Fault root cause positioning method and device and readable storage medium

Abstract

This application provides a method, apparatus, and readable storage medium for locating the root cause of a fault. The method includes: determining node topology information based on a network topology graph; obtaining fault weight information, wherein the fault weight information represents the probability distribution of fault root causes in the network topology graph; and determining the probability of each fault root cause in the network topology graph based on the node topology information and the fault weight information, wherein the node with the highest fault root cause probability is the predicted fault root cause.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of communication, in particular to a fault root cause positioning method, device and readable storage medium. BACKGROUND

[0002] Industry field network is a general term for network access technologies of industry field end nodes. They connect various terminals, machines, sensors and systems at the end of the industry field, and meet the diverse business needs of the industry field for sensing, data, positioning, control, management, etc. The main research objects of industry field network include passive communication (such as Radio Frequency Identification (RFID)), short-range communication (such as star flash) and Time Sensitive Networking (TSN). As industry field network penetrates into scenarios such as buildings, hospitals, supermarkets and industrial parks, the problems of heterogeneous network management difficulty and low network operation efficiency are increasingly prominent, and enterprise users have stronger demand for unified and convenient network management.

[0003] For example, Figure 1 As shown in the RFID of the industry field network scenario, the complete node topology includes electronic tags, reader antennas, readers and host computers. When there is a fault, dozens of alarms will occur at the same time. How to accurately and quickly locate the fault cause from the numerous alarms is a problem to be solved. SUMMARY

[0004] Embodiments of the present application provide a fault root cause positioning method, device and readable storage medium, which solve the problem of how to accurately and quickly locate the fault cause.

[0005] In a first aspect, a fault root cause positioning method is provided, comprising:

[0006] determining node topology information according to a network topology graph;

[0007] obtaining fault weight information, the fault weight information indicating the probability distribution of fault root causes in the network topology graph;

[0008] determining fault root cause probabilities of each node in the network topology graph according to the node topology information and the fault weight information, wherein the node with the highest fault root cause probability is the predicted fault root cause.

[0009] Optionally, the step of determining node topology information according to a network topology graph comprises:

[0010] Based on the path information in the network topology graph, determine the topology matrix and the integral filter for each node in the network topology graph;

[0011] A global correlation matrix is ​​generated based on the fault type and / or the fault-related path. The elements in the global correlation matrix are used to represent the fault correlation relationships of each node in the network topology graph for a specific fault.

[0012] The node topology matrix is ​​obtained by multiplying the topology matrix and the global correlation matrix.

[0013] Optionally, the step of obtaining fault weight information includes:

[0014] Determine whether the local fault database of the node user or the fault root cause data model of the node manufacturer has been updated;

[0015] If the local fault database of the node user or the fault root cause data model of the node manufacturer is updated, the updated local fault database or the updated fault root cause data model is used to update the fault weight information; otherwise, the existing fault weight information is used.

[0016] Optionally, the method further includes:

[0017] The local fault database is reported to the node manufacturer. The local fault database is generated by the node user based on a digital twin.

[0018] The updated root cause data model is obtained from the node manufacturer. The root cause data model is obtained by the node manufacturer based on the local fault database reported by multiple node users.

[0019] Optionally, the step of determining the root cause probability of each node in the network topology graph based on the node topology information and the fault weight information includes:

[0020] pass

[0021]

[0022] The objective function X' is calculated, where n represents the number of fault alarms, and n is greater than 1;

[0023] Where A is the topological matrix, D error H is the global correlation matrix corresponding to a fault alarm. error This is the probability distribution matrix of the root causes of a fault corresponding to a fault alarm.

[0024] After normalizing the objective function X', the target root cause prediction matrix X is obtained;

[0025] by prop i =∑∑f i * X, the failure root cause probability of the computing node i;

[0026] where f i represents the integral filter of node i, and * is Hadamard product.

[0027] Optionally, the method further comprises:

[0028] updating the failure root cause probability distribution matrix according to local samples, the local samples comprising: a failure alarm and a root cause of the failure alarm.

[0029] Optionally, the step of updating the failure root cause probability distribution matrix according to the local samples comprises:

[0030] generating a target failure root cause probability distribution matrix within a specified time range according to samples within a preset range;

[0031] updating the failure root cause probability distribution matrix according to a preset weight parameter and the target failure root cause probability distribution matrix, to obtain an updated failure root cause probability distribution matrix;

[0032] obtaining a prediction probability after normalization processing according to a product of the updated failure root cause probability distribution matrix and a node topology matrix;

[0033] constructing a loss function according to the prediction probability;

[0034] training the failure root cause probability distribution matrix according to the local samples and the loss function, and optimizing the weight parameter.

[0035] In a second aspect, a failure root cause positioning device is provided, comprising:

[0036] a node topology module configured to determine node topology information according to a network topology graph;

[0037] a failure weight module configured to obtain failure weight information, the failure weight information indicating a failure root cause probability distribution in the network topology graph;

[0038] a root cause positioning module configured to determine a failure root cause probability of each node in the network topology graph according to the node topology information and the failure weight information, wherein a node with the highest failure root cause probability is a predicted failure root cause.

[0039] Optionally, the node topology module is further configured to:

[0040] determine a topology matrix and an integral filter of each node in the network topology graph according to path information in the network topology graph;

[0041] generating a global correlation matrix according to the fault type and / or the path related to the fault, an element in the global correlation matrix being used to represent a fault correlation relationship of each node in the network topology graph for a specific fault;

[0042] obtaining a node topology matrix according to a product of the topology matrix and the global correlation matrix.

[0043] Optionally, the fault weight module is further used for:

[0044] judging whether a local fault database of a node user or a fault root cause data model of a node manufacturer is updated;

[0045] if the local fault database of the node user or the fault root cause data model of the node manufacturer is updated, updating the fault weight information using the updated local fault database or the updated fault root cause data model; otherwise, using the existing fault weight information.

[0046] Optionally, the apparatus further comprises:

[0047] an updating module, configured to report a local fault database to a node manufacturer, the local fault database being generated by a node user based on digital twinning;

[0048] obtaining an updated fault root cause data model from the node manufacturer, the fault root cause data model being obtained by the node manufacturer based on local fault databases reported by a plurality of node users.

[0049] Optionally, the root cause positioning module is further used for:

[0050] by

[0051]

[0052] obtaining a target function X’ by calculation, n representing a number of fault alarms, n being greater than 1;

[0053] wherein, A is a topology matrix, D error is a global correlation matrix corresponding to a fault alarm, H error is a fault root cause probability distribution matrix corresponding to a fault alarm;

[0054] after normalizing the target function X’, obtaining a target root cause prediction matrix X;

[0055] calculating a fault root cause probability of a node i by prop i =∑∑f i *X,

[0056] wherein, fi This represents the integral filter for node i, and * represents the Hadamard product.

[0057] Thirdly, an electronic device is provided, comprising: a processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, performs the steps of the method described in the first aspect.

[0058] Fourthly, a readable storage medium is provided, on which a program is stored, which, when executed by a processor, implements the steps of the method described in the first aspect.

[0059] In this embodiment of the application, the root cause probability of each node in the network topology graph is determined based on node topology information and fault weight information. The node with the highest root cause probability is the predicted root cause of the fault, which reduces the dependence on specific mechanism models and enables accurate and rapid location of fault causes among numerous alarms. Attached Figure Description

[0060] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0061] Figure 1 This is a schematic diagram of RFID services in the field network;

[0062] Figure 2 This is a flowchart of a fault root cause localization method provided in an embodiment of this application;

[0063] Figure 3 This is a schematic diagram of a fault root cause location device provided in an embodiment of this application;

[0064] Figure 4 This is a schematic diagram of the execution flow of each module in the fault root cause localization device provided in the embodiments of this application;

[0065] Figure 5 This is a topological diagram of the nodes;

[0066] Figure 6 This is a topology diagram of nodes in an RFID scenario;

[0067] Figure 7 This is a schematic diagram of the electronic device provided in the embodiments of this application. Detailed Implementation

[0068] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0069] The term "comprising," and any variations thereof, used in the specification and claims of this application, is intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or apparatus. Furthermore, the use of "and / or" in the specification and claims indicates at least one of the connected objects, such as A and / or B, indicating the inclusion of A alone, B alone, or both A and B.

[0070] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0071] It is worth noting that the technologies described in this application are not limited to Long Term Evolution (LTE) / LTE-Advanced (LTE-A) systems, but can also be used in other wireless communication systems, such as Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Frequency Division Multiple Access (FDMA), Orthogonal Frequency Division Multiple Access (OFDMA), Single-carrier Frequency-Division Multiple Access (SC-FDMA), and other systems. The terms "system" and "network" in this application are often used interchangeably, and the described technologies can be used with the systems and radio technologies mentioned above, as well as with other systems and radio technologies. However, the following description describes New Radio (NR) systems for illustrative purposes, and the term NR is used in most of the following description, although these technologies can also be applied to applications other than NR systems, such as 6th Generation (6G) communication systems.

[0072] Current root cause localization solutions for field network node failures in the industry fall into the following four categories:

[0073] 1. Node detection: Based on the alarm notification, check each node that may be related to the alarm one by one, and eliminate the faulty node one by one.

[0074] 2. Mechanism Modeling: Based on node mechanisms, simulate faults and their characteristics, establish a fault feature library, and then compare it with actual faults to determine the source of the fault.

[0075] 3. Data Analysis: Based on historical data, statistical analysis methods are used to infer the cause of the failure.

[0076] 4. Digital Twin: Fully utilize data such as physical models, sensor updates, and operational history to integrate simulation processes involving multiple disciplines, physical quantities, scales, and probabilities, and complete mapping in virtual space to reflect the entire life cycle of the corresponding physical equipment.

[0077] When a fault occurs in an industry field network, the existing platform often generates dozens of related alarms or errors. The efficiency of troubleshooting depends on how quickly the root cause of the fault is identified from the numerous alarms or errors.

[0078] The drawbacks of existing root cause localization techniques:

[0079] 1. Disadvantages of node detection technology: For industry field networks, a single node failure may trigger dozens of abnormal alarms for related business indicators. Testing the relevant nodes for each alarm will consume a lot of time and effort, resulting in low efficiency and a poor user experience.

[0080] 2. Disadvantages of mechanistic modeling technology: It requires a very high level of expertise in the specific field, typically mastered by only a few high-quality node providers. This expertise is their competitive advantage, and they are unwilling to share it. Therefore, it restricts node procurement, forcing users to purchase nodes from a limited number of vendors at high prices. Furthermore, industry field networks involve multiple areas (passive, short-range, TSN), and the quality of suppliers varies across these areas, resulting in different operation and maintenance systems for different field network nodes, increasing system management costs.

[0081] 3. Disadvantages of data analysis technology: The sample size of a single enterprise is limited, and insufficient samples may lead to statistical bias; sufficient data accumulation is required to conduct data analysis, so root cause localization is difficult to achieve in real time.

[0082] 4. Disadvantages of existing digital twin technology: Currently, digital twins are rarely used across all lifecycle stages of a node; most are single-stage digital twins. Different lifecycle stages of a node often belong to different vendors, making information flow difficult. Digital twins can help and promote the flow of information across lifecycle stages through methods such as access control.

[0083] See Figure 2 This application provides a method for locating the root cause of a fault, which includes steps 201, 202 and 203.

[0084] Step 201: Determine the node topology information based on the network topology diagram;

[0085] Taking the topology graph constructed by industry field network nodes as an example, a topology subgraph is constructed based on the node topology graph to obtain multiple topology subgraphs that can complete basic business. The combination of topology subgraphs can express the topology subgraphs corresponding to complex business, thereby realizing the abstract expression of industry field network nodes and constructing the correspondence between business and topology subgraphs (nodes).

[0086] Step 202: Obtain fault weight information, which represents the probability distribution of fault root causes in the network topology graph;

[0087] Fault weight information includes: for a given type of fault, the probability of each node in the network topology graph being the root cause of the fault, for example, see [link to relevant documentation]. Figure 5For fault A, the root cause probability of node 1 is 30%, node 2 is 20%, node 5 is 20%, node 3 is 60%, node 6 is 30%, and node 4 is 0%. The root cause probabilities of each node can be determined based on historical data. It is understandable that, in the absence of historical data, the root cause probability of each node can be an average value. Figure 5 For example, the root cause probability of faults in nodes 1 to 6 can be set to 20%.

[0088] Fault weight information can be determined based on the probability of root causes of various types of business faults on each relevant node, provided by the node user, or the initial function can be an average distribution function. As the industry field network is maintained, the fault weight information is adjusted based on accumulated data to achieve fault weight information more suitable for the local environment and improve the accuracy of root cause localization.

[0089] Step 203: Based on the node topology information and the fault weight information, determine the root cause probability of each node in the network topology graph, wherein the node with the highest root cause probability is the predicted root cause of the fault.

[0090] In one embodiment of this application, the step of determining the node topology matrix based on the network topology graph includes:

[0091] Based on the path information in the network topology graph, determine the topology matrix and the integral filter for each node in the network topology graph;

[0092] Generate a global correlation matrix based on the fault type and the fault-related paths;

[0093] The node topology matrix is ​​obtained by multiplying the topology matrix and the global correlation matrix.

[0094] In one embodiment of this application, the step of obtaining the fault weight function includes:

[0095] Determine whether the local fault database of the node user or the fault root cause data model of the node manufacturer has been updated;

[0096] If the node user's local fault database or the node manufacturer's fault root cause data model is updated, the updated local fault database or the updated fault root cause data model is used to update the fault weight function; otherwise, the existing fault weight function is used.

[0097] Based on a continuous learning mechanism, the method periodically trains fault weight information using a constantly updated training dataset. This method has self-learning and adaptive capabilities, achieving "the more you use it, the smarter it becomes".

[0098] In one embodiment of this application, the method further includes:

[0099] The node user reports the local fault database generated based on the digital twin to the node manufacturer;

[0100] The updated root cause data model is obtained from the node manufacturer. The root cause data model is obtained by the node manufacturer based on the local fault database reported by multiple node users.

[0101] In one embodiment of this application, the step of determining the probability that each node in the network topology graph is a root cause of a fault based on the node topology matrix and the fault weight function includes:

[0102] The objective function X' for fault alarms 1 to n is calculated using the following formula;

[0103]

[0104] Where A is the topological matrix, D error H is the global correlation matrix corresponding to a fault alarm. error This is the probability distribution matrix of the root causes of a fault corresponding to a fault alarm.

[0105] After normalizing the objective function X', the target root cause prediction matrix X is obtained;

[0106] The root cause probability of node i is calculated using the following formula;

[0107] prop i =∑∑f i *X

[0108] Among them, f i This represents the integral filter for node i, and * represents the Hadamard product.

[0109] The dot product of the fault root cause probability distribution matrix and the topology matrix is ​​a probability matrix. After normalization, it represents the probability that each node in the industry field network is the root cause of a business fault. By superimposing the probability matrices of multiple business faults within the same time period, the true root cause of the fault will be traced back to multiple business attributions. Other nodes may be involved in a certain business fault attribution, but are unlikely to be traced back to most business faults, thus making the true root cause stand out.

[0110] Optionally, the method further includes:

[0111] The fault root cause probability distribution matrix is ​​updated based on local samples, whereby the local samples include fault alarms and the root causes of the fault alarms.

[0112] Optionally, the step of updating the root cause probability distribution matrix based on local samples includes:

[0113] Generate a target fault root cause probability distribution matrix within a specified time range based on samples within a preset range;

[0114] The fault root cause probability distribution matrix is ​​updated according to the preset weight parameters and the target fault root cause probability distribution matrix to obtain the updated fault root cause probability distribution matrix.

[0115] The predicted probability is obtained by normalizing the product of the updated root cause probability distribution matrix and the node topology matrix.

[0116] Based on the predicted probabilities, a loss function is constructed;

[0117] The root cause probability distribution matrix of the fault is trained based on local samples and the loss function, and the weight parameters are optimized.

[0118] In this embodiment of the application, the root cause probability of each node in the network topology graph is determined based on node topology information and fault weight information. The node with the highest root cause probability is the predicted root cause of the fault, which reduces the dependence on specific mechanism models and enables accurate and rapid location of fault causes among numerous alarms.

[0119] See Figure 3 This application provides a fault root cause localization device, which includes a node topology module 301, a fault weight module 302, and a root cause localization module 303.

[0120] The node topology module 301 is used to determine node topology information based on the network topology diagram.

[0121] The fault weight module 302 is used to obtain fault weight information, which represents the probability distribution of fault root causes in the network topology graph.

[0122] The root cause localization module 303 is used to determine the root cause probability of each node in the network topology graph based on the node topology information and the fault weight information, wherein the node with the highest root cause probability is the predicted root cause of the fault.

[0123] In one embodiment of this application, the node topology module is further configured to:

[0124] Based on the path information in the network topology graph, determine the topology matrix and the integral filter for each node in the network topology graph;

[0125] A global correlation matrix is ​​generated based on the fault type and / or the fault-related path. The elements in the global correlation matrix are used to represent the fault correlation relationships of each node in the network topology graph for a specific fault.

[0126] The node topology matrix is ​​obtained by multiplying the topology matrix and the global correlation matrix.

[0127] In one embodiment of this application, the fault weighting module is further configured to:

[0128] Determine whether the local fault database of the node user or the fault root cause data model of the node manufacturer has been updated;

[0129] If the local fault database of the node user or the fault root cause data model of the node manufacturer is updated, the updated local fault database or the updated fault root cause data model is used to update the fault weight information; otherwise, the existing fault weight information is used.

[0130] In one embodiment of this application, the apparatus further includes:

[0131] The update module is used to report the local fault database to the node manufacturer. The local fault database is generated by the node user based on a digital twin.

[0132] The updated root cause data model is obtained from the node manufacturer. The root cause data model is obtained by the node manufacturer based on the local fault database reported by multiple node users.

[0133] In one embodiment of this application, the root cause localization module is further configured to:

[0134] pass

[0135]

[0136] The objective function X' is calculated, where n represents the number of fault alarms, and n is greater than 1;

[0137] Where A is the topological matrix, D error H is the global correlation matrix corresponding to a fault alarm. error This is the probability distribution matrix of the root causes of a fault corresponding to a fault alarm.

[0138] Normalize the objective function X' to obtain the target root cause prediction matrix X;

[0139] via prop i =∑∑f i *X, calculates the root cause probability of failure at node i;

[0140] Among them, f i This represents the integral filter for node i, and * represents the Hadamard product.

[0141] Optionally, the device further includes:

[0142] The update module is used to update the fault root cause probability distribution matrix based on local samples, wherein the local samples include fault alarms and the root causes of the fault alarms.

[0143] Optionally, the update module is further configured to:

[0144] Generate a target fault root cause probability distribution matrix within a specified time range based on samples within a preset range;

[0145] The fault root cause probability distribution matrix is ​​updated according to the preset weight parameters and the target fault root cause probability distribution matrix to obtain the updated fault root cause probability distribution matrix.

[0146] The predicted probability is obtained by normalizing the product of the updated root cause probability distribution matrix and the node topology matrix.

[0147] Based on the predicted probabilities, a loss function is constructed;

[0148] The root cause probability distribution matrix of the fault is trained based on local samples and the loss function, and the weight parameters are optimized.

[0149] Furthermore, the node topology module 301 is responsible for constructing and monitoring the topology relationships between each node. It collects the relationships between nodes through the transmission paths of tasks on each node, thereby constructing a global node topology matrix. The node topology module 301 is also responsible for monitoring changes in nodes. If any node is added or removed, the node topology module will update the node topology matrix in a timely manner.

[0150] Furthermore, the fault weighting module 302 is responsible for assigning root cause weights to alarms from various nodes. The fault weighting module 302 can obtain root cause weight assignments from the local fault databases of both the equipment manufacturer and the equipment user. Based on digital twins, equipment fault data flows between the equipment manufacturer and the equipment user, creating additional value.

[0151] Furthermore, the root cause localization module 303 is responsible for retrieving relevant data from the node topology module 301 and the fault weight module 302 to calculate and infer faults when the system receives a large number of device alarms in a short period of time.

[0152] The apparatus provided in this application embodiment can achieve... Figure 2 The various processes implemented in the method embodiments shown achieve the same technical effects, and will not be described again here to avoid repetition.

[0153] See Figure 4 The execution flow of the node topology module is as follows:

[0154] Step a: The node topology module can monitor whether the node connection topology has changed.

[0155] If not, continue using the original node topology matrix.

[0156] If so, update the node topology matrix. The update method can be found in the section on node topology matrix construction described below.

[0157] Step b: Push the updated node topology matrix to the root cause localization module.

[0158] See also Figure 4 The execution flow of the fault weight module is as follows:

[0159] Step a: The fault weight module monitors the device and checks if there are any updates to the manufacturer's local fault database.

[0160] If so, the fault weight module retraining function will be activated to update the fault weight function;

[0161] Step b: The fault weighting module monitors the fault root cause data model of the equipment manufacturer and checks if there are any updates.

[0162] If so, the fault weight module retraining function will be activated to update the fault weight function;

[0163] Step c: Push the updated fault weight function to the root cause localization module upon request.

[0164] Before step a, there is also step d: based on digital twins, the local fault database of the equipment user is periodically reported to the equipment manufacturer after being de-identified; the equipment manufacturer updates the fault root cause data model in a timely manner based on data from multiple equipment users and pushes the new fault root cause data model to the equipment users.

[0165] See also Figure 4 The root cause localization module executes the following process:

[0166] Step a: Determine whether the system's fault alarm frequency (number of alarms per unit time) exceeds a certain threshold.

[0167] Step b: If the error exceeds the limit, start the root cause localization module, call the stored node topology matrix, and call the fault weight function based on the collected alarm information to infer the root cause of the fault.

[0168] To facilitate a better understanding of the embodiments of this application, the following technical points are introduced:

[0169] I. On the construction of the node topology matrix.

[0170] A graph search-based method is used to determine the distribution of all nodes for a given task. Based on the business node paths involved in the current task, an overall topology matrix and a global correlation matrix are constructed for calculating the node topology matrix.

[0171] The specific steps are as follows:

[0172] Step a: Label the operator nodes in the computation graph based on a unified rule. For example, first label the nodes in the branch with the most nodes in the computation graph, then label all the nodes in the branch with the second most nodes, and so on, until all operator nodes in the computation graph are labeled.

[0173] Step b: For each terminal node (nodes 1 and 6), starting from that node, search for the path to the platform node (node ​​4) and record it, until every path is recorded. For each path, perform normalization processing, padding omitted nodes with 0, so that the vector length of each path remains consistent.

[0174] by Figure 5 For example, the recorded path is: 1-2-3-4, 1-5-3-4, 6-0-3-4.

[0175] Step c: Generate the overall topology matrix A based on the path. p*q and the integral filter f of each node p*q .

[0176] by Figure 5 For example, the overall topology matrix is The integral filter f1 of node 1 is The integral filter f4 of node 4 is This integral filter is used to superimpose the attribution steps.

[0177] Step d: Based on the fault type, determine the fault-related paths to generate the global correlation matrix D. q*p .

[0178] For example, the service unresponsiveness fault 1 corresponding to node 1

[0179] In summary, the topology matrix of each node is E = D·A.

[0180] II. Learning about the root cause probability distribution matrix of faults

[0181] One of the problems this invention aims to solve is establishing a root cause analysis method that does not rely entirely on mechanistic models or historical data. Therefore, it employs the superposition method, commonly used in signal processing. The advantage of the superposition method is that even with very few mechanistic models and no historical data, the averaging method can be used to set fault weights, and good results can still be achieved.

[0182] For a single, specific fault type, its attribution can be simulated based on the mechanism or historical data, thereby improving the accuracy of the attribution. Two related designs are provided for this purpose:

[0183] (1) The initial root cause probability distribution matrix can be periodically updated by the equipment manufacturer through digital twin technology. Some manufacturers also possess state-of-the-art mechanistic models, thus providing a valuable root cause probability distribution matrix. The equipment manufacturer provides the root cause probability distribution matrix H based on the provided data format. 1*q , as the initial value.

[0184] When the manufacturer is unable or refuses to provide the root cause probability distribution matrix, the initial function H 1*q It can be defined as an average function.

[0185] (2) The probability distribution matrix of the root cause of the fault is learned and updated based on local small samples.

[0186] The root cause probability distribution matrix provided by the equipment manufacturer is derived from various operating environments and operators, and therefore does not consider the actual local environment and operator skill level. Local samples contain this information and can therefore be used to update the root cause probability distribution matrix. The specific steps are as follows:

[0187] Step a: Generate the recent failure root cause probability distribution matrix J based on a small sample. 1*q ;

[0188] Step b: Set weight parameters The initial value is 0.5. The updated root cause probability distribution matrix is ​​generated as follows:

[0189]

[0190] Step c: Generate the predicted probability T′=H·E, and then perform L1 standardization (normalization) to calculate the standardized predicted probability:

[0191]

[0192] Step d: Construct the loss function based on the predicted probabilities:

[0193]

[0194] Where t i This is the prediction result, t′ i It is the ground truth; correctly labeled data is called ground truth.

[0195] Step e: Based on local samples (each fault alarm and its root cause constitutes one sample) and the loss function, train the fault root cause probability distribution matrix and optimize the weight parameters.

[0196] III. On the attribution of superposition

[0197] Based on the node topology matrix and the fault root cause probability distribution matrix, the superposition method is used to predict the root cause of the fault time and achieve fast and efficient root cause prediction.

[0198]

[0199] Where A is the overall topology matrix (constructed from the node topology matrix), and D... error H is the global correlation matrix corresponding to the fault (constructed from the node topology matrix). error The root cause probability distribution matrix is ​​derived from the node fault root cause probability distribution matrix learning module. After L1 normalization of X', the target root cause prediction matrix X is obtained. p*q Then, the integral filter f for each node is used. p*q (Based on the node topology matrix construction), calculate the root cause probability of node i:

[0200] prop i =∑∑f i *X p*q

[0201] Note: * indicates Hadama product.

[0202] Among them, the node with the highest probability is the predicted root cause of the failure.

[0203] See Figure 6 The application scenarios of this application embodiment are illustrated using RFID as an example.

[0204] In RFID scenarios within industry field networks, the node paths are generally similar, consisting of (D-exciter, C-tag environmental interference, B-receiver, A-gateway). For example... Figure 6 As shown, the RFID scenario it demonstrates includes 5 node paths, namely A1-B1-C1-D1, A2-B2-C2-D2, A3-B3-C3-D3, A4-B4-C4-D4, and A5-B5-C5-D5.

[0205] For counting services of a single tag, only one topology subgraph is needed; for tag positioning services, which typically require 3 / 4 point positioning, 3-4 topology subgraphs are required.

[0206] The objective function consists of the node topology matrix and the root cause probability distribution matrix.

[0207]

[0208] Overall topology matrix because Figure 6 The described RFID-based positioning scenario has 5 node paths, and the maximum number of nodes in a single node path is 4.

[0209] Global correlation matrix D error It is a 5x4 matrix, and its location depends on the alarm fault. Locating an RFID tag requires three links (assuming xy-dimensional positioning). The positioning accuracy of RFID tag t1 corresponds to these three links (A1-B1-C1-D1, A2-B2-C2-D2, A4-B4-C4-D4), and the positioning accuracy of RFID tag t2 corresponds to these three links (A1-B1-C1-D1, A3-B3-C3-D3, A5-B5-C5-D5). The topology function differs depending on the error type.

[0210] Table 1 shows an example of the topology function for a fault-corresponding industry field network node.

[0211]

[0212]

[0213] Fault Root Cause Probability Distribution Matrix H error It is a 1x4 matrix, representing the weight function of the industry field network nodes corresponding to the fault, where 4 is the maximum number of nodes on a single node path. Still using... Figure 6 Taking the RFID-based positioning scenario as an example, there are 4 nodes on each path.

[0214] The initial values ​​of the functions shown in Table 2 should be provided by the hardware manufacturer and optimized based on the data during runtime. If the relevant data from the hardware manufacturer is unavailable, the following should be used:

[0215]

[0216] When one of these nodes fails, there are typically dozens of related fault alarms, which may include, but are not limited to: fault alarms related to the node itself, fault alarms from associated nodes, and fault alarms related to Quality of Experience (QoE). Each fault alarm is calculated using the following formula:

[0217]

[0218] L1 normalization is performed on X' to obtain the target root cause prediction matrix X. 4*5 Then, the integral filter f for each node is used. 4*5 Calculate the root cause probability of each node:

[0219] prop = ∑∑f 4*5 *X 4*5

[0220] It should be noted that * represents Hadama product.

[0221] For example, three fault alarms are received in a short period of time: fault alarm 1 is that the positioning deviation of Tag-t1 is too large, fault alarm 2 is that the positioning deviation of Tag-t2 is too large, and fault alarm 3 is that there is no state data for D1.

[0222] For the fault alarm error=1,

[0223]

[0224] For fault alarm error=2,

[0225]

[0226] For fault alarm error=3,

[0227]

[0228]

[0229] After normalizing X', we get

[0230] The root cause probability of a fault at each node is calculated sequentially using its own integral filter. Taking node D1 as an example, its integrator filter is... The calculation process for the root cause probability of the failure is as follows:

[0231]

[0232] The root cause probabilities of all nodes are sorted, and the node with the highest root cause probability is D1. Therefore, the root cause of this series of faults is determined to be the D1 exciter.

[0233] like Figure 7As shown, this application embodiment also provides a communication device 700, including a processor 701, a memory 702, and a program or instructions stored in the memory 702 and executable on the processor 701. When the program or instructions are executed by the processor 701, they implement the above-mentioned... Figure 2 The various processes in the method embodiments can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0234] This application embodiment also provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the above-described functionality. Figure 2 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0235] The processor is the processor in the first or second communication device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as a computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0236] The steps of the methods or algorithms described in this application can be implemented in hardware or by executing software instructions on a processor. The software instructions can consist of corresponding software modules, which can be stored in RAM, flash memory, ROM, EPROM, EEPROM, registers, hard disk, portable hard disk, read-only optical disk, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor, enabling the processor to read information from and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and storage medium can be housed in an ASIC. Alternatively, the ASIC can be housed in a core network interface device. Of course, the processor and storage medium can also exist as discrete components in the core network interface device.

[0237] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this application can be implemented using hardware, software, firmware, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transfer of a computer program from one place to another. Storage media can be any available medium accessible to a general-purpose or special-purpose computer.

[0238] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above description is only a specific embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of this application should be included within the scope of protection of this application.

[0239] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, embodiments of this application can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of this application can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0240] This application describes embodiments of methods, apparatus (systems), and computer program products according to embodiments of this application with reference to flowchart illustrations and / or block diagrams. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0241] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0242] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0243] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A method of fault root cause localization, the method comprising: The method comprises the following steps: determining node topology information according to a network topology graph; obtaining fault weight information, wherein the fault weight information represents a fault root cause probability distribution in the network topology graph; determining a fault root cause probability of each node in the network topology graph according to the node topology information and the fault weight information, wherein a node with the highest fault root cause probability is a predicted fault root cause; the step of determining the fault root cause probability of each node in the network topology graph according to the node topology information and the fault weight information comprises: obtaining a target function X' by calculating the target function X', wherein n is greater than 1; wherein A is a topology matrix, D error is a global correlation matrix corresponding to a fault alarm, H error is a fault root cause probability distribution matrix corresponding to a fault alarm, an element in the fault root cause probability distribution matrix representing a probability of each node being a fault root cause, and norm() represents a norm function. obtaining a target root cause prediction matrix X after normalizing the target function X'; By prop i =∑∑f i X, the failure root cause probability of a computing node i, which is any node in the network topology graph; where f i denotes the integral filter of node i, * is the Hadamard product.

2. The method of claim 1, wherein, the step of determining the node topology information according to the network topology graph comprises: determining a topology matrix and an integral filter of each node in the network topology graph according to path information in the network topology graph; generating a global correlation matrix according to fault types and / or paths related to faults, wherein elements in the global correlation matrix are used to represent fault correlation relationships of each node in the network topology graph for a specific fault; obtaining a node topology matrix according to a product of the topology matrix and the global correlation matrix.

3. The method of claim 1, wherein, the step of obtaining the fault weight information comprises: judging whether a local fault database of a node user or a fault root cause data model of a node manufacturer is updated; if the fault root cause data model of the node manufacturer is updated, updating the fault weight information using the updated fault root cause data model; otherwise, using the existing fault weight information.

4. The method of claim 1, wherein, The method further comprises: reporting the local fault database to the node manufacturer, wherein the local fault database is generated by the node user based on digital twinning; obtaining the updated fault root cause data model from the node manufacturer, wherein the fault root cause data model is obtained by the node manufacturer based on the local fault databases reported by multiple node users.

5. The method of claim 1, wherein, The method further comprises: updating the fault root cause probability distribution matrix according to a local sample, wherein the local sample comprises a fault alarm and a root cause of the fault alarm.

6. The method of claim 5, wherein, The step of updating the fault root cause probability distribution matrix according to the local sample comprises: generating a target fault root cause probability distribution matrix within a specified time range according to a sample in a preset range; updating the fault root cause probability distribution matrix according to a preset weight parameter and the target fault root cause probability distribution matrix to obtain an updated fault root cause probability distribution matrix; obtaining a prediction probability after normalizing a product of the updated fault root cause probability distribution matrix and a node topology matrix; constructing a loss function according to the prediction probability; training the fault root cause probability distribution matrix according to the local sample and the loss function to optimize the weight parameter.

7. A fault root cause localization apparatus characterized by, The method comprises the following steps: a node topology module is configured to determine node topology information according to a network topology graph; a fault weight module is configured to obtain fault weight information, wherein the fault weight information represents a fault root cause probability distribution in the network topology graph; The root cause positioning module is configured to determine a fault root cause probability of each node in the network topology graph according to the node topology information and the fault weight information, wherein a node with the highest fault root cause probability is a predicted fault root cause. The root cause positioning module is further configured to: obtain a target function X' by calculating the target function X', wherein n represents a number of fault alarms, and n is greater than 1; Wherein, A is a topology matrix, D error is a global correlation matrix corresponding to a fault alarm, H error is a fault root cause probability distribution matrix corresponding to a fault alarm, and norm() represents a norm function. perform normalization processing on the target function X' to obtain a target root cause prediction matrix X; By prop i =∑∑f i X, the failure root cause probability of a computing node i, which is any node in the network topology graph; where f i denotes the integral filter of node i, * is the Hadamard product.

8. The apparatus of claim 7, wherein, The node topology module is further configured to: determine a topology matrix and an integral filter of each node in the network topology graph according to path information in the network topology graph; generate a global correlation matrix according to a fault type and / or a path related to the fault, wherein an element in the global correlation matrix is used to represent a fault correlation relationship of each node in the network topology graph for a specific fault; obtain a node topology matrix according to a product of the topology matrix and the global correlation matrix.

9. The apparatus of claim 7, wherein, The apparatus further includes: an updating module configured to report a local fault database to a node manufacturer, wherein the local fault database is generated by a node user based on digital twinning; obtain an updated fault root cause data model from the node manufacturer, wherein the fault root cause data model is obtained by the node manufacturer based on the local fault databases reported by a plurality of node users.

10. The apparatus of claim 7, wherein, The fault weight module is further configured to: determine whether the local fault database of the node user or the fault root cause data model of the node manufacturer is updated; if the fault root cause data model of the node manufacturer is updated, update the fault weight information using the updated fault root cause data model; otherwise, use the existing fault weight information.

11. An electronic device, comprising: include: a processor, a memory, and a program stored on the memory and executable on the processor, wherein the program is executed by the processor to implement the steps of the method according to any one of claims 1 to 6.

12. A readable storage medium, characterized by, The readable storage medium stores a program, and the program is executed by the processor to implement the steps of the method according to any one of claims 1 to 6.

Images