in time, find out the cause of the failure, replace the problematic components, and minimize the total risk of the system when the failure occurs. The internal systems of rotating machinery devices are precise and complex, and to identify the key propagation path often requires hierarchical analysis of the internal systems, but the existing hierarchical methods cannot well analyze the causal relationship between systems. For example, the explanatory structural modeling method only considers the relationship between the reachable set and the common set of systems for hierarchical analysis, and lacks the influence of the precedent set on the hierarchical results. In addition, the existing methods for judging the importance or relevance between different systems are difficult to adjust according to the relationship between systems.

[0004] In summary, how to effectively hierarchize components, sort out the causal relationship between systems, and update the importance or relevance of each system according to the hierarchical results is a major problem in identifying the failure propagation path of key components in rotating machinery devices. SUMMARY

[0005] The present application provides a device failure propagation analysis method and system based on unstable point detection, aiming to solve the technical problems that the existing hierarchical methods cannot well analyze the causal relationship between systems, and the existing methods for judging the importance or relevance between different systems are difficult to adjust according to the relationship between systems.

[0006] One aspect of the present application relates to a device failure propagation analysis method based on unstable point detection, comprising the following steps:

[0007] constructing a failure propagation graph according to the failure propagation relationship between industrial complex systems;

[0008] The BRISM method under different rules is used to detect unstable points, and the PageRank method is used to evaluate the importance of the original fault propagation graph, and the weight of the unstable point is updated.

[0009] The weight of the edge in the hierarchical result is calculated, and the key fault propagation path between different systems is identified according to the weight of the edge.

[0010] Further, the step of constructing a fault propagation graph according to the fault propagation relationship between industrial complex systems comprises:

[0011] Analyze the influence relationship of a single component of an industrial complex system on other components, and construct a fault propagation subgraph;

[0012] Combine multiple propagation subgraphs to form a complete fault propagation graph, and convert the fault propagation graph into an adjacency matrix.

[0013] Further, the method of using the BRISM method under different rules to detect unstable points, and using the PageRank method to evaluate the importance of the original fault propagation graph, and updating the weight of the unstable point comprises:

[0014] The BRISM method under different rules is used to perform layering on the adjacency matrix;

[0015] Compare the layering results of the BRISM method under different initial layer extraction rules, find unstable points, and update the PageRank value using the hierarchical relationship of the unstable points.

[0016] Further, the step of calculating the weight of the edge in the hierarchical result and identifying the key fault propagation path between different systems according to the weight of the edge comprises:

[0017] The edge weight existing in the skeleton matrix is calculated using weighted PageRank;

[0018] According to the result obtained by the weighted PageRank algorithm, starting from the fault point, find the point with large in-edge weight as the upstream fault point, and perform tracing, finally find the tracing fault point, and form the final key fault propagation path.

[0019] Further, the PageRank value of the node after updating the weight is:

[0020]

[0021] wherein layer(Node i in M′1) refers to the layer level of the i-th node in M′1, layer(Node i in M′2) refers to the layer level of the i-th node in M′2, and PR(Nodei PR'(Node i) refers to the PageRank value of node i after using PageRank algorithm i PR'(Node i) refers to the PageRank value of node i after using PageRank algorithm

[0022] The edge weight existing in the skeleton matrix calculated by the weighted PageRank calculation is:

[0023]

[0024] Wherein, PR(i) refers to the PageRank value of the i th node, PR(j) refers to the PageRank value of the j th node, PR(e i→j PR(e i,j) refers to the weight value of the edge between node i and node j.

[0025] Another aspect of the application relates to a device fault propagation analysis system based on unstable point detection, comprising:

[0026] A construction module is configured to construct a fault propagation graph according to the fault propagation relationship between industrial complex systems.

[0027] An evaluation module is configured to detect unstable points by using BRISM methods under different rules, and to update the weights of the unstable points by using PageRank methods to evaluate the importance of the original fault propagation graph.

[0028] An identification module is configured to calculate the weights of the edges in the hierarchical results, and to identify the key fault propagation paths between different systems according to the weights of the edges.

[0029] Further, the construction module comprises:

[0030] A construction unit is configured to analyze the influence relationship of a single component of an industrial complex system on other components, and to construct a fault propagation subgraph.

[0031] A conversion unit is configured to combine multiple propagation subgraphs to form a complete fault propagation graph, and to convert the fault propagation graph into an adjacency matrix.

[0032] Further, the evaluation module comprises:

[0033] A hierarchical unit is configured to use BRISM methods under different rules to perform hierarchical processing on the adjacency matrix.

[0034] A comparison unit is configured to compare the hierarchical results of the BRISM methods under different initial layer extraction rules, to find unstable points, and to update the PageRank values by using the hierarchical relationship of the unstable points.

[0035] Further, the identification module comprises:

[0036] A calculation unit is configured to calculate the edge weight existing in the skeleton matrix by using the weighted PageRank;

[0037] A searching unit is configured to search the point with large in-edge weight as the upstream point of the fault from the fault point according to the result obtained by the weighted PageRank algorithm, trace back to the source, finally find the source fault point, and form the final critical fault propagation path.

[0038] Further, in the comparison unit, the PageRank value of the node after updating the weight is:

[0039]

[0040] Wherein, layer(Node i in M1) refers to the layer level of the i-th node in M1, layer(Node i in M2) refers to the layer level of the i-th node in M2, PR(Node i ) refers to the PageRank value of the node i after using the PageRank algorithm, and PR'(Node i ) refers to the PageRank value of the node i after updating the weight.

[0041] In the calculation unit, the edge weight existing in the skeleton matrix calculated by using the weighted PageRank calculation is:

[0042]

[0043] Wherein, PR(i) refers to the PageRank value of the i-th node, PR(j) refers to the PageRank value of the j-th node, and PR(e i→j ) refers to the weight value of the edge between the node i and the node j.

[0044] The present application has the following beneficial effects:

[0045] This invention provides a method for analyzing equipment fault propagation based on unstable point detection. It constructs a fault propagation graph according to the fault propagation relationship between complex industrial systems; uses the BRISM ​​method under different rules to detect unstable points; uses the PageRank method to evaluate the importance of the original fault propagation graph and update the weights of unstable points; calculates the weights of edges in the hierarchical results; and identifies key fault propagation paths between different systems based on the edge weights. This invention provides a device fault propagation analysis method based on instability point detection. It comprehensively considers the shortcomings of traditional complex system component correlation analysis techniques and improves upon the traditional ISM (Interpretative Structural Modeling) layering technique. By alternately adding different extraction rules, it proposes the BRISM ​​method to clarify the causal relationships between different components, fully considering the influence of the prior set on the layering results. The PageRank algorithm is used to evaluate the importance of each component in the original fault propagation graph and calculate the PageRank value. By comparing the layering results of the BRISM ​​method with different initial layer extraction rules, unstable points are identified. The PageRank value is updated using the hierarchical relationship of the unstable points, and a weighted PageRank algorithm is proposed. This effectively reduces the impact of unstable points on critical propagation paths, further improving the accuracy of fault propagation path identification, and ultimately effectively identifying the fault propagation paths of different faulty components. Attached Figure Description

[0046] Figure 1 A flowchart illustrating an embodiment of the equipment fault propagation analysis method based on unstable point detection provided by the present invention;

[0047] Figure 2 for Figure 1 The diagram shows a detailed process flow of an embodiment of the steps for constructing a fault propagation graph based on the fault propagation relationships between complex industrial systems.

[0048] Figure 3 for Figure 1 The diagram shows a detailed flowchart of an embodiment of the steps for detecting unstable points using the BRISM ​​method under different rules, evaluating the importance of the original fault propagation graph using the PageRank method, and updating the weights of unstable points.

[0049] Figure 4 for Figure 1 The diagram shows a detailed flowchart of an embodiment of the step of identifying critical fault propagation paths between different systems based on the weights of the edges in the calculation results of the hierarchical structure shown.

[0050] Figure 5A general flow chart of an embodiment of the device fault propagation analysis method based on unstable point detection provided by the present application;

[0051] Fig. 6(a) is a fault propagation result diagram using traditional ISM of the present application;

[0052] Fig. 6(b) is a fault propagation result diagram using BRISM of the present application;

[0053] Fig. 7(a) is a fault propagation result diagram using traditional ISM of the present application after injection of the fault propagation result;

[0054] Fig. 7(b) is a fault propagation result diagram using BRISM of the present application after injection of the fault propagation result;

[0055] Figure 8 A functional block diagram of an embodiment of the device fault propagation analysis system based on unstable point detection provided by the present application;

[0056] Figure 9 For Figure 8 A functional module schematic diagram of an embodiment of the construction module shown in the present application;

[0057] Figure 10 For Figure 8 A functional module schematic diagram of an embodiment of the evaluation module shown in the present application;

[0058] Figure 11 For Figure 8 A functional module schematic diagram of an embodiment of the identification module shown in the present application.

[0059] BRIEF DESCRIPTION OF DRAWINGS

[0060] 10, construction module; 20, evaluation module; 30, identification module; 11, construction unit; 12, conversion unit; 21, layering unit; 22, comparison unit; 31, calculation unit; 32, search unit. DETAILED DESCRIPTION

[0061] In order to better understand the above technical solutions, the above technical solutions will be described in detail below in conjunction with the drawings of the specification and specific embodiments.

[0062] As Figure 1 and Figure 2 shown, the first embodiment of the present application proposes a device fault propagation analysis method based on unstable point detection, comprising the following steps:

[0063] Step S100, constructing a fault propagation graph according to the fault propagation relationship between industrial complex systems.

[0064] Industrial complex system includes magnetic drive pump system. The magnetic drive pump system is roughly divided into 12 systems according to functional system, which are mechanical transmission system, sealing system, filtering system, cooling system, compression system, electrical control system, adjustment system, lubrication system, detection system, oil injection system, support system and safety system. The influence relationship of single component of industrial complex system on other components is analyzed to build a fault propagation subgraph, and multiple propagation subgraphs are combined to form a complete fault propagation graph.

[0065] Step S200, BRISM method under different rules is used to detect unstable points, PageRank method is used to evaluate the importance of the original fault propagation graph, and the weight of the unstable point is updated.

[0066] BRISM (Bilateral Rotation Interpretative Structural Modeling) method is used to divide the system into layers and detect unstable points, and PageRank (Weighted PageRank Algorithm) method is used to evaluate the importance of the original fault propagation graph, and the weight of the unstable point is updated.

[0067] Step S300, the weight of the edge in the hierarchical result is calculated, and the key fault propagation path between different systems is identified according to the weight of the edge.

[0068] The edge weight existing in the skeleton matrix is calculated by using the weighted PageRank, and according to the result obtained by the weighted PageRank algorithm, the source fault point is found from the fault point to form the final key fault propagation path.

[0069] The device fault propagation analysis method based on unstable point detection provided in the embodiment is to construct a fault propagation graph according to the fault propagation relationship between industrial complex systems; BRISM method under different rules is used to detect unstable points, and the importance of the original fault propagation graph is evaluated by using the PageRank method, and the weight of the unstable point is updated; the weight of the edge in the hierarchical result is calculated, and the key fault propagation path between different systems is identified according to the weight of the edge. The device fault propagation analysis method based on unstable point detection provided in the embodiment can comb the causal relationship between different components by adding different extraction rules in turn, fully considers the influence of the precedent set on the hierarchical result, evaluates the importance of each component in the original fault propagation graph by using the PageRank algorithm, calculates the PageRank value, compares the hierarchical results of the BRISM method under different initial layer extraction rules, finds the unstable point, updates the PageRank value by using the hierarchical relationship of the unstable point, and further proposes a weighted PageRank algorithm, which effectively reduces the influence of the unstable point on the key propagation path and further improves the accuracy of the fault propagation path identification, and finally effectively identifies the fault propagation path of different fault components.

[0070] Further, see Figure 2 , Figure 2 for Figure 1 an embodiment of the step S100 shown in the flowchart, in the embodiment, the step S100 includes:

[0071] The step S110 is to analyze the influence relationship of a single component of an industrial complex system on other components, and construct a fault propagation subgraph.

[0072] The relationship between components of an industrial complex system is analyzed, and a fault propagation subgraph is constructed through the influence relationship of a single component on other components.

[0073] The step S120 is to combine a plurality of propagation subgraphs to form a complete fault propagation graph, and convert the fault propagation graph into an adjacency matrix.

[0074] The propagation subgraphs are integrated, a plurality of propagation subgraphs are combined to form a complete fault propagation graph, and the fault propagation graph is converted into an adjacency matrix. In the adjacency matrix, if two components have a direct influence relationship, the row and column representing the two components are set to 1, and if two components have no direct influence relationship, the row and column representing the two components are set to 0.

[0075] The device fault propagation analysis method based on unstable point detection provided in the embodiment is to analyze the influence relationship of a single component of an industrial complex system on other components, construct a fault propagation subgraph, combine multiple propagation subgraphs to form a complete fault propagation graph, and convert the fault propagation graph into an adjacency matrix. The device fault propagation analysis method based on unstable point detection provided in the embodiment is to use the PageRank algorithm to evaluate the importance of each component in the original fault propagation graph and calculate the PageRank value. The layered results of the BRISM method in different initial layer extraction rules are compared to find unstable points, the PageRank value is updated by using the hierarchical relationship of the unstable points, and then the weighted PageRank algorithm is proposed, which effectively reduces the influence of unstable points on the key propagation path and further improves the accuracy of fault propagation path identification, and finally effectively identifies the fault propagation path of different fault components.

[0076] Preferably, see Figure 3 , Figure 3 for Figure 1 an embodiment of the step S100 shown in the figure, in this embodiment, the step S200 includes:

[0077] Step S210, using the BRISM method under different rules to layer the adjacency matrix.

[0078] The BRISM (Bilateral Rotation Interpretative Structural Modeling) method is used to layer the adjacency matrix, and the specific process is:

[0079] a. The adjacency matrix X n*n formed by the fault propagation graph is normalized to obtain the normalized matrix Y n*n , that is:

[0080]

[0081] In formula (1), X n*n is the adjacency matrix, Y n*n is the normalized matrix, denotes the maximum value of the row sum of the adjacency matrix X n*n .

[0082] b. The direct influence relationship and the indirect influence relationship between nodes are accumulated to construct a comprehensive influence matrix S n*n , that is:

[0083] S n*n = lim k→∞ (Y 1 + Y 2+... + Y k ) = Y(I-Y) -1 (2)

[0084] In formula (2), S n*n is the comprehensive influence matrix, I is the unit matrix; Y is the normalized matrix.

[0085] c. The comprehensive influence matrix S n*n is added to the unit matrix I to obtain the overall influence matrix H n*n , that is:

[0086] H n*n = S + I (3)

[0087] In formula (3), H n*n is the overall influence matrix, I is the unit matrix, and S is the comprehensive influence matrix.

[0088] d. The overall influence matrix H n*n is set to 1 if the element is greater than the set threshold σ, and the reachable matrix M n*n is obtained, that is:

[0089]

[0090] In formula (4), I is the unit matrix, and h ij is the element value of the i-th row and j-th column in the matrix.

[0091] e. The reachable set R n*n , the predecessor set S i , and the common set C i of the reachable matrix M i are calculated, that is:

[0092] R i = {v j |v j ∈ V, m ij = 1}, (i = 1, 2,..., n) (5)

[0093] S i = {v j |v j ∈ V, m ji = 1}, (i = 1, 2,..., n) (6)

[0094] C i = R i ∩ S i , (i = 1, 2,..., n) (7)

[0095] In formulas (5)-(7), v j is the j-th node in the matrix, V is the node set, and m ijis the value of the element in the i-th row and j-th column of the reachable matrix, m ji is the value of the element in the j-th row and i-th column of the reachable matrix.

[0096] f.Each layer alternately judges whether the reachable set R i , the predecessor set S i and the common set C i are equal, determines the hierarchical components, that is:

[0097] R i =C i , (i=1, 2, …, n) (8)

[0098] S i =C i , (i=1, 2, …, n) (9)

[0099] In formulas (8) and (9), R i =C i and S i =C i Two rules are alternated after extracting each layer of nodes.

[0100] g.Analyzing the relationship of each layer of components, removing self-relationship and cross-level binary relationship to obtain the skeleton matrix.

[0101] Step S220, comparing the hierarchical results of the BRISM method in different initial layer extraction rules, finding unstable points, and updating the PageRank value using the hierarchical relationship of the unstable points.

[0102] The importance of each node in PageRank (Weighted PageRank Algorithm, web page ranking algorithm) is evaluated, and the specific process is:

[0103] a.Using the BRISM algorithm proposed above twice to layer the adjacency matrix, and the initial layer rules are exchanged in the two times of layering, different skeleton matrices, that is:

[0104] M′1=BRISM(first layer: R i =C i ) (10)

[0105] M′2=BRISM(first layer: S i =C i ) (11)

[0106] In formulas (10) and (11), M′1 refers to the rule that R i =C i is used as the initial layer to extract nodes, and M′2 refers to the rule that Si =C i Rules for extracting nodes as initial layers.

[0107] b. Compare M'1 and M'2 to find the different points of each layer, called unstable points, calculate the PageRank value of each node using the PageRank algorithm, and update the unstable points in the different layers where the two hierarchical results are located, that is:

[0108]

[0109] In formula (12), layer(Node i in M'1) refers to the layer where the i-th node is located in M'1, layer(Node i in M'2) refers to the layer where the i-th node is located in M'2, PR(Node i ) refers to the PageRank value of node i after using the PageRank algorithm, and PR'(Node i ) refers to the PageRank value of node i after updating the weight.

[0110] The device fault propagation analysis method based on unstable point detection provided in the embodiment uses the BRISM method under different rules to layer the adjacency matrix; compares the hierarchical results of the BRISM method in different initial layer extraction rules to find unstable points, and updates the PageRank value using the hierarchical relationship of the unstable points. The device fault propagation analysis method based on unstable point detection provided in the embodiment adds different extraction rules in turn, proposes the BRISM method to sort out the causal relationship between different components, fully considers the influence of the precedent set on the hierarchical result; uses the PageRank algorithm to evaluate the importance of each component in the original fault propagation graph, calculates the PageRank value; compares the hierarchical results of the BRISM method in different initial layer extraction rules to find unstable points, and updates the PageRank value using the hierarchical relationship of the unstable points, and further proposes a weighted PageRank algorithm, which effectively reduces the influence of unstable points on the key propagation path and further improves the accuracy of fault propagation path identification, and finally effectively identifies the fault propagation path of different fault components.

[0111] Further, see Figure 4 , Figure 4 as a refinement of an embodiment of the step S300 shown in Figure 1 , in this embodiment, the step S300 comprises:

[0112] Step S310, calculate the edge weight existing in the skeleton matrix using the weighted PageRank.

[0113] The edge weights in the skeleton matrix are calculated using weighted PageRank, i.e.:

[0114]

[0115] In formula (13), PR(i) refers to the PageRank value of the i-th node, PR(j) refers to the PageRank value of the j-th node, and PR(e) refers to the PageRank value of the j-th node. i→j The weight of the edge between node i and node j is denoted as .

[0116] Step S320: Based on the results obtained from the weighted PageRank algorithm, starting from the fault point, find the point with the larger incoming edge weight as the upstream of the fault, and continue to trace the source until the source fault point is found, forming the final critical fault propagation path.

[0117] According to the weighted PageRank algorithm, when a component fails, the failure propagates along edges with higher weights. Therefore, starting from the failure point, we search for points with high incoming edge weights as upstream points of the failure, tracing the source until we find the root failure point, thus forming the final critical failure propagation path.

[0118] The equipment fault propagation analysis method based on unstable point detection provided in this embodiment utilizes weighted PageRank to calculate the edge weights in the skeleton matrix. Based on the results obtained from the weighted PageRank algorithm, starting from the fault point, it searches for points with large incoming edge weights as upstream points of the fault, tracing the source until the originating fault point is found, forming the final critical fault propagation path. This method, by alternately adding different extraction rules, proposes the BRISM ​​method to clarify the causal relationships between different components, fully considering the impact of the antecedent set on the hierarchical results. The PageRank algorithm is used to evaluate the importance of each component in the original fault propagation graph, calculating the PageRank value. By comparing the hierarchical results of the BRISM ​​method under different initial layer extraction rules, unstable points are identified, and the PageRank value is updated using the hierarchical relationship of the unstable points. This leads to the proposed weighted PageRank algorithm, effectively reducing the impact of unstable points on the critical propagation path, further improving the accuracy of fault propagation path identification, and ultimately effectively identifying the fault propagation paths of different faulty components.

[0119] like Figure 8 As shown, Figure 8The function block diagram of an embodiment of the device fault propagation analysis system based on unstable point detection provided by the present application, in this embodiment, the device fault propagation analysis system based on unstable point detection comprises a construction module 10, an evaluation module 20 and an identification module 30, wherein the construction module 10 is used to construct a fault propagation graph according to the fault propagation relationship between industrial complex systems; the evaluation module 20 is used to detect unstable points by using the BRISM method under different rules, and at the same time, the importance of the original fault propagation graph is evaluated by using the PageRank method, and the weight of the unstable points is updated; and the identification module 30 is used to calculate the weight of the edges in the hierarchical result, and identify the key fault propagation paths between different systems according to the weight of the edges.

[0120] The industrial complex system comprises a magnetic drive pump system. The magnetic drive pump system is roughly divided into 12 systems according to functional systems, which are respectively a mechanical transmission system, a sealing system, a filtering system, a cooling system, a compression system, an electrical control system, an adjustment system, a lubrication system, a detection system, an oil injection system, a support system and a safety system. The construction module 10 analyzes the influence relationship of a single component of the industrial complex system on other components to construct a fault propagation subgraph, and combines multiple propagation subgraphs to form a complete fault propagation graph.

[0121] The evaluation module 20 uses the BRISM (Bilateral Rotation Interpretative Structural Modeling) method to stratify the system and detect unstable points, and at the same time, the importance of the original fault propagation graph is evaluated by using the PageRank (Weighted PageRank Algorithm) method, and the weight of the unstable points is updated.

[0122] The identification module 30 uses the weighted PageRank to calculate the edge weight existing in the skeleton matrix, and according to the result obtained by the weighted PageRank algorithm, the source fault point is found from the fault point to form the final key fault propagation path.

[0123] Compared with the prior art, the device fault propagation analysis system based on unstable point detection provided in the embodiment adopts a construction module 10, an evaluation module 20 and an identification module 30, constructs a fault propagation graph according to the fault propagation relationship between industrial complex systems, detects unstable points by using a BRISM method under different rules, simultaneously evaluates the importance of the original fault propagation graph by using a PageRank method, and updates the weight of the unstable points, calculates the weight of the edges in the hierarchical result, and identifies the key fault propagation paths between different systems according to the weight of the edges. The device fault propagation analysis system based on unstable point detection provided in the embodiment combs the causal relationship between different components by using the BRISM method by taking turns to add different extraction rules, fully considers the influence of the precedent set on the hierarchical result, evaluates the importance of each component in the original fault propagation graph by using a PageRank algorithm, calculates the PageRank value, compares the hierarchical results of the BRISM method in different initial layer extraction rules, finds unstable points, updates the PageRank value by using the hierarchical relationship of the unstable points, and further proposes a weighted PageRank algorithm, which effectively reduces the influence of the unstable points on the key propagation paths and further improves the accuracy of fault propagation path identification, and finally effectively identifies the fault propagation paths of different fault components.

[0124] Further, see Figure 9 , Figure 9 For Figure 8 The function module schematic diagram of an embodiment of the construction module shown in the embodiment, in the embodiment, the construction module 10 includes a construction unit and a conversion unit 12, wherein the construction unit 11 is used to analyze the influence relationship of a single component of an industrial complex system on other components, and construct a fault propagation subgraph; the conversion unit 12 is used to combine multiple propagation subgraphs to form a complete fault propagation graph, and convert the fault propagation graph into an adjacency matrix.

[0125] The construction unit 11 analyzes the relationship between the components of the industrial complex system, and constructs a fault propagation subgraph through the influence relationship of a single component on other components.

[0126] The conversion unit 12 integrates the propagation subgraphs, combines multiple propagation subgraphs to form a complete fault propagation graph, and converts the fault propagation graph into an adjacency matrix. In the adjacency matrix, if two components have a direct influence relationship, the position is 1, and if two components have no direct influence relationship, the position is 0.

[0127] The device fault propagation analysis system based on unstable point detection provided by the embodiment comprises a construction module 10, a construction unit and a conversion unit 12, a fault propagation subgraph is constructed by analyzing the influence relationship of a single component of an industrial complex system on other components, a plurality of propagation subgraphs are combined to form a complete fault propagation graph, and the fault propagation graph is converted into an adjacency matrix. The device fault propagation analysis method based on unstable point detection provided by the embodiment uses a PageRank algorithm to evaluate the importance of each component in the original fault propagation graph and calculate a PageRank value. The layered results of the BRISM method in different initial layer extraction rules are compared to find unstable points, the PageRank value is updated by using the hierarchical relationship of the unstable points, a weighted PageRank algorithm is further proposed, the influence of the unstable points on the key propagation path is effectively reduced, the accuracy of fault propagation path identification is further improved, and finally the fault propagation path of different fault components is effectively identified.

[0128] Preferably, see Figure 10 , Figure 10 for Figure 8 an embodiment of the evaluation module shown in the function block diagram of the evaluation module 20 in the embodiment, the evaluation module 20 comprises a hierarchical unit 21 and a comparison unit 22, wherein the hierarchical unit 21 is used to layer the adjacency matrix by using the BRISM method under different rules; the comparison unit 22 is used to compare the layered results of the BRISM method in different initial layer extraction rules, find unstable points, and update the PageRank value by using the hierarchical relationship of the unstable points.

[0129] The hierarchical unit 21 layers the adjacency matrix by using the BRISM (Bilateral Rotation Interpretative Structural Modeling) method, and the specific process is as follows:

[0130] a. The adjacency matrix X n*n formed by the fault propagation graph is normalized to obtain a normalized matrix Y n*n , that is:

[0131]

[0132] In formula (14), X n*n is the adjacency matrix, Y n*n is the normalized matrix, denotes the maximum value of the row sum of the adjacency matrix X n*n .

[0133] b. The direct influence relationship and the indirect influence relationship between nodes are accumulated to construct a comprehensive influence matrix S n*n , that is:

[0134] S n*n = lim k→∞ (Y 1 + Y 2 + … + Y k ) = Y(I-Y) -1 (15)

[0135] In formula (15), S n*n is the comprehensive influence matrix, I is the unit matrix; Y is the normalized matrix.

[0136] c. Add the comprehensive influence matrix S n*n to the unit matrix I to obtain the overall influence matrix H n*n , that is:

[0137] H n*n = S + I (16)

[0138] In formula (16), H n*n is the overall influence matrix, I is the unit matrix, and S is the comprehensive influence matrix.

[0139] d. The overall influence matrix H n*n is greater than the set threshold σ, and the element is set to 1 to obtain the reachable matrix M n*n , that is:

[0140]

[0141] In formula (17), I is the unit matrix, and h ij is the element value of the i row and j column in the matrix.

[0142] e. Calculate the reachable set R n*n , the predecessor set S i and the common set C i of the reachable matrix M i , that is:

[0143] R i = {v j |v j ∈ V, m ij = 1}, (i = 1, 2, …, n) (18)

[0144] S i = {v j |v j ∈ V, m ji = 1}, (i = 1, 2, …, n) (19)

[0145] C i = R i ∩ S i(i = 1, 2, …, n) (20)

[0146] In the formula (18) ~ (20), v j is the jth node in the matrix, V is the node set, m ij is the element value of the ith row and jth column in the reachable matrix, m ji is the element value of the jth row and ith column in the reachable matrix.

[0147] f. Each layer alternately judges whether the reachable set R i , the predecessor set S i and the common set C i are equal, determines the hierarchical components, that is:

[0148] R i = C i , (i = 1, 2, …, n) (21)

[0149] S i = C i , (i = 1, 2, …, n) (22)

[0150] In the formula (21) and (22), R i = C i and S i = C i Two rules are alternated after extracting each layer node.

[0151] g. Analyze the relationship of each layer component, remove self relationship and cross-level binary relationship to obtain the skeleton matrix.

[0152] The comparison unit 22 uses PageRank (Weighted PageRank Algorithm, web page ranking algorithm) to evaluate the importance of each node, and the specific process is:

[0153] a. The adjacency matrix is layered by using the BRISM algorithm proposed above twice, and the initial layer rules are exchanged in the two times of layering, and different skeleton matrices, that is:

[0154] M′1 = BRISM (first layer: R i = C i ) (23)

[0155] M′2 = BRISM (first layer: S i = C i ) (24)

[0156] In the formula (23) and (24), M′1 refers to the use of R i = C iAs the rule of initial layer extraction node, M'2 refers to using S i = C i As the rule of initial layer extraction node.

[0157] b. Comparing M'1 and M'2, finding the different points of each layer, called unstable points, using PageRank algorithm to calculate the PageRank value of each node, and using unstable points to update the different number of layers where the two hierarchical results are located, that is:

[0158]

[0159] In formula (25), layer(Node i in M'1) refers to the layer where the i-th node is located in M'1, layer(Node i in M'2) refers to the layer where the i-th node is located in M'2, PR(Node i ) refers to the PageRank value of node i after using PageRank algorithm; PR'(Node i ) refers to the PageRank value of node i after updating the weight.

[0160] The device fault propagation analysis system based on unstable point detection provided by the embodiment compares the hierarchical results of BRISM method under different initial layer extraction rules, finds unstable points, and updates the PageRank value by using the hierarchical relationship of unstable points. Compared with the prior art, the evaluation module 20 adopts the hierarchical unit 21 and the comparison unit 22, and uses the BRISM method under different rules to layer the adjacency matrix; the comparison unit 22 compares the hierarchical results of the BRISM method in different initial layer extraction rules, finds unstable points, and updates the PageRank value by using the hierarchical relationship of the unstable points. The device fault propagation analysis system based on unstable point detection provided by the embodiment adds different extraction rules in turn, proposes the BRISM method to sort out the causal relationship between different components, fully considers the influence of the precedent set on the hierarchical result; uses the PageRank algorithm to evaluate the importance of each component in the original fault propagation graph, calculates the PageRank value; compares the hierarchical results of the BRISM method in different initial layer extraction rules, finds unstable points, and updates the PageRank value by using the hierarchical relationship of the unstable points, and further proposes the weighted PageRank algorithm, which effectively reduces the influence of unstable points on the key propagation path, further improves the accuracy of fault propagation path identification, and finally effectively identifies the fault propagation path of different fault components.

[0161] Preferably, referring to Figure 11 , Figure 11 is Figure 8A function module schematic diagram of an embodiment of the identification module shown in the figure, in this embodiment, the identification module 30 includes a calculation unit 31 and a lookup unit 32, wherein the calculation unit 31 is configured to calculate the edge weight existing in the skeleton matrix by using the weighted PageRank; the lookup unit 32 is configured to find the point upstream of the fault as the fault according to the result obtained by the weighted PageRank algorithm, and trace back to find the trace fault point, and form the final key fault propagation path.

[0162] The calculation unit 31 calculates the edge weight existing in the skeleton matrix by using the weighted PageRank, that is:

[0163]

[0164] In formula (26), PR(i) refers to the PageRank value of the i th node, PR(j) refers to the PageRank value of the j th node, and PR(e i→j ) refers to the weight value of the edge between the i th node and the j th node.

[0165] The lookup unit 32 finds the point upstream of the fault as the fault according to the result obtained by the weighted PageRank algorithm, and traces back to find the trace fault point, and form the final key fault propagation path.

[0166] Compared with the prior art, the identification module 30 of the equipment fault propagation analysis system based on unstable point detection provided in this embodiment adopts the calculation unit 31 and the lookup unit 32, calculates the edge weight existing in the skeleton matrix by using the weighted PageRank, finds the point upstream of the fault as the fault according to the result obtained by the weighted PageRank algorithm, and traces back to find the trace fault point, and form the final key fault propagation path. The equipment fault propagation analysis system based on unstable point detection provided in this embodiment uses the BRISM method to sort out the causal relationship between different components by adding different extraction rules in turn, fully considers the influence of the precedent set on the hierarchical result, uses the PageRank algorithm to evaluate the importance of each component in the original fault propagation graph, calculates the PageRank value, compares the hierarchical results of the BRISM method in different initial layer extraction rules, finds the unstable point, updates the PageRank value by using the hierarchical relationship of the unstable point, and further proposes the weighted PageRank algorithm, which effectively reduces the influence of the unstable point on the key propagation path, further improves the accuracy of fault propagation path recognition, and finally effectively identifies the fault propagation path of different fault components.

[0167] In order to prove the effectiveness of the device fault propagation analysis method based on unstable point detection provided in the embodiment, the method is compared with the traditional ISM method, and further described. The fault propagation results of the method of the embodiment and the traditional ISM method are compared and analyzed, and it is found that after fault injection, the traditional ISM method modifies the key fault propagation path, and the fault propagation path between the systems adopting the method of the embodiment does not change, which shows that the method of the embodiment has robustness and can effectively identify the key fault propagation path of different components. The results of the two methods before fault injection are shown in FIG. 6(a) and FIG. 6(b), and the comparison results of the two methods after fault injection are shown in FIG. 7(a) and FIG. 7(b).

[0168] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Therefore, the appended claims are intended to encompass within their scope all possible variations and modifications of the preferred embodiments. It is apparent that those skilled in the art can, without departing from the spirit or scope of the application, make various changes and modifications of the application. Thus, the application is intended to encompass all such changes and modifications as fall within the scope of the claims, together with all equivalents thereof.

Claims

1. A method for device fault propagation analysis based on unstable point detection, characterized in that, The method comprises the following steps: constructing a fault propagation graph according to a fault propagation relationship between industrial complex systems; detecting unstable points by using a BRISM method under different rules, performing importance evaluation on the original fault propagation graph by using a PageRank method, and updating weights of the unstable points; calculating weights of edges in a hierarchical result, and identifying a key fault propagation path between different systems according to the weights of the edges; wherein the detecting of the unstable points by using the BRISM method under the different rules, the performing of the importance evaluation on the original fault propagation graph by using the PageRank method, and the updating of the weights of the unstable points comprise: Step 1: performing hierarchical division on an adjacency matrix by using the BRISM method under different rules; the hierarchical division is performed on the adjacency matrix by using a BRISM (Bilateral Rotation Interpretative Structural Modeling) method, and the specific process is as follows: a. adjacency matrix formed by the fault propagation graph Normalization processing is performed to obtain a normalized matrix That is, （1） In equation (1), is the adjacency matrix, is the normalization matrix, refers to the adjacency matrix the maximum value of the row sums; b. Accumulate the direct and indirect influence relations between nodes to build a comprehensive influence matrix That is: （2） In equation (2), is the comprehensive influence matrix, is the unit matrix; is the normalization matrix; c. The overall impact matrix is obtained by adding the integrated impact matrix to the unit matrix , i.e.: ​ （3） In equation (3), is the identity matrix, is the comprehensive influence matrix; d. Overall influence matrix Set elements greater than a set threshold to 1 to obtain the reachability matrix i.e.: （4） In equation (4), is the identity matrix, is the element value in the matrix row column. e. the reachable matrix of the reachable set , the predecessor set and the co-set i.e.: （5） （6） （7） In formulas (5) to (7), is the node in the matrix, is the node set, is the element value in the reachable matrix, is the element value in the reachable matrix; and is the element value in the reachable matrix.​​​​ f. Each layer in turn judges the reachable sets , the predecessor sets and the co-sets for equality, determines the hierarchical components, namely: （8） （9） In equations (8) and (9), and The two rules are alternated after extracting each layer of nodes. g. analyzing a relationship of each layer component, removing self relationship and cross-level binary relationship to obtain a skeleton matrix; Step 2: comparing hierarchical results of the BRISM method under different initial layer extraction rules to find unstable points, and updating a PageRank value according to a hierarchical relationship of the unstable points; each node is evaluated for importance by using a PageRank (Weighted PageRank Algorithm) method, and the specific process is as follows: a. performing hierarchical division on the adjacency matrix by using the BRISM algorithm twice, and changing the initial layer rule in the two times of hierarchical division, and different skeleton matrices are obtained, namely: （10） （11） In formulas (10) and (11), means that the as the initial layer extraction node, means that the as the initial layer extraction node; b.Comparison and , find the different points of each layer, called unstable points, calculate the PageRank value of each node using PageRank algorithm, and update the unstable points in different layers where the two hierarchical results are located, that is: （12） In Equation (12), refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the PageRank value of the node after using the PageRank algorithm; refers to the PageRank value of the node after updating the weight.

2. The method of claim 1, wherein, the step of constructing the fault propagation graph according to the fault propagation relationship between the industrial complex systems comprises: analyzing an influence relationship of a single component of an industrial complex system on other components to construct a fault propagation subgraph; combining multiple propagation subgraphs to form a complete fault propagation graph, and converting the fault propagation graph into an adjacency matrix.

3. The method of claim 2, wherein, the method of detecting the unstable points by using the BRISM method under the different rules, performing the importance evaluation on the original fault propagation graph by using the PageRank method, and updating the weights of the unstable points comprises: performing hierarchical division on the adjacency matrix by using the BRISM method under different rules; comparing hierarchical results of the BRISM method under different initial layer extraction rules to find unstable points, and updating a PageRank value according to a hierarchical relationship of the unstable points.

4. The method of claim 3, wherein, the step of calculating weights of edges in a hierarchical result, and identifying a key fault propagation path between different systems according to the weights of the edges comprises: calculating edge weights existing in the skeleton matrix by using a weighted PageRank; according to a result obtained by using the weighted PageRank algorithm, starting from a fault point, finding a point with a large in-edge weight as an upstream fault point, and tracing back to source, finally finding a trace-back fault point, and forming a final key fault propagation path.

5. The method of claim 4, wherein, a PageRank value of a node after the weights are updated is: wherein, refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the PageRank value of the node after using the PageRank algorithm, refers to the PageRank value of the node after updating the weight; edge weights existing in the skeleton matrix calculated by using the weighted PageRank are: wherein, denotes the PageRank value of the th node, denotes the PageRank value of the th node, denotes the weight value of the edge between the nodes and the node .

6. An unstable point detection based system for analyzing device fault propagation, the system comprising: comprise: The construction module (10) is configured to construct a fault propagation graph according to a fault propagation relationship between industrial complex systems; The evaluation module (20) is configured to detect unstable points by using BRISM methods under different rules, and to update weights of the unstable points by using a PageRank method to evaluate importance of the original fault propagation graph; The identification module (30) is configured to calculate weights of edges in a hierarchical result, and to identify a key fault propagation path between different systems according to the weights of the edges; The evaluation module includes: A first evaluation submodule is configured to perform hierarchical analysis on the adjacency matrix by using BRISM methods under different rules; The BRISM (Bilateral Rotation Interpretative Structural Modeling) method is used to perform hierarchical analysis on the adjacency matrix, and the specific process is as follows: a. adjacency matrix formed by the fault propagation graph Normalization is performed to obtain a normalized matrix That is: （1） In equation (1), is the adjacency matrix, is the normalization matrix, refers to the adjacency matrix the maximum value of the row sums. b. Accumulate the direct and indirect influence relations between nodes to build a comprehensive influence matrix That is: （2） In equation (2), is the comprehensive influence matrix, is the unit matrix; is the normalization matrix; c. The overall impact matrix is obtained by adding the combined impact matrix to the unit matrix , i.e.: ​ （3） In equation (3), is the identity matrix, is the comprehensive influence matrix; d. Overall influence matrix Set elements greater than a set threshold to 1 to obtain the reachability matrix i.e.: （4） In equation (4), is the identity matrix, is the element value in the matrix row column. e. the reachability matrix the reachability set the predecessor set and the co-set i.e.: （5） （6） （7） In formulas (5) to (7), It is the first in the matrix 1 node It is a set of nodes. It is the th in the reachability matrix Line number The element values ​​of the column, It is the th in the reachability matrix Line number The element values ​​of the column; f. Each layer in turn judges the reachable sets , the predecessor sets and the co-sets for equality, determines the hierarchical components, namely: （8） （9） In equations (8) and (9), and The two rules are alternated after extracting each layer of nodes. g. Analyzing component relationships in each layer, removing self-relationships and cross-level binary relationships to obtain a skeleton matrix; A second evaluation submodule is configured to compare hierarchical results of the BRISM method under different initial layer extraction rules, to find unstable points, and to update PageRank values by using hierarchical relationships of the unstable points; The PageRank (Weighted PageRank Algorithm) method is used to evaluate importance of each node, and the specific process is as follows: a. The BRISM algorithm is used twice to perform hierarchical analysis on the adjacency matrix, and initial layer rules are exchanged in the two times of hierarchical analysis, and different skeleton matrices are obtained, that is: （10） （11） In formulas (10) and (11), means that the As a rule of extracting nodes of the initial layer, means that the As a rule of extracting nodes of the initial layer; b.Comparison and , find the different points of each layer, called unstable points, calculate the PageRank value of each node using PageRank algorithm, and update the unstable points in different layers where the two hierarchical results are located, that is: （12） In formula (12), It refers to the first Each node The level at which they are located It refers to the first Each node The level at which they are located This refers to the node after using the PageRank algorithm. PageRank value; The node after weight update PageRank value.

7. The system for device fault propagation analysis based on unstable point detection of claim 6, wherein, The construction module (10) includes: A construction unit (11) is configured to analyze influence relationships of a single component of an industrial complex system on other components, and to construct a fault propagation subgraph; A conversion unit (12) is configured to combine a plurality of fault propagation subgraphs to form a complete fault propagation graph, and to convert the fault propagation graph into an adjacency matrix.

8. The system for device fault propagation analysis based on unstable point detection of claim 7, wherein, The evaluation module (20) includes: A hierarchical analysis unit (21) is configured to perform hierarchical analysis on the adjacency matrix by using BRISM methods under different rules; A comparison unit (22) is configured to compare hierarchical results of the BRISM method under different initial layer extraction rules, to find unstable points, and to update PageRank values by using hierarchical relationships of the unstable points.

9. The system for device fault propagation analysis based on unstable point detection of claim 8, wherein, The identification module (30) includes: A calculation unit (31) is configured to calculate edge weights existing in the skeleton matrix by using a weighted PageRank; A finding unit (32) is configured to find, from a fault point, a point with a large edge weight as an upstream fault point, to perform tracing, and to finally find a traced fault point to form a final key fault propagation path.

10. The system for device fault propagation analysis based on unstable point detection of claim 9, wherein, In the comparison unit (22), a PageRank value of a node after the weight is updated is: wherein, refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the level at which the th node is located, refers to the PageRank value of the node after using the PageRank algorithm, refers to the PageRank value of the node after updating the weight; In the calculation unit (31), an edge weight existing in the skeleton matrix calculated by using the weighted PageRank is: wherein, denotes the PageRank value of the th node, denotes the PageRank value of the th node, denotes the weight value of the edge between the nodes and the node .

Images