Device fault diagnosis model establishment method and apparatus

Abstract

A device fault diagnosis model establishment method and apparatus. The method comprises: constructing a fault diagnosis knowledge graph for devices; using same as a device fault diagnosis model in a fault diagnosis process; determining a fault according to a symptom of a device; and then determining a root cause of the fault. Since the knowledge graph describes the hierarchical structure of the devices and root cause-type nodes describing root causes of faults, a Bayesian network (BN) formed by fault-type nodes describing the faults and symptom-type nodes describing symptoms is embedded in the knowledge graph and is associated with device-type nodes and the root cause-type nodes, so as to perform two-stage diagnosis by means of the knowledge graph, that is, the BN is first used for diagnosing the symptom of the device and then the fault thereof, and then the knowledge graph is used for inference, such that the root cause is obtained by means of the fault and a fault-associated device structure-type node, and the device structure-type node such as a part-type node is positioned according to the root cause. The present application can comprehensively and objectively perform device fault diagnosis, and accurately determine faults and root causes of the faults.

Description

A method and device for establishing an equipment fault diagnosis model Technical Field

[0001] The present application relates to equipment fault diagnosis technology in the industrial field, and in particular to a method and device for establishing an equipment fault diagnosis model. Background Art

[0002] Industrial systems consist of multiple devices of varying types, working together to complete a specific production task. A failure in any device within an industrial system can halt production, hindering normal operations. Therefore, when a device within an industrial system fails, quickly diagnosing the fault based on its symptoms and effectively restoring the device to normal operation to minimize the duration of the failure are crucial to maintaining normal operation.

[0003] Currently, equipment fault diagnosis can be performed by combining equipment operating data with expert knowledge. However, expert knowledge cannot cover all equipment faults and their symptoms, and may be subjective, making it difficult to accurately diagnose equipment faults and their symptoms.

[0004] Therefore, how to comprehensively and objectively diagnose equipment faults, accurately determine the faults and their symptoms, and further determine the root cause of the faults is a technical problem that needs to be solved urgently.

[0005] Summary of the Invention

[0006] In view of this, an embodiment of the present application proposes, on one hand, a method for establishing a device fault diagnosis model, which can comprehensively and objectively diagnose device faults based on the established model, accurately determine the fault and its symptoms, and the root cause of the fault.

[0007] On the other hand, an embodiment of the present application proposes a device for establishing a device fault diagnosis model, which can comprehensively and objectively diagnose device faults based on the established model, accurately determine the fault and its symptoms, and the root cause of the fault.

[0008] In one embodiment of the present application, a method for establishing a device fault diagnosis model is provided, the method comprising:

[0009] A knowledge graph for fault diagnosis of a device is set, wherein the knowledge graph includes a device class node describing the device, a Bayesian network (BN), and a root cause class node describing the root cause of the fault, wherein the BN is a two-layer structure network formed by the fault class node describing the fault and the symptom class node describing the symptom, the fault class node in the BN is associated with the device class node and the root cause class node, and the root cause class node is associated with the device structure class node based on a preset diagnostic approach;

[0010] The knowledge graph is used as a device fault diagnosis model in the fault diagnosis of the device, and the fault is determined according to the symptoms of the device, and the root cause of the fault is determined.

[0011] In the above method, determining a fault based on a symptom of the device and determining a root cause of the fault includes:

[0012] The BN in the knowledge graph is used to diagnose the equipment from symptoms to faults, and then the knowledge graph is used for reasoning to derive the root cause from the fault described by the fault class node and the equipment structure class node related to the fault, and the root cause is located on the equipment structure class node based on the root cause.

[0013] In the above method, the knowledge graph describes a hierarchical structure of devices, and the knowledge graph also includes a system class node that describes the system to which the device belongs. The device class node is associated with the system class node, and the knowledge graph describes a connection edge with affiliation attributes between the device class node and the system class node;

[0014] The knowledge graph also includes a component class node that describes a device component, which is associated with the device class node, and a connection edge described in the knowledge graph as having affiliation attributes between the component class node and the device class node.

[0015] In the above method, the knowledge graph also includes:

[0016] A PV-type node describing a process variable PV and a physical principle-type node describing a physical principle, wherein the relationship between the PV-type node and the physical principle-type node is determined, and the knowledge graph is described as a connecting edge with determined attributes between the PV-type node and the physical principle-type node;

[0017] There is an application relationship between the PV node and the symptom node, which is described in the knowledge graph as a connection edge with application attributes between the PV node and the symptom node;

[0018] There is an application relationship between the physical principle node and the symptom node, which is described in the knowledge graph as a connection edge having application attributes between the physical principle node and the symptom node;

[0019] The process variables described by the process variable class nodes are used to calculate the symptoms described by the symptom class nodes, and in the process of calculating the process variables, the physical principles described by the physical principle class nodes are applied.

[0020] In the above method, the knowledge graph also includes:

[0021] A sensor node describing a sensor has a monitored relationship with the PV node, and is described in the knowledge graph as a connection edge having a monitored attribute between the sensor node and the PV node;

[0022] The PV described by the PV class node is monitored by the sensor described by the sensor class node.

[0023] In the above method, the BN included in the knowledge graph is a trained BN:

[0024] Preset the initial BN based on expert knowledge of the equipment;

[0025] The running data of the equipment is used as training data for training to obtain the trained BN.

[0026] In the above method, the root cause class node also has a connection edge with the root cause association attribute between the symptom class node in the BN, and the association between the root cause class node and the equipment structure class node is described as a connection edge with the association attribute in the knowledge graph.

[0027] In the above method, the device structure class node includes:

[0028] Component class nodes that describe device components, controller class nodes that describe controllers, maintenance method class nodes that describe maintenance methods, and troubleshooting item class nodes that describe troubleshooting items;

[0029] The root cause node is associated with the device structure node based on a preset diagnostic approach, including:

[0030] When associated with the component class node, deriving the root cause of the component failure described by the component class node;

[0031] When associated with the controller class node, obtaining the root cause of the controller fault described by the controller class node;

[0032] When associated with the maintenance method class node, the maintenance method described in the maintenance method class is executed to obtain the root cause of the fault;

[0033] When associated with the troubleshooting item class node, the troubleshooting items described in the troubleshooting item class node are executed to obtain the root cause of the fault.

[0034] In another embodiment of the present application, a device for establishing a device fault diagnosis model includes: an establishing unit and an applying unit, wherein:

[0035] The establishing unit is used to set a knowledge graph for fault diagnosis of the device, wherein the knowledge graph includes a device class node describing the device, a Bayesian network (BN), and a root cause class node describing the root cause of the fault, wherein the BN is a two-layer structure network formed by the fault class nodes describing the fault and the symptom class nodes describing the symptom, the fault class nodes in the BN are associated with the device class nodes and the root cause class nodes, and the root cause class nodes are associated with the device structure class nodes based on a preset diagnostic approach;

[0036] The application unit is used to use the knowledge graph as a device fault diagnosis model in device fault diagnosis, determine the fault based on the symptoms of the device, and determine the root cause of the fault.

[0037] In yet another embodiment of the present application, an electronic device is provided, including:

[0038] processor;

[0039] A memory stores a program, wherein the program is configured to implement any of the above methods for establishing a device fault diagnosis model when executed by the processor.

[0040] In an embodiment of the present application, the present application constructs a knowledge graph for fault diagnosis for the equipment, which is used as an equipment fault diagnosis model in the fault diagnosis process to determine the fault based on the symptoms of the equipment and then determine the root cause of the fault. Since the hierarchical structure of the equipment and the root cause class node describing the root cause of the fault are described in the knowledge graph, a Bayesian network (BN) formed by the fault class node describing the fault and the symptom class node describing the symptom is embedded therein, associated with the equipment class node and the root cause class node, and a two-level diagnosis is performed through the knowledge graph, that is, first using BN to diagnose from the symptoms of the equipment to the fault, and then using the knowledge graph reasoning to derive the root cause through the fault and the equipment structure class node associated with the fault, and locate the equipment structure class node such as parts according to the root cause. In this way, the embodiment of the present application can comprehensively and objectively diagnose the equipment and accurately determine the fault and the root cause of the fault. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The preferred embodiments of the present application will be described in detail below with reference to the accompanying drawings, so that those skilled in the art can better understand the above and other features and advantages of the present application. In the accompanying drawings:

[0042] FIG1 is a schematic structural diagram of an existing initial BN provided in an embodiment of the present application;

[0043] FIG2 is a schematic diagram of the structure of a conventional trained BN according to an embodiment of the present application;

[0044] FIG3 is a schematic diagram of a conventional process for establishing a complex structure BN according to an embodiment of the present application;

[0045] FIG4 is a flow chart of a method for establishing a device fault diagnosis model provided by an embodiment of the present application;

[0046] FIG5 is a schematic diagram of the structure of a device fault diagnosis model provided in an embodiment of the present application;

[0047] FIG6 is a schematic diagram of the structure of a trained BN according to an embodiment of the present application;

[0048] FIG7 is a schematic diagram of a hierarchical structure of a device provided in an embodiment of the present application;

[0049] FIG8 is a schematic diagram of a specific example structure of a device fault diagnosis model provided in an embodiment of the present application;

[0050] FIG9 is a flowchart of equipment fault diagnosis using a knowledge graph according to an embodiment of the present application;

[0051] FIG10 is a schematic diagram of the structure of a device for establishing a device fault diagnosis model provided in this embodiment.

[0052] The accompanying drawings are numerals as follows: DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of this application clearer, the following embodiments are given to further illustrate this application in detail.

[0054] Currently, when diagnosing equipment faults in industrial systems, a combination of equipment operating data and expert knowledge is typically used. There are two approaches to this, each described below.

[0055] The first method is to use BN to diagnose equipment faults

[0056] First, an initial BN is constructed based on expert knowledge. The BN has a two-tier structure: one tier consists of fault-class nodes describing faults, and the other tier consists of symptom-class nodes describing symptoms. Directed edges between fault-class and symptom-class nodes describe the causal relationship between the fault and symptom classes—that is, the fault described by the fault-class node causes the symptom described by the symptom-class node. Second, the initial BN is trained using the operational data of the industrial system as training data, yielding a trained BN. Finally, when diagnosing equipment faults, the BN is used to determine the device fault based on its symptoms.

[0057] Figure 1 is a structural diagram of the existing initial BN provided in an embodiment of the present application. As shown in the figure, the initial BN adopts a two-layer structure. The different fault class nodes in the first and third layers in the figure describe different types of faults, respectively. The different symptom class nodes in the second and fourth layers in the figure describe different symptoms, respectively. The connecting edges between the fault class nodes and the symptom class nodes describe the causal relationship between the fault and the symptoms. In the figure, the fault nodes are represented by gray dots, and the symptom nodes are represented by white dots.

[0058] The initial BN is trained, and the operating data of the equipment is used as training data to train the initial BN, and the structural diagram of the trained BN is obtained through training as shown in Figure 2. Figure 2 is a structural diagram of the existing trained BN provided in an embodiment of the present application. During the initial BN training, the association probability parameters of the connection edges between the fault class nodes and the symptom class nodes can also be calculated. In the figure, different linear connection edges between the nodes are used to describe the connection edges with different association probability parameters. Since the symptoms described by some of the symptom class nodes are determined based on the operating data obtained by sensor monitoring, the faults described by the fault class nodes associated with these symptom class nodes and the correlation between the symptoms described by these symptom class nodes may be inaccurate or unexplainable, which are described as dotted line connections in the figure, indicating that the association probability parameters are set to be relatively small.

[0059] A trained BN can identify faults based on device symptoms. However, because the BN lacks information about the cause of the fault and its relationship to other nodes, it cannot identify the root cause, especially the underlying cause. Consequently, it cannot diagnose the cause of the fault.

[0060] The second method is to train a complex structure BN based on the knowledge graph of the industrial system to perform fault diagnosis.

[0061] As shown in Figure 3, Figure 3 is a schematic diagram of the existing process of establishing a complex structure BN provided by an embodiment of the present application. In the upper half of the accompanying figure, a knowledge graph based on failure mode and effects analysis (FMEA) is generated based on the expert knowledge of the equipment. The knowledge graph includes the association relationship between multiple devices (described as the first device 30 and the second device 31 in the figure), the association relationship between the device and the fault tree 32, and the association relationship between the sensor 33 and the fault tree 32. The fault tree 32 is described in the figure with a virtual box, including symptom nodes that describe symptoms, fault nodes that describe faults, and cause nodes that describe the causes of faults. After the symptom nodes are associated with the fault nodes, the fault nodes are associated with the cause nodes. The knowledge graph also includes a maintenance method node 34 that describes the micro-repair method of the fault, which is associated with the cause node.

[0062] Based on the knowledge graph shown in Figure 3, a BN is constructed, as shown in the lower half of Figure 3, to form a BN with a multi-layer structure. For the connection edges between different nodes in the BN, its association probability parameters are also calculated to reflect the association strength between different nodes.

[0063] The constructed BN shown in FIG3 is applied to implement equipment fault diagnosis.

[0064] When diagnosing equipment faults using this method, due to the complex multi-layered structure of the BN, the cause of the fault can be determined and the corresponding fault handling method can be implemented based on the fault cause. However, since the knowledge graph is based on expert knowledge, the subsequent BN is also based on expert knowledge. This lacks objectivity when diagnosing equipment faults. In addition, it is not feasible to use the equipment operating data to train the BN shown in Figure 3 to enhance its objectivity in equipment fault diagnosis. This is because, on the one hand, due to the complex structure of the BN, the amount of computation required for training will increase exponentially during the training process, making training time-consuming and labor-intensive. On the other hand, in the knowledge graph used to generate the BN, for the part-level fault-related nodes, there is a lack of sensor-related nodes deployed on-site, so it is impossible to monitor the operating data related to this level for BN training.

[0065] It can be seen that no matter which of the above methods is adopted, BN is used to implement equipment fault diagnosis, which will bring many problems: the limited depth of sensor deployment leads to a small number of BN levels, which can only diagnose the component level and cannot perform in-depth diagnosis of faults at deeper levels; if there are too many BN levels, the model training will be time-consuming.

[0066] To address the above issues, the device constructs a knowledge graph for fault diagnosis as a device fault diagnosis model. This graph is used during the fault diagnosis process to determine the fault based on the device's symptoms and then determine the root cause of the fault. Since the knowledge graph describes the device's hierarchical structure and the root cause node describing the root cause of the fault, a BN formed by the fault node describing the fault and the symptom node describing the symptom is embedded in it and associated with the device node and the root cause node. This allows for two-level diagnosis using the knowledge graph: first, the BN is used to diagnose the device from symptoms to the fault. Then, using the knowledge graph for reasoning, the root cause is derived from the fault and the device structure node associated with the fault. The root cause is then located on the device structure node, such as a component.

[0067] In this way, the embodiments of the present application can comprehensively and objectively diagnose equipment faults and accurately determine the faults and their root causes.

[0068] FIG4 is a flow chart of a method for establishing a device fault diagnosis model provided in an embodiment of the present application, wherein the specific steps include:

[0069] Step 401: Setting a knowledge graph for device fault diagnosis. The knowledge graph includes a device class node describing the device, a Bayesian network (BN), and a root cause class node describing the root cause of the fault. The BN is a two-layer structure network formed by the fault class nodes describing the fault and the symptom class nodes describing the symptom. The fault class nodes in the BN are associated with the device class nodes and the root cause class nodes. The root cause class nodes are associated with the device structure class nodes based on a preset diagnostic approach.

[0070] Step 402: Use the knowledge graph as a device fault diagnosis model in device fault diagnosis to determine the fault based on the device symptoms and the root cause of the fault.

[0071] In the above method, determining a fault based on a symptom of the device and determining a root cause of the fault includes:

[0072] The BN in the knowledge graph is used to diagnose the equipment from symptoms to faults, and then the knowledge graph is used for reasoning to derive the root cause from the fault described by the fault class node and the equipment structure class node related to the fault, and the root cause is located on the equipment structure class node based on the root cause.

[0073] In conjunction with Figure 5, the method for establishing the equipment fault diagnosis model provided by the embodiment of the present application is described in detail. Figure 5 is a structural diagram of the equipment fault diagnosis model provided by the embodiment of the present application. The connection edges between the nodes in the knowledge graph are all directional, representing the association relationship between the two nodes connected by the connection edge. In the upper half of Figure 5, an ontology model in the knowledge graph set for equipment fault diagnosis is shown. Among them, a device class node 510 is included, which is associated with a system class node 511 that describes an industrial system. The knowledge graph describes a connection edge 10 with affiliation attributes between the device class node 510 and the system class node 511. The ontology model also includes a two-layer structure BN 5. BN5 is a two-layer structure network formed by a fault class node 512 and a symptom class node 513. A connection edge 11 with causal attributes is set between the fault class node 512 and the symptom class node 513, and a connection edge 13 with association attributes is set between the fault class node 512 and the symptom class node 513. The ontology model also includes a root cause node 514, and a connection edge 12 with a root cause attribute is set between the root cause node 514 and the symptom node 513 and the fault node 512, respectively, indicating that the described root cause is the root cause of the corresponding symptom and the corresponding fault.

[0074] The lower half of Figure 5 describes the structure of BN. BN adopts a two-layer structure network. The multiple nodes in the upper layer are fault class nodes 512 that describe different faults, and the multiple nodes in the lower layer are symptom class nodes 513 that describe different symptoms. There is a causal connection edge 11 between the two, indicating that there is a causal relationship between the fault and the symptom.

[0075] The connection edge mentioned in the knowledge graph of the embodiment of the present application actually describes the association relationship between two nodes. The specific association relationship is identified by attribute values, and this connection edge is directional.

[0076] Here, BN can be used as a sub-model for causal analysis from fault to symptom. It is initially set based on expert knowledge and can subsequently be trained using the equipment's operating data to determine the probability parameter value of the connection edge 11 with cause attributes between the fault-type node 512 and the symptom-type node 513. The probability parameter value reflects the strength of the causal relationship between the fault-type node 512 and the symptom-type node 513.

[0077] Specifically, in the embodiment of the present application, the BN included in the knowledge graph is a trained BN, and the process of training the BN is:

[0078] Preset the initial BN based on expert knowledge of the equipment;

[0079] The operating data of the equipment is used as training data and input into the initial BN for training to obtain the trained BN.

[0080] Here, as shown in FIG6 , FIG6 is a schematic diagram of the structure of the trained BN provided in the embodiment of the present application. The trained BN is a BN model obtained by training a BN established based on expert knowledge according to the operating data of the equipment. The structure of the BN provided in the embodiment of the present application is two-layer, that is, there are only two types of nodes, the fault class nodes in the first layer as shown in FIG6 , and the symptom class nodes in the second layer. The fault class node describes the fault occurring at the equipment level, and the connection edge between the fault class node and the symptom class node describes the causal relationship between the fault and the symptom, which is described by the connection edge 11 with causal attributes, and the probability parameter value representing the causal relationship is obtained by training. For example, when the probability of no fault described by a fault class node is 0.86%, the corresponding symptom omission probability is 99.14%, which indicates the degree of causal relationship between the fault class node and the symptom class node. The probability parameter value of the fault class node is shown in a list in the figure.

[0081] As can be seen from FIG5 , the knowledge graph also includes a system class node 511 describing an industrial system, and the device class node 510 is subordinate to the system class node 511 , which is described in the knowledge graph as a connection edge 10 having a subordination attribute between the system class node 511 and the device node 510 ;

[0082] The knowledge graph also includes: a component class node 515 describing a device component, associated with the device class node 510, and a connection edge 10 described in the knowledge graph as having affiliation attributes between the component class node 515 and the device class node.

[0083] In the embodiment of the present application, the knowledge graph also includes:

[0084] A PV node 516 describing a process variable (PV) and a physical principle node 517 describing a physical principle, wherein the PV node 516 and the physical principle node 517 have a determined relationship, which is described in the knowledge graph as a connection edge 22 having a determined attribute between the PV node 516 and the physical principle node 517; and an application relationship between the PV node 516 and the symptom node 513, which is described in the knowledge graph as a connection edge 15 having an application attribute between the PV node 516 and the symptom node 513.

[0085] There is an application relationship between the physical principle node 517 and the symptom node 513, which is described in the knowledge graph as a connection edge 15 with application attributes between the physical principle node 517 and the symptom node 513;

[0086] The process variables described by the process variable node 516 are used to calculate the symptoms described by the symptom node 513 , and the physical principles described by the physical principle node 517 are applied during the process of calculating the process variables.

[0087] When implementing the symptom calculation described by symptom node 513, the knowledge graph also includes: a sensor node 518 describing a sensor, a connection edge 10 with the system node 511 having a membership attribute, and a monitored relationship between the PV node 516. The knowledge graph describes a connection edge 16 with the monitored attribute between the sensor node 518 and the PV node 516, indicating that the PV described by the PV node 516 is obtained by monitoring the sensor described by the sensor node 518. The physical principle described by the physical principle node 517 describes the calculation method used for the PV calculation, such as the mathematical function used for the PV calculation. The PV node 516 and the physical principle node 517 serve as the context for the symptom node 513, making the symptoms described by the symptom node 513 easier to understand.

[0088] The knowledge graph provided in the embodiments of this application is a knowledge graph that describes domain knowledge related to fault diagnosis. This domain knowledge comes from fault trees or expert knowledge of equipment in industrial systems, which is not limited here. The content includes a BN as a submodel, a hierarchical structure of equipment in industrial systems, such as systems, devices, controllers, and components, an ontology model consisting of root cause nodes and device structure nodes. The device structure nodes include troubleshooting item nodes, controller nodes, and maintenance method nodes.

[0089] As shown in Figure 6, when the root cause of the fault is diagnosed, the root cause class node 514 also has a connection edge 12 with the root cause association attribute with the symptom class node 513 in the BN, and the association between the root cause class node 514 and the equipment structure class node is described as a connection edge 13 with the association attribute in the knowledge graph.

[0090] In the above process, the diagnosis-related nodes include: a component class node 515 describing a device component, a controller class node 519 describing a controller, a maintenance method class node 520 describing a maintenance method, and a troubleshooting item class node 521 describing a troubleshooting item;

[0091] The root cause node is associated with the device structure node based on a preset diagnostic approach, including:

[0092] When associated with the component class node 515, the root cause of the component failure described by the component class node is obtained;

[0093] When associated with the controller class node 519, the root cause of the controller fault described by the controller class node is obtained;

[0094] When associated with the maintenance method node class 520, the maintenance method described in the maintenance method class is executed to obtain the root cause of the fault;

[0095] When associated with the troubleshooting item node class 521, the troubleshooting item described in the troubleshooting item class node is executed to obtain the root cause of the fault.

[0096] In this embodiment of the present application, the root cause node 514 is based on a preset diagnostic pathway provided by a preset rule base 522. The diagnostic pathway is used to determine the root cause of the fault. The diagnostic pathway includes: devices associated with the fault, components and controllers associated with the devices, troubleshooting items for the components, determining whether symptoms are present when troubleshooting the items, and determining the root cause of the fault corresponding to the symptoms.

[0097] It can be seen that the knowledge graph includes the ontology model and the sub-model formed by BN. When the knowledge graph is used for fault diagnosis, it is checked layer by layer from top to bottom whether the equipment described by the device class node in the ontology model has faults and symptoms, as well as the root cause of the fault that causes the fault and / or symptoms. The root cause of the fault involves components or controllers such as component class nodes.

[0098] Figure 7 is a schematic diagram of the hierarchical structure of a device provided in an embodiment of the present application. As shown in the figure, in Figure 7, the hierarchical structure between the various devices of the industrial system is described, which includes, in order according to the hierarchical structure: the system class node describing the system to which the device belongs is defined as the system class node, the device class node, the controller class node and the sensor class node are in a subordinate relationship with the system class node, and can be connected to the system class node through a connection edge with a subordinate attribute. The device is composed of components, so the component class node and the device class node are connected by a connection edge with a subordinate attribute. The device is controlled by the controller, and the device class node and the controller class node are connected by a connection edge with a control attribute. The bottom layer in Figure 7 represents the troubleshooting item class node of each specific troubleshooting item, such as a high voltage troubleshooting item, a low voltage troubleshooting item, a controller overheating troubleshooting item or / and a controller overflow troubleshooting item, etc.

[0099] Based on the structural diagram provided in FIG7 , we can obtain the ontology model of the knowledge graph shown in FIG5 , that is, the left part shown in FIG5 .

[0100] In this embodiment of the present application, the sub-model formed by the BN in the knowledge graph can be considered a fault tree, in which fault class nodes describing different types of faults are connected to device class nodes describing the corresponding devices with associated attributes, and fault class nodes and symptom class nodes are connected with causal attributes. The top nodes in the fault tree correspond to device class nodes and fault class instances.

[0101] When a device has symptoms of a fault, the knowledge graph can be used to check each component and controller of the device to find the cause of the fault. Therefore, the knowledge graph includes a troubleshooting item class node, and a connection edge with the proposed troubleshooting attribute between the controller class node and the component class node. Here, the troubleshooting items involved in the troubleshooting item can be: temperature, current or voltage value, etc., which are not restricted here. If the inspection exceeds the preset threshold, it is a symptom, and the troubleshooting item class node and the symptom class node have a connection edge with an association attribute. Assuming that the symptom described by the symptom class node associated with the identified troubleshooting item class node is derived from the root cause of the fault, it is associated with the corresponding maintenance method in the maintenance method class node. If the execution of the maintenance method can eliminate the fault of the equipment, the result of the maintenance method execution can be used as the root cause of the fault.

[0102] As can be seen, the embodiments of the present application use a knowledge graph as a model for device fault diagnosis, rather than using a fault tree constructed by CN as in the prior art. Because a fault tree only describes the diagnostic path based on a preset fault hierarchy, it is a rule-based diagnostic approach and does not consider the hierarchical structure of the device. Therefore, its use requires multiple human-computer interactions. The knowledge graph provided by the embodiments of the present application, however, includes not only sub-models formed by the BN but also an ontology model based on the device's hierarchical structure, root cause nodes, and device structure nodes. Therefore, during device fault diagnosis, accurate device fault diagnosis can be achieved without multiple human-computer interactions, improving diagnostic efficiency. Furthermore, the BN in the knowledge graph can add sub-graphs in real time based on new device faults and corresponding symptoms, and embed them into the knowledge graph without affecting the ontology model in the knowledge graph. During diagnosis, the BN can also quickly locate the device symptoms and their corresponding faults based on the similarity between the added sub-graphs and historical sub-graphs, and derive the root cause, thereby improving diagnostic efficiency and avoiding multiple rounds of human-computer interactions.

[0103] The following is a specific example to illustrate the embodiments of the present application in detail.

[0104] FIG8 is a schematic structural diagram of a specific example of a device fault diagnosis model provided in an embodiment of the present application. The specific example will be described in detail with reference to FIG8 .

[0105] The knowledge graph shown in Figure 8 is a model for diagnosing abnormal rotation failures in equipment. In this example, the equipment has experienced an abnormal rotation failure, and the diagnosed symptom is abnormal air supply temperature.

[0106] In Figure 8, there is a connection edge 20 with the type attribute between the device class node 510 and the blower node 811 describing the blower, indicating that the blower is the device to be tested. The blower node 811 has connection edges 10 with the component a node 812 describing component a and the component b node 813 describing component b, respectively, and has a connection edge 18 with the controller c node 818 describing controller c. The component b node 813 has connection edges 19 with the troubleshooting item 3 node 814 and the troubleshooting item 4 node 815, respectively, with the proposed troubleshooting attribute. There is a connection edge 20 with the component class node 515 between the component a node 812 and the component b node 813. The component a described by the component a node 812 and the component b described by the component b node 813 are both one of the component class nodes 515. There is an edge 13 with an association attribute between the blower node 811 and the fault node 817 describing the abnormal rotation symptom. There is an edge 11 with a causal attribute between the fault node 817 and the symptom node 818 describing the abnormal air supply temperature. There is an edge 20 with a type attribute between the fault node 817 describing the abnormal rotation symptom and the fault class node 512. There is an edge with a type attribute between the symptom node 818 describing the abnormal air supply temperature and the symptom class node 513. There is an edge 20 with a type attribute between the controller c node 814 and the controller class node 519. There are also edges 19 with a proposed troubleshooting attribute between the controller node 820 and the temperature detection node 821 describing temperature detection, respectively. There is also an edge 13 with an association attribute between the controller c node 814 and the troubleshooting item 1 node 820 and the temperature detection node 821 describing temperature detection, respectively. There is also an edge 13 with an association attribute between the controller fault cause node 822 describing the controller fault as the cause. Controller Fault Cause node 822 has an edge 13 with an association attribute between it and Temperature Detection node 821. It also has an edge 21 with a repair method attribute between it and Repair Method node 823, which describes the repair method for replacing the controller. It also has an edge 20 with a Type attribute between it and Root Cause node 514. It also has an edge 11 with a Causal attribute between it and Overtemperature Symptom node 824, which describes the overtemperature symptom. It also has an edge 20 with a Type attribute between it and Fault node 512. It also has an edge 20 with a Type attribute between Overtemperature Symptom node 824 and Symptom node 513, and an edge 23 with a Symptom Association attribute between it and Temperature Detection node 821. It also has an edge 20 with a Type attribute between Repair Method node 823, which describes the repair method for replacing the controller, and Repair Method node 520.

[0107] When using the above knowledge graph to diagnose equipment faults, it includes:

[0108] In the first step, based on the connection edge with type association attributes, the device node is located at the fan node, and the device to be diagnosed is determined to be the fan.

[0109] In the second step, based on the connection edges of the affiliation attributes, the components of the blower are searched to locate the nodes of component a and component b. Furthermore, based on the connection edges of the control attributes, the controller of the blower is searched to determine the controller c.

[0110] The third step is to detect the operating status of component a, component b and controller c;

[0111] In this step, the knowledge graph is searched for items that need to be tested for each component or controller, and the items (for example, the temperature of controller C) are provided to the maintenance personnel. When the maintenance personnel check, the value is fed back to the knowledge graph, and the detected temperature is compared with the preset temperature threshold. If it exceeds, the symptom node corresponding to controller C is located through the connection edge of the symptom association attribute, and then the connection edge of the causal attribute is used to locate the controller fault cause node to obtain the fault cause;

[0112] In this step, the association between the cause described by the controller fault cause node and the symptom described by the overtemperature symptom node is predefined based on professional knowledge.

[0113] Step 4: Determine whether the cause of the abnormal equipment rotation detected - controller failure is the root cause.

[0114] In this step, after locating the controller failure cause node, equipment maintenance personnel will perform a repair method based on the cause, replacing the controller, and confirm whether this repair method is effective. If so, a valid flag is set in the knowledge graph, confirming that the root cause of the abnormal equipment rotation is a controller failure. Otherwise, the controller failure cause node and its associated edges are deleted from the knowledge graph, retaining only the failure node. The third step is continued for other component nodes until edges with root cause attributes and root cause nodes are created in the knowledge graph.

[0115] When performing subsequent fault detection on the device, the knowledge graph established above is used.

[0116] FIG9 is a flowchart of equipment fault diagnosis using a knowledge graph according to an embodiment of the present application, wherein the specific steps include:

[0117] Step 901: For a device in an industrial system, symptoms are input into a knowledge graph for device fault diagnosis. The knowledge graph performs causal reasoning based on the BN therein to obtain the fault described by the fault class node with the highest probability parameter.

[0118] Step 902: Based on the knowledge graph, the fault class node is merged into the main body of the knowledge graph of the equipment fault diagnosis, and reasoning based on the fault diagnosis approach is performed to locate the root cause class node, and obtain the root cause of the fault described by the root cause class node.

[0119] FIG10 is a schematic diagram of a system structure for establishing a device fault diagnosis model according to an embodiment of the present application, including: an establishing unit 1001 and an applying unit 1002, wherein:

[0120] The establishing unit 1001 is used to set a knowledge graph for fault diagnosis of a device, wherein the knowledge graph includes a device class node describing the device, a Bayesian network (BN), and a root cause class node describing the root cause of the fault, wherein the BN is a two-layer structure network formed by the fault class nodes describing the fault and the symptom class nodes describing the symptom, the fault class nodes in the BN are associated with the device class nodes and the root cause class nodes, and the root cause class nodes are associated with the device structure class nodes based on a preset diagnostic approach;

[0121] The application unit 1002 is used to use the knowledge graph as a device fault diagnosis model in device fault diagnosis, determine the fault based on the symptoms of the device, and determine the root cause of the fault.

[0122] In another embodiment of the invention, an electronic device is provided, comprising: a processor; and a memory storing a program configured to implement the aforementioned method for establishing a device fault diagnosis model when executed by the processor. The electronic device specifically comprises a processor configured to execute the aforementioned steps of establishing a device fault diagnosis model. The electronic device provided in the aforementioned embodiment also includes a non-transitory computer-readable storage medium storing a computer program configured to execute the aforementioned steps of establishing a device fault diagnosis model when executed by the processor.

[0123] Specifically, the non-transitory computer-readable storage medium can be a general storage medium, such as a mobile disk, a hard disk, FLASH, a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), or a portable compact disk read-only memory (CD-ROM), etc. When the computer program on the non-transitory computer-readable storage medium is executed by the processor, it can trigger the processor to execute the various steps of establishing the above-mentioned equipment fault diagnosis model.

[0124] In practical applications, the non-transitory computer-readable storage medium may be included in the device / apparatus / system described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, can perform the steps of establishing the aforementioned device fault diagnosis model.

[0125] In another embodiment of the present application, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores instructions, which, when executed by a processor, cause the processor to execute a method for establishing a device fault diagnosis model in the aforementioned embodiment.

[0126] Yet another embodiment of the present application provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements the various steps in establishing the aforementioned equipment fault diagnosis model.

[0127] The flowcharts and block diagrams in the accompanying drawings of the present application show the possible implementation architecture, functions and operations of the systems, methods and computer program products according to the various embodiments disclosed in the present application. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in the order of the standards in different figures. For example, the boxes described in two connections can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of the boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0128] Those skilled in the art will appreciate that the features described in the various embodiments and / or claims disclosed in this application may be combined and / or coupled in various ways, even if such combinations and / or couplings are not explicitly disclosed in this application. In particular, without departing from the spirit and teachings of this application, the features described in the various embodiments and / or claims of this application may be combined and / or coupled in various ways, and all such combinations and / or couplings fall within the scope disclosed in this application.

[0129] Specific embodiments are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application and is not intended to limit the present application. For those skilled in the art, changes can be made in the specific implementation methods and application scope based on the ideas, spirit and principles of the present application. Any modifications, equivalent replacements, improvements, etc. made by them should be included in the scope of protection of this application.

[0130] Nouns and pronouns referring to persons in this patent application are not limited to a specific gender.

Claims

1. A method for establishing a device fault diagnosis model, characterized in that: The method comprises: A knowledge graph for fault diagnosis of a device is set, wherein the knowledge graph includes a device class node describing the device, a Bayesian network BN, and a root cause class node describing the root cause of the fault, wherein the BN is a two-layer structure network formed by the fault class node describing the fault and the symptom class node describing the symptom, the fault class node in the BN is associated with the device class node and the root cause class node, and the root cause class node is associated with the device structure class node based on a preset diagnostic approach; The knowledge graph is used as a device fault diagnosis model in the fault diagnosis of the device, the fault is determined according to the symptoms of the device, and the root cause of the fault is determined.

2. The method according to claim 1, characterized in that Determining a fault according to a symptom of the device and determining a root cause of the fault includes: The BN in the knowledge graph is used to diagnose the equipment from symptoms to faults, and then the knowledge graph is used for reasoning to derive the root cause through the fault described by the fault class node and the equipment structure class node related to the fault, and the root cause is located on the equipment structure class node according to the root cause.

3. The method according to claim 1, characterized in that The knowledge graph describes a hierarchical structure of devices, and the knowledge graph also includes a system class node describing a system to which the device belongs, the device class node is associated with the system class node, and is described in the knowledge graph as a connection edge having a subordination association attribute between the device class node and the system class node; The knowledge graph also includes a component class node describing a device component, which is associated with the device class node, and a connection edge described in the knowledge graph as having affiliation attributes between the component class node and the device class node.

4. The method according to claim 1, characterized in that The knowledge graph also includes: A PV-type node describing a process variable PV and a physical principle-type node describing a physical principle, wherein the PV node and the physical principle-type node are in a determined relationship, and a connection edge having determined attributes between the PV-type node and the physical principle-type node is described in the knowledge graph; There is an application relationship between the PV class node and the symptom class node, which is described in the knowledge graph as a connection edge with application attributes between the PV class node and the symptom class node; There is an application relationship between the physical principle node and the symptom node, which is described in the knowledge graph as a connection edge having application attributes between the physical principle node and the symptom node; The process variables described by the process variable class nodes are used to calculate the symptoms described by the symptom class nodes, and in the process of calculating the process variables, the physical principles described by the physical principle class nodes are applied.

5. The method according to claim 4, characterized in that The knowledge graph also includes: A sensor node describing a sensor has a monitored relationship with the PV node, and is described in the knowledge graph as a connection edge having a monitored attribute between the sensor node and the PV node; The PV described by the PV class node is monitored by the sensor described by the sensor class node.

6. The method according to claim 1, characterized in that The BN included in the knowledge graph is the trained BN: Preset the initial BN based on the expert knowledge of the equipment; The operation data of the equipment is used as training data for training to obtain the trained BN.

7. The method according to claim 1, characterized in that The root cause class node also has a connection edge with the symptom class node in the BN having a root cause association attribute, and the association between the root cause class node and the equipment structure class node is described as a connection edge with an association attribute in the knowledge graph.

8. The method according to claim 7, characterized in that The device structure class node includes: A component class node describing a device component, a controller class node describing a controller, a maintenance method class node describing a maintenance method, and a troubleshooting item class node describing a troubleshooting item; The root cause node is associated with the equipment structure node based on the preset diagnostic approach, including: When associated with the component class node, deriving the root cause of the component failure described by the component class node; When associated with the controller class node, the root cause of the controller fault described by the controller class node is obtained; When associated with the maintenance method class node, the maintenance method described in the maintenance method class is executed to obtain the root cause of the fault; When associated with the troubleshooting item class node, the troubleshooting items described in the troubleshooting item class node are executed to obtain the root cause of the fault.

9. A device for establishing a device fault diagnosis model, characterized in that: include: Establishing unit (1001) and applying unit (1002), wherein, The establishment unit (1001) is used to set a knowledge graph for fault diagnosis of the device, wherein the knowledge graph includes a device class node describing the device, a Bayesian network BN, and a root cause class node describing the root cause of the fault, wherein the BN is a two-layer structure network formed by the fault class node describing the fault and the symptom class node describing the symptom, the fault class node in the BN is associated with the device class node and the root cause class node, and the The root cause class nodes are associated with the equipment structure class nodes based on the preset diagnostic pathways; The application unit (1002) is used to use the knowledge graph as a device fault diagnosis model in device fault diagnosis, determine the fault based on the symptoms of the device, and determine the root cause of the fault.

10. An electronic device, characterized in that: include: processor; A memory stores a program, wherein the program is configured to implement a method for establishing a device fault diagnosis model according to any one of claims 1 to 7 when executed by the processor.

Images