Device operation and maintenance method, device operation and maintenance device, device operation and maintenance equipment and storage medium

Abstract

The application relates to the technical field of computer operation maintenance, and specifically discloses a device operation and maintenance method, a device operation and maintenance device, a device operation and maintenance equipment and a storage medium. A triple structure including an alarm state value, a conversion condition and a normal state value is adopted as training data, the training data is extracted from device resource information and device alarm configuration information, a training data set is constructed, a text information label processing framework based on a TransE model is trained, and a device fault diagnosis model describing the correlation of data such as device resource information and device alarm configuration information in the form of a vector is obtained. When receiving device alarm information, based on the device alarm information and the device fault diagnosis model, a fault feedback result corresponding to the device alarm information can be automatically generated for reference by an operation and maintenance personnel, and the operation and maintenance personnel needs to go to the scene to troubleshoot the device fault.

Description

Technical Field

[0001] This application relates to the field of computer operation and maintenance technology, and in particular to a device operation and maintenance method, device operation and maintenance apparatus, device operation and maintenance equipment, and storage medium. Background Technology

[0002] Current computer equipment maintenance requires monitoring equipment resource information and equipment alarm information. Taking servers as an example, the equipment resource information to be collected mainly includes server name, server serial number, server CPU utilization, server port traffic data, server memory utilization, and server hard disk specifications. As for equipment alarm information, all information originates from the device itself and is defined by the manufacturer. After the device generates an alarm, it is pushed to the maintenance platform via Trap (Linux command) or Hypertext Transfer Protocol (HTTP). The maintenance platform parses the alarm message and displays it in text form, and also supports sending equipment alarm information to maintenance personnel.

[0003] After receiving equipment alarm information, maintenance personnel need to troubleshoot the cause of the equipment failure and perform maintenance based on the equipment resource information and alarm information. However, since equipment alarm information can only provide the symptoms of the equipment alarm, such as a parameter exceeding the standard, maintenance personnel often cannot remotely determine the cause of the equipment failure based solely on equipment resource information and alarm information. They often need to go to the site to troubleshoot the equipment failure in order to repair it. Summary of the Invention

[0004] The purpose of this application is to provide an equipment operation and maintenance method, equipment operation and maintenance device, equipment operation and maintenance equipment and storage medium, which are used to expand the equipment fault diagnosis capability, provide fault feedback results for equipment alarm information for operation and maintenance personnel to refer to, and reduce the need for operation and maintenance personnel to go to the site to diagnose equipment faults.

[0005] To address the aforementioned technical problems, this application provides a method for equipment operation and maintenance, comprising:

[0006] A training dataset is constructed based on equipment resource information and equipment alarm configuration information; the training data of the training dataset is a triplet structure including alarm status value, transition condition, and normal status value.

[0007] Using the training dataset, a text information labeling framework based on the TransE model is trained to obtain a device fault diagnosis model.

[0008] When a device alarm message is received, a fault feedback result corresponding to the device alarm message is obtained based on the device alarm message and the device fault diagnosis model.

[0009] The fault feedback result and the equipment alarm information are output together.

[0010] Optionally, the text information labeling framework specifically includes the TransE model, the multilayer perceptron model, and the confidence learning module;

[0011] The step of training the text information labeling framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model specifically includes:

[0012] The triple containing the newly added instance entity, the ontology concept of each newly added instance entity in the triple, and the set of neighboring instance entities of each newly added instance entity are input into the TransE model.

[0013] For each newly added instance entity in the triple, a representation of the ontology concept is generated through a two-layer attention mechanism;

[0014] The template representation of the newly added instance entity is generated based on the representations of all the ontology concepts and the triples.

[0015] The final representation vector of the newly added instance entity is generated by combining the template representation and the set of neighboring instance entities;

[0016] The validity of the triple is evaluated based on the final representation vector;

[0017] If the legality requirement is met, the parameters of the text information label processing framework are updated using the final representation vector;

[0018] The ontology concept refers to the alarm status value or the normal status value, and the neighbor instance entity set includes the association information of the alarm status value and / or the association information of the normal status value.

[0019] Optionally, the alarm status value is used as the head entity in the triplet, and the normal status value is used as the tail entity in the triplet.

[0020] Optionally, the step of using the training dataset to train the text information labeling framework built on the TransE model to obtain the equipment fault diagnosis model specifically includes:

[0021] The alarm status value is converted into a head entity vector, the conversion condition is converted into a relation vector, and the normal status value is converted into a tail entity vector.

[0022] Substitute the head entity vector, the relation vector, and the tail entity vector into the distance function to calculate the vector distance between the sum of the head entity vector and the relation vector and the tail entity vector;

[0023] The triplet consisting of the head entity vector, the relation vector, and the tail entity vector corresponding to the smallest vector spacing is selected as the correct triplet for the current iteration number.

[0024] The parameters of the text information label processing framework are updated using the correct triples until the loss value obtained by substituting the correct triples calculated in two adjacent iterations into the loss function is less than the preset loss value, thus obtaining the equipment fault diagnosis model.

[0025] Optionally, the distance function is specifically expressed by the following equation:

[0026] f r (h,t)=||h+rt|| L1 / L2 ;

[0027] Among them, f r (h,t) is the vector spacing, h is the head entity vector, r is the relation vector, t is the tail entity vector, L1 is the Manhattan distance, and L2 is the Euclidean distance.

[0028] Optionally, the loss function is specifically expressed by the following equation:

[0029]

[0030] Where L is the loss function value, h is the head entity vector, r is the relation vector, t is the tail entity vector, and ∑ (h,r,t) The outputs of the three vectors h, r, and t The output is the unit vector in the directions of the three vectors h, r, and t. Let h be the unit vector in the direction of the vector. Let r be the unit vector in the direction of vector r. f is the unit vector in the direction of vector t. r (h,t) is the vector spacing. The unit vector spacing.

[0031] Optionally, the fault feedback results include fault diagnosis results and / or fault repair solutions.

[0032] Optionally, obtaining the fault feedback result corresponding to the device alarm information based on the device alarm information and the device fault diagnosis model specifically includes:

[0033] Using the device alarm information as the alarm status value, query the candidate fault feedback results from the device information database established based on the device resource information and the device alarm configuration information;

[0034] If the candidate fault feedback result cannot be found, then the fault feedback result shall be that no fault cause was found.

[0035] If only one candidate fault feedback result is obtained from the query, then the candidate fault feedback result shall be taken as the fault feedback result.

[0036] If multiple candidate fault feedback results are obtained from the query, each candidate fault feedback result is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate fault feedback result; one or more candidate fault feedback results with the highest confidence ranking result are taken as the fault feedback result.

[0037] Optionally, obtaining the fault feedback result corresponding to the device alarm information based on the device alarm information and the device fault diagnosis model specifically includes:

[0038] Using the device alarm information as the alarm status value, query the candidate intermediate fault causes in the device information database established based on the device resource information and the device alarm configuration information;

[0039] If the candidate intermediate fault cause cannot be found, the fault feedback result shall be that no fault cause was found.

[0040] If only one candidate intermediate fault cause is found, then the candidate intermediate fault cause shall be used as the intermediate fault cause.

[0041] If multiple candidate intermediate fault causes are obtained from the query, each candidate intermediate fault cause is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate intermediate fault cause; one or more candidate intermediate fault causes with the highest confidence ranking result are taken as the intermediate fault causes.

[0042] Using the intermediate fault cause as the alarm status value, return to the step of querying the candidate intermediate fault cause in the device information database established based on the device resource information and the device alarm configuration information, until the candidate intermediate fault cause cannot be found, and use the final alarm status value as the fault feedback result.

[0043] To address the aforementioned technical problems, this application also provides an equipment operation and maintenance device, comprising:

[0044] The construction unit is used to construct a training dataset based on device resource information and device alarm configuration information; the training data of the training dataset is a triplet structure including alarm status value, transition condition, and normal status value;

[0045] The training unit is used to train the text information label processing framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model.

[0046] The calculation unit is used to obtain a fault feedback result corresponding to the device alarm information based on the device alarm information and the device fault diagnosis model when the device alarm information is received.

[0047] The output unit is used to correlate and output the fault feedback result with the device alarm information.

[0048] To address the aforementioned technical problems, this application also provides equipment operation and maintenance equipment, comprising:

[0049] Memory, used to store computer programs;

[0050] A processor for executing the computer program, which, when executed by the processor, implements the steps of the device operation and maintenance method as described in any of the above.

[0051] To address the aforementioned technical problems, this application also provides a storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the device operation and maintenance method described in any of the above claims.

[0052] The equipment operation and maintenance method provided in this application uses a triplet structure including alarm status values, transition conditions, and normal status values ​​as training data. Training data is extracted from equipment resource information and equipment alarm configuration information to construct a training dataset. This training dataset is then used to train a text information labeling processing framework based on the TransE model, resulting in an equipment fault diagnosis model that describes the correlation between equipment resource information, equipment alarm configuration information, and other data in vector form. When equipment alarm information is received, based on the alarm information and the equipment fault diagnosis model, a fault feedback result corresponding to the alarm information can be automatically generated for operation and maintenance personnel to refer to, rather than simply providing the alarm information. Therefore, the equipment operation and maintenance method provided in this application expands the equipment fault diagnosis capabilities of the operation and maintenance platform and reduces the need for operation and maintenance personnel to conduct on-site equipment fault diagnosis.

[0053] This application also provides an equipment operation and maintenance device, equipment operation and maintenance equipment, and storage medium, which have the above-mentioned beneficial effects, and will not be described in detail here. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1 A flowchart illustrating a device operation and maintenance method provided in this application embodiment;

[0056] Figure 2 This is a schematic diagram of the structure of an equipment operation and maintenance device provided in an embodiment of this application;

[0057] Figure 3 This is a schematic diagram of the structure of an equipment maintenance device provided in an embodiment of this application. Detailed Implementation

[0058] The core of this application is to provide an equipment operation and maintenance method, equipment operation and maintenance device, equipment operation and maintenance equipment and storage medium, which are used to expand the equipment fault diagnosis capability, provide fault feedback results for equipment alarm information for operation and maintenance personnel to refer to, and reduce the need for operation and maintenance personnel to go to the site to diagnose equipment faults.

[0059] 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, and 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.

[0060] Example 1

[0061] Figure 1 This is a flowchart illustrating a device operation and maintenance method provided in an embodiment of this application.

[0062] like Figure 1 As shown, the equipment operation and maintenance method provided in this application embodiment includes:

[0063] S101: Construct a training dataset based on device resource information and device alarm configuration information.

[0064] The training dataset consists of a triplet structure that includes alarm state values, transition conditions, and normal state values.

[0065] S102: Using the training dataset, train the text information labeling framework built on the TransE model to obtain the equipment fault diagnosis model.

[0066] S103: When a device alarm message is received, the fault feedback result corresponding to the device alarm message is obtained based on the device alarm message and the device fault diagnosis model.

[0067] S104: Associate the output of fault feedback results and equipment alarm information.

[0068] Existing equipment alarm configurations can only provide fault symptoms, indicating that a fault has occurred, specifically an abnormal indicator. Even with fault codes and corresponding maintenance plans, only fault types exceeding underlying equipment parameters can be detected and addressed, but more complex equipment fault diagnosis results cannot be provided. Especially in the operation and maintenance of large data centers, whenever equipment issues an alarm, maintenance personnel often need to go to the site to troubleshoot and perform maintenance.

[0069] To address this issue, the equipment operation and maintenance method provided in this application utilizes the advantages of the TransE algorithm model for data identification to reorganize the resource monitoring, alarm monitoring, and information feedback processes.

[0070] The equipment operation and maintenance method provided in this application embodiment can be implemented on the software based on an equipment monitoring platform or an operation and maintenance platform, and on the hardware based on an equipment monitoring server, to provide operation and maintenance personnel and users with further troubleshooting results based on equipment alarm information.

[0071] Specifically, for S101, device resource information and device alarm configuration information can be created as a knowledge graph.

[0072] Device resource information may include server name, server serial number, server CPU utilization, server port traffic data, server memory utilization, and server hard drive specifications. Device alarm configuration information mainly includes the alarm mechanisms configured by the device manufacturer corresponding to fault signals, and may also include alarm configuration information customized by the user after the device leaves the factory.

[0073] In the library and information science field, knowledge graphs are referred to as knowledge domain visualization or knowledge domain mapping maps. They are a series of various graphics that display the development process and structural relationships of knowledge. They use visualization techniques to describe knowledge resources and their carriers, and to mine, analyze, construct, draw, and display knowledge and the interrelationships between them. In this embodiment, a knowledge graph for equipment monitoring is constructed from two aspects: equipment resource information and equipment alarm configuration information. Different display formats are provided to maintenance personnel based on the different types of information.

[0074] Equipment asset information can be broken down into static asset information and dynamic performance information. Static asset information refers to data that remains largely unchanged, such as device name and serial number. Dynamic performance information refers to data that changes in real time, such as port traffic and the number of bytes received. Dynamic performance information can be processed into trend charts by the platform to facilitate a more subjective capture of device performance change trends. Static asset information can be further broken down into text information and video information. Video information refers to real-time monitoring recordings of the device from surveillance cameras. These recordings primarily display the device's video and audio information, allowing maintenance personnel to intuitively view and hear information such as fan noise and appearance, providing a more comprehensive understanding of the device's status and facilitating remote maintenance.

[0075] When new devices are added to the monitoring objects or during the monitoring of existing monitoring objects, the process of reporting device asset information is continuously executed to update the knowledge graph.

[0076] In this embodiment of the application, the equipment asset information reporting process may include:

[0077] Regularly access and obtain device resources to get device information or receive device information actively pushed by the device;

[0078] If the equipment information is equipment resource information, the text information is stored in the database in the standardized format of text asset information, and the non-text information is saved in the standardized format of video information.

[0079] If the device information is not device resource information, then the received device information is determined to be device performance information. Based on time, a trend chart of the device performance information is generated and displayed on the monitoring page of the operation and maintenance personnel.

[0080] Maintenance personnel can view the monitoring page to obtain device performance information or access the device information database to obtain device resource information.

[0081] Similarly, the corresponding device alarm configuration information will also be pushed to the device operation and maintenance platform so that operation and maintenance personnel can remotely view the device alarm status.

[0082] When creating a knowledge graph from device resource information and device alarm configuration information, it is necessary not only to establish a retrieval mechanism but also to mine the relationships between data, performing connection processing and classification on multiple data sets. TransE is a method that uses vector form to describe knowledge graphs, which is more concise than commonly used ontology languages ​​such as OWL and RDF. By translating and expressing the knowledge graph through TransE, not only can device resource information and device alarm configuration information be concisely represented, but the relationships between the data can also be shown.

[0083] Therefore, to address the problem of expanding the equipment fault diagnosis capability of the equipment operation and maintenance platform to be solved in this application embodiment, the TransE algorithm model is trained to learn the correlation between the data involved in the equipment resource information and equipment alarm configuration information. Based on the requirements of the TransE algorithm model and the equipment fault diagnosis problem, training data is constructed using a triplet structure including alarm status values, transition conditions, and normal status values ​​related to equipment fault diagnosis, forming a training dataset. For example, port disconnection is a fault manifestation, i.e., an alarm status value, while port connectivity is the corresponding normal state, i.e., a normal status value. The reasons for the transition from port connectivity to port disconnection may include port removal, remote control power failure, server power failure, the impact of other port disconnections, insufficient bandwidth, etc., all of which can serve as transition relationships from port connectivity to port disconnection. Conversely, port insertion can also serve as a transition relationship from port disconnection to port connectivity.

[0084] In addition to providing training datasets, device resource information and device alarm configuration information can also be provided to maintenance personnel for data querying. Therefore, device resource information and device alarm configuration information can be created as a device information database, and SQL Server 2005 can be used to store massive amounts of data.

[0085] For S102, the text information labeling processing framework built on the TransE model is trained using the training dataset mentioned above. That is, the knowledge graph of the device is translated and expressed by the TransE model to obtain the device fault diagnosis model.

[0086] For S103, when a device alarm message is received, a fault feedback result corresponding to the device alarm message is obtained based on the device alarm message and the device fault diagnosis model. The fault feedback result is an expansion of the device alarm message; it can be a fault diagnosis result, a fault repair solution, or both. For example, for the transition from a port disconnected state to a port connected state, the port being removed can be considered a fault diagnosis result, while the port being inserted can be considered a fault repair solution.

[0087] When a device alarm message is received, the reference fault feedback result can be obtained by first querying the device information database based on the device alarm message. Then, the reference fault feedback result can be input into the trained device fault diagnosis model to quickly obtain the reference value of each reference fault feedback result.

[0088] Therefore, in S103, based on the equipment alarm information and the equipment fault diagnosis model, the fault feedback result corresponding to the equipment alarm information is obtained, which may specifically include:

[0089] Using equipment alarm information as the alarm status value, query the candidate fault feedback results from the equipment information database established based on equipment resource information and equipment alarm configuration information;

[0090] If no fault feedback result can be found, the fault feedback result will be "no fault cause found".

[0091] If only one candidate fault feedback result is obtained from the query, then the candidate fault feedback result shall be used as the fault feedback result;

[0092] If multiple candidate fault feedback results are obtained from the query, each candidate fault feedback result is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate fault feedback result; one or more candidate fault feedback results with the highest confidence ranking result are taken as the fault feedback result.

[0093] To increase the probability of successfully retrieving candidate fault feedback results from the equipment information database, it is preferable to run the equipment information database on a high-traffic, high-bandwidth, and high-speed monitoring device. At the same time, a wide range of indexes are established and fuzzy search methods are used to retrieve candidate fault feedback results. Then, the equipment fault diagnosis model is used for optimal matching to obtain one or more fault diagnosis results or fault repair solutions for maintenance personnel to refer to.

[0094] If the result of a keyword fuzzy search in the device information database is empty, the device alarm information will be reported directly to the maintenance personnel, or "no cause of failure found" will be reported at the same time.

[0095] If a result is found in the device information database through a keyword fuzzy search, the executable instructions corresponding to the obtained fault feedback result can be used for automatic repair.

[0096] If multiple results are found through a fuzzy keyword search in the device information database, the device fault diagnosis model trained using a text information labeling framework based on the TransE model needs to process these multiple results. Association rules are used to find the relationships between data through data mining and analysis. For example, when the device alarm message is "server fan speed exceeds threshold," all matching data for "server fan speed exceeds threshold" are analyzed, the support of the association rule is calculated, the confidence of the association rule is calculated after determining the association, and finally, the lift of the association rule is analyzed. The basis for judging the recommendation effect of association rules is the lift of the association rule. This lift depends on whether the obtained fault diagnosis results are accurate or whether the obtained fault repair methods are executable, and whether the service is restored after execution.

[0097] To further accelerate equipment failure recovery, a cache can be created specifically for the top-ranked failure repair solutions selected and executed by maintenance personnel from the failure repair solutions obtained through the equipment failure diagnosis model, as well as failure repair solutions set by maintenance personnel, thereby improving retrieval efficiency.

[0098] For S104, after receiving the fault feedback result, the fault feedback result is associated with the equipment alarm information and output to provide more reference information for operation and maintenance personnel.

[0099] Based on the equipment operation and maintenance method provided in this application embodiment, the traditional operation and maintenance mode of the original device uploading device alarm information, the device monitoring platform reporting device alarm information in text form, and operation and maintenance personnel going to the site to troubleshoot and repair equipment faults is changed. After the device uploads device alarm information to the device monitoring platform, the device monitoring platform reports monitoring information to the operation and maintenance personnel in various forms such as text and video. On the other hand, it repairs the device according to the preset automatic repair plan or the repair instructions issued by the operation and maintenance personnel.

[0100] The equipment operation and maintenance method provided in this application uses a triplet structure including alarm status values, transition conditions, and normal status values ​​as training data. Training data is extracted from equipment resource information and equipment alarm configuration information to construct a training dataset. This training dataset is then used to train a text information labeling processing framework based on the TransE model, resulting in an equipment fault diagnosis model that describes the correlation between equipment resource information, equipment alarm configuration information, and other data in vector form. When equipment alarm information is received, based on the alarm information and the equipment fault diagnosis model, a fault feedback result corresponding to the alarm information can be automatically generated for operation and maintenance personnel to refer to, rather than simply providing the alarm information. Therefore, the equipment operation and maintenance method provided in this application expands the equipment fault diagnosis capabilities of the operation and maintenance platform, reducing the need for operation and maintenance personnel to conduct on-site equipment fault diagnosis.

[0101] Example 2

[0102] A text information labeling framework built on the TransE model can take various forms, but it mainly needs to include the TransE model, a classifier, and a confidence learning module. In the equipment operation and maintenance method provided in this application embodiment, a text information labeling framework using a multilayer perceptron model as the classifier is provided.

[0103] In the equipment operation and maintenance method provided in this application embodiment, the text information tag processing framework specifically includes the TransE model, the multilayer perceptron model, and the confidence learning module.

[0104] S102: Using the training dataset, train the text information labeling framework built on the TransE model to obtain the equipment fault diagnosis model, specifically including:

[0105] Input the triple containing the newly added instance entity, the ontology concept of each newly added instance entity in the triple, and the set of neighboring instance entities of each newly added instance entity into the TransE model.

[0106] For each newly added instance entity in the triple, a representation of the ontology concept is generated through a two-layer attention mechanism;

[0107] Template representations for generating new instance entities based on the representations of all ontology concepts and triples;

[0108] The final representation vector of the newly added instance entity is generated by combining the template representation and the set of neighboring instance entities;

[0109] The validity of triples is evaluated based on the final representation vector;

[0110] If the legality requirements are met, the parameters of the text information label processing framework are updated using the final representation vector;

[0111] The ontology concept is either an alarm status value or a normal status value, and the set of neighbor instance entities includes the association information of alarm status values ​​and / or the association information of normal status values.

[0112] Specifically, in the device operation and maintenance method provided in this application embodiment, when training the text information label processing framework built on the TransE model, a triple containing a new entity instance (which can be a head entity, a tail entity, or both) is taken from the training dataset. At the same time, the ontology concept (alarm status value, normal status value, or both) of each new entity instance in the triple is given, as well as the set of neighboring instance entities (i.e., entities that have been determined to be associated with the new entity instance).

[0113] For each new entity instance in the triplet, a representation of the ontology concept is generated through a two-layer attention mechanism, which yields all transformation relationships associated with alarm or normal state values.

[0114] Then, based on the representations of all ontology concepts and the triples containing the new entity instance, a template representation of the new entity instance is generated. That is, the candidate transformation relationship of the new entity instance is obtained by summarizing the transformation relationships of each association. Then, combined with the set of neighboring entity instances (the entities that have been determined to be associated with the new entity instance), the final representation vector of the new entity instance is obtained, which is the final transformation relationship.

[0115] The validity of the triples of the newly added entity instance is evaluated based on the final representation vector of the newly added entity instance, that is, whether the alarm state value and the normal state value can be converted through the final conversion relationship. If the final conversion relationship is satisfied, the parameters of the text information tag processing framework built on the TransE model are updated using the triple containing the newly added entity instance and the final conversion relationship, that is, the knowledge graph of the device is updated.

[0116] Based on the model training method provided in this application embodiment, the triplet data in the training dataset can be learned quickly, and finally an equipment fault diagnosis model that can be used for equipment fault investigation and equipment maintenance plan push can be obtained.

[0117] The conversion between the head entity and the tail entity is mutual. In this embodiment, the head entity in the triplet can be used as the alarm state value and the tail entity in the triplet as the normal state value, so as to obtain the final representation vector of the newly added entity instance, that is, the final conversion relationship is the conversion relationship between the alarm state value and the normal state value.

[0118] Example 3

[0119] As mentioned in the above embodiment one, when multiple candidate fault feedback results are obtained from the self-device information database, it is necessary to use the equipment fault diagnosis model trained by the text information label processing framework based on the TransE model to process the multiple results. Specifically, this involves mining the association rules of equipment alarm information, determining the confidence of each transformation relationship associated with the equipment alarm information, and analyzing whether each transformation relationship can obtain the corresponding normal state value after acting on the alarm state value corresponding to the equipment alarm information.

[0120] A fundamental starting point for TransE model research is to define a distance function to measure the distance between the sum of the head entity vector h and the relation vector r and the tail entity vector t, and then further adjust the specific relationship between the head entity vector h and the tail entity vector t to make (h+r) and t as equivalent as possible.

[0121] In the equipment operation and maintenance method provided in this application embodiment, S102: using the training dataset, the text information label processing framework built based on the TransE model is trained to obtain the equipment fault diagnosis model, specifically including:

[0122] Convert alarm status values ​​into head entity vectors, conversion conditions into relation vectors, and normal status values ​​into tail entity vectors.

[0123] Substitute the head entity vector, relation vector, and tail entity vector into the distance function to calculate the vector distance between the sum of the head entity vector and the relation vector and the tail entity vector.

[0124] The triplet consisting of the head entity vector, relation vector, and tail entity vector corresponding to the smallest vector spacing is selected as the correct triplet for the current iteration number.

[0125] The parameters of the text information label processing framework are updated using the correct triples until the loss value obtained by substituting the correct triples calculated in two adjacent iterations into the loss function is less than the preset loss value, thus obtaining the equipment fault diagnosis model.

[0126] Furthermore, embodiments of this application provide a distance function, specifically expressed by the following equation:

[0127] f r (h,t)=||h+rt|| L1 / L2 ;

[0128] Among them, f r (h,t) represents the vector spacing, where h is the head entity vector, r is the relation vector, t is the tail entity vector, L1 is the Manhattan distance, and L2 is the Euclidean distance.

[0129] For correct triples that appear in the training dataset, the distance function result should be as low as possible; while for problematic triples that do not appear in the training dataset, the distance function value must be larger than that of correct triples.

[0130] TransE then uses a loss function to test and describe the effectiveness of the algorithm. In this embodiment, the loss function can be specifically expressed by the following equation:

[0131]

[0132] Where L is the loss function value, h is the head entity vector, r is the relation vector, t is the tail entity vector, and ∑ (h,r,t) The result of the loss function is the output of the three vectors h, r, and t. The output is the unit vector in the directions of the three vectors h, r, and t. Let h be the unit vector in the direction of the vector. Let r be the unit vector in the direction of vector r. f is the unit vector in the direction of vector t. r (h,t) represents the vector spacing. The unit vector spacing.

[0133] Example 4

[0134] In the first embodiment described above, a method is provided for handling situations where different numbers of candidate fault feedback results are found in the device information database. In practical applications, device alarm information is often only a manifestation of a fault, and the direct cause of this manifestation is not the root cause. For example, when the device alarm information is "port disconnected," the transition from "port disconnected" to "port connected" could be insufficient bandwidth. However, simply knowing that insufficient bandwidth is the cause of the fault cannot solve the port disconnection problem; it is necessary to explore a more fundamental cause of the fault. Therefore, insufficient bandwidth is an intermediate fault cause. It is understandable that intermediate fault causes can also be used as alarm status values ​​to further explore the transition relationship between them and normal status values ​​until actionable fault diagnosis results or fault repair methods are obtained.

[0135] Therefore, in the equipment operation and maintenance method provided in this application embodiment, S103: based on the equipment alarm information and the equipment fault diagnosis model, obtain the fault feedback result corresponding to the equipment alarm information, specifically including:

[0136] Using equipment alarm information as alarm status values, query the possible intermediate fault causes in the equipment information database established based on equipment resource information and equipment alarm configuration information;

[0137] If the cause of the intermediate failure cannot be found, the failure feedback result will be "no cause found".

[0138] If only one candidate intermediate fault cause is found, then the candidate intermediate fault cause shall be used as the intermediate fault cause.

[0139] If multiple candidate intermediate fault causes are obtained from the query, each candidate intermediate fault cause is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate intermediate fault cause; one or more candidate intermediate fault causes with the highest confidence ranking result are taken as intermediate fault causes.

[0140] Using the intermediate fault cause as the alarm status value, return to the steps of querying the candidate intermediate fault cause in the device information database established based on device resource information and device alarm configuration information, until no candidate intermediate fault cause can be found, and use the final alarm status value as the fault feedback result.

[0141] In other words, after retrieving the fault cause associated with the device alarm information from the device information database, the system continues to explore the deeper fault causes by using the fault cause as an intermediate node, until no related fault cause can be retrieved from the device information database. Then, the fault generation logic from the device alarm information and intermediate fault causes to the final fault cause can be presented to the maintenance personnel.

[0142] The above details various embodiments of the equipment operation and maintenance method. Based on this, this application also discloses equipment operation and maintenance devices, equipment operation and maintenance equipment, and storage media corresponding to the above methods.

[0143] Example 5

[0144] Figure 2 This is a schematic diagram of the structure of an equipment operation and maintenance device provided in an embodiment of this application.

[0145] like Figure 2 As shown, the equipment maintenance device provided in this application embodiment includes:

[0146] The construction unit 201 is used to construct a training dataset based on the device resource information and device alarm configuration information; the training data of the training dataset is a triplet structure including alarm status value, transition condition, and normal status value;

[0147] Training unit 202 is used to train the text information label processing framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model.

[0148] The calculation unit 203 is used to obtain the fault feedback result corresponding to the equipment alarm information based on the equipment alarm information and the equipment fault diagnosis model when the equipment alarm information is received.

[0149] Output unit 204 is used to associate the output fault feedback results with the equipment alarm information.

[0150] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0151] Example 6

[0152] Figure 3 This is a schematic diagram of the structure of an equipment maintenance device provided in an embodiment of this application.

[0153] like Figure 3 As shown, the equipment maintenance equipment provided in this application embodiment includes:

[0154] Memory 310 is used to store computer program 311;

[0155] The processor 320 is configured to execute a computer program 311, which, when executed by the processor 320, implements the steps of the device operation and maintenance method as described in any of the above embodiments.

[0156] The processor 320 may include one or more processing cores, such as a 3-core processor or an 8-core processor. The processor 320 may be implemented using at least one hardware form selected from Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 320 may also include a main processor and a coprocessor. The main processor, also known as a Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 320 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 320 may also include an Artificial Intelligence (AI) processor, which handles computational operations related to machine learning.

[0157] The memory 310 may include one or more storage media, which may be non-transitory. The memory 310 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 310 is used to store at least the following computer program 311, wherein, after being loaded and executed by the processor 320, the computer program 311 can implement the relevant steps in the device operation and maintenance method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 310 may also include an operating system 312 and data 313, and the storage method may be temporary storage or permanent storage. The operating system 312 may be Windows. The data 313 may include, but is not limited to, the data involved in the above methods.

[0158] In some embodiments, the equipment maintenance equipment may further include a display screen 330, a power supply 340, a communication interface 350, an input / output interface 360, a sensor 370, and a communication bus 380.

[0159] Those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the equipment operation and maintenance equipment, and may include more or fewer components than shown.

[0160] The device maintenance equipment provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the device maintenance method described above, with the same effect.

[0161] Example 7

[0162] It should be noted that the device and equipment embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or modules may be electrical, mechanical, or other forms. Modules described as separate components may or may not be physically separate, and components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0163] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0164] If the integrated modules are implemented as software functional modules and sold or used as independent products, they can be stored in a storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application.

[0165] Therefore, this application embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements steps such as those of a device operation and maintenance method.

[0166] The storage medium can include various media that can store program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0167] The computer program contained in the storage medium provided in this embodiment can implement the steps of the device operation and maintenance method described above when executed by the processor, with the same effect.

[0168] The foregoing has provided a detailed description of the equipment operation and maintenance method, equipment operation and maintenance device, equipment operation and maintenance equipment, and storage medium provided in this application. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the devices, equipment, and storage medium disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple, and relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0169] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for equipment operation and maintenance, characterized in that, include: A training dataset is constructed based on equipment resource information and equipment alarm configuration information; The training dataset consists of a triplet structure including alarm status values, transition conditions, and normal status values. Using the training dataset, a text information labeling framework based on the TransE model is trained to obtain a device fault diagnosis model. When a device alarm message is received, a fault feedback result corresponding to the device alarm message is obtained based on the device alarm message and the device fault diagnosis model. The fault feedback result and the equipment alarm information are correlated and output; The text information labeling framework specifically includes the TransE model, the multilayer perceptron model, and the confidence learning module. The step of training the text information labeling framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model specifically includes: The triple containing the newly added instance entity, the ontology concept of each newly added instance entity in the triple, and the set of neighboring instance entities of each newly added instance entity are input into the TransE model. For each newly added instance entity in the triple, a representation of the ontology concept is generated through a two-layer attention mechanism; The template representation of the newly added instance entity is generated based on the representations of all the ontology concepts and the triples. The final representation vector of the newly added instance entity is generated by combining the template representation and the set of neighboring instance entities; The validity of the triple is evaluated based on the final representation vector; If the legality requirement is met, the parameters of the text information label processing framework are updated using the final representation vector; The ontology concept refers to the alarm status value or the normal status value, and the neighbor instance entity set includes the association information of the alarm status value and / or the association information of the normal status value.

2. The equipment operation and maintenance method according to claim 1, characterized in that, The alarm status value is the head entity in the triplet, and the normal status value is the tail entity in the triplet.

3. The equipment operation and maintenance method according to claim 1, characterized in that, The step of training the text information labeling framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model specifically includes: The alarm status value is converted into a head entity vector, the conversion condition is converted into a relation vector, and the normal status value is converted into a tail entity vector. Substitute the head entity vector, the relation vector, and the tail entity vector into the distance function to calculate the vector distance between the sum of the head entity vector and the relation vector and the tail entity vector; The triplet consisting of the head entity vector, the relation vector, and the tail entity vector corresponding to the smallest vector spacing is selected as the correct triplet for the current iteration number. The parameters of the text information label processing framework are updated using the correct triples until the loss value obtained by substituting the correct triples calculated in two adjacent iterations into the loss function is less than the preset loss value, thus obtaining the equipment fault diagnosis model.

4. The equipment operation and maintenance method according to claim 3, characterized in that, The distance function is specifically expressed by the following equation: ； in, The vector spacing is... For the head entity vector, Let the relation vector be... For the tail entity vector, For Manhattan distance, It is a Euclidean distance.

5. The equipment operation and maintenance method according to claim 3, characterized in that, The loss function is specifically expressed by the following equation: ； in, The value of the loss function, For the head entity vector, Let the relation vector be... For the tail entity vector, for , , The outputs of the three vectors, for , , The output of the unit vectors in the three vector directions. for The unit vector in the direction of the vector. for The unit vector in the direction of the vector. for The unit vector in the direction of the vector. The vector spacing is... The unit vector spacing.

6. The equipment operation and maintenance method according to claim 1, characterized in that, The fault feedback results include fault diagnosis results and / or fault repair solutions.

7. The equipment operation and maintenance method according to claim 1, characterized in that, The step of obtaining fault feedback results corresponding to the equipment alarm information based on the equipment alarm information and the equipment fault diagnosis model specifically includes: Using the device alarm information as the alarm status value, query the candidate fault feedback results from the device information database established based on the device resource information and the device alarm configuration information; If the candidate fault feedback result cannot be found, then the fault feedback result shall be that no fault cause was found. If only one candidate fault feedback result is obtained from the query, then the candidate fault feedback result shall be taken as the fault feedback result. If multiple candidate fault feedback results are obtained from the query, each candidate fault feedback result is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate fault feedback result; one or more candidate fault feedback results with the highest confidence ranking result are taken as the fault feedback result.

8. The equipment operation and maintenance method according to claim 1, characterized in that, The step of obtaining fault feedback results corresponding to the equipment alarm information based on the equipment alarm information and the equipment fault diagnosis model specifically includes: Using the device alarm information as the alarm status value, query the candidate intermediate fault causes in the device information database established based on the device resource information and the device alarm configuration information; If the candidate intermediate fault cause cannot be found, the fault feedback result shall be that no fault cause was found. If only one candidate intermediate fault cause is found, then the candidate intermediate fault cause shall be used as the intermediate fault cause. If multiple candidate intermediate fault causes are obtained from the query, each candidate intermediate fault cause is input into the equipment fault diagnosis model to obtain the confidence ranking result of each candidate intermediate fault cause; one or more candidate intermediate fault causes with the highest confidence ranking result are taken as the intermediate fault causes. Using the intermediate fault cause as the alarm status value, return to the step of querying the candidate intermediate fault cause in the device information database established based on the device resource information and the device alarm configuration information, until the candidate intermediate fault cause cannot be found, and use the final alarm status value as the fault feedback result.

9. An equipment operation and maintenance device, characterized in that, include: The construction unit is used to build a training dataset based on device resource information and device alarm configuration information; The training dataset consists of a triplet structure including alarm status values, transition conditions, and normal status values. The training unit is used to train the text information label processing framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model. The calculation unit is used to obtain a fault feedback result corresponding to the device alarm information based on the device alarm information and the device fault diagnosis model when receiving device alarm information; The output unit is used to correlate and output the fault feedback result with the equipment alarm information; The text information labeling framework specifically includes the TransE model, the multilayer perceptron model, and the confidence learning module. The step of training the text information labeling framework based on the TransE model using the training dataset to obtain the equipment fault diagnosis model specifically includes: The triple containing the newly added instance entity, the ontology concept of each newly added instance entity in the triple, and the set of neighboring instance entities of each newly added instance entity are input into the TransE model. For each newly added instance entity in the triple, a representation of the ontology concept is generated through a two-layer attention mechanism; The template representation of the newly added instance entity is generated based on the representations of all the ontology concepts and the triples. The final representation vector of the newly added instance entity is generated by combining the template representation and the set of neighboring instance entities; The validity of the triple is evaluated based on the final representation vector; If the legality requirement is met, the parameters of the text information label processing framework are updated using the final representation vector; The ontology concept refers to the alarm status value or the normal status value, and the neighbor instance entity set includes the association information of the alarm status value and / or the association information of the normal status value.

10. An equipment operation and maintenance device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program, which, when executed by the processor, implements the steps of the device operation and maintenance method as described in any one of claims 1 to 8.

11. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the equipment operation and maintenance method as described in any one of claims 1 to 8.

Images