CN120217158A - Asset operation and maintenance decision-making management platform and management method based on data fusion - Google Patents
Asset operation and maintenance decision-making management platform and management method based on data fusion Download PDFInfo
- Publication number
- CN120217158A CN120217158A CN202510336469.0A CN202510336469A CN120217158A CN 120217158 A CN120217158 A CN 120217158A CN 202510336469 A CN202510336469 A CN 202510336469A CN 120217158 A CN120217158 A CN 120217158A
- Authority
- CN
- China
- Prior art keywords
- asset
- data
- maintenance
- risk
- knowledge
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
- G06N5/025—Extracting rules from data
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本申请涉及运维管理技术领域，具体为基于数据融合的资产运维决策管理平台及管理方法，方法包括：采集资产全生命周期数据，构建标准化资产数据集；利用本体建模和知识抽取技术，构建包含资产属性、拓扑关系和运行状态的动态资产知识图谱；训练深度学习模型实现资产故障诊断，并构建资产运维风险检测算法，对资产多维度数据进行实时分析，识别资产安全风险等级，触发相应的运维策略；利用自然语言生成技术，将诊断出的资产故障和生成的运维策略转化为自然语言，推送给运维管理人员执行。本申请方法融合了多源异构数据，实现了资产全生命周期的智慧运维决策。
The present application relates to the field of operation and maintenance management technology, specifically to an asset operation and maintenance decision management platform and management method based on data fusion, the method including: collecting data of the entire life cycle of assets and constructing a standardized asset data set; using ontology modeling and knowledge extraction technology to construct a dynamic asset knowledge graph containing asset attributes, topological relationships and operating status; training a deep learning model to realize asset fault diagnosis, and construct an asset operation and maintenance risk detection algorithm to perform real-time analysis of multi-dimensional asset data, identify asset safety risk levels, and trigger corresponding operation and maintenance strategies; using natural language generation technology to convert the diagnosed asset faults and generated operation and maintenance strategies into natural language, and push them to operation and maintenance managers for execution. The present application method integrates multi-source heterogeneous data to realize intelligent operation and maintenance decision-making throughout the entire life cycle of assets.
Description
Technical Field
The application relates to the technical field of operation and maintenance management, in particular to an asset operation and maintenance decision management platform and an asset operation and maintenance decision management method based on data fusion.
Background
The traditional IT operation and maintenance mode mainly depends on manual experience and manual operation, and has the problems of low efficiency, slow response, easy error and the like. The operation and maintenance personnel need to log in different management systems, inquire asset information, monitor performance indexes, check fault alarms and the like, and then judge and decide and take operation and maintenance measures according to experience. The artificial operation and maintenance mode is difficult to cope with the increasing size and complexity of IT assets, and quick and accurate fault diagnosis and risk prediction cannot be realized.
In addition, full life cycle management of IT assets involves various links such as asset procurement, deployment, configuration, monitoring, maintenance, optimization, retirement, etc., producing massive amounts of multi-source heterogeneous data. However, these data often lack correlation and fusion, and it is difficult to form a global asset view and knowledge base. The fracturing of multi-source heterogeneous data causes the lack of data support for IT operation and maintenance decisions, and the full life cycle intelligent management of the assets is difficult to realize.
In view of the above, the application provides an asset operation and maintenance decision management platform and a management method based on data fusion.
Disclosure of Invention
In order to achieve the above purpose, the application provides an asset operation and maintenance decision management platform and a management method based on data fusion, and the specific technical scheme is as follows:
the asset operation and maintenance decision management method based on data fusion comprises the following steps:
Acquiring multi-source heterogeneous data of the full life cycle of the asset through a unified operation and maintenance platform, and performing data cleaning, conversion and fusion to form a standardized asset data set;
Constructing a dynamically updated asset knowledge graph based on the fused asset data set by utilizing an ontology modeling and knowledge extraction technology, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information;
Training an asset fault diagnosis model based on a deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether a fault exists in the current asset according to an asset operation and maintenance strategy by the asset fault diagnosis model;
Constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the asset multidimensional data, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level;
And converting the asset faults and the asset operation and maintenance strategies identified based on the knowledge graph into natural language by utilizing a natural language generation technology, pushing the natural language to an operation and maintenance manager, and executing operation and maintenance measures by the operation and maintenance manager.
Preferably, the assets comprise switches, routers, network security devices, whole network terminals, data center proxy servers, virtual machines, operating systems, databases, middleware, containers, IP addresses and domain name information;
Collecting various parameter information of an asset, including asset static attribute data, asset dynamic monitoring data and asset operation and maintenance operation data;
Carrying out data preprocessing on the acquired data, wherein the data preprocessing comprises missing value processing, noise removal, outlier detection and data standardization;
and summarizing all the acquired data, and solving the problem of the same index data conflict among a plurality of data sources by calculating JS divergence among the data as data priority order.
Preferably, modeling an asset ontology to construct an ontology model covering the whole life cycle management of the asset, wherein the ontology model comprises asset concepts, asset relationships and asset attributes, and associating asset attribute values to asset entities;
Based on the constructed IT asset ontology model, extracting structured asset knowledge from the multi-source heterogeneous data set, wherein the structured asset knowledge comprises entity identification, relation extraction, attribute association and ontology matching;
the asset knowledge graph is constructed through knowledge representation learning, knowledge fusion, knowledge graph storage and knowledge graph updating steps.
Preferably, the constructed asset knowledge graph is converted into a low-dimensional dense vector representation, and the low-dimensional dense vector representation is used as the input of a fault diagnosis model, and simultaneously the acquired asset dynamic data is also used as the input of the fault diagnosis model;
And constructing a fault diagnosis model based on the graph attention network GAT model and the long-short-term memory network LSTM model, wherein the fault diagnosis model comprises an asset embedding layer, a graph attention layer, an LSTM layer and a fault classification layer.
Preferably, the fault type of the asset is marked based on the asset risk assessment result and the historical operation and maintenance record to form a training sample for training a fault diagnosis model;
and performing end-to-end supervised learning on the fault diagnosis model through a training sample, learning model parameters, and when the training of the fault diagnosis model is completed, predicting the fault type by using the trained fault diagnosis model, and outputting a fault diagnosis result according to the asset risk level and the predicted fault type.
Preferably, the asset data is subjected to multidimensional data analysis, asset static data is acquired, asset dynamic data is acquired, and asset knowledge graph data is acquired;
Combining the characteristics of the asset static data and the asset dynamic data to construct an asset security risk assessment characteristic matrix;
based on the asset relationship adjacency matrix, calculating the association risk coefficient between the assets by adopting a random walk algorithm.
Preferably, an asset operation and maintenance risk detection algorithm is constructed based on the associated risk coefficients among the assets, and comprehensive security risk scores of the assets are calculated;
Normalizing the comprehensive security risk score to obtain a risk probability value of the asset;
And setting a risk level dividing threshold according to the comprehensive risk probability value of the asset, dividing the asset into four risk levels of low risk, medium risk, high risk and serious risk, and distributing different operation and maintenance strategies to each risk level.
Preferably, the method comprises the steps of carrying out semantic analysis on an asset fault diagnosis result, representing the asset fault result as a fault knowledge vector, carrying out semantic analysis on an asset risk assessment result, representing the asset risk as a risk knowledge vector, inquiring the fault asset and the risk asset from an asset knowledge graph, and representing the inquiry result by an asset comprehensive knowledge vector;
And outputting the operation and maintenance decision text scheme by taking the fault knowledge vector, the risk knowledge vector and the asset comprehensive knowledge vector as input through the pre-training language model.
Preferably, the decision text scheme is pushed to an operation and maintenance manager, and the operation and maintenance manager edits and modifies the scheme through a human-computer interaction interface and executes operation and maintenance measures on the asset.
The asset operation and maintenance decision management platform based on data fusion is used for realizing the asset operation and maintenance decision management method based on data fusion and comprises a data acquisition module, a knowledge graph module, an asset fault diagnosis module, an operation and maintenance risk detection module and an operation and maintenance execution module;
The data acquisition module acquires multi-source heterogeneous data of the full life cycle of the asset through the unified operation and maintenance platform, and performs data cleaning, conversion and fusion to form a standardized asset data set;
The knowledge graph module utilizes ontology modeling and knowledge extraction technology to construct a dynamically updated asset knowledge graph based on the fused asset data set, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information;
The asset fault diagnosis module is used for training an asset fault diagnosis model based on a deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether a fault exists in the current asset according to an asset operation and maintenance strategy by the asset fault diagnosis model;
The operation and maintenance risk detection module is used for constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the multi-dimensional data of the asset, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level;
The operation and maintenance execution module converts the asset faults and the asset operation and maintenance strategies identified based on the knowledge graph into natural language by utilizing a natural language generation technology and pushes the natural language to operation and maintenance management staff, and the operation and maintenance management staff executes operation and maintenance measures.
The method has the beneficial effects that a standardized and high-quality asset data set is established through collecting, cleaning, converting and fusing the full life cycle data of the asset, and a reliable data base is provided for subsequent intelligent operation and maintenance.
The application constructs the dynamic knowledge graph containing the multi-dimensional information of the assets, forms the global view of the assets, and provides a comprehensive, accurate and real-time asset semantic knowledge base for intelligent decision-making.
The application trains the asset fault diagnosis model by utilizing the knowledge graph and the deep learning technology, and can automatically, quickly and accurately discover the asset fault according to the asset state and the historical fault mode.
According to the application, an asset operation and maintenance risk detection algorithm is constructed, and the asset security risk level is continuously evaluated by analyzing the multi-dimensional data of the asset in real time, so that the active identification and early warning of the operation and maintenance risk are realized.
The application automatically generates fault diagnosis and operation strategy report with strong readability by using natural language generation technology, and pushes the report to operation staff, thereby improving operation efficiency and executable performance.
Drawings
FIG. 1 is a flow chart of an asset operation and maintenance decision management method based on data fusion provided by the application;
FIG. 2 is a flow chart of an asset fault diagnosis method of the asset operation and maintenance decision management method based on data fusion provided by the application;
FIG. 3 is a flowchart of an operation and maintenance risk assessment method of the asset operation and maintenance decision management method based on data fusion provided by the application;
FIG. 4 is a flow chart of an operation and maintenance decision generation method of the asset operation and maintenance decision management method based on data fusion provided by the application;
FIG. 5 is a block diagram of an asset operation and maintenance decision management platform based on data fusion provided by the application.
Detailed Description
In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application, but the present application may be practiced in other ways other than those described herein, and persons skilled in the art will be able to make a similar generalization without departing from the spirit of the application, so that the present application is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the application. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Example 1
Referring to fig. 1 to 4, for a first embodiment of the present application, an asset operation and maintenance decision management method based on data fusion is provided.
And step 1, acquiring multi-source heterogeneous data of the full life cycle of the asset through a unified operation and maintenance platform, and performing data cleaning, conversion and fusion to form a standardized asset data set.
The assets comprise a switch, a router, network security equipment, a whole network terminal, a data center proxy server, a virtual machine, an operating system, a database, middleware, a container, an IP address and domain name information.
The method comprises the steps of deploying various sensors, intelligent gateways and other Internet of things equipment on an asset, collecting various parameter information of the asset, wherein the asset static attribute data comprise information of an asset ID, a type, a specification, a manufacturer, purchase time and a position, the asset dynamic monitoring data comprise working states, performance indexes (such as efficiency, yield and energy consumption), environment parameters (such as temperature, humidity and vibration) and health metrics (such as failure rate and maintenance rate) of the asset, the asset dynamic monitoring data comprise start-stop records, parameter setting, maintenance and defect processing operation and maintenance events of the asset, and the data collecting frequency can be set to be real-time, second-level, minute-level or hour-level according to the characteristics of different asset types and monitoring parameters.
And simultaneously recording metadata information of the data, including data sources, acquisition time and original formats, and providing data traceability for subsequent data processing.
Because the multi-source heterogeneous data has the problems of missing values, outliers and inconsistent data quality, data preprocessing is needed, and the method comprises missing value processing, noise removal, outlier detection and data normalization.
The method comprises the steps of missing value processing, noise removal, outlier detection, data normalization, data interoperability improvement, wherein the missing value is estimated through a record deletion mode, an interpolation filling mode and a model prediction mode, the noise removal comprises the steps of removing high-frequency noise interference through signal processing technologies such as Kalman filtering and wavelet transformation, the outlier detection comprises the step of identifying abnormal values through a statistical method or a machine learning algorithm such as clustering based on distance and density analysis, and the data normalization comprises the step of identifying data with different sources and different formats according to unified naming standards and coding rules.
The method comprises the steps of integrating, storing and managing the standardized data after cleaning and converting by utilizing a database technology and a big data platform to form a database of an asset theme domain, associating the same asset data from different sources by key attributes of asset IDs and time stamps to form a complete asset state record, summarizing and merging the same index data from a plurality of sources to improve data quality and reliability, solving conflict data among a plurality of data sources by defining a consistency rule, and setting the consistency rule to be realized by adopting a priority ordering mode.
In the data fusion process, the consistency rule of the multi-source heterogeneous data is measured by calculating JS divergence among data sources, and the formula is as follows:
Wherein, the Representing probability distributionAndThe JS divergence between the two,Is a probability distributionAndIs provided with a uniform mixing distribution of the components,,Is a probability distributionRelative toThe KL divergence between the two is calculated as follows:, As a function of the parameters, Is a probability distributionThe probability value on the first category,Is a mixed probability distributionProbability values on the ith category, and so onRepresenting probability distributionRelative toIs used for the distribution of KL of the formula (I),For calculating distributionRelative toIs provided.
The JS divergence can measure the similarity degree of two probability distributions, the value is between 0 and 1, the smaller the JS divergence is, the closer the distribution among the data sources is, the higher the data consistency is, the data indexes with high priority are selected as the representative data of the same index data by using the forward sequence of the JS divergence values as a priority sequence, and the problem of data conflict among a plurality of data sources is solved.
And 2, constructing a dynamically updated asset knowledge graph based on the fused asset data set by utilizing an ontology modeling and knowledge extraction technology, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information.
The method comprises the steps of modeling an asset ontology, wherein the ontology is formed and explicitly described by specific domain knowledge and consists of elements such as concepts, relationships, attributes and axioms, constructing an ontology model for covering the whole life cycle management of assets aiming at the IT asset operation and maintenance domain, wherein the asset concepts are used for abstracting IT assets with different granularities and different types into uniform ontology concepts to form an asset classification system, the asset classification system comprises switch and router network equipment concepts, server and virtual machine computing equipment concepts and software, middleware and database application concepts, the asset relationships are used for defining semantic association relationships between the assets such as composition relationships, connection relationships, dependency relationships and position relationships, and the asset attributes are used for describing the inherent characteristics of the assets, including brands, models, configuration parameters and health states of the assets.
And formally defining asset concepts, asset relationships and asset attributes by using an ontology description language (such as OWL and RDF) to construct an IT asset ontology model.
Based on the constructed IT asset ontology model, extracting structured asset knowledge from the multi-source heterogeneous data set, wherein the structured asset knowledge comprises entity identification, relation extraction, attribute association and ontology matching;
The entity identification includes identifying an asset entity from unstructured text data (such as device log and operation and maintenance work order) by using named entity identification (NER) technology and mapping the asset entity to an ontology concept, wherein NER tasks can be realized based on a machine learning model of a Conditional Random Field (CRF) or a cyclic neural network (RNN), the relation extraction includes extracting semantic relations between asset entities from the text by using a rule, template or machine learning based method, such as connection topology of devices and relativity of faults, etc., a commonly used relation extraction model such as PCNN or a Transformer model, the attribute association includes associating asset monitoring data streams with the asset entities to form real-time attribute value update, such as mapping CPU utilization rate and memory occupation monitoring index of a server to corresponding attributes of the server entity, the ontology matching includes semantically associating the extracted asset instances with the ontology concept by using an ontology matching technology, realizing data and ontology fusion, extracting obtained asset entity and relation knowledge elements, storing the extracted asset and relation knowledge elements in a triplet form, forming a triplet RepresentingThe head entity is provided with a control unit,Represents the tail entity of the plant,Representing relationships between entities, an entity is a mapping of an asset in a triplet.
Because the asset knowledge is derived from multi-source heterogeneous data and dynamically changes in real time, knowledge fusion and reasoning are needed to form a high-quality and consistent asset knowledge graph.
The asset knowledge graph is constructed through knowledge representation learning, knowledge fusion, knowledge graph storage and knowledge graph updating steps.
Knowledge representation learning, namely embedding asset entities and relations into a low-dimensional semantic space by adopting a knowledge representation model, such as TransE model or TransR model, so that the similar entities have similar distances in the embedded space, knowledge fusion, namely comprehensively utilizing knowledge representation learning, predicate logic reasoning and rule-based reasoning to carry out conflict resolution, consistency check and new knowledge reasoning on RDF knowledge under the constraint of a domain ontology, and storing a knowledge map by adopting a map database, wherein the knowledge map is stored, and the attribute values in the asset knowledge map are updated in real time according to the state parameter change in the asset operation process, and meanwhile, the knowledge map also supports periodic batch updating, and fusion of main data of the asset, an externally introduced industry knowledge base and the like.
The asset knowledge graph obtained by fusion reasoning formally represents the static attribute, the dynamic state and the topological semantic relation among the assets, provides rich-semantic and high-quality knowledge support for intelligent operation and maintenance decision, and can remarkably improve the accuracy and efficiency of tasks such as defect cause analysis, fault diagnosis, risk prediction and the like through the construction of the knowledge graph.
In the knowledge graph construction process, the semantic similarity between asset entities is calculated through TransE model, and for any triplet in the asset knowledge graphLearning triples by TransE modelLow-dimensional vector representation in embedding spaceSo that the header entity embeds the vectorWarp relationship vectorPost-translational and tail entity embedding vectorsAs close as possible, namely:。
TransE model employs an interval-based ordering loss function And (3) performing optimization training:
Wherein the method comprises the steps of Is the set of all real triples in the asset knowledge graph,For a negative sample triplet set obtained by randomly replacing the head or tail entity,Is the negative sample header entity embedding vector,Is the embedded vector of the negative sample tail entity,Is the separation distance between the positive and negative samples,The euclidean distance between entity embeddings is measured,Represents a ReLU operation for obtaining the value of the non-negative part.
Asset entities and relation embedded vectors obtained through TransE model learning can be used for calculating semantic similarity among entities, and the semantic similarity is calculated by cosine similarity of the embedded vectors:
Based on the triplet And carrying out low-dimensional vector representation in the embedded space, so as to support knowledge services such as similar asset retrieval, fault propagation analysis and the like on the knowledge graph.
And 3, training an asset fault diagnosis model based on the deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether the current asset has faults or not according to an asset operation and maintenance strategy by the asset fault diagnosis model.
Converting the constructed asset knowledge graph into a low-dimensional dense vector representation as input of a fault diagnosis model, adopting a TransE model knowledge graph representation learning algorithm to learn the embedded vectors of each asset node and the relation edge, and for any asset node in the knowledge graphObtaining the embedded vectorRelationship edges for asset nodesObtaining the embedded vectorWhereinIs the dimension of the embedded vector.
The collected dynamic data of the assets are also used as the input of a fault diagnosis model, and for any asset node, the monitoring index of the asset node is extractedWhereinIn order to monitor the number of indicators,Is a time stamp.
Constructing a fault diagnosis model based on a graph attention network (GAT) model and a long and short time memory network (LSTM) model, wherein the model comprises:
asset embedding layer, namely embedding asset knowledge graph into vector As an initial characteristic representation of the asset node.
The attention layer of the graph is used for aggregating neighbor information of asset nodes through an attention mechanism, updating characteristic representation of the nodes and for the asset nodesAttention coefficient thereofRepresenting asset nodesFor a pair ofThe importance of (2) is calculated by the following formula:
Wherein, the Is an asset nodeIs defined by a set of neighboring nodes of the network,As a matrix of weights, the weight matrix,AndRepresenting asset nodesAndIs used to determine the embedded vector of (c),In order for the attention vector to be of interest,The transpose of the representation vector is performed,The vector concatenation operation is represented by a vector,Representing the computed stitched feature vector to the attention vectorThe similarity between the two is set to be similar,In order to activate the function,Is any node different from the assetAndIs a vector representation of neighboring asset nodes.
LSTM layer real-time monitoring data for assetsPerforming time sequence modeling to capture dynamic change of asset state, wherein the input of the LSTM layer is an asset nodeSplicing the updated characteristic representation and the real-time monitoring data, and outputting the updated characteristic representation and the real-time monitoring data as the hidden state of the asset node:
Wherein, the Is a nodeThe feature representation updated by the attention layer of the graph,Indicating LSTM layer atA hidden state of the moment.
Fault classification layer based on output of LSTM layerOutput asset nodes through fully connected layers and softmax functionsAt the moment of timeFault type of (a):
Wherein, the Representing asset nodesThe probability distribution of different types of faults occurring in time,AndThe weight matrix and the bias vector of the classification layer respectively,Representation fetchThe index with the highest probability is used as the predicted fault category.
Labeling the fault type of the asset based on the asset risk assessment result and the historical operation and maintenance record to form a training sampleThe fault types can be divided into a number of granularities, such as hardware faults, software faults, network faults and security faults.
Performing end-to-end supervised learning of the fault diagnosis model on a training set by using a cross entropy loss function and an Adam optimizer, and learning model parameters、、And。
And outputting fault diagnosis results including fault asset ID, fault type, fault time and fault position information according to the asset risk level and the predicted fault type.
Feeding back the fault diagnosis result to the constructed operation and maintenance strategy, and adopting corresponding maintenance operations, such as fault isolation, abnormality analysis, configuration adjustment and safety reinforcement, on the assets predicted to have faults; and meanwhile, updating the diagnosis result into the asset knowledge graph to form a knowledge closed loop of the asset fault.
The method comprises the steps of embedding asset knowledge graphs into real-time monitoring data, combining a graph attention network and an LSTM, and constructing an asset fault diagnosis model, wherein the model can fully utilize static attribute, dynamic state and topology association information of the asset, realize real-time prediction and positioning of the asset fault and provide basis for intelligent operation and maintenance decision of the asset.
And 4, constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the multi-dimensional data of the asset, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level.
Acquiring asset static data, extracting asset static data information of the asset from an asset knowledge graph, wherein the asset static data information comprises asset types, models, specifications, deployment positions, configuration parameters and the like, and forming asset static data characteristicsWhereinIs an asset static data dimension.
Acquiring asset dynamic data, and acquiring various performance monitoring index data of the asset in real time through an asset data acquisition interface, wherein the performance monitoring index data comprises CPU (Central processing Unit) utilization rate, memory occupancy rate, disk IO (input/output), network flow, abnormal logs and the like, and a time sequence data set is formedWhereinFor the amount of dynamic data of an asset,Is a time stamp.
Acquiring asset knowledge graph data, acquiring topological connection relations, business dependency relations, historical fault relations and the like among assets from the asset knowledge graph, and constructing an asset relation adjacency matrixWhereinFor the total number of assets,Representing asset nodesAsset nodeAnd the association relation exists between the two, otherwise, the association relation is 0.
Based on asset static dataAnd asset dynamic dataAn asset security risk assessment index system is constructed, including multiple dimensions of asset importance, configuration compliance, performance anomalies, and event severity.
Calculating importance weight of each index by utilizing index association relation and historical risk event in asset knowledge graphThe degree centrality and the medium number centrality graph characteristics of the index nodes and the association strength of the index and the historical risk event are considered.
For the dynamic monitoring index, a time sequence analysis method can be adopted to predict the future trend and abnormal condition of the index and dynamically adjust the weight.
Static data of assetsAnd asset dynamic dataFeature combination is carried out, and an asset security risk assessment feature matrix is constructed。
Adjacency matrix based on asset relationshipsCalculating associated risk coefficients between assets using a random walk algorithmThe step of the random walk algorithm comprises the steps of defining a transition probability matrixThe related edge semantic weight in the knowledge graph is considered to obtain a weighted transition probability matrixObtaining the steady-state distribution of the asset nodes through repeated iterative computationBy usingRepresenting asset nodesAndIs a risk factor associated with the system.
Comprehensively considering the characteristics and associated risks of the asset, constructing an asset operation and maintenance risk detection algorithm, and calculating comprehensive security risk scores of the asset:
Wherein, the Representing the static data dimension of the asset,Representing the amount of dynamic data of the asset,Evaluating feature matrices for asset security riskFirst, theThe characteristic value of the column,For the index parameter of the subscript,AndIs an asset nodeAndIs embedded in the vector of the knowledge graph,As a weighting factor for the associated risk,For measuring the influence degree of the association risk of the assets on the overall risk score, for example, if the association between the assets is weak, such as independent equipment, a smaller beta value (such as 0.1-0.3) is set, and if the association between the assets is strong (such as network equipment and production equipment with tight dependency relationship), a larger beta value (such as 0.5-0.9) is set.
Scoring comprehensive security riskNormalizing to obtain the risk probability value of the assetMapping is carried out by adopting a Sigmoid function:
Wherein, the To normalize the steepness of the curve,For the threshold of risk score, the historically failed asset may be calculated based on historical data statistics settingsValue asExemplary if in the past 1000 pieces of data, the failed assets that have occurred are averagedA value of 65, thenSetting 65.
Based on comprehensive risk probability values for assetsSetting a risk level dividing thresholdAssets are classified into four classes of low risk, medium risk, high risk and severe risk.
The dividing criteria include whenClassified as low risk whenThe time is classified as stroke risk whenClassified as high risk whenThe time is classified as a serious risk,The probability threshold value for the risk level can be adjusted according to the security requirements and risk preferences of the business system.
And (3) aiming at different asset security risk levels, presetting corresponding operation and maintenance strategies to form a mapping relation between the risk levels and the strategies.
The operation and maintenance strategy comprises, but is not limited to, monitoring frequency adjustment, alarm threshold setting, isolation protection, access control, data backup, emergency plan and the like, and the establishment of the strategy needs to comprehensively consider factors such as importance of assets, risk influence degree, service continuity requirement and the like, and balance safety and usability.
And monitoring the change of the security risk level of the asset in real time, and automatically triggering a corresponding operation and maintenance strategy when the risk level exceeds a preset threshold value or a mutation occurs.
And the policy execution is issued to the asset management domain through the unified automatic operation and maintenance platform, related interfaces are called to realize the execution of the policy, and the execution state and the result are monitored.
And according to feedback information of policy execution, evaluating the effectiveness of the policy and the improvement condition of the asset risk state, and forming a closed loop feedback mechanism of policy optimization.
The method fully utilizes multi-source data such as asset static data, dynamic monitoring and knowledge graph, provides a comprehensive risk scoring model based on index weight dynamic distribution and random walk, can comprehensively describe the risks and associated propagation risks of the asset, realizes risk-driven self-adaptive operation and maintenance decision through risk classification and strategy matching on the basis, introduces a continuous optimization mechanism, and continuously improves the intelligent level of the safe operation and maintenance of the asset.
And 5, fusing the asset knowledge graph information and fault diagnosis and risk assessment results, generating a comprehensive asset operation and maintenance decision scheme by using a natural language generation technology, and pushing the comprehensive asset operation and maintenance decision scheme to an operation and maintenance manager for execution.
The method comprises the steps of carrying out semantic analysis on an asset fault diagnosis result, extracting asset fault information of fault assets, fault types, fault symptoms and influence ranges, representing the fault information as fault knowledge vectors, carrying out semantic analysis on asset risk assessment results, extracting asset risk information of risk assets, risk grades, risk factors and potential influences, representing the asset risk information as risk knowledge vectors, inquiring attribute, relationship and context information related to the fault assets and the risk assets from an asset knowledge graph, and fusing asset static attribute and dynamic monitoring data to form an asset comprehensive knowledge vector.
And outputting an operation and maintenance decision scheme by taking the fault knowledge vector, the risk knowledge vector and the asset comprehensive knowledge vector as inputs through a pre-training language model.
And the operation and maintenance manager edits and modifies the scheme through a human-computer interaction interface, and inputs supplementary explanation to form an executable operation and maintenance scheme.
The operation and maintenance manager decomposes the operation and maintenance scheme into a plurality of specific operation and maintenance tasks, and distributes the specific operation and maintenance tasks to relevant technicians for execution.
In the operation and maintenance task execution process, asset state data are collected in real time through means such as Internet of things equipment and monitoring tools, and information such as operation logs and change records of each task is recorded and used for evaluating the execution effect of a decision scheme.
Example 2
Referring to FIG. 5, a second embodiment of the present application provides an asset operation and maintenance decision management platform based on data fusion.
The platform comprises a data acquisition module, a knowledge graph module, an asset fault diagnosis module, an operation and maintenance risk detection module and an operation and maintenance execution module.
The data acquisition module acquires multi-source heterogeneous data of the full life cycle of the asset through the unified operation and maintenance platform, and performs data cleaning, conversion and fusion to form a standardized asset data set.
The knowledge graph module utilizes ontology modeling and knowledge extraction technology to construct a dynamically updated asset knowledge graph based on the fused asset data set, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information.
The asset fault diagnosis module is used for training an asset fault diagnosis model based on a deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether a fault exists in the current asset according to an asset operation and maintenance strategy.
The operation and maintenance risk detection module is used for constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the multi-dimensional data of the asset, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level.
The operation and maintenance execution module converts the asset faults and the asset operation and maintenance strategies identified based on the knowledge graph into natural language by utilizing a natural language generation technology and pushes the natural language to operation and maintenance management staff, and the operation and maintenance management staff executes operation and maintenance measures.
Example 3
The present application provides a storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method as described above. By the technical scheme, when the computer program is executed by the processor, the method in any optional implementation mode of the embodiment is executed, so that the functions of acquiring multiple groups of historical electricity utilization data, dividing the historical electricity utilization data, outputting specific electricity utilization groups, calculating the historical electricity utilization data, outputting a moving average value, analyzing the moving average value, outputting the equipment type and the peak time, acquiring the current time and the real-time electricity utilization amount, analyzing and calculating the real-time electricity utilization amount based on the current time, the peak time and the equipment type, and outputting abnormal electricity utilization information based on a calculation result are realized.
In the foregoing embodiments of the present application, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media having computer-usable program code embodied therein. The storage medium may be implemented by any type or combination of volatile or nonvolatile memory devices, such as static random access memory (StaticRandomAccessMemory, SRAM), electrically erasable programmable Read-only memory (ElectricallyErasableProgrammableRead-only memory, EEPROM), erasable programmable Read-only memory (ErasableProgrammableReadOnlyMemory, EPROM), programmable Read-only memory (ProgrammableRead-only memory, PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
In the embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
The embodiments of the present application have been described above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and variations, modifications, substitutions and alterations of the above-described embodiments may be made by those having ordinary skill in the art without departing from the spirit of the present application and the scope of the present application as defined by the claims, which are all within the scope of the present application.
Claims (10)
1. The asset operation and maintenance decision management method based on data fusion is characterized by comprising the following steps of:
Acquiring multi-source heterogeneous data of the full life cycle of the asset through a unified operation and maintenance platform, and performing data cleaning, conversion and fusion to form a standardized asset data set;
Constructing a dynamically updated asset knowledge graph based on the fused asset data set by utilizing an ontology modeling and knowledge extraction technology, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information;
Training an asset fault diagnosis model based on a deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether a fault exists in the current asset according to an asset operation and maintenance strategy by the asset fault diagnosis model;
Constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the asset multidimensional data, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level;
And converting the asset faults and the asset operation and maintenance strategies identified based on the knowledge graph into natural language by utilizing a natural language generation technology, pushing the natural language to an operation and maintenance manager, and executing operation and maintenance measures by the operation and maintenance manager.
2. The asset operation and maintenance decision management method based on data fusion according to claim 1, wherein the asset comprises a switch, a router, a network security device, a whole network terminal, a data center proxy server, a virtual machine, an operating system, a database, middleware, a container, an IP address and domain name information;
Collecting various parameter information of an asset, including asset static attribute data, asset dynamic monitoring data and asset operation and maintenance operation data;
Carrying out data preprocessing on the acquired data, wherein the data preprocessing comprises missing value processing, noise removal, outlier detection and data standardization;
and summarizing all the acquired data, and solving the problem of the same index data conflict among a plurality of data sources by calculating JS divergence among the data as data priority order.
3. The method for managing the operation and maintenance decisions of the asset based on the data fusion of claim 2, wherein modeling the asset ontology and constructing an ontology model covering the whole life cycle management of the asset comprises asset concepts, asset relationships and asset attributes, associating the asset attributes to the asset entity;
Based on the constructed IT asset ontology model, extracting structured asset knowledge from the multi-source heterogeneous data set, wherein the structured asset knowledge comprises entity identification, relation extraction, attribute association and ontology matching;
the asset knowledge graph is constructed through knowledge representation learning, knowledge fusion, knowledge graph storage and knowledge graph updating steps.
4. The asset operation and maintenance decision management method based on data fusion according to claim 3, wherein the constructed asset knowledge graph is converted into a low-dimensional dense vector representation as input of a fault diagnosis model and the acquired asset dynamic data is also used as input of the fault diagnosis model;
And constructing a fault diagnosis model based on the graph attention network GAT model and the long-short-term memory network LSTM model, wherein the fault diagnosis model comprises an asset embedding layer, a graph attention layer, an LSTM layer and a fault classification layer.
5. The asset operation and maintenance decision management method based on data fusion according to claim 4, wherein the fault type of the asset is marked based on the asset risk assessment result and the historical operation and maintenance record to form a training sample for training a fault diagnosis model;
and performing end-to-end supervised learning on the fault diagnosis model through a training sample, learning model parameters, and when the training of the fault diagnosis model is completed, predicting the fault type by using the trained fault diagnosis model, and outputting a fault diagnosis result according to the asset risk level and the predicted fault type.
6. The method for managing the operation and maintenance of the asset based on the data fusion according to claim 5, wherein the method is characterized in that the asset data is subjected to multidimensional data analysis, asset static data is obtained, asset dynamic data is obtained and asset knowledge graph data is obtained;
Combining the characteristics of the asset static data and the asset dynamic data to construct an asset security risk assessment characteristic matrix;
based on the asset relationship adjacency matrix, calculating the association risk coefficient between the assets by adopting a random walk algorithm.
7. The asset operation and maintenance decision management method based on data fusion according to claim 6, wherein an asset operation and maintenance risk detection algorithm is constructed based on the associated risk coefficients among the assets, and comprehensive security risk scores of the assets are calculated;
Normalizing the comprehensive security risk score to obtain a risk probability value of the asset;
And setting a risk level dividing threshold according to the comprehensive risk probability value of the asset, dividing the asset into four risk levels of low risk, medium risk, high risk and serious risk, and distributing different operation and maintenance strategies to each risk level.
8. The method for managing the operation and maintenance decisions of the asset based on the data fusion according to claim 7, wherein the method is characterized by carrying out semantic analysis on the diagnosis result of the asset fault, and representing the diagnosis result of the asset as a fault knowledge vector;
And outputting the operation and maintenance decision text scheme by taking the fault knowledge vector, the risk knowledge vector and the asset comprehensive knowledge vector as input through the pre-training language model.
9. The method for managing asset operation and maintenance decision based on data fusion according to claim 8, wherein the decision text scheme is pushed to an operation and maintenance manager, and the operation and maintenance manager edits and modifies the scheme through a man-machine interface and performs operation and maintenance measures on the asset.
10. The asset operation and maintenance decision management platform based on data fusion is used for realizing the asset operation and maintenance decision management method based on data fusion as claimed in any one of claims 1 to 9, and is characterized by comprising a data acquisition module, a knowledge graph module, an asset fault diagnosis module, an operation and maintenance risk detection module and an operation and maintenance execution module;
The data acquisition module acquires multi-source heterogeneous data of the full life cycle of the asset through the unified operation and maintenance platform, and performs data cleaning, conversion and fusion to form a standardized asset data set;
The knowledge graph module utilizes ontology modeling and knowledge extraction technology to construct a dynamically updated asset knowledge graph based on the fused asset data set, wherein the dynamically updated asset knowledge graph comprises asset attributes, topological relations and operation state multidimensional information;
The asset fault diagnosis module is used for training an asset fault diagnosis model based on a deep learning model, learning an asset fault mode from an asset knowledge graph, and detecting whether a fault exists in the current asset according to an asset operation and maintenance strategy by the asset fault diagnosis model;
The operation and maintenance risk detection module is used for constructing an asset operation and maintenance risk detection algorithm, identifying an asset security risk coefficient through real-time analysis of the multi-dimensional data of the asset, judging an asset security risk level, and triggering a corresponding asset operation and maintenance strategy according to the asset security risk level;
The operation and maintenance execution module converts the asset faults and the asset operation and maintenance strategies identified based on the knowledge graph into natural language by utilizing a natural language generation technology and pushes the natural language to operation and maintenance management staff, and the operation and maintenance management staff executes operation and maintenance measures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510336469.0A CN120217158A (en) | 2025-03-21 | 2025-03-21 | Asset operation and maintenance decision-making management platform and management method based on data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202510336469.0A CN120217158A (en) | 2025-03-21 | 2025-03-21 | Asset operation and maintenance decision-making management platform and management method based on data fusion |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN120217158A true CN120217158A (en) | 2025-06-27 |
Family
ID=96107310
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202510336469.0A Pending CN120217158A (en) | 2025-03-21 | 2025-03-21 | Asset operation and maintenance decision-making management platform and management method based on data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN120217158A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120449020A (en) * | 2025-07-10 | 2025-08-08 | 北京安帝科技有限公司 | Industrial asset fingerprint feature dynamic learning and abnormal behavior recognition method and system |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230237404A1 (en) * | 2022-01-21 | 2023-07-27 | Honeywell International Inc. | Performance metric assurance for asset management |
| CN118278778A (en) * | 2024-05-30 | 2024-07-02 | 湖南索思科技开发有限公司 | System operation and maintenance decision support system based on artificial intelligence |
| CN118842176A (en) * | 2024-06-27 | 2024-10-25 | 白银银珠电力(集团)有限责任公司 | Operation and maintenance management system and method based on intelligent substation |
| CN119579142A (en) * | 2024-11-15 | 2025-03-07 | 夏斯韵 | A knowledge graph-based intelligent diagnosis and maintenance system and method for power equipment |
-
2025
- 2025-03-21 CN CN202510336469.0A patent/CN120217158A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230237404A1 (en) * | 2022-01-21 | 2023-07-27 | Honeywell International Inc. | Performance metric assurance for asset management |
| CN118278778A (en) * | 2024-05-30 | 2024-07-02 | 湖南索思科技开发有限公司 | System operation and maintenance decision support system based on artificial intelligence |
| CN118842176A (en) * | 2024-06-27 | 2024-10-25 | 白银银珠电力(集团)有限责任公司 | Operation and maintenance management system and method based on intelligent substation |
| CN119579142A (en) * | 2024-11-15 | 2025-03-07 | 夏斯韵 | A knowledge graph-based intelligent diagnosis and maintenance system and method for power equipment |
Non-Patent Citations (2)
| Title |
|---|
| LIQIAO XIA: ""Maintenance planning recommendation of complex industrial equipment based on knowledge graph and graph neural network"", 《RELIABILITY ENGINEERING AND SYSTEM SAFETY》, 3 January 2023 (2023-01-03), pages 1 - 12 * |
| 韩建友: ""行业软件定义运维平台的研究与设计"", 《计算机科学与应用》, 17 January 2025 (2025-01-17), pages 1 - 9 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN120449020A (en) * | 2025-07-10 | 2025-08-08 | 北京安帝科技有限公司 | Industrial asset fingerprint feature dynamic learning and abnormal behavior recognition method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN117473571A (en) | Data information security processing method and system | |
| CN118539441A (en) | Power grid topology optimization method and system based on search ordering | |
| CN117675691B (en) | Remote fault monitoring method, device, equipment and storage medium of router | |
| CN120217158A (en) | Asset operation and maintenance decision-making management platform and management method based on data fusion | |
| Xie et al. | Logm: Log analysis for multiple components of hadoop platform | |
| CN118840140A (en) | Enterprise network sales data early warning management system based on artificial intelligence | |
| CN118861683A (en) | Power grid historical message processing method, device and power grid monitoring system | |
| CN119691576A (en) | Self-adaptive operation and maintenance root cause positioning method and system based on deep learning | |
| CN120162213A (en) | Risk warning method, system and electronic equipment | |
| CN117354251A (en) | Automatic extraction method for electric power Internet of things terminal characteristics | |
| CN117834202A (en) | A SDWAN abnormal traffic identification method based on XGBoost model | |
| CN120197957A (en) | An abnormal behavior analysis system for data centers based on artificial intelligence | |
| CN118965247B (en) | Power plant data management method and system based on multi-source data | |
| CN119623775A (en) | Task chain self-healing transfer prediction method and system based on parameter fingerprint holographic perception | |
| CN120256196A (en) | Data processing method, system and storage medium for cloud computing | |
| CN120085885A (en) | A method for updating an operating system based on cloud services | |
| CN118396606A (en) | Manufacturing industry product production optimization method based on multi-mode large model | |
| CN118885354A (en) | Log anomaly detection method based on LSTM network and process status inspection | |
| CN117311777A (en) | Automatic operation and maintenance platform and method | |
| CN117349744A (en) | A terminal-side alarm traceability graph filtering method based on graph neural network clustering | |
| CN120257113B (en) | An intelligent data management system and method based on multi-source data acquisition | |
| CN119579147B (en) | Substation network equipment fault elimination and verification method and system based on deep learning | |
| CN119441800B (en) | Enterprise management data management method based on big data analysis | |
| CN119051996B (en) | Training method and device for abnormal flow detection model, monitoring method and equipment | |
| CN120811863A (en) | Operation and maintenance delivery abnormal event early warning method and system based on cloud platform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |