An early fault warning method and system based on dynamic evolution of complex industrial maps

Abstract

The application relates to the technical field of fault early warning, in particular to an early fault early warning method and system based on dynamic evolution of a complex industrial graph. The method comprises the following steps: constructing a multilevel industrial system knowledge graph based on obtained industrial equipment parameters; performing self-adaptive knowledge updating on the multilevel industrial system knowledge graph based on a graph dynamic evolution mechanism; constructing a joint reasoning framework based on a large model and the knowledge graph; optimizing the joint reasoning framework based on the self-adaptive knowledge updating and a continuous learning mechanism; and performing early fault early warning on industrial equipment by using the optimized joint reasoning framework. The early fault early warning system based on dynamic evolution of a complex industrial system graph and joint reasoning of a large model fuses a dynamic knowledge graph and a large language model, and realizes early, accurate and interpretable industrial fault early warning.

Description

Technical Field

[0001] The present invention relates to the field of fault warning technology, and in particular to an early fault warning method and system based on the dynamic evolution of complex industrial graphs. Background Art

[0002] Traditional failure maintenance strategies are mainly divided into three categories: post-failure maintenance, preventive maintenance, and predictive maintenance. Post-failure maintenance involves repairs after a device fails. While this avoids unnecessary maintenance costs, it often results in greater losses. Preventive maintenance involves inspecting and maintaining equipment at fixed intervals, disregarding the equipment's actual condition and easily leading to over- or under-maintenance. Predictive maintenance is based on the equipment's actual operating status, continuously monitoring key equipment parameters to predict potential failures and taking action before they occur. It is currently the most advanced maintenance strategy. Research shows that compared to traditional post-failure maintenance strategies, effective early warning systems can reduce equipment downtime by 50%-70%, lower maintenance costs by 25%-30%, and extend equipment life by 20%-40%.

[0003] In early fault warning technology, rule-based approaches use pre-set expert rules or thresholds to determine whether a system is abnormal. These methods are simple to implement, computationally efficient, and offer highly interpretable results. However, they rely on expert experience, have poor generalization capabilities, and struggle to handle dynamic changes and uncertainty. Statistical methods, such as multivariate statistical process control and principal component analysis, can handle correlations between multivariate data and offer some robustness to data noise. However, they typically assume that data follows a specific distribution, have limited ability to model nonlinear relationships, and struggle to handle data with strong time-series dependencies.

[0004] Large language models (LLMs), with their powerful natural language processing and knowledge representation capabilities, have been widely used in the industrial field for applications such as intelligent document analysis, expert knowledge acquisition, fault diagnosis assistance, and anomaly description generation. However, in terms of fault warning, large language models still have limitations, such as limited numerical computing capabilities, insufficient domain knowledge, real-time challenges, reliability concerns, and a lack of structured representation for industrial systems.

[0005] Overall, current industrial fault warning technology faces core challenges, including system complexity and dynamics, the separation of knowledge and data, the trade-off between explainability and accuracy, the weakness of early warning signs, the formalization and updating of domain knowledge, the integration of multi-source heterogeneous data, computing efficiency and resource consumption, the scarcity of fault samples, the lack of integration between system knowledge structure and data-driven models, and the lack of operability of warning results. Specifically, they include:

[0006] The complexity of the structure and component relationship of the industrial system makes it difficult for traditional fault early warning methods to fully grasp the overall state and internal interaction mechanism of the system, resulting in inaccurate prediction of the fault propagation path and the influence range, which seriously affects the timeliness and effectiveness of the early warning.

[0007] Industrial knowledge and real-time data have been in a state of separation for a long time, and existing technologies cannot realize the organic integration of the two. Knowledge graph can represent structured knowledge but lacks combination with dynamic data, while data-driven model can process real-time data but is difficult to integrate domain knowledge. This separation limits the fault early warning capability.

[0008] Fault early signals are weak and submerged in a large amount of normal data and background noise. Existing technologies lack the ability to identify such weak signals, resulting in failure to issue early warning at the initial stage of the fault and missing the best maintenance opportunity.

[0009] Large language models face the problem of field adaptability in the industrial field. General large models lack sufficient industrial expertise and have limited capabilities in numerical calculation and time series analysis, making it difficult to be directly applied to professional industrial fault early warning tasks.

[0010] In order to overcome the limitations of existing technologies, it is necessary to develop a new type of industrial fault early warning system that can organically integrate system structure knowledge, expert experience and data-driven models, dynamically capture system evolution and achieve early, accurate and interpretable fault early warning. SUMMARY

[0011] To solve the above-mentioned problems, the present application provides an early fault early warning method and system based on dynamic evolution of complex industrial graph.

[0012] In the first aspect, the present application provides an early fault early warning method based on dynamic evolution of complex industrial graph, which adopts the following technical solution:

[0013] An early fault early warning method based on dynamic evolution of complex industrial graph, comprising:

[0014] Obtaining industrial equipment parameters;

[0015] Constructing a multi-level industrial system knowledge graph based on the obtained industrial equipment parameters;

[0016] Adaptive knowledge updating of the multi-level industrial system knowledge graph based on the graph dynamic evolution mechanism;

[0017] Constructing a joint reasoning framework based on a large model and a knowledge graph;

[0018] Optimizing the joint reasoning framework based on adaptive knowledge updating and continuous learning mechanism;

[0019] Early fault warning of industrial equipment is performed by using the optimized joint inference framework.

[0020] Further, the multi-level industrial system knowledge graph is constructed based on the obtained industrial equipment parameters, including defining the overall architecture of the graph, constructing a basic graph representing the physical structure of the industrial system and the relationship between equipment components through physical entity nodes and parameter characteristics to form a physical layer subgraph, constructing an abstract representation of system function modules and their interaction relationships through function unit node function relationships to form a function layer subgraph, and establishing a network representation of monitoring parameters and their associated relationships through monitoring node data and monitoring devices to form a monitoring layer subgraph.

[0021] ,

[0022] wherein, represents the physical layer subgraph, including equipment, component physical entities and their connection relationships; represents the function layer subgraph, describing the function units and their interactions of the system; represents the monitoring layer subgraph, including sensor and monitoring point information; represents the cross-layer relationship between the physical layer and the function layer, the function layer and the monitoring layer, and the physical layer and the monitoring layer, respectively.

[0023] Further, the multi-level industrial system knowledge graph is constructed based on the obtained industrial equipment parameters, including establishing mapping relationships between the three layers of the graph based on the physical layer, the monitoring layer and the function layer, respectively, to ensure the completeness and consistency of the knowledge representation, wherein the importance of physical components to function implementation is quantified through contribution degree analysis to establish mapping weights between physical components and function units; the reflection ability of monitoring parameters to function status is determined through sensitivity and coverage analysis to establish associated weights between function status and monitoring parameters; finally, the overall semantic consistency is ensured through cross-layer constraints, represented as:

[0024] ,

[0025] wherein, represents the semantic consistency measure of the graph; represents the path set that meets the semantic consistency, i.e. the valid path from the physical entity through the function unit to the monitoring point; represents the set of all possible paths.

[0026] Furthermore, the adaptive knowledge update of the multi-level industrial system knowledge graph based on the dynamic evolution mechanism of the graph includes establishing a mathematical representation of the graph changing over time, expanding the static graph into a dynamic time series structure, and establishing a mathematical model of the graph evolution to describe the driving factors and change laws of the graph state change; designing a dynamic update mechanism for the node and edge attributes in the graph so that the graph can reflect the system state changes in real time, wherein a dynamic update model for the health status of physical components is established, comprehensively considering monitoring data, historical status and external environmental factors; establishing an integral update model for cumulative attributes such as usage time to accurately record the cumulative operating status of the equipment. For cumulative attributes such as usage time, an integral update model is adopted, which is expressed as:

[0027] ,

[0028] in, Presentation Component i Deadline t The cumulative running time of Indicates that the component is at time running status.

[0029] Furthermore, the adaptive knowledge update of the multi-level industrial system knowledge graph based on the dynamic evolution mechanism of the graph also includes the design of a dynamic adjustment mechanism for the graph topology structure, including node addition and deletion, edge addition and deletion, and weight update, to adapt to changes in the system structure, wherein the node addition and deletion operation is expressed as:

[0030] ,

[0031] in, V ( t ) means at time t The set of all nodes in the graph; the addition and deletion of edges is handled by establishing a dynamic management mechanism for relationship edges to handle the establishment and disconnection of connection relationships in the system, which is expressed as:

[0032] ,

[0033] in, Indicates time t The edge set of represents a new edge set, Indicates the removal of edge sets.

[0034] Further, the joint inference framework based on large models and knowledge graphs is constructed, which includes designing a three-layer enhancement strategy to improve the understanding ability and inference ability of large models in the industrial field. The first layer is domain knowledge enhancement, which improves the understanding ability of large models in the industrial field by professional knowledge injection. Parameter efficient fine-tuning technology is adopted to avoid the high computational cost of full parameter fine-tuning. The second layer is multi-modal information fusion, which enhances the ability of large models to process multiple data types. Multi-modal encoder and attention mechanism are used to realize the effective fusion of different modal information. The third layer is graph structure perception, which enhances the understanding ability of large models to knowledge graph structure. Graph neural network is used for graph structure coding, and cross-modal attention mechanism is established, which is represented as:

[0035] ,

[0036] wherein, is the node feature matrix of the first layer, l is the adjacency matrix with self-loop added, is the corresponding degree matrix, is the learnable weight matrix, is the nonlinear activation function.

[0037] Further, the joint inference framework based on large models and knowledge graphs also includes designing a three-level joint inference mechanism, combining graph reasoning, statistical reasoning and large model reasoning to form a multi-level collaborative early warning mechanism. The first level is the graph reasoning layer, which performs fault propagation analysis and impact assessment based on the structure of the knowledge graph and uses graph structure information for logical reasoning. The second level is the statistical reasoning layer, which uses multivariate anomaly detection and degradation trend analysis methods to detect anomalies and analyze trends based on monitoring data. The third level of inference mechanism is the large model reasoning layer, which uses large language models for deep semantic understanding and complex reasoning to realize multi-source information integration and chain thinking reasoning. Finally, a weighted decision fusion mechanism is designed to integrate the results of the three reasoning layers, which is represented as:

[0038] ,

[0039] wherein, represents the final decision, , and represent the decision results of the graph reasoning layer, the statistical reasoning layer and the large model reasoning layer respectively, is the weight coefficient, which satisfies .

[0040] ​Furthermore, the optimization of the joint reasoning framework based on adaptive knowledge updating and continuous learning mechanism includes establishing a multi-source knowledge acquisition framework, continuously collecting and integrating new knowledge from multiple channels, automatically extracting knowledge patterns from historical monitoring data, and using unsupervised learning methods to mine data regularities and implicit patterns; extracting empirical knowledge from historical failure cases, analyzing recorded failure cases and summarizing failure modes and solutions; using large models to extract structured knowledge from technical documents, and converting unstructured technical information into usable knowledge representations; converting feedback from engineers and maintenance personnel into structured knowledge, integrating manual experience and system automatic learning results, and finally designing a knowledge quality evaluation mechanism, including knowledge consistency evaluation, reliability evaluation and practicality evaluation, through which the accuracy and practicality of knowledge in the system are effectively ensured.

[0041] Furthermore, the optimized joint reasoning framework is used to perform early fault warning for industrial equipment, including the design of a fault warning verification and false alarm optimization mechanism to solve the problems of high false alarm rate and low credibility in traditional warning systems. The accuracy and practicality of warning results are improved through multi-source verification, false alarm control and feedback learning mechanisms. Among them, a multi-source data cross-validation mechanism is used to reduce the misjudgment that may be caused by a single data source, and the credibility of the warning is improved through consistency verification of multiple data sources. The cross-validation score is defined as:

[0042] ,

[0043] in, Validation_Score Indicates the cross-validation score of the warning (0-1), Indicator i Indicates the i The indication results of the data source, It represents the weight of the i-th data source, reflecting the reliability of the data source.

[0044] The second aspect is an early fault warning system based on the dynamic evolution of complex industrial graphs, including:

[0045] The data acquisition module is configured to acquire industrial equipment parameters;

[0046] The knowledge graph module is configured to construct a multi-level industrial system knowledge graph based on the acquired industrial equipment parameters;

[0047] The updating module is configured to adaptively update the knowledge graph of the multi-level industrial system based on the dynamic evolution mechanism of the graph;

[0048] The reasoning module is configured to build a joint reasoning framework based on the large model and the knowledge graph;

[0049] An optimization module configured to optimize the joint inference framework based on an adaptive knowledge update and a continuous learning mechanism.

[0050] An early warning module configured to perform early fault warning of the industrial equipment using the optimized joint inference framework.

[0051] In a third aspect, the present application provides a computer-readable storage medium having stored therein a plurality of instructions adapted to be loaded and executed by a processor of a terminal device to implement the early fault warning method based on dynamic evolution of a complex industrial graph.

[0052] In a fourth aspect, the present application provides a terminal device comprising a processor and a computer-readable storage medium, the processor being configured to implement the instructions, and the computer-readable storage medium being configured to store a plurality of instructions adapted to be loaded and executed by the processor to implement the early fault warning method based on dynamic evolution of a complex industrial graph.

[0053] In summary, the present application has the following beneficial technical effects:

[0054] The early fault warning system based on dynamic evolution of a complex industrial system graph and joint inference of a large model of the present application fuses a dynamic knowledge graph and a large language model, and realizes early, accurate and interpretable industrial fault warning. The system uses a multi-level knowledge graph to represent the structure of a complex industrial system, and combines a graph dynamic evolution mechanism to capture system state changes in real time, thereby significantly improving the modeling capability of the complex system. The attention enhancement and sparse pattern learning mechanism designed for early and weak signs of faults can effectively identify fault signals submerged in background noise, and provide reliable basis for maintenance decisions.

[0055] The system overcomes multiple limitations of traditional warning methods, and realizes deep fusion of knowledge and data. The adaptive knowledge update mechanism can automatically extract new knowledge from multiple sources of data, so that the system can adapt to changes in the industrial environment and avoid the problem of warning failure caused by outdated knowledge. The joint inference framework of the large model and the graph provides highly interpretable warning results, including fault cause analysis, development trend prediction and processing suggestions, which greatly improves the practical value of the warning results. Through the organic combination of these innovative technologies, the system provides a comprehensive solution for predictive maintenance of industrial equipment, effectively reduces maintenance costs, prolongs the service life of equipment, and has significant economic benefits and social value. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a schematic diagram of an early fault warning method based on dynamic evolution of a complex industrial graph according to an embodiment of the present application. DETAILED DESCRIPTION

[0057] The present application will be further described in detail below with reference to the accompanying drawings.

[0058] Embodiment 1

[0059] With reference to Figure 1 The early fault warning method based on complex industrial graph dynamic evolution of the embodiment comprises:

[0060] Step S1, multi-level industrial system knowledge graph construction:

[0061] This step aims to construct a multi-level industrial system knowledge graph as a basic knowledge representation framework for the fault warning system. Unlike traditional single flat structure, this knowledge graph adopts a three-layer architecture design, including a physical layer, a functional layer, and a monitoring layer, to comprehensively express the physical structure, functional relationship, and monitoring parameters of the industrial system.

[0062] Step S1.1, overall architecture definition of the graph:

[0063] The multi-level industrial system knowledge graph can be represented as:

[0064] ,

[0065] Among them, represents the physical layer subgraph, including physical entities such as equipment, components, and their connection relationships; represents the functional layer subgraph, describing the functional units of the system and their interactions; represents the monitoring layer subgraph, including sensor and monitoring point information; represents the cross-layer relationship between the physical layer and the functional layer, the functional layer and the monitoring layer, and the physical layer and the monitoring layer, respectively.

[0066] Step S1.1.1, physical layer subgraph construction:

[0067] Construct a basic graph representing the physical structure and equipment component relationship of the industrial system to form a complete representation of the system physical topology. The physical layer subgraph is defined as:

[0068] ,

[0069] Among them, represents the set of physical entity nodes, such as pump, valve, bearing, and other equipment components; represents the set of physical relationship edges, such as mechanical connection, electrical connection, etc.; represents the set of attributes, including the characteristics of each physical entity.

[0070] Physical layer node hierarchy relationship establishment:

[0071] The hierarchy relationship between the physical layer nodes is represented by the containing edge, forming a multi-level hierarchy of equipment-component-component-element. The fault propagation probability between physical components can be represented as:

[0072] ,

[0073] where, represents the probability of failure propagating from component i to component j ; represents the normalized value of physical distance between components (0-1), the closer the value is, the greater the distance is; represents the degree of functional dependence (0-1), the stronger the value is, the stronger the dependence is; represents the historical failure propagation evidence strength (0-1); is the weight coefficient, satisfying .

[0074] Step S1.1.2, function layer subgraph construction:

[0075] The function layer subgraph is an abstract representation of system function modules and their interaction relationships, and establishes a function-based system understanding framework. The function layer subgraph is defined as:

[0076] ,

[0077] where, represents the set of functional unit nodes, such as power transmission, signal processing, etc. functional modules; represents the set of functional relationship edges, describing the interaction between functions; represents the set of functional attributes, including performance indicators and other information.

[0078] Step S1.1.3, monitoring layer subgraph construction:

[0079] A network representation of monitoring parameters and their associated relationships is established to form a complete description of system observability. The monitoring layer subgraph is defined as:

[0080] ,

[0081] where, represents the set of monitoring point nodes, including various sensors and monitoring devices; represents the set of parameter association edges, describing the correlation between monitoring parameters; represents the set of monitoring attributes, including measurement parameters, sampling frequency, and other characteristics. The correlation between parameters can be calculated by the correlation coefficient:

[0082] ,

[0083] where, represents the correlation coefficient of parameters i and j ; represents the correlation coefficient of parameters iand j covariance of and i standard deviation of j and

[0084] Step S1.1.3.1, monitor layer subgraph anomaly degree definition:

[0085] Establish a quantitative index for monitoring the degree of parameter anomaly, considering both single parameter anomaly and multi-parameter correlation anomaly, to provide a basis for early fault sign identification. The anomaly degree of monitoring parameters can be defined as:

[0086] ,

[0087] wherein, represents the anomaly degree of parameter i at time t represents the measured value of parameter i at time t and represent the historical mean and standard deviation of parameter i respectively; represents the parameter set related to parameter i represents the correlation degree (0-1) between parameters i and j wiw_i and represent the weights of parameter i and related parameter j

[0088] Step S1.2, establish cross-layer relationship mapping:

[0089] Establish the mapping relationship between the three-layer atlas to ensure the completeness and consistency of knowledge representation.

[0090] S1.2.1, physical layer and functional layer association relationship:

[0091] Establish the mapping weight between physical components and functional units, and quantify the importance of physical components to functional implementation through contribution degree analysis. The association relationship between the physical layer and the functional layer can be represented as:

[0092] ,

[0093] wherein, represents the association weight of physical component i to functional unit j represents the association weight of physical component i to functional unit j ​​​​​the contribution degree of the physical component to the functional unit; the denominator represents the sum of the contribution degrees of all physical components to the functional unit j for normalization.

[0094] S1.2.2, the association relationship between the functional layer and the monitoring layer:

[0095] Establish the association weight between the functional state and the monitoring parameter, and determine the reflection ability of the monitoring parameter to the functional state through sensitivity and coverage analysis. The association relationship between the functional layer and the monitoring layer can be expressed as:

[0096] ,

[0097] wherein, represents the association weight of the monitoring parameter j to the functional unit i ; represents the sensitivity (0-1) of the monitoring parameter j to the state change of the functional unit i ; represents the coverage (0-1) of the monitoring parameter j to the functional unit i .

[0098] 1.2.3, semantic consistency constraint:

[0099] The graph ensures the consistency of the overall semantics through cross-layer constraints, and the formula is as follows:

[0100] ,

[0101] wherein, represents the semantic consistency measure of the graph; represents the path set satisfying the semantic consistency, i.e. the effective path from the physical entity to the monitoring point through the functional unit; represents the set of all possible paths.

[0102] S2, design of graph dynamic evolution mechanism:

[0103] The second key step is to design the dynamic evolution mechanism of the graph, so that the knowledge graph can be continuously and adaptively updated as the state of the industrial system changes. Unlike traditional static knowledge graphs, the knowledge graph of the system is a dynamic knowledge base that can capture the dynamic changes of the system over time, reflecting the impact of factors such as component aging, working condition changes, and system upgrades.

[0104] S2.1, overall framework of graph dynamic evolution:

[0105] The overall framework of the dynamic evolution of the graph is designed to extend the static graph to a dynamic time sequence structure. The dynamic evolution of the graph can be formally expressed as a time sequence graph sequence:

[0106] ,

[0107] where, G ( t ) denotes the system graph state at time t , containing three layers of sub-graphs and their cross-layer relationships, which continuously changes over time t;

[0108] S2.2 Dynamic updating mechanism of graph attributes,

[0109] The dynamic updating mechanism of node and edge attributes in the graph is designed to enable the graph to reflect the changes in system state in real time.

[0110] S2.2.1 Updating of node health state,

[0111] A dynamic updating model of physical component health state is established, taking into account monitoring data, historical state and external environmental factors. The attributes of nodes and edges in the graph will be dynamically updated according to real-time data. Taking the health state attribute of a physical component as an example, its updating model is as follows:

[0112] ,

[0113] where, denotes the health state index (0-1) of component i at time t , is a smoothing factor (0-1), is a health state evaluation function; denotes the monitoring data related to component i ; denotes the age, maintenance history and other attributes of component i ; denotes external environmental conditions.

[0114] S2.2.2 Cumulative attribute updating model,

[0115] An integral updating model of cumulative attributes such as usage time is established to accurately record the cumulative operating conditions of equipment. For cumulative attributes such as usage time, an integral updating model is adopted, and the formula is as follows:

[0116] ,

[0117] where, denotes the cumulative operating time of component i up to time t , denotes the operating state (0 or 1) of component at time .

[0118] S2.3 Evolution of graph topology,

[0119] The dynamic adjustment mechanism of the design graph topology is designed, including node addition and deletion, edge addition and deletion, and weight update, to adapt to the system structure changes.

[0120] S2.3.1 Node addition and deletion operation,

[0121] A node life cycle management mechanism is established to handle the addition and removal operations of devices and components in the system. The topology of the graph is adjusted according to the changes in the system, mainly including three operations: node addition and deletion, edge addition and deletion, and weight update. The node addition and deletion operation can be represented as:

[0122] ,

[0123] where, V ( t ) represents the set of all nodes in the graph at time t , which may represent physical devices, functional units, or monitoring points in the industrial system. represents the set of nodes in the graph at the next time (t+Δt), i.e., the updated node set. represents the set of new nodes that need to be added to the graph within the time interval , representing newly added devices, functions, or monitoring points. represents the set of nodes that need to be removed from the graph within the time interval , representing devices, functions, or monitoring points that are eliminated or disabled. is the set of new nodes added to the existing node set . ∖ is the set difference operator, which means removing nodes in from the merged set.

[0124] S2.3.2 Edge addition and deletion operation,

[0125] A dynamic management mechanism for relationship edges is established to handle the establishment and disconnection of connection relationships in the system. The edge addition and deletion operation formula is as follows:

[0126] ,

[0127] where, represents the edge set at time t , represents the new edge set, represents the removed edge set.

[0128] S2.3.3 Relationship weight dynamic update,

[0129] An incremental learning model for relationship weights is established to dynamically adjust the relationship strength based on the system operation status. For the dynamic update of relationship weights, an incremental learning model is used, and the formula is as follows:

[0130] ,

[0131] in, At the moment t The weight of is the learning rate (0-1), is a weight adjustment amount that can be calculated based on correlation changes, fault samples, or expert feedback.

[0132] S2.4 Parameter-related dynamic modeling,

[0133] A dynamic modeling mechanism for monitoring the correlation between parameters is established to capture the changes in parameter correlation strength with system status.

[0134] S2.4.1 Dynamic Mutual Information Calculation,

[0135] The correlation strength between parameters in the monitoring layer is not static, but changes dynamically with the system state. Dynamic mutual information can be used to represent time-varying correlations:

[0136] ,

[0137] in, Representation parameters i and j At the moment t The mutual information of . Indicates at time t Nearby time window and The joint probability distribution of . and Respectively represent the corresponding marginal probability distribution;

[0138] S2.4.2 Sliding time window mechanism,

[0139] A sliding time window mechanism is introduced to capture the dynamic changes of parameter associations and adapt to the time-varying characteristics of the system operation state. The formula is as follows:

[0140] ,

[0141] in, Representation parameters i and j At the moment t correlation. Indicates parameters Indicates parameters i exist[ t - W ,t data sequence within a time window, W is the window length;

[0142] S2.5.1 Dynamic updating of failure propagation probability,

[0143] A dynamic updating model of failure propagation probability is established to adjust the propagation probability parameters according to the latest observation data, and the dynamic updating model of failure propagation probability is as follows:

[0144] S2.5.2 Dynamic updating of failure propagation time,

[0145] A dynamic updating model of failure propagation time is established to reflect the change of propagation speed with system conditions, and the dynamic updating model of failure propagation time is as follows:

[0146] ,

[0147] wherein, represents the probability of failure propagation from component i to component j , is an updating coefficient (0-1); represents the propagation probability calculated based on the latest observation data.

[0148] S2.5.2 Dynamic updating of failure propagation time,

[0149] A dynamic updating model of failure propagation time is established to reflect the change of propagation speed with system conditions, and the dynamic updating model of failure propagation time is as follows:

[0150] ,

[0151] wherein, represents the expected time of failure propagation from component i to component j , is an updating coefficient (0-1), represents the propagation time calculated based on the latest observation data.

[0152] S2.6 System evolution state monitoring,

[0153] A monitoring mechanism of system overall evolution state is established to quantify the system evolution trend through multi-dimensional indexes.

[0154] S2.6.1 Definition of graph evolution state vector,

[0155] In order to capture the overall evolution trend of the system, a graph evolution state vector is introduced, and the formula is as follows:

[0156] ,

[0157] wherein, is the evolution state vector of the system at time t , which contains multiple indicators reflecting the changes of different aspects of the system. Each component Si(t) represents a specific evolution indicator. The system designs three key indicators: system stability index, system stability index, and correlation structure change rate.

[0158] S2.6.2 System stability index calculation,

[0159] A quantitative indicator of system stability is established to measure the stability of the system relationship structure. The formula is as follows:

[0160] ,

[0161] This indicator measures the stability of the system structure. When the system relationship (edge weight) changes less, the stability index is close to 1, indicating that the system state is stable; when the system relationship changes dramatically, the stability index is close to 0, indicating that the system state is unstable, and there may be abnormalities or failures.

[0162] S2.6.3 System degradation rate calculation,

[0163] The overall degradation rate of the system is calculated to reflect the performance decline trend of the system. The formula is as follows:

[0164] ,

[0165] This indicator calculates the average decline speed of all physical component health states. A positive value indicates that the system is recovering (such as after maintenance), and a negative value indicates that the system is degrading. The larger the absolute value of the degradation rate, the faster the system performance changes, and more close monitoring is needed.

[0166] S2.6.4 Correlation structure change rate calculation,

[0167] The correlation structure change rate is calculated to quantify the speed of graph structure adjustment. The formula is as follows:

[0168] ,

[0169] This indicator measures the change speed of the graph structure. High change rate indicates that the system relationship is adjusting rapidly, which may reflect changes in working conditions, configuration updates, or the occurrence of abnormal states.

[0170] Step S3, large model and graph joint reasoning framework construction:

[0171] The step constructs a large model and atlas combined reasoning framework, combines the powerful reasoning ability of a large language model with the structured knowledge of an industrial knowledge graph, and designs an innovative fault early warning reasoning mechanism. This framework aims to overcome the limitations of single methods and achieve deep integration of knowledge-driven and data-driven.

[0172] S3.1 Large model field enhancement strategy,

[0173] In order to adapt the general large language model to the industrial fault early warning task, the present application designs a three-layer enhancement strategy to improve the industrial field understanding ability and reasoning ability of the large model.

[0174] S3.1.1 Field knowledge enhancement,

[0175] The first layer of enhancement strategy is field knowledge enhancement, which improves the industrial field understanding ability of the large model through professional knowledge injection, and adopts parameter efficient fine tuning technology to avoid the high computational cost of full parameter fine tuning. The specific formula is as follows:

[0176] ,

[0177] Wherein, represents the basic large language model, represents the industrial field knowledge dataset, and the knowledge injection adopts parameter efficient fine tuning (PEFT) technology to avoid the high computational cost of full parameter fine tuning.

[0178] S3.1.2 Multi-modal information fusion enhancement,

[0179] The second layer of enhancement strategy is multi-modal information fusion, which enhances the ability of the large model to process multiple data types, and realizes the effective fusion of different modal information through multi-modal encoder and attention mechanism. The formula is as follows:

[0180] ,

[0181] Wherein, is a multi-modal encoder, are text, time series, numerical and spectral data encoders, is a multi-modal fusion function. The fusion method adopts attention mechanism:

[0182] ,

[0183] Wherein, Q is a query vector, and are the key vector and value vector of the first i modal.

[0184] S3.1.3 Graph structure perception enhancement,

[0185] The third layer enhancement strategy is graph structure perception, which enhances the understanding ability of the large model to the knowledge graph structure. Graph neural network is used for graph structure coding, and cross-modal attention mechanism is established. The formula is as follows:

[0186] ,

[0187] wherein, is the node feature matrix of the l layer, is the adjacency matrix with self-loop added, is the corresponding degree matrix, is the learnable weight matrix, is a nonlinear activation function. The graph structure features are fused with the text representation through the cross-modal attention mechanism.

[0188] S3.2 Three-level joint inference mechanism,

[0189] A three-level joint inference mechanism is designed to combine the advantages of graph reasoning, statistical reasoning and large model reasoning to form a multi-level collaborative early warning mechanism.

[0190] S3.2.1 Graph reasoning layer,

[0191] The first level of inference mechanism is the graph reasoning layer, which performs fault propagation analysis and impact assessment based on the structure of the knowledge graph, and uses graph structure information for logical reasoning. The specific formula is as follows:

[0192] ,

[0193] wherein, represents the probability of fault propagation from node to node , represents the path from to , represents the propagation probability between adjacent nodes.

[0194] ,

[0195] wherein, represents the impact of node fault, represents the importance weight of node .

[0196] S3.2.2 Statistical reasoning layer,

[0197] The second level of inference mechanism is the statistical reasoning layer, which performs anomaly detection and trend analysis based on monitoring data. Multivariate anomaly detection and degradation trend analysis methods are used. The specific formula is as follows:

[0198] ,

[0199] wherein, denotes the multivariate monitoring data at time t , and denote the mean vector and covariance matrix in normal state, respectively, denotes the Mahalanobis distance calculation function.

[0200] ,

[0201] wherein, denotes the remaining useful life, denotes the monitoring data in the time window [t-W, t], denotes the current health state, denotes the degradation rate.

[0202] S3.2.3 Large model inference layer,

[0203] The third-level inference mechanism is the large model inference layer, which uses large language models for deep semantic understanding and complex reasoning, realizes multi-source information integration and chain thinking reasoning, and the specific formula is as follows:

[0204] ,

[0205] wherein, denotes the integrated context, denotes the graph state, denotes the real-time data, denotes the event record, denotes the historical case, denotes the relevant document.

[0206] ,

[0207] wherein, denotes the prompt template for guiding chain thinking, denotes the reasoning task description.

[0208] S3.2.4 Weighted decision fusion,

[0209] The system designs a weighted decision fusion mechanism to integrate the results of the three inference layers, and the formula is as follows:

[0210] ,

[0211] wherein, denotes the final decision, , and respectively represent the decision results of the graph reasoning layer, the statistical reasoning layer and the large model reasoning layer, is a weight coefficient, satisfying .

[0212] S3.3 Early fault symptom capture mechanism,

[0213] In view of the problem that early faults are not easy to find, the application specially designs a capture mechanism for early and weak symptoms of faults. The attention mechanism is used to amplify weak signal features and improve the recognition ability of early fault symptoms. The formula is as follows:

[0214] ,

[0215] wherein, represents the attention score of the parameter at time t , and is an attention network, represents the current data context, represents the current graph state;

[0216] S4. Adaptive knowledge update and continuous learning mechanism,

[0217] S4.1 Multi-source knowledge acquisition framework,

[0218] A multi-source knowledge acquisition framework is established to continuously collect and integrate new knowledge from various channels.

[0219] S4.1.1 Monitoring data mining,

[0220] Knowledge patterns are automatically extracted from historical monitoring data. Unsupervised learning methods are used to mine data rules and implicit patterns. The formula is as follows:

[0221] ,

[0222] wherein, represents a mode set extracted from data in a time period , and Clustering, Association and Anomaly respectively represent clustering analysis, association rule mining and anomaly detection methods.

[0223] S4.1.2 Case base learning,

[0224] Case base learning extracts experience knowledge from historical fault cases. The system analyzes recorded fault cases, summarizes fault patterns and solutions, and the formula is as follows:

[0225] ,

[0226] where, Experience ( Case ) denotes the extracted experience knowledge from cases, including fault symptoms, root causes, propagation paths, solutions, and effect assessments.

[0227] S4.1.3 Document Knowledge Extraction,

[0228] Extract structured knowledge from technical documents using large models, transforming unstructured technical information into usable knowledge representations, as shown in the following formula:

[0229] ,

[0230] where, denotes the extracted knowledge from documents, denotes the knowledge extraction function of the large model, denotes the predefined knowledge schema.

[0231] S4.1.4 Expert Feedback Integration,

[0232] Convert the feedback of engineers and maintenance personnel into structured knowledge, integrate artificial experience and system automatic learning results, as shown in the following formula:

[0233] ,

[0234] where, denotes the extracted knowledge from expert feedback, denotes the original feedback of experts, Context denotes the relevant context, G ( t ) denotes the current graph state.

[0235] S4.2 Knowledge Quality Evaluation Mechanism,

[0236] Design a knowledge quality evaluation mechanism, including knowledge consistency evaluation, reliability evaluation, and practicality evaluation, which can effectively ensure the accuracy and practicality of knowledge in the system.

[0237] S4.2.1 Knowledge Consistency Evaluation,

[0238] Check the consistency of new knowledge with existing knowledge, identify and handle knowledge conflicts, and ensure the logical consistency of the knowledge system, as shown in the following formula:

[0239] ,

[0240] where, denotes the consistency score of new knowledge with the current graph , Function identification conflict knowledge, Indicates a subset of the atlas related to new knowledge.

[0241] S4.2.2 Knowledge reliability assessment,

[0242] According to the evaluation of knowledge sources, the reliability of knowledge is evaluated, considering the reliability of the source, the reliability of the extraction method and the confidence, and the formula is as follows:

[0243] ,

[0244] Where, represents the knowledge reliability score, represents the knowledge source reliability, represents the extraction method reliability, ConfidenceConfidence represents the confidence.

[0245] S4.2.3 Knowledge utility assessment,

[0246] Mainly for the actual value of existing knowledge assessment for fault early warning, considering the potential influence, novelty and applicability, the formula is as follows:

[0247]

[0248] Where, represents the knowledge utility score, represents the potential influence, represents the novelty, represents the applicability.

[0249] S4.3 Knowledge fusion strategy,

[0250] Design knowledge fusion strategy, handle heterogeneous knowledge source and potential conflict, including incremental knowledge fusion and knowledge conflict detection and resolution.

[0251] S4.3.1 Incremental knowledge fusion,

[0252] The new knowledge is integrated into the existing knowledge atlas in an incremental way, maintaining the continuity and stability of the knowledge base. Incremental knowledge fusion integrates new knowledge into existing knowledge atlas in an incremental way, and the formula is as follows:

[0253] ,

[0254] Where, represents the updated atlas, Fusion represents the fusion function, represents the set of verified new knowledge.

[0255] S4.3.2 Knowledge conflict detection and resolution,

[0256] Identify and resolve knowledge conflicts, handle contradictions between knowledge through quality assessment, expert arbitration, etc. The formula is as follows:

[0257] ,

[0258] Among them, Resolve represents the conflict resolution function, Quality the function evaluates the quality of knowledge, is a threshold parameter, and Merge represents the merge operation.

[0259] S4.4 Model adaptive optimization,

[0260] Continuously adjust model parameters and structure according to running feedback, including early warning effect record, parameter self-optimization and model structure optimization.

[0261] S4.4.1 Early warning effect record and evaluation,

[0262] The system records and evaluates the effect of each early warning, establishes a performance evaluation system for early warning, and provides a basis for model optimization. The formula is as follows:

[0263] ,

[0264] Among them, represents the early warning performance score at time t , represents the F1 score calculation function, TP , FP , FN , TN respectively represent the number of true failures, false failures, false failures and true failures. True failure is that the system predicts "will fail" and actually fails. False failure is that the system predicts "will fail" but actually does not fail.

[0265] S4.4.2 Parameter adaptive adjustment,

[0266] Based on the early warning effect, automatically adjust the model parameters, and use gradient optimization and other methods to realize the automatic optimization of parameters. The formula is as follows:

[0267] ,

[0268] Among them, represents the model parameter set at time t, η is the learning rate, represents the gradient of performance to parameters.

[0269] S4.4.3 Model structure optimization,

[0270] Dynamically adjust the model structure based on performance feedback to adapt to different task requirements and changes in system characteristics. The formula is as follows:

[0271] ,

[0272] in, Indicates time t The model structure, Evolution Represents a structural evolution function that adjusts model complexity and focus based on historical performance and task distribution.

[0273] Step S5: Fault warning verification and false alarm optimization:

[0274] This step designs a fault warning verification and false alarm optimization mechanism to solve the problems of high false alarm rate and low credibility in traditional warning systems. Through mechanisms such as multi-source verification, false alarm control and feedback learning, the accuracy and practicality of warning results are improved.

[0275] S5.1 Multi-source data cross-validation mechanism,

[0276] A multi-source data cross-validation mechanism is used to reduce the possibility of misjudgment caused by a single data source and improve the credibility of warnings through consistency verification of multiple data sources. The warning trigger threshold is set based on the cross-validation score. When the verification score exceeds the preset threshold, the system triggers a warning. The cross-validation score is defined as:

[0277] ,

[0278] in, Validation_Score Indicates the cross-validation score of the warning (0-1), Indicator i Indicates the i The result of the data source indication (0 or 1), It represents the weight of the i-th data source, reflecting the reliability of the data source.

[0279] S5.2 Early warning feedback learning mechanism,

[0280] Establish an early warning feedback learning mechanism to continuously improve the accuracy of early warnings through feedback, achieving self-optimization and improvement of the system. This mechanism can collect multi-dimensional feedback information on early warning effects and automatically adjust model parameters and early warning strategies based on the feedback results, forming a closed-loop optimization system.

[0281] S5.2.1 Feedback information collection,

[0282] The system has established a comprehensive feedback information collection mechanism, collecting feedback data on early warning effectiveness from multiple dimensions. Feedback information is mainly divided into three categories: early warning accuracy feedback, timeliness feedback, and expert evaluation feedback.

[0283] The early warning accuracy feedback evaluates the accuracy of the early warning by comparing the early warning results with the actual fault occurrence, including true positive (TP, correct early warning and fault occurrence), false positive (FP, false alarm), true negative (TN, correct non-early warning), and false negative (FN, missed alarm).

[0284] The timeliness feedback evaluates the timeliness of the early warning, analyzes the relationship between the early warning time and the actual fault occurrence time, and ensures that the early warning is issued within the effective time window.

[0285] The expert evaluation feedback collects the subjective evaluation of the field experts on the relevance of the early warning, severity evaluation, root cause analysis, and recommended measures.

[0286] The comprehensive score calculation formula of the feedback information is:

[0287] ,

[0288] Among them, is the weight coefficient, which satisfies ; represents the accuracy score (0-1); represents the timeliness score (0-1); represents the expert evaluation score (0-1).

[0289] The accuracy score is calculated based on the confusion matrix:

[0290] ,

[0291] Among them, TP represents the number of true positives; TN represents the number of true negatives; FP represents the number of false positives; FN represents the number of false negatives.

[0292] S5.2.2 Early warning strategy dynamic optimization,

[0293] According to the feedback results, the early warning strategy is optimized, including early warning level setting and early warning content customization optimization, to improve the practicality and effectiveness of the early warning.

[0294] The early warning level setting optimization dynamically adjusts the division standard of the early warning level according to the historical development law and actual impact degree of different fault types. The system analyzes the matching situation of the early warning level and the actual fault severity in historical fault cases, identifies the deviation in level setting and adjusts it.

[0295] The early warning content customization optimization dynamically adjusts the detail level and expression method of the early warning information according to the needs of different user roles. The system collects user feedback on the early warning content, analyzes which information is most helpful for decision-making, and optimizes the structure and content of the early warning report accordingly.

[0296] Embodiment 2

[0297] The embodiment provides an early fault warning system based on dynamic evolution of a complex industrial graph, comprising:

[0298] The data acquisition module is configured to:

[0299] A computer readable storage medium, wherein a plurality of instructions are stored, the instructions are suitable for being loaded by a processor of a terminal device and performing the early fault warning method based on dynamic evolution of a complex industrial graph.

[0300] A terminal device, comprising a processor and a computer readable storage medium, the processor is used for implementing instructions; the computer readable storage medium is used for storing a plurality of instructions, the instructions are suitable for being loaded by the processor and performing the early fault warning method based on dynamic evolution of a complex industrial graph.

[0301] The above are preferred embodiments of the present application, not limited to the protection scope of the present application, therefore: all equivalent changes made according to the structure, shape, principle of the present application should be covered in the protection scope of the present application.

Claims

1. An early fault warning method based on the dynamic evolution of complex industrial graphs, characterized by: include: Obtain industrial equipment parameters; Construct a multi-level industrial system knowledge graph based on the acquired industrial equipment parameters; Adaptive knowledge update of multi-level industrial system knowledge graph based on graph dynamic evolution mechanism; Build a joint reasoning framework based on large models and knowledge graphs; Optimize the joint reasoning framework based on adaptive knowledge updating and continuous learning mechanisms; Use the optimized joint reasoning framework for early warning of industrial equipment failures; The multi-level industrial system knowledge graph based on the acquired industrial equipment parameters is constructed, including defining the overall architecture of the graph, constructing a basic graph representing the physical structure of the industrial system and the relationship between equipment components through physical entity node components and parameter characteristics, forming a physical layer subgraph; constructing an abstract representation describing the system functional modules and their interactive relationships through the functional relationship of functional unit nodes, forming a functional layer subgraph; and establishing a network representation of monitoring parameters and their associations through monitoring node data and monitoring devices, forming a monitoring layer subgraph. The multi-level industrial system knowledge graph is represented as follows: ， in, Represents the physical layer subgraph, including the physical entities of devices and components and their connection relationships; Represents the functional layer subgraph, describing the functional units of the system and their interactions; Represents the monitoring layer subgraph, which contains sensor and monitoring point information; Respectively represent the cross-layer relationships between the physical layer and the functional layer, the functional layer and the monitoring layer, and the physical layer and the monitoring layer; The multi-level industrial system knowledge graph constructed based on the acquired industrial equipment parameters also includes establishing mapping relationships between the three-level graphs based on the physical layer, monitoring layer, and functional layer to ensure the integrity and consistency of knowledge representation. Specifically, the importance of physical components to functional realization is quantified through contribution analysis, and mapping weights between physical components and functional units are established; the ability of monitoring parameters to reflect functional status is determined through sensitivity and coverage analysis, and the association weights between functional status and monitoring parameters are established; finally, cross-layer constraints are used to ensure the consistency of overall semantics, which is expressed as: ， in, A measure of semantic consistency of the representation graph; Represents a set of paths that satisfy semantic consistency, i.e., a valid path from the physical entity through the functional unit to the monitoring point; Represents the set of all possible paths; The adaptive knowledge update of the multi-level industrial system knowledge graph based on the dynamic evolution mechanism of the graph includes establishing a mathematical representation of the graph's changes over time, expanding the static graph into a dynamic time series structure, and establishing a mathematical model of the graph evolution to describe the driving factors and change laws of the graph state changes; designing a dynamic update mechanism for the node and edge attributes in the graph so that the graph can reflect the system state changes in real time. Among them, a dynamic update model for the health status of physical components is established, which comprehensively considers monitoring data, historical status and external environmental factors; and an integral update model for the cumulative usage time attribute is established to accurately record the cumulative operating status of the equipment. For the cumulative usage time attribute, the integral update model is adopted, which is expressed as: ， in, Presentation Component i Deadline t The cumulative running time of Indicates that the component is at time running status.

2. The early fault warning method based on the dynamic evolution of complex industrial graphs according to claim 1 is characterized in that: The adaptive knowledge update of the multi-level industrial system knowledge graph based on the dynamic evolution mechanism of the graph also includes the design of a dynamic adjustment mechanism for the graph topology structure, including node addition and deletion, edge addition and deletion, and weight update, to adapt to changes in the system structure. The node addition and deletion operation is expressed as: ， Among them, edge addition and deletion handles the establishment and disconnection of connection relationships in the system by establishing a dynamic management mechanism for relationship edges, which is expressed as: ， in, Indicates time t The edge set of represents a new edge set, Indicates the removal of edge sets.

3. The early fault warning method based on the dynamic evolution of complex industrial graphs according to claim 2 is characterized in that: The joint reasoning framework based on the large model and knowledge graph is constructed. First, a three-layer enhancement strategy is designed to improve the industrial domain understanding and reasoning capabilities of the large model. The first layer is domain knowledge enhancement, which improves the industrial domain understanding capabilities of the large model by injecting professional knowledge. The efficient parameter fine-tuning technology is used to avoid the high computational cost of full parameter fine-tuning. The second layer is multimodal information fusion, which enhances the large model's ability to process multiple data types and achieves effective fusion of different modal information through multimodal encoders and attention mechanisms. The third layer is graph structure perception, which enhances the large model's ability to understand the knowledge graph structure. Graph structure encoding is performed through graph neural networks, and a cross-modal attention mechanism is established, which can be expressed as: ， in, For the l The node feature matrix of the layer, is the adjacency matrix with self-loops added, is the corresponding degree matrix, is the learnable weight matrix, is a non-linear activation function.

4. The early fault warning method based on the dynamic evolution of complex industrial graphs according to claim 3 is characterized in that: The joint reasoning framework based on the large model and knowledge graph is constructed, and further includes designing a three-level joint reasoning mechanism, combining graph reasoning, statistical reasoning, and large model reasoning to form a multi-level collaborative early warning mechanism. The first level is the graph reasoning layer, which performs fault propagation analysis and impact assessment based on the structure of the knowledge graph and uses graph structure information for logical reasoning; the second level is the statistical reasoning layer, which uses multivariate anomaly detection and degradation trend analysis methods based on monitoring data for anomaly detection and trend analysis; the third level reasoning mechanism is the large model reasoning layer, which uses a large language model for deep semantic understanding and complex reasoning, realizing multi-source information integration and chain thinking reasoning. Finally, a weighted decision fusion mechanism is designed to integrate the results of the three reasoning layers, expressed as: ， in, Indicates the final decision, 、 and Represent the decision results of the graph reasoning layer, statistical reasoning layer and large model reasoning layer respectively, is the weight coefficient, satisfying .

5. The early fault warning method based on the dynamic evolution of complex industrial graphs according to claim 4 is characterized in that: The optimization of the joint reasoning framework based on adaptive knowledge updating and continuous learning mechanisms includes establishing a multi-source knowledge acquisition framework, continuously collecting and integrating new knowledge from multiple channels, automatically extracting knowledge patterns from historical monitoring data, and using unsupervised learning methods to mine data regularities and implicit patterns; extracting empirical knowledge from historical failure cases, analyzing recorded failure cases and summarizing failure modes and solutions; and using large models to extract structured knowledge from technical documents and transform unstructured technical information into usable knowledge representations. The feedback from engineers and maintenance personnel is converted into structured knowledge, integrating human experience and system automatic learning results. Finally, a knowledge quality assessment mechanism is designed, including knowledge consistency assessment, reliability assessment and practicality assessment. Through this mechanism, the accuracy and practicality of the knowledge in the system are effectively ensured.

6. The early fault warning method based on the dynamic evolution of complex industrial graphs according to claim 5 is characterized in that: The optimized joint reasoning framework is used to perform early fault warning for industrial equipment, including the design of fault warning verification and false alarm optimization mechanisms to address the high false alarm rate and low credibility issues in traditional warning systems. The accuracy and practicality of warning results are improved through multi-source verification, false alarm control, and feedback learning mechanisms. A multi-source data cross-validation mechanism is used to reduce possible misjudgments caused by a single data source, and consistency checks on multiple data sources are used to improve warning credibility. The cross-validation score is defined as: ， in, Validation_Score Indicates the cross-validation score of the warning (0-1), Indicator i Indicates the i The indication results of the data source, It represents the weight of the i-th data source, reflecting the reliability of the data source.

7. An early fault warning system based on the dynamic evolution of complex industrial graphs, which implements the early fault warning method based on the dynamic evolution of complex industrial graphs as claimed in claim 1, characterized in that: include: The data acquisition module is configured to acquire industrial equipment parameters; The knowledge graph module is configured to construct a multi-level industrial system knowledge graph based on the acquired industrial equipment parameters; The updating module is configured to adaptively update the knowledge graph of the multi-level industrial system based on the dynamic evolution mechanism of the graph; The reasoning module is configured to build a joint reasoning framework based on the large model and the knowledge graph; The optimization module is configured to optimize the joint reasoning framework based on adaptive knowledge updating and continuous learning mechanisms; The early warning module is configured to use the optimized joint reasoning framework to provide early warning of industrial equipment failures.

Images