CN110213087B - Complex system fault positioning method based on dynamic multilayer coupling network - Google Patents

Abstract

The invention provides a complex system fault positioning method based on a dynamic multilayer coupling network, which comprises the following steps: step A: constructing a dynamic multi-layer coupling network under multiple modes; and B: positioning a fault time period of a complex system; and C: and positioning the fault node of the complex system. The invention provides a complex system fault positioning method based on a dynamic multilayer coupling network by applying a complex network theory, aiming at the difficulty that the traditional reliability analysis means is often applied to the scene of single fault source and static analysis and the fault positioning of a complex system with the characteristics of dynamic propagation, cascade failure, multiple faults, uncertainty and the like is difficult to realize.

Description

A Fault Location Method for Complex Systems Based on Dynamic Multilayer Coupling Networks

technical field

The invention provides a complex system fault location method based on dynamic multi-layer coupling network, which relates to a complex system fault location method based on dynamic multi-layer coupling network, and belongs to the field of complexity science and reliability technology.

Background technique

Since the 21st century, the progress of information science led by the Internet has not only driven the rapid development of the world's industrial and economic level, but also brought an impact on systems science and systems engineering. The traditional simple system cannot solve the increasingly diversified application requirements. Emerging technologies such as embedded technology, Internet of Things, and artificial intelligence have greatly improved the system performance while making the system structure more and more complex. Taking the information physical system as an example, the information layer, physical layer, network layer and decision-making layer of the internal structure of the system have multi-level nonlinear coupling relationships, and functions emerge when the system runs dynamically. When a complex system fails, its increasingly complex topological structure and nonlinear coupling dynamic characteristics make the fault have the characteristics of dynamic propagation, cascading failure, and multiple faults. In addition, due to the complexity of coupled faults, the failure of complex systems is uncertain even when the same test case is executed. Traditional reliability analysis methods are often used in single-fault source and static analysis scenarios. For these types of faults with complex mechanisms and structural coupling, it is difficult to use traditional reliability methods to effectively implement fault location analysis. How to achieve complex system fault location becomes a problem. An urgent problem to be solved.

As a model method that abstractly describes the overall structure of complex systems and the interaction rules between individuals within the system, complex network theory provides a new solution to this problem. The dynamic multi-layer coupled network based on the complex network theory is an abstract model of the dynamic operating mechanism at different levels within the complex system. Its form is a set of static multi-layer coupled networks under a set of time slices. It consists of single-layer networks and coupled edge sets between single-layer networks. Compared with the traditional reliability model, this model can more realistically reflect the internal dynamic evolution mechanism of the complex system under normal operation and fault operation. Combined with the statistical indicators of the network model, it can more efficiently analyze the fault propagation path and explore the cascading failure mechanism. , Realize reliability analysis and improvement tasks such as multi-fault source location.

The present invention proposes an effective solution to the above problems. This scheme mainly proposes a set of solutions to the problem of how to locate faults in complex systems. First, test cases with high failure probability are executed, and dynamic coupled network models for normal and faulty complex systems are established respectively, and then complex systems are implemented based on complex network statistical indicators. Roughly locate faults, locate the time period when the performance of the complex system is degraded or even collapsed, and then realize the fine location of the complex system faults in this time period, and analyze the sudden changes of statistical indicators in the static multi-layer coupling network under each time slice in this time period in turn. nodes, and finally realize fault location of complex systems. The complex system fault location method based on the dynamic multi-layer coupling network adopted by the present invention solves the problem that it is difficult to locate complex faults such as dynamic propagation, cascading failure and multiple faults of complex systems by using traditional reliability analysis methods.

SUMMARY OF THE INVENTION

The invention mainly provides a complex system fault location method based on a dynamic multi-layer coupling network. Traditional reliability analysis methods are often used in single-fault source and static analysis scenarios. They have limitations in implementing complex system fault location tasks with dynamic propagation, cascading failures, multiple faults, and uncertainty, making it difficult to achieve fault location. analyze. Therefore, we propose a method based on complex network theory that can effectively realize fault location of complex systems.

In view of the above technical problems and the purpose of the present invention, this paper proposes a complex system fault location method based on a dynamic multi-layer coupled network. The scheme includes the following parts:

(1) Purpose of the invention

In view of the defects existing in the prior art, the purpose of the present invention is to provide a complex system fault location method based on a dynamic multi-layer coupled network, taking into account the complex structure of the complex system, dynamic propagation of faults, cascading failures, multi-layer faults Due to the characteristics of coupling and other characteristics, the traditional reliability analysis methods often used in static fault and single-fault scenarios are difficult to locate the fault of the complex system. By using the complex system fault location method based on the dynamic multi-layer coupling network proposed by the present invention, the complex network can be used. Theoretical statistical indicators compare and analyze the similarities and differences between complex system faults and normal dynamic multi-layer coupled networks from macro and micro perspectives, so as to achieve clearer and more accurate complex system fault location, which is convenient for system designers to understand system performance and failure mechanism, and quickly troubleshoot Fault causes, optimize system reliability indicators.

(2) Technical solutions

In order to achieve the above object, the technical solution adopted by the method of the present invention is: a method for locating faults in complex systems based on a dynamic multi-layer coupling network.

The present invention is a complex system fault location method based on dynamic multi-layer coupling network, and the steps are as follows:

Step A: Construct a dynamic multi-layer coupling network under multi-mode;

Step B: locate the fault period of the complex system;

Step C: Locate the faulty node of the complex system.

The specific meaning of the "dynamic multi-layer coupled network" described in step A is: a dynamic network model that can abstractly reflect the internal dynamic running state of a complex system, and the dynamic network model is a set of static networks based on time slices. The system often has multiple layers, and each layer has cross-layer coupling in addition to the same-layer interaction. For example, there is a module mapping relationship between the software layer and the hardware layer of the embedded system. Therefore, in order to describe the system state more realistically, the dynamic network model contains Multi-layer coupling network of multi-layer nodes, same-layer connection edges and coupled connection edges; "Constructing a dynamic multi-layer coupling network under multi-mode" described in step A, including the following steps:

Step A1: Quantify the normal mode and failure mode of the complex system;

Step A2: construct a dynamic multi-layer coupling network of normal mode;

Step A3: Build a dynamic multi-layer coupling network of failure modes;

Wherein, the "normal mode" described in step A1 specifically means: after the complex system executes a certain fault triggering test case, the complex system is in the form of normal operation;

The specific meaning of the "failure mode" described in step A1 is: after the complex system executes a certain fault excitation test case, the complex system is in the form of fault operation;

Among them, the specific method of "quantifying the normal mode and failure mode of a complex system" described in step A1 is as follows: quantify the performance indicators of the complex system state monitoring tool corresponding to the normal mode and the failure mode, that is, to standardize the corresponding performance indicators of different failure phenomena. The monitoring indicators of the system, as well as the monitoring indicators when the system is running normally; take a typical integrated software and hardware system such as an embedded system as an example, after the operating system performs a certain operation, the response delay exceeds 10s, and the system is considered to have a blocking fault;

Among them, the specific method of "building a dynamic multi-layer coupling network in normal mode" described in step A2 is as follows: according to the characteristics of the complex system, select an appropriate dynamic operation data sampling tool, execute a fault excitation test case, and collect the data online. Based on the dynamic operation data of the system during normal operation under the fault excitation test case, then based on the complex network theory, the node interaction relationship within the operation data is extracted, and a dynamic multi-layer coupling network that can map the operation mechanism of the normal mode of the complex system is constructed offline;

Among them, the specific method of "building a dynamic multi-layer coupling network of fault mode" described in step A3 is as follows: according to the characteristics of the complex system, select an appropriate dynamic operation data sampling tool, execute a fault excitation test case, and collect the data online. Based on the dynamic operation data of the system during the fault operation under the fault excitation test case, then based on the complex network theory, the node interaction relationship within the operation data is extracted, and a dynamic multi-layer coupling network that can map the operation mechanism of the normal mode of the complex system is constructed offline;

Among them, the "dynamic multi-layer coupling network" described in step A2 and step A3, its specific meaning is: a set of static multi-layer coupling networks in time order, each static multi-layer coupling network corresponds to a Time slicing of dynamic operation data, suppose there are m slices of dynamic operation data after segmentation, then the dynamic multi-layer coupling network should contain m groups of static multi-layer coupling networks; due to the calling relationship between nodes, especially the calling relationship between nodes in the coupling layer There is directionality, so the connected edges of the dynamic coupling network are directed edges, and the dynamic coupling network is a directed network.

Wherein, the specific meaning of "locating the fault period of a complex system" described in step B is: by comparing the statistical characteristics of the dynamic multi-layer coupling network under the normal mode constructed in step A and the fault mode, find the period of fault occurrence, and respectively Extracting the dynamic multi-layer coupling network slice of the fault mode and the normal mode under the fault period, providing support for the subsequent step C to locate the fault node of the complex system; including the following steps:

Step B1: Calibrate the dynamic multilayer coupling network in different modes;

Step B2: Calculate the statistical characteristics of the network under different modes;

Step B3: Quantify the fault index;

Step B4: Extract the dynamic multi-layer coupling network slice of the fault occurrence period;

Among them, the "calibration of dynamic multi-layer coupling networks in different modes" described in step B1 is as follows: take the execution of a fault excitation test case as the criterion, and set the start execution time of a fault excitation test case as At time 0, the dynamic multi-layer coupling network time in different modes is calibrated, and the dynamic multi-layer coupling network data before and after the execution of a fault excitation test case is eliminated, and only the dynamic multi-layer coupling related to the execution of the fault excitation test case is retained. network;

Among them, the specific meaning of "network statistical characteristics" described in step B2 is: a set of statistical indicators that can reflect the overall nature of the network under different time intervals of the dynamic multi-layer coupled network. Common indicators are: network scale, average degree, average distance, etc. The network scale is the total number of active nodes in the network under a certain time interval _Tk , which can measure the work intensity of the complex system at this moment; the average degree is the average degree of all nodes in the network under a certain time interval _Tk , and the average degree can be An index to measure the overall connectivity of the complex system at this time interval; the average distance is the average distance between any two nodes in the network at a certain time interval _Tk , and the average distance can measure the information inside the complex system at this time interval Or the transfer efficiency of matter. The statistical characteristics of the network are not limited to the three examples mentioned above, because they belong to the well-known part, they will not be described in detail;

Among them, the specific method of "calculating the statistical characteristics of the network in different modes" described in step B2 is as follows: respectively calculating the network of the static multi-layer coupling network at each time interval in the dynamic multi-layer coupling network in the normal mode and the fault mode Statistical characteristics such as scale, average degree, average distance and various combinations, and save the calculation results in time series;

当α₁≥k₁，即故障指标不小于某个设定的常数阈值时，认定在此时间间隔内发生故障，由于在故障模式下复杂系统并不是全时段都处于故障状态，且复杂系统具有故障发生后自我修复的特性，因此故障发生时段的总时长应不大于故障激发测试用例的总时长；Among them, the specific method of "quantifying the fault index" described in step B3 is as follows: quantifying the abnormality standard of statistical features, set the value of a certain statistical feature of the normal mode at the time interval T _k as _Pr , and at the same time interval The value of the statistical feature with the same failure mode is P _f , and the failure index is set

When α ₁ ≥ k ₁ , that is, the fault index is not less than a certain constant threshold, it is determined that a fault occurs within this time interval, because the complex system is not in a fault state all the time in the fault mode, and the complex system has The self-healing feature after a fault occurs, so the total duration of the fault occurrence period should not be greater than the total duration of the fault triggering test case;

Among them, the specific method of "extracting the dynamic multi-layer coupling network slice of the fault occurrence period" described in step B4 is as follows: sequentially compare the statistical feature combination of the dynamic multi-layer coupling network slice at each time interval in the normal mode and the fault mode. , calculate the fault index at each time interval, find the abnormal time segment of the statistical feature of the fault mode relative to the normal mode, and extract the dynamic multi-layer coupling network slice under the abnormal time segment of the statistical feature, which provides the detailed fault location of the complex system in the subsequent step C. data support.

Wherein, the specific meaning of "locating the complex system fault node" described in step C is: by analyzing the dynamic multi-layer coupling network slice of the fault occurrence period saved in step B4, the node where the fault occurs is found, and the complex system fault is realized. Fine positioning is convenient for system designers to improve the system structure and improve the reliability of complex systems; it includes the following steps:

Step C1: Calculate the statistical characteristics of nodes during the fault occurrence period;

Step C2: quantify the fault index;

Step C3: extract the faulty node and output the fault information;

Step C4: simulation verification;

Among them, the specific meaning of "node statistical characteristics" described in step C1 is: a set of statistical indicators that can reflect the characteristics of each node of the dynamic multi-layer coupled network during the fault occurrence period. Common indicators are: degree and Betweenness; the degree of a network node is an index that can effectively reflect the direct call interaction state between a node and other nodes in the network, which is defined as the total number of nodes that a node v _i is directly connected to other nodes in a certain time interval Tk _, For a directed network, it is divided into out-degree and in-degree; the betweenness of a network node is an index that can reflect the influence and importance of a node in the overall network, which defines all the shortest paths in the network under a certain time interval _Tk The proportion of the number of passing through a node v _i to the total number of shortest paths;

Among them, the specific method of "calculating the statistical characteristics of nodes during the fault occurrence period" described in step C1 is as follows: successively calculate the degree, betweenness and other node statistics of the fault mode and normal mode dynamic multi-layer coupled network slice during the fault occurrence period. Feature combination, and save the index distribution of each dynamic multi-layer coupled network slice;

设故障指标

当α₂≥k₂，即故障指标不小于某个设定的常数阈值时，认定在此时间间隔内该节点发生异常；Among them, the specific method of "quantifying the fault index" described in step C2 is as follows: quantifying the abnormality standard of statistical features, it is assumed that in the slice under the time interval _Tk , a node v _i is in the normal mode of a certain statistical feature The value is P _r ^v , and the value of the statistical characteristics of the same failure mode in the same time interval is the value of

Set failure indicators

When α ₂ ≥ k ₂ , that is, the fault index is not less than a certain constant threshold, it is determined that the node is abnormal within this time interval;

Among them, the specific method of "extracting the faulty node and outputting the fault information" described in step C3 is as follows: calculating the fault indicators in all the dynamic coupling network slices under the fault occurrence period of each node in turn, and extracting each dynamic coupling network slice If the node exceeds the set threshold, the extracted node set is the fault node, and the fault report such as the occurrence frequency of the fault node is counted;

Among them, the "simulation verification" described in step C4 is as follows: input the fault node extracted in step C3 into the dynamic multi-layer coupled network model of the complex system, in the order of fault frequency, based on the seepage theory for the network. The model implements a deliberate attack, observes whether the failure effect of the network model after the attack matches the actual state, and outputs the failure risk assessment according to the simulation results, which is convenient for system designers to optimize the weak nodes of the system.

Through the above steps, this method proposes a fault location method for complex systems based on dynamic multi-layer coupled networks, which solves the problems of traditional reliability analysis methods for the dynamic propagation, cascading failures, multiple faults, uncertainties and other characteristics of complex systems. Due to the limitations of the fault location task, based on the complex network theory, the fault location analysis of the complex system is realized. The complex system fault location method can quickly find faulty nodes, and has been double verified by simulation analysis, which has good practical application value.

(3) Merit innovation

The present invention has the following innovations:

1. Rapid positioning: The method for locating complex system faults based on dynamic multi-layer coupled networks adopted in the present invention first analyzes the statistical indicators that can reflect the overall nature of the network, so as to locate the fault period of the complex system, and then analyzes the dynamic fault period under the fault period. The statistical indicators of each node in the layer-coupling network slice realize the fine location of the faulty node, avoiding the disadvantage of slow location of the traditional fault location method caused by the bottleneck of computer computing power;

2. Easy to transplant: The complex system fault location method designed by the present invention is suitable for multiple types of complex systems. When applied to different types of complex systems, it is only necessary to perform step A "Building a multi-mode fault location" according to the characteristics of the complex system requiring fault location. “Dynamic multi-layer coupling network” can be adapted to quickly realize the transplantation of the fault location method proposed in this patent, without large-scale modification, the construction method is highly portable, and reduces the workload of designers;

3. The method is advanced: according to the complex system fault location method based on the dynamic multi-layer coupling network proposed by the present invention, the statistical analysis method of the complex network can solve the problem that the traditional reliability analysis method is difficult to realize the dynamic propagation and cascading failure of the complex system. The problem of locating complex faults such as multiple faults and multiple faults has certain advanced methods.

In conclusion, this complex system fault location method based on dynamic multi-layer coupled network provides a good solution for complex system fault location in engineering applications.

Description of drawings

Figure 1 is a flow chart of the method of the present invention.

Detailed ways

In order to make the technical problems and technical solutions to be solved by the present invention clearer, the following will describe in detail with reference to the accompanying drawings and specific implementation cases. It should be understood that the embodiments described herein are only used to illustrate and explain the present invention, but not to limit the present invention.

The purpose of the present invention is to solve the problem that the traditional reliability analysis method is often used in the scenario of single fault source and static analysis, and it is difficult to locate complex faults such as dynamic propagation, cascading failure, and multiple faults of complex systems. Here we propose a fault location method that can efficiently locate the complex fault mechanism of complex system faults. This method does not need to calculate and compare the characteristic information of all nodes, and can quickly locate the faulty nodes of complex systems under the premise of less computer computing power. Has good practical application value.

The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

The embodiments of the present invention illustrate the method of the present invention by taking the fault location of a certain type of embedded development platform as an example. Specifically, there is an unknown driver fault in the underlying software layer of the embedded development platform. When a high-intensity computing load is applied to the embedded development platform and the system hardware layer falls into a computing power bottleneck, the underlying software driver is abnormal. The computer system is stuck for a long time and cannot complete the set task, so it is necessary to perform fault location analysis on the embedded development platform based on the dynamic multi-layer coupling network, which is convenient for system designers to improve the reliability and other performance of the system.

In order to achieve the above object, the technical solution adopted by the method of the present invention is: a fault location method for realizing efficient location of complex fault mechanisms of complex system faults. Its process is shown in Figure 1:

Step A: Construct a dynamic multi-layer coupling network under multi-mode;

Step B: locate the fault period of the complex system;

Step C: Locate the faulty node of the complex system.

Wherein, the "dynamic multi-layer coupling network" described in step A, its specific meaning is: a dynamic network model that can abstractly reflect the internal dynamic operating state of a complex system, in the specific implementation mode, can describe a certain type of embedded development platform execution The network model of the internal operating state of the system after the fault triggers the test case. The dynamic network model is a set of static networks based on time slices. In order to describe the system state more realistically, the dynamic network model includes multi-layer nodes, same-layer edges and coupled edges. The multi-layer coupling network; the "Constructing a dynamic multi-layer coupling network under multi-mode", including the following steps:

Step A1: Quantify the normal mode and failure mode of the complex system;

Step A2: construct a dynamic multi-layer coupling network of normal mode;

Step A3: Build a dynamic multi-layer coupling network of failure modes;

Wherein, the "normal mode" described in step A1 has a specific meaning: after a certain type of embedded development platform executes the fault-stimulated test case, a certain type of embedded development platform is in the form of normal operation;

Wherein, the specific meaning of the "failure mode" described in step A1 is: after a certain type of embedded development platform executes the fault-stimulated test case, a certain type of embedded development platform is in the form of failure operation;

Among them, the specific method of "quantifying the normal mode and failure mode of a complex system" described in step A1 is as follows: quantify the performance indicators corresponding to the normal mode and the failure mode of a certain type of embedded development platform state monitoring tool, that is, the specification The monitoring indicators corresponding to different fault phenomena and the monitoring indicators when the system is running normally, the fault indicators under the test cases imposed by the embedded development platform can be set as the system's response delay to the test cases exceeds 10s, and the system is determined to have a blocking fault;

Among them, the specific method of "building a dynamic multi-layer coupling network in normal mode" described in step A2 is as follows: according to the characteristics of a certain type of embedded development platform, select an appropriate dynamic operation data sampling tool, and execute the fault excitation test case , collect the dynamic operation data of the system under the normal operation of the fault excitation test case online, and then based on the complex network theory, extract the node interaction inside the operation data, and construct offline to map the dynamic operation mechanism of a certain type of embedded development platform in the normal mode. Multilayer coupling network;

Among them, the specific method of "building a dynamic multi-layer coupling network of failure mode" described in step A3 is as follows: according to the characteristics of a certain type of embedded development platform, select an appropriate dynamic operation data sampling tool, and execute the fault excitation test case , collect the dynamic operation data of the system during the fault operation under the fault excitation test case online, and then extract the node interaction relationship within the operation data based on the complex network theory, and build the offline construction that can map the dynamic operation mechanism of a certain type of embedded development platform in the normal mode. Multilayer coupling network;

Among them, the "dynamic multi-layer coupling network" described in step A2 and step A3, its specific meaning is: a set of static multi-layer coupling networks in time order, each static multi-layer coupling network corresponds to a The time slice of dynamic operation data, there are 2210 pieces of dynamic operation data after segmentation, and the dynamic multi-layer coupling network should contain 2,210 groups of static multi-layer coupling networks; due to the call relationship between nodes, especially the call relationship between nodes in the coupling layer, there is a direction Therefore, the connected edges of the dynamic coupling network are directed edges, and the dynamic coupling network is a directed network.

Wherein, the specific meaning of "locating the fault period of a complex system" described in step B is: by comparing the statistical characteristics of the dynamic multi-layer coupling network under the normal mode constructed in step A and the fault mode, find the period of fault occurrence, and respectively Extracting the dynamic multi-layer coupling network slice of the fault mode and the normal mode under the fault period, providing support for the subsequent step C to locate the fault node of the complex system; including the following steps:

Step B1: Calibrate the dynamic multilayer coupling network in different modes;

Step B2: Calculate the statistical characteristics of the network under different modes;

Step B3: Quantify the fault index;

Step B4: Extract the dynamic multi-layer coupling network slice of the fault occurrence period;

Among them, the "calibration of dynamic multi-layer coupling networks in different modes" described in step B1 is as follows: The execution of the fault excitation test case shall prevail, and the moment when the execution of the fault excitation test case starts is set to be time 0. , calibrate the dynamic multi-layer coupling network time in different modes, and remove the dynamic multi-layer coupling network data before and after the execution of the fault excitation test case, and only retain the dynamic multi-layer coupling network related to the execution of the fault excitation test case. After processing, only 1950 of the original 2210 slices remained;

Among them, the specific meaning of "network statistical characteristics" described in step B2 is: a set of statistical indicators that can reflect the overall nature of the network under different time intervals of the dynamic multi-layer coupled network. Common indicators are: network scale, average degree, average distance, etc. For the specific implementation, it is necessary to calculate the network scale, average degree and average distance of the dynamic multi-layer coupling network of a certain type of embedded development platform. Three network statistical characteristics;

Among them, the specific method of "calculating the statistical characteristics of the network in different modes" described in step B2 is as follows: respectively calculating the network of the static multi-layer coupling network at each time interval in the dynamic multi-layer coupling network in the normal mode and the fault mode The combination of statistical features such as scale, average degree, and average distance, and the calculation results are saved in time series;

当α₁≥k₁，即故障指标不小于某个设定的常数阈值时，认定在此时间间隔内发生故障，对于具体实施方式，将常数设为k₁＝0.3；Among them, the specific method of "quantifying the fault index" described in step B3 is as follows: quantifying the abnormality standard of statistical features, set the value of a certain statistical feature of the normal mode at the time interval T _k as _Pr , and at the same time interval The value of the statistical feature of the failure mode is P _f , and the failure index is set

When α ₁ ≥ k ₁ , that is, the failure index is not less than a certain constant threshold, it is determined that a failure occurs within this time interval, and for a specific implementation, the constant is set as k ₁ =0.3;

Among them, the specific method of "extracting the dynamic multi-layer coupling network slice of the fault occurrence period" described in step B4 is as follows: sequentially compare the statistical feature combination of the dynamic multi-layer coupling network slice at each time interval in the normal mode and the fault mode. , calculate the fault index at each time interval, find the abnormal time segment of the statistical feature of the failure mode relative to the normal mode, and extract the dynamic multi-layer coupling network slice under the abnormal time segment of the statistical feature. For two groups of different modes, each mode has a total of 1950 After the calibrated dynamic multilayer coupling network is extracted in step B4, it is found that there are 42 slices in the fault mode that meet the extraction criteria, so the 42 slices of the dynamic multilayer coupling network in the fault mode are extracted, as well as the 42 slices corresponding to their time. Dynamic multi-layer coupled network slicing in normal mode, the subsequent step C provides data support for fine-grained fault location of complex systems.

The specific meaning of "locating the complex system fault node" described in step C is: by analyzing the dynamic multi-layer coupling network slice of the fault occurrence period saved in step B4, find the node where the fault occurs, and realize a certain type of embedded The fine location of faults on the development platform is convenient for system designers to improve the system structure and improve the reliability of complex systems; it includes the following steps:

Step C1: Calculate the statistical characteristics of nodes during the fault occurrence period;

Step C2: quantify the fault index;

Step C3: extract the faulty node and output the fault information;

Step C4: simulation verification;

Among them, the specific meaning of "node statistical characteristics" described in step C1 is: a set of statistical indicators that can reflect the characteristics of each node of the dynamic multi-layer coupled network during the fault occurrence period. Common indicators are: degree and Betweenness, for the specific implementation, it is necessary to calculate the two statistical characteristics of the degree and betweenness of the dynamic multi-layer coupled network slice of a certain type of embedded development platform;

Among them, the specific method of "calculating the statistical characteristics of nodes during the fault occurrence period" described in step C1 is as follows: successively calculate the degree and betweenness node statistical characteristics of the fault mode and normal mode dynamic multi-layer coupled network slices during the fault occurrence period Combine and save the index distribution of each dynamic multi-layer coupled network slice;

设故障指标

当α₂≥k₂，即故障指标不小于某个设定的常数阈值时，认定在此时间间隔内该节点发生异常，对于具体实施方式，将常数设为k₂＝0.25；Among them, the specific method of "quantifying the fault index" described in step C2 is as follows: quantifying the abnormality standard of statistical features, it is assumed that in the slice under the time interval _Tk , a node v _i is in the normal mode of a certain statistical feature The value is P _r ^v , and the value of the statistical characteristics of the same failure mode in the same time interval is the value of

Set failure indicators

When α ₂ ≥k ₂ , that is, the fault index is not less than a certain constant threshold, it is determined that the node is abnormal within this time interval. For the specific implementation, the constant is set to k ₂ =0.25;

Among them, the specific method of "extracting the faulty node and outputting the fault information" described in step C3 is as follows: calculating the fault indicators in all the dynamic coupling network slices under the fault occurrence period of each node in turn, and extracting each dynamic coupling network slice If the node exceeds the set threshold, the extracted node set is the fault node, and the fault report such as the occurrence frequency of the fault node is counted;

Among them, the "simulation verification" described in step C4 is as follows: input the fault node extracted in step C3 into the dynamic multi-layer coupled network model of the complex system, in the order of fault frequency, based on the seepage theory for the network. The model implements a deliberate attack, observes the failure effect of the network model after being attacked and fits the actual state, and outputs a failure risk assessment based on the simulation results.

The parts of the present invention that are not described in detail belong to the well-known technology in the art.

The above description is only a part of the specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person familiar with the art within the technical scope disclosed by the present invention can easily think of changes or substitutions. Included within the scope of protection of the present invention.

Claims (1)

1. A complex system fault positioning method based on a dynamic multilayer coupling network is characterized in that: the method comprises the following steps:

step A: constructing a dynamic multi-layer coupling network under multiple modes;

and B: positioning a fault time period of a complex system;

and C: positioning a fault node of the complex system;

the method for constructing the dynamic multi-layer coupling network under the multi-mode in the step A comprises the following steps:

step A1: quantifying a normal mode and a fault mode of the complex system;

step A2: constructing a dynamic multi-layer coupling network in a normal mode;

step A3: constructing a dynamic multi-layer coupling network of a failure mode;

the "normal mode" described in step a1 has the specific meaning: after the complex system executes a fault excitation test case, the complex system is in an expression form of normal operation;

the "failure mode" described in step a1 has the specific meaning: after the complex system executes a fault excitation test case, the complex system is in a fault operation expression form;

the "quantifying the normal mode and the failure mode of the complex system" described in step a1 is implemented as follows: quantifying performance indexes of a complex system state monitoring tool corresponding to a normal mode and a fault mode, namely standardizing monitoring indexes corresponding to different fault phenomena and monitoring indexes when a system operates normally;

the "building a normal-mode dynamic complex-layer coupling network" described in step a2 is implemented as follows: selecting a proper dynamic operation data sampling tool according to the characteristics of the complex system, executing a fault excitation test case, collecting system dynamic operation data in normal operation under the fault excitation test case on line, extracting node interaction relation in the operation data based on a complex network theory, and constructing a dynamic complex layer coupling network capable of mapping the normal mode operation mechanism of the complex system off line;

the "build failure mode dynamic complex-layer coupling network" described in step a3 is implemented as follows: selecting a proper dynamic operation data sampling tool according to the characteristics of the complex system, executing a fault excitation test case, collecting system dynamic operation data in fault operation under the fault excitation test case on line, extracting node interaction relation in the operation data based on a complex network theory, and constructing a dynamic complex layer coupling network capable of mapping a complex system fault mode operation mechanism off line;

the dynamic multilayer coupling network described in step a2 and step A3 has the following specific meanings: a set of static multi-layer coupling networks in time sequence, wherein each static multi-layer coupling network corresponds to a time slice of dynamic operation data, and the total number of the dynamic operation data after being divided is m, so that the dynamic multi-layer coupling networks comprise m sets of static multi-layer coupling networks; because the calling relation among the nodes of the coupling layer has directionality, the connecting edge of the dynamic coupling network is a directed connecting edge, and the dynamic coupling network is a directed network;

the step B of positioning the fault time interval of the complex system has the specific meanings as follows: b, finding the time period of the occurrence of the fault by comparing the statistical characteristics of the dynamic multiple-layer coupling network in the normal mode and the fault mode constructed in the step A, and respectively extracting the dynamic multiple-layer coupling network slices in the fault mode and the normal mode in the fault time period to provide support for positioning the fault node of the complex system in the subsequent step C; the method comprises the following steps:

step B1: calibrating the dynamic multi-layer coupling network under different modes;

step B2: calculating network statistical characteristics under different modes;

step B3: quantifying a fault index;

step B4: extracting dynamic multilayer coupling network slices in a fault occurrence period;

the "calibrating the dynamic multi-layer coupling network in different modes" described in step B1 is implemented as follows: setting the starting execution time of a fault excitation test case as 0 time based on the executed fault excitation test case, calibrating the time of the dynamic multi-layer coupling networks in different modes, removing the data of the dynamic multi-layer coupling networks before and after the execution of the fault excitation test case, and only reserving the dynamic multi-layer coupling networks related to the execution of the fault excitation test case;

the "network statistical characteristics" described in step B2 have the following specific meanings: a group of statistical indexes capable of reflecting the overall network properties of the dynamic multi-layer coupling network at different time intervals includes the following common indexes: network size, mean and mean distance; the network size is in a time interval T_kLower theThe total number of active nodes in the network can measure the working strength of the complex system at the moment; the average being the average time interval T_kThe degrees of all nodes in the network are obtained, and the average degree can measure the index of the overall connectivity degree in the complex system at the time interval; the average distance is the average time interval T_kThe average distance of the distance between any two nodes in the network can measure the transmission efficiency of information and substances in the complex system at the time interval;

the step B2 of "calculating network statistical characteristics under different modes" includes the following steps: respectively calculating the network scale, the average degree, the average distance statistical characteristics and various combinations of the static multi-layer coupling network at each time interval in the dynamic multi-layer coupling network in the normal mode and the fault mode, and storing the calculation results according to the time sequence;

the "quantitative fault indicator" described in step B3 is implemented as follows: quantifying the criteria for statistical characteristic anomalies by setting a time interval T_kOne statistical characteristic of the lower normal mode has a value of P_rThe value of the statistical feature with the same fault mode at the same time interval is P_fSetting a fault indicator

When α₁≥k₁When the fault index is not less than a set constant threshold value, the fault is determined to occur in the time interval, and as the complex system is not in a fault state in the fault mode in all time periods and has the characteristic of self-repairing after the fault occurs, the total time length of the fault occurrence time period is not more than the total time length of the fault excitation test case;

the "extracting the dynamic multi-layer coupling network slice of the failure occurrence period" in step B4 is implemented as follows: sequentially comparing the statistical characteristic combinations of the plurality of layers of coupling network slices of the normal mode and the fault mode at each time interval, calculating fault indexes at each time interval, searching a time segment of the fault mode corresponding to the abnormal statistical characteristic of the normal mode, and extracting the plurality of layers of coupling network slices of the fault mode at the time segment of the abnormal statistical characteristic, thereby providing data support for the fault fine positioning of the complex system in the subsequent step C;

the step C of positioning the fault node of the complex system has the specific meanings as follows: the method for finding the node with the fault by analyzing the fault occurrence period dynamic multi-layer coupling network slice stored in the step B4 comprises the following steps:

step C1: calculating statistical characteristics of nodes in a fault occurrence period;

step C2: quantifying a fault index;

step C3: extracting fault nodes and outputting fault information;

step C4: simulation verification;

the "node statistical characteristics" described in step C1 mean specifically: a group of statistical indexes capable of reflecting the characteristics of each node of a dynamic multi-layer coupling network in a fault occurrence period include: degree and betweenness; the degree of a network node is an index which can effectively reflect the direct calling interactive state of a node and other nodes in the network, and is defined as a time interval T_kNext node v_iThe total number of nodes directly connected with other nodes is divided into out degree and in degree for the directed network; the betweenness of network nodes is an index which can reflect the influence and importance degree of a node in the whole network and defines a certain time interval T_kPassing through a certain node v in all shortest paths in lower network_iThe number of the shortest paths accounts for the proportion of the total number of the shortest paths;

the "calculating the statistical characteristics of the nodes in the failure occurrence period" described in step C1 is specifically performed as follows: sequentially calculating the degree and betweenness node statistical characteristic combinations of the dynamic multiple-layer coupling network slices in the fault mode and the normal mode in the fault occurrence period, and storing the index distribution of each dynamic multiple-layer coupling network slice;

the "quantitative fault indicator" described in step C2 is implemented as follows: quantifying the criteria for statistical characteristic anomalies by setting a time interval T_kIn the slice below, a node v_iOne statistical characteristic in the normal mode has a value ofP_r ^vThe statistical characteristics of the same failure mode at the same time interval have a value of

Set fault index

When α₂≥k₂When the fault index is not less than a set constant threshold value, the node is determined to be abnormal in the time interval;

the "extracting a failure node and outputting failure information" described in step C3 is specifically performed as follows: sequentially calculating fault indexes in all dynamic coupling network slices in the fault occurrence period of each node, extracting nodes exceeding a set threshold value in each dynamic coupling network slice, wherein the extracted node set is a fault node, and counting fault reports of the occurrence frequency of the fault node;

the "simulation verification" described in step C4 is implemented as follows: inputting the fault node extracted in the step C3 into the complex system dynamic complex layer coupling network model, performing deliberate attack on the network model based on the seepage theory by taking the fault frequency as an order, observing whether the fault effect of the network model after the attack is fitted with the actual state, and outputting fault risk assessment according to the simulation result.

Images