Equipment predictive maintenance system based on knowledge graph, construction method, equipment and medium

Abstract

The invention relates to the technical field of predictive maintenance, in particular to an equipment predictive maintenance system based on a knowledge graph, a construction method, equipment and a medium, and the method comprises a fault tree extraction module, a mode layer construction module, a data layer construction module and a fault tree filling module. The fault tree extraction module extracts a longitudinal connection relation of equipment fault types; the mode layer construction module is used for constructing and standardizing the relation among the equipment, the equipment fault type and the residual service life of the equipment; the data layer construction module is used for storing and updating instance data by taking the abnormal event as a center; and the fault tree filling module is used for realizing the updating of the longitudinal relation and the connection of the transverse relation of the equipment fault tree. And finally, comprehensive analysis, accurate prediction and efficient maintenance of equipment faults are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of predictive maintenance, and in particular to a knowledge graph-based equipment predictive maintenance system and construction method, equipment, and medium. Background Art

[0002] In the context of the Internet of Things, data has become a fuel for predictive maintenance. Trustworthy data can provide great support for the implementation of predictive maintenance. However, existing data is widely distributed and difficult to unify in format. Therefore, it is particularly important to manage data effectively.

[0003] Fault tree analysis is an analytical method that deduce equipment failures from top to bottom. In predictive maintenance, it is usually used to describe the vertical relationship from the undesirable equipment failure state at the bottom to the abnormal event of the top equipment. But in fact, a single equipment abnormality corresponds to multiple equipment failure modes, and there are causal and accompanying relationships between equipment failure modes; there are also causal and accompanying relationships between equipment failure modes corresponding to multiple equipment abnormalities. Therefore, the fault tree of vertical relationship is no longer suitable for equipment data management of predictive maintenance.

[0004] In addition, for a single device anomaly, each device failure mode has a remaining useful life (RUL); as time goes on, the value of the device life is in a state of dynamic change. On this basis, there is also a competitive relationship between the RULs of each device failure mode. In this case, a dual dynamic competition relationship of the remaining useful life of the equipment is formed. There is a question of how to select the final RUL value. Summary of the invention

[0005] The purpose of the present invention is to provide a knowledge graph-based equipment predictive maintenance system and construction method, equipment, and medium to solve the above-mentioned problems in the prior art.

[0006] The present invention is achieved through the following technical solutions: In a first aspect, the present invention provides a method for constructing a predictive maintenance system for equipment based on a knowledge graph, comprising: Extracting the vertical connection relationship of the equipment fault type, wherein the vertical connection relationship is the fault relationship evolution formed from bottom to top for a single equipment abnormality; Construct a relationship model between equipment, equipment failure types, and equipment remaining useful life; Centered on abnormal events, a mesh structure is formed with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and a single event knowledge graph is constructed through the mesh structure; Output the update of vertical relationships and the connection of horizontal relationships of the device fault tree.

[0007] Preferably, the construction of a single event knowledge graph includes: Obtain the contention of the equipment's remaining service life of a single fault type and the contention of the equipment's remaining service life between fault types.

[0008] Preferably, the competition for the remaining service life of the equipment of the single fault type itself includes: Establish a Bayesian model for the sub-fault type and predict the RUL of the fault type through the Bayesian model; Get the RUL of all fault types, generate the predicted value of the total fault type, extract samples from the posterior distribution of each sub-fault type, and generate the predicted distribution; Get the extracted samples, calculate the minimum RUL value among all sub-fault types, and obtain the overall RUL distribution of the fault type; The mean of the overall RUL distribution of the device is selected as the final RUL.

[0009] Preferably, the competition for remaining service life of equipment between the fault types includes: The risk function of the fault type is defined by the occurrence rate of the current fault type, and the survival function of the fault type is defined by the probability that the equipment does not have any faults before time t, and a function model of the survival function and the risk function is established; The minimum RUL is extracted from all the fault types of the equipment as the RUL distribution of the entire equipment.

[0010] Preferably, the function model includes:

[0011] In the formula, is the survival function, is the risk function, t is the time, and u is the type of fault.

[0012] Preferably, the updating of the vertical relationship and the connection of the horizontal relationship of the output device fault tree include: Determine new causal relationships between faults through fault trees; Setting a frequency threshold for simultaneous occurrence of faulty nodes, and finding a target faulty node, wherein the target faulty node is a faulty node whose frequency of simultaneous occurrence with other faulty nodes during the operation of the device reaches the frequency threshold and has no causal relationship with the faulty node that appears simultaneously; Get new causal relationships and add nodes to the fault tree in causal order; Add association lines or new nodes at the corresponding level of the fault tree to represent the horizontal relationship.

[0013] In a second aspect, the present invention also provides a system for building a predictive maintenance system for equipment based on a knowledge graph, comprising: A fault tree extraction module is configured to extract a vertical connection relationship of equipment fault types, wherein the vertical connection relationship is a bottom-up evolution of fault relationships oriented to a single equipment anomaly; A pattern layer building module, configured to build a relationship model between equipment, equipment failure type and equipment remaining useful life; The data layer construction module is configured to be centered on abnormal events, forming a mesh structure with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and constructing a single event knowledge graph through the mesh structure; A fault tree filling module is configured to output updates of vertical relationships and connections of horizontal relationships of a device fault tree; The main control module is connected to the fault tree extraction module, the pattern layer construction module, the data layer construction module and the fault tree filling module, and is used to execute the above-mentioned knowledge graph-based equipment predictive maintenance system construction method.

[0014] In a third aspect, the present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for constructing a knowledge graph-based equipment predictive maintenance system when executing the computer program.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned method for constructing a knowledge graph-based equipment predictive maintenance system.

[0016] The technical solution of the present invention has at least the following advantages and beneficial effects: The knowledge graph effectively integrates the relevant data of equipment predictive maintenance; specifically, the knowledge graph integrates the scattered equipment predictive maintenance related data and stores them in the same format. This solves the problem that the equipment predictive maintenance related data is widely distributed and difficult to unify in format.

[0017] The top-down model layer construction of the knowledge graph belongs to planned maintenance, and the bottom-up data layer construction belongs to situation-based maintenance. The combination of the systematic framework of planned maintenance and the real-time perception of situation-based maintenance enables equipment maintenance to develop from simple fault warning to accurate fault prediction, and can plan maintenance resources and schedule maintenance time in advance, and take effective measures before a fault occurs, thereby significantly improving equipment reliability and reducing maintenance costs.

[0018] The knowledge graph is centered on events, forming a network structure with equipment failure types and equipment service life as the main lines. The single-event knowledge graph is expanded to a multi-event knowledge graph, which expands the scope of equipment failure type prediction and equipment remaining service life prediction. It can also dig out the potential chain reactions and synergy rules between events in a complex system with multiple devices and multiple scenarios, providing data support for comprehensive assessment of system operation risks and optimization of overall maintenance strategies.

[0019] In the knowledge graph, the remaining service life of the equipment is in a dynamically updated state; the dual competitive relationship between the remaining service life of the equipment failure type itself and the remaining service life between equipment failures ensures the accurate assessment of the equipment's operating status and the advance planning of equipment maintenance activities.

[0020] Before the knowledge graph was constructed, the fault tree mainly came from planned maintenance, with a single fault type and relationship. After the knowledge graph was constructed, the association of multiple fault types was realized, the vertical relationship between fault types was supplemented, and the horizontal relationship between fault types was expanded, providing more complete data support for the formulation of maintenance decision plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is an overall flow chart of a method for constructing a knowledge graph-based equipment predictive maintenance system of the present invention; Figure 2 This is an example of a single event knowledge graph of the present invention; Figure 3 A flow chart of the competition relationship calculation of the remaining service life of the equipment of the single fault type itself of the present invention; Figure 4 A flow chart for calculating the dynamic competition relationship between fault types of the present invention; Figure 5 This is a flow chart of the fault tree filling module of the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0024] The modules or submodules described independently may be physically separated or not: they may be implemented by software or hardware, and some modules or submodules may be implemented by software, and the processor may call the software to implement the functions of these modules or submodules, and other modules or submodules may be implemented by hardware, such as by hardware circuits. In addition, some or all of the modules may be selected according to actual needs to achieve the purpose of the present application.

[0025] Please refer to Figure 1-Figure 5 The present invention provides a method for constructing a predictive maintenance system for equipment based on a knowledge graph, comprising: S101: extracting a vertical connection relationship of device fault types, where the vertical connection relationship is a bottom-up evolution of fault relationships for a single device anomaly; Among them, the vertical connection relationship of equipment failure types is extracted from planned maintenance data. Planned maintenance can plan maintenance activities in advance, but there is excessive maintenance and it is difficult to accurately predict the actual failure point.

[0026] The vertical relationship of equipment fault types can be extracted. For a single equipment anomaly, the equipment fault set is represented as ,in Indicates different fault types. If the device has a fault type and fault type , , , and the fault type Is the fault type The reason for this can be expressed by the following relationship: ; In order to intuitively determine whether there is a causal relationship between fault types, a function can be defined To make a judgment; when the fault type Is the fault type There is a correlation hour, ; When the fault type Is the fault type No correlation exists hour, ; S102: Construct a relationship model between equipment, equipment failure type and equipment remaining service life; Specifically, it includes the construction of data quality ontology and the visualization of knowledge graph. The construction of data quality ontology can rely on the modification of existing ontology or the creation of new data quality ontology; the visualization of knowledge graph is realized by using Protégé.

[0027] S103: With the abnormal event as the center, a mesh structure is formed with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and a single event knowledge graph is constructed through the mesh structure; It takes equipment abnormal events as the center, forming a mesh structure with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes.

[0028] Specifically expressed as: ,in Indicates abnormal device events. Indicates the type of device failure. Indicates the location of the device. Indicates the function of the device. Indicates the remaining service life of the device. is a mesh structure set. The relationship between them is expressed as: , It should be noted that the event knowledge graph constructed here is a single event knowledge graph.

[0029] Specifically, an abnormal state of a device (indicator light on) will display the working location of the device and the function of the device. At the same time, an abnormal state of a device (indicator light on) may indicate that the device contains multiple fault types. , Fault type , ..., fault type , ; These types of failures directly affect the remaining service life of the equipment. In addition, equipment failure types It can also cause equipment failure types occurs.

[0030] S104: Output the update of the vertical relationship and the connection of the horizontal relationship of the device fault tree.

[0031] Among them, the horizontal connection of the fault tree discovers the potential association between different equipment failure types and integrates the influencing factors of different entities on equipment failure types, which are derived from the condition-based maintenance data; the condition-based maintenance is carried out according to the actual operation status of the equipment, which can reduce the situation of poor maintenance and excessive maintenance. The combination of planned maintenance and condition-based maintenance can effectively identify the equipment failure type and provide effective support for equipment predictive maintenance.

[0032] In an exemplary embodiment of the present invention, constructing a single event knowledge graph includes: Obtain the contention of the equipment's remaining service life of a single fault type and the contention of the equipment's remaining service life between fault types.

[0033] Specifically, in the single event knowledge graph, an abnormal device event usually leads to multiple types of device failures. For a single failure type, affected by the actual operation of the device, the remaining service life of the device is in a dynamic competitive state; for example, At this moment, the remaining service life of the equipment is ;exist At this moment, the equipment service life is ,like Greater than ,but Will replace The remaining service life of the equipment is finally selected. For multiple fault types, multiple remaining service lives of equipment will be generated, and these remaining service lives are also in a dynamic competition relationship, that is, a dynamic competition relationship among multiple fault types is formed.

[0034] Specifically, the competition for the remaining service life of equipment of the single fault type itself includes: S201: Establishing a Bayesian model of the sub-fault type, and predicting the RUL of the fault type through the Bayesian model; A device exception type There may be multiple sub-fault types, denoted as , , ..., , For each sub-fault type, a remaining service life prediction model is established. , establish independent Bayesian models to predict their RUL. The posterior distribution of each model is recorded as ;in, Model parameters representing sub-fault types, Indicates at a point in time The data collected is related to the sub-fault type.

[0035] S202: Obtain the RUL of all fault types, generate a predicted value of the total fault type, extract samples from the posterior distribution of each sub-fault type, and generate a predicted distribution; From the posterior distribution of each sub-fault type Draw samples from the dataset to generate a predicted distribution ;in, Indicates the fault type of the slave The first sample drawn from the posterior distribution of sample values, which can be used to approximate the sub-fault type At the point in time The parameter distribution of the fault type is obtained through subsequent calculation and analysis based on these sample values ​​to obtain the predicted value of the total fault type.

[0036] S203: Obtain the extracted samples, calculate the minimum RUL value of all sub-fault types, and obtain the overall RUL distribution of the fault type; For each sample value, calculate the minimum RUL value among all sub-fault types. This way, the fault type Overall RUL distribution: ,in, express Type of fault at time point No. sample values. At a specific point in time Under this condition, the minimum remaining service life value among the sampling values ​​corresponding to various sub-fault types is obtained, thus reflecting the fault type. The lower limit of the overall remaining useful life at that point in time. This collection of minimum RUL values ​​constitutes the fault type The overall RUL distribution provides a basis for further analysis of the overall status of fault types.

[0037] By calculating the minimum RUL value, the fault type can be obtained Overall RUL distribution ,in For the Type of fault at time point No. A minimum RUL value.

[0038] S204: Select the mean of the overall RUL distribution of the device as the final RUL.

[0039] It should be noted that as time goes by, the operating status of the equipment changes in real time, and the remaining service life of the equipment is also in dynamic change. The data length for storing the final RUL value is , the updated RUL value will be at the rightmost end of the data and overwrite the leftmost RUL value.

[0040] In an exemplary embodiment of the present invention, the competition for remaining service life of equipment between fault types includes: S301: defining a risk function of the fault type by the occurrence rate of the current fault type, defining a survival function of the fault type by the probability that no fault occurs in the device before time t, and establishing a function model of the survival function and the risk function; In order to consider the RUL competition between different fault types, the competitive risk model is used to describe the impact of multiple competing risk events on the service life of the equipment. In the competitive risk, the risk function of each fault type can be used to describe its occurrence rate. Assume that The risk function of the failure type is , the total risk function is the sum of the risk functions of all fault types: . The device is at time The survival function (SurvivalFunction) represents the device at time The probability that no failure occurred before is denoted as The relationship between the survival function and the hazard function is: , The fault type.

[0041] S302: Extract the minimum RUL from all the fault types of the equipment respectively as the RUL distribution of the whole equipment.

[0042] Equipment failure type The minimum RUL is extracted from each of them to form the RUL distribution of the entire device, which is expressed as ,in, Indicates the type of device failure The corresponding equipment remaining life minimum value. Extract the minimum RUL value as the final RUL value after competition, expressed as .

[0043] In a single event knowledge graph, the relationship between faults is simple; in reality, there are causal, accompanying and other relationships between equipment failures; a single event knowledge graph is difficult to cope with the dynamic and complex equipment operation status; therefore, single events are merged to form a multi-domain knowledge graph, forming a more comprehensive and complete knowledge graph, that is, realizing fault tree filling.

[0044] The multi-domain knowledge graph is represented as ,in Represents the links between multiple single event knowledge graphs. Through these links, the information in different single events can be integrated to form a more comprehensive and complete knowledge graph to better cope with dynamic and complex equipment operation status.

[0045] In an exemplary embodiment of the present invention, updating the vertical relationship and connecting the horizontal relationship of the output device fault tree includes: S401: Determine new causal relationships between faults through the fault tree; Determine new causal relationships between faults through decision trees. By training a decision tree model, the states of different components are used as features and equipment failures are used as target variables. In the decision tree, the abnormal state of component A is used as an important branch condition and eventually leads to the failure of component B. Then it can be inferred that the abnormal state of component A is a potential cause of the failure of component B.

[0046] S402: Setting a frequency threshold of simultaneous occurrence of faulty nodes, and searching for a target faulty node, wherein the target faulty node is a faulty node that has a frequency reaching the frequency threshold of simultaneous occurrence of other faulty nodes during the operation of the device and has no causal relationship with the faulty node that appears simultaneously; Looking at the horizontal relationship, we look for fault nodes that have no obvious cause and effect but often appear at the same time during the operation of the equipment. For example, we find that fault nodes C and D have a high frequency of abnormalities in multiple equipment operation records. This may be due to common external factors (such as power supply fluctuations, ambient temperature changes, etc.), or they have interdependent parts in the functional structure of the equipment.

[0047] S403: Obtain new causal relationships and add nodes to the fault tree in causal order; For newly discovered causal relationships, add nodes to the fault tree in causal order. If it is found in the knowledge graph that fault E is caused by both fault F and fault G, but there is no such relationship in the original fault tree, then find the appropriate position in the fault tree (such as the position corresponding to fault E) and add fault F and fault G as new cause branches of fault E.

[0048] S404: Adding association lines or new nodes at corresponding levels of the fault tree to represent the horizontal relationship.

[0049] For horizontal relationships, you can add association lines or new nodes at the corresponding level of the fault tree to represent this relationship. For example, if it is found that fault H and fault I are horizontally related, you can add a virtual node (such as "related fault group") at their corresponding level in the fault tree, and connect fault H and fault I to this virtual node to reflect the horizontal relationship between them.

[0050] In a second aspect, the present invention also provides a system for building a predictive maintenance system for equipment based on a knowledge graph, comprising: A fault tree extraction module is configured to extract a vertical connection relationship of equipment fault types, wherein the vertical connection relationship is a bottom-up evolution of fault relationships oriented to a single equipment anomaly; A pattern layer building module, configured to build a relationship model between equipment, equipment failure type and equipment remaining useful life; The data layer construction module is configured to be centered on abnormal events, forming a mesh structure with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and constructing a single event knowledge graph through the mesh structure; A fault tree filling module is configured to output updates of vertical relationships and connections of horizontal relationships of a device fault tree; The main control module is connected to the fault tree extraction module, the pattern layer construction module, the data layer construction module and the fault tree filling module, and is used to execute the above-mentioned knowledge graph-based equipment predictive maintenance system construction method.

[0051] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0052] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. The computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0053] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing a predictive maintenance system for equipment based on knowledge graph, characterized in that: include: Extracting the vertical connection relationship of the equipment fault type, wherein the vertical connection relationship is the fault relationship evolution formed from bottom to top for a single equipment abnormality; Construct a relationship model between equipment, equipment failure types, and equipment remaining useful life; Centered on abnormal events, a mesh structure is formed with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and a single event knowledge graph is constructed through the mesh structure; Output the update of vertical relationships and the connection of horizontal relationships of the device fault tree.

2. The method for constructing a predictive maintenance system for equipment based on a knowledge graph according to claim 1, characterized in that: The construction of the single event knowledge graph includes: Obtain the contention of the equipment's remaining service life of a single fault type and the contention of the equipment's remaining service life between fault types.

3. The method for constructing a predictive maintenance system for equipment based on a knowledge graph according to claim 2 is characterized in that: The competition for the remaining service life of the equipment of the single fault type itself includes: Establish a Bayesian model for the sub-fault type and predict the RUL of the fault type through the Bayesian model; Get the RUL of all fault types, generate the predicted value of the total fault type, extract samples from the posterior distribution of each sub-fault type, and generate the predicted distribution; Get the extracted samples, calculate the minimum RUL value among all sub-fault types, and obtain the overall RUL distribution of the fault type; The mean of the overall RUL distribution of the device is selected as the final RUL.

4. The method for constructing a predictive maintenance system for equipment based on a knowledge graph according to claim 3 is characterized in that: Competition between the types of faults for equipment remaining useful life includes: The risk function of the fault type is defined by the occurrence rate of the current fault type, and the survival function of the fault type is defined by the probability that the equipment does not have any faults before time t, and a function model of the survival function and the risk function is established; The minimum RUL is extracted from all the fault types of the equipment as the RUL distribution of the entire equipment.

5. The method for constructing a predictive maintenance system for equipment based on a knowledge graph according to claim 4 is characterized in that: The function model includes: In the formula, is the survival function, is the risk function, t is the time, and u is the type of fault.

6. The method for constructing a predictive maintenance system for equipment based on a knowledge graph according to claim 5 is characterized in that: The updating of the vertical relationship and the connection of the horizontal relationship of the output device fault tree include: Determine new causal relationships between faults through fault trees; Setting a frequency threshold for simultaneous occurrence of faulty nodes, and finding a target faulty node, wherein the target faulty node is a faulty node whose frequency of simultaneous occurrence with other faulty nodes during the operation of the device reaches the frequency threshold and has no causal relationship with the faulty node that appears simultaneously; Get new causal relationships and add nodes to the fault tree in causal order; Add association lines or new nodes at the corresponding level of the fault tree to represent the horizontal relationship.

7. A system for building a predictive maintenance system for equipment based on knowledge graph, characterized in that: include: A fault tree extraction module is configured to extract a vertical connection relationship of equipment fault types, wherein the vertical connection relationship is a bottom-up evolution of fault relationships oriented to a single equipment anomaly; A pattern layer building module, configured to build a relationship model between equipment, equipment failure type and equipment remaining useful life; The data layer construction module is configured to be centered on abnormal events, forming a mesh structure with equipment failure type, equipment remaining service life, equipment location, and equipment function as key nodes, and constructing a single event knowledge graph through the mesh structure; A fault tree filling module is configured to output updates of vertical relationships and connections of horizontal relationships of a device fault tree; A main control module, connected to the fault tree extraction module, the pattern layer construction module, the data layer construction module and the fault tree filling module, is used to execute the equipment predictive maintenance system construction method based on the knowledge graph as described in any one of claims 1-6.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for constructing a equipment predictive maintenance system based on a knowledge graph as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for constructing a knowledge graph-based equipment predictive maintenance system as described in any one of claims 1 to 6 is implemented.

Images