A method for generating a drilling pump knowledge chain based on component topology and fault propagation

Abstract

This invention discloses a method for generating a drilling pump knowledge chain based on component topology and fault propagation, belonging to the fields of natural language processing, industrial equipment fault information processing, and knowledge graph construction. This method aims to solve the problems of lack of structural constraints in component associations, difficulty in determining the direction of fault propagation, and difficulty in matching response actions in fault descriptions. It establishes topological relationships between drilling pump components, fault propagation rules, and response association rules; it aggregates state change items, impact representation items, and response action items using component nodes as anchor objects to form component-anchored event units; it generates candidate fault chain segments based on causal connection markers and verifies them; it constructs valid fault chain segments into a fault knowledge chain, and attaches response action items according to the response association rules, outputting a structured fault knowledge chain. This invention can improve the reliability and traceability of the generated fault knowledge chain and can be used for drilling pump fault diagnosis, maintenance decision-making, and knowledge graph construction.

Description

Technical Field

[0001] This invention relates to the fields of natural language processing, industrial equipment fault information processing, and knowledge graph construction, specifically to a method for generating a knowledge chain for drilling pumps based on component topology and fault propagation. Background Technology

[0002] Drilling pumps are critical power equipment in oil and gas drilling operations, operating under high pressure, high load, and high wear conditions for extended periods. The hydraulic, power, and transmission components of a drilling pump have assembly connections, power transmission, media flow, and inclusion relationships. When one component malfunctions, the impact of the failure can gradually propagate along these connections and operating mechanisms to other components or equipment. Drilling pump fault information is typically stored in unstructured text formats such as maintenance records, operation logs, fault reports, and historical case studies. A single fault description often includes information on equipment components, changes in condition, operational impacts, and maintenance measures. For example, a fault description might simultaneously record component wear, pump pressure drop, abnormal vibration, and component replacement. If associations are based solely on text co-occurrence or semantic similarity, it's easy to link content that doesn't belong to the same fault propagation process, making it difficult to form a fault knowledge chain with practical engineering significance.

[0003] Existing entity relationship extraction methods mostly use pipeline extraction or joint extraction to identify entities and relationships in text. These methods can usually identify components, fault phenomena and maintenance measures from text, but they mainly rely on word position, contextual semantics or attention relationships to make candidate association judgments, and lack constraints on the actual component structure of drilling pumps, fault propagation direction and corresponding handling actions.

[0004] Patent 1 (Zhang Xiaoqin, Nie Shuhan, Lu Yanjun, Zhu Xiaofei. A joint entity relation extraction method and system [P]. China: CN121980014A, 2026.05.05.) discloses a triple extraction scheme based on whole-sentence semantic focus and subject-object directional interaction. By introducing entity pair geometric prior and directional semantic interaction mechanism, it jointly identifies entities and their relations in the text. However, the geometric prior mainly reflects the positional distribution of entities in the text and is difficult to characterize the actual assembly connection, power transmission or medium flow relationship between drilling pump components. Therefore, it is difficult to verify the fault association results based on the component connection path. Patent 2 (Zhang Hanyue, Pan Chao, Zhang Kun). A method, apparatus, device, storage medium and program product for entity relation extraction [P]. China: CN121581177A, 2026.02.27.) discloses an entity relation extraction method based on graph convolutional networks. This method constructs a pseudo-entity relation graph and uses graph convolutional networks for feature propagation to achieve joint identification of entity types and relation types. However, the graph connection edges in this scheme are mainly constructed based on text entities and semantic associations, and do not form a clear correspondence with the actual component topology of the drilling pump, making it difficult to determine whether candidate fault chain segments have reasonable component reachable paths; Patent 3 (Zhang Chao, Liao Haoran, Kong Mingming, Wang Pinyan, Wang Xianjun, Joint Extraction Method of Relationship Triples Based on Information Augmentation and Bidirectional Modeling [P]. China: CN120764530A, 2025.10.10.) A structured triple generation scheme for complex relationship scenarios is proposed. It mainly improves the relationship recognition ability in complex text by enhancing the interaction between entity information and relationship information. However, this method focuses on entity relationship matching at the text semantic level and does not constrain the engineering correspondence between fault state, impact representation and maintenance treatment. It is difficult to determine whether the treatment action matches the corresponding component and state change item.

[0005] Reference 1 (Zheng H., Wen R., Chen X., et al. PRGC: Potential Relation and Global Correspondence Based Joint Relational Triple Extraction[C] / / Proceedings of ACL, 2021.) proposes a joint triple extraction method based on potential relation judgment and global correspondence modeling, achieving entity relation matching through relation filtering and entity correspondence mechanisms; Reference 2 (Zhang C., Gao S., Wang H., et al. Position-aware Joint Entity and Relation Extraction with Attention Mechanism[C] / / Proceedings of IJCAI, 2022.) proposes a position-aware joint entity relation extraction method, utilizing position encoding and attention mechanisms to enhance the modeling of relative positions between entities; Reference 3 (Chen Z., Zheng Y., Ge J., et al. PRE-Span: A Parallel Model for Jointly Extracting Entities and Relations[J]. Neural Processing Letters, (2024.) proposed a parallel entity-relation joint extraction method, which generates entity and relation results synchronously through a multi-branch structure. The above methods mainly generate candidate associations based on text semantics, positional relationships, or entity correspondence, and are applicable to general text scenarios. However, they lack constraints on the actual connection structure of equipment components, the direction of fault state propagation, and maintenance and handling logic, making it difficult to directly form a drilling pump fault knowledge chain with engineering interpretability.

[0006] Existing entity relation joint extraction methods mainly focus on semantic association modeling in general text scenarios. Although they have achieved certain results on public datasets, they still have the following shortcomings in the scenario of constructing knowledge about oil and gas drilling pump faults: (1) Relationships are often established based on text context, word position or semantic similarity, making it difficult to determine whether there is an actual assembly connection, power transmission, medium flow or inclusion relationship between the drilling pump components that appear together, which can easily lead to association results that do not conform to the equipment structure. (2) Existing methods usually output state change terms and impact characterization terms as mutually independent relational results, lacking directional constraints on the occurrence, propagation and impact processes of faults, making it difficult to form a fault propagation chain with a clear evolutionary order; (3) Some methods do not verify the correspondence between fault status and maintenance actions, which may easily generate handling association results that are inconsistent with the actual maintenance logic, making it difficult to directly support drilling pump fault tracing, maintenance decision-making and knowledge graph updates.

[0007] Therefore, there is an urgent need for a method that can collect, verify, and connect fault descriptions based on the topological relationships of drilling pump components, fault propagation rules, and handling association rules, thereby generating a structured fault knowledge chain with component reachable paths, propagation directions, and textual location information. Summary of the Invention

[0008] To address the challenges of existing fault text processing methods in determining whether components in a text have actual structural relationships, influencing the direction of fault state propagation, and ensuring that maintenance actions match the fault state, this invention provides a drilling pump knowledge chain generation method based on component topology and fault propagation.

[0009] 1. The present invention adopts the following technical solution: a method for generating a drilling pump knowledge chain based on component topology and fault propagation, comprising the following steps: S1: Establish knowledge constraint information for oil and gas drilling pump failures, including the topological relationships of drilling pump components, failure propagation rules, and handling association rules; S2: Using the component nodes in the topological relationship of oil and gas drilling pump components as anchoring objects, the component association fragments of the fault description to be analyzed are collected to form component anchoring event units containing state change items, impact characterization items and handling action items, and their original positioning information is retained. S3: Generate candidate fault chains based on the causal connection markers between the state change items and the influence characterization items within the component anchored event unit; determine the connectable candidate fault chains based on the topological associations between component nodes of different component anchored event units. S4: Based on the topological relationship of drilling pump components and the fault propagation rules, the reachable paths between component nodes involved in the candidate fault chain segment and the propagation direction from the state change item to the influence characterization item are checked to obtain the effective fault chain segment; S5: Based on the connection anchor points between valid fault chain segments, the continuity of the reachable path of the component, and the text positioning order, construct the valid fault chain segments into a fault knowledge chain, and attach the handling action items in the component anchor event unit to the corresponding fault knowledge chain according to the handling association rules. S6: For fault knowledge chains, associate the reachable paths of components, text location information, and rule matching results. Retain fault knowledge chains where the reachable paths of components match the content of the chain segments, the text location order is continuous, and the rule matching results meet the requirements, and output the structured fault knowledge chain.

[0010] 2. Further, step S1 includes the following steps: S11: The hydraulic end components, power end components, and transmission connection components of the oil and gas drilling pump are used as component nodes, and the topological relationship of the drilling pump components is established according to the assembly connection relationship, power transmission relationship, medium flow relationship, or inclusion relationship between the components. S12: For different types of component connection relationships, set the allowed propagation direction from state change items to influence characterization items between corresponding component nodes, as well as the allowed consecutive state change items and influence characterization items, to form fault propagation rules; S13: For each component node and its corresponding state change item, set the allowed handling action items to form handling association rules.

[0011] 3. Further, step S2 includes the following steps: S21: Use the standard name, alias or abbreviation of each component node in the component topology as component anchor words, and locate the component association fragment containing the component anchor words in the fault description to be parsed. S22: Centered on the component anchor word, the text content associated with the component anchor word through state change markers, causal connection markers, or disposal action markers is collected into component anchor event units; S23: In the component anchoring event unit, determine the component node, state change item, impact characterization item and handling action item, and record the location information of each item in the original fault description.

[0012] 4. Further, step S3 includes the following steps: S31: Determine the state change item corresponding to the component node based on the state change flag in the component anchoring event unit; S32: Based on the causal connection marker between the state change item and the influence characterization item, establish the candidate transmission relationship between the state change item and the influence characterization item, and retain the disposal action item and its positioning information in the component anchoring event unit; S33: Combine component nodes, state change items, impact characterization items, and candidate transmission relationships into candidate fault chain segments; S34: For candidate fault chains in different component anchor event units, when there is a topological association between their component nodes, and the influence characterization term of the previous candidate fault chain is the same as the state change term of the next candidate fault chain after terminology normalization, or when the two constitute a combination that is allowed to be continued in the fault propagation rules, the candidate fault chain is marked as a continueable candidate fault chain.

[0013] 5. Further, step S4 includes the following steps: S41: Query the corresponding node of the component node in the candidate fault chain segment in the drilling pump component topology, and determine the component reachable path between the component nodes when multiple component nodes are involved; S42: According to the fault propagation rules, determine whether the connection direction between the state change item and the influence characterization item in the candidate fault chain segment is consistent with the preset propagation direction, and determine whether the positioning order of the state change item in the original fault description is before the influence characterization item. S43: When the component nodes in the candidate fault chain segment are the same, or there is a component reachable path between multiple component nodes corresponding to the connection direction, and the candidate fault chain segment satisfies the fault propagation rule, the candidate fault chain segment is determined as a valid fault chain segment.

[0014] 6. Further, step S5 includes the following steps: S51: Use the state change items or influence characterization items in the effective fault chain segment as the chain segment continuation anchor point. S52: When two valid fault chain segments have the same chain segment connection anchor point, or when the influence characterization item of the previous valid fault chain segment and the state change item of the next valid fault chain segment satisfy the fault propagation rule, the two valid fault chain segments are connected in series to form a fault knowledge chain, taking into account the continuity of the reachable path of the component and the positioning order in the original fault description. S53: For the action items in the component anchored event units that constitute the fault knowledge chain, determine whether the action item forms an allowed combination with the corresponding component node and state change item according to the action association rules; when an allowed combination is formed, attach the action item to the corresponding fault knowledge chain.

[0015] 7. Further, step S6 includes the following steps: S61: Form a link verification record by associating the fault knowledge chain with the component node identifier, component reachable path, location information in the original fault description, fault propagation rule matching result, and handling association rule matching result; S62: For multiple fault knowledge chains with the same starting component node, state change item, and impact representation item, determine whether the component reachable path in each fault knowledge chain matches the chain segment content, and whether the location order of each valid fault chain segment is continuous. S63: Delete fault knowledge chains that contain reachable paths for mismatched components, violate fault propagation rules, have action items that do not meet action association rules, or have discontinuous location sequences; S64: Output a structured fault knowledge chain containing link verification records.

[0016] Beneficial effects: 1. Using drilling pump component nodes as anchor objects, the status change items, impact characteristics items, and handling action items in the fault description are collected, reducing invalid associations caused solely by text co-occurrence; 2. Verify candidate fault chains using component topology relationships and fault propagation rules, so that the retained fault chains have component reachable paths and propagation directions corresponding to the actual structure of the drilling pump; 3. Based on the chain segment connection anchor points, the continuity of the reachable path of the component and the text positioning order, a fault knowledge chain is constructed, and disposal action items are attached using disposal association rules, which reduces unreasonable matching between state change items, impact representation items and disposal action items; 4. It can output a structured fault knowledge chain containing component reachability paths, propagation directions, text location information, and link verification records, providing data support for drilling pump fault tracing, maintenance decisions, and knowledge graph construction. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall process for generating the drilling pump knowledge chain used in the example. Figure 2 A schematic diagram illustrating the verification of fault knowledge constraint information and candidate fault chain segments used in this embodiment; Figure 3 This is a schematic diagram illustrating the construction, link verification, and structured output of the fault knowledge chain used in the example. Detailed Implementation

[0018] The technical solution of the present invention will be further described below with reference to specific embodiments. It should be understood that these embodiments are only used to explain the present invention and are not intended to limit the scope of protection of the present invention. Any equivalent substitutions or modifications made by those skilled in the art without departing from the technical concept of the present invention should fall within the scope of protection of the present invention.

[0019] Example: This example discloses a method for generating a drilling pump knowledge chain based on component topology and fault propagation. The method is performed according to the following steps: S1. Establishment of Fault Knowledge Constraint Information Based on the drilling pump structural diagram, equipment manual, maintenance procedures, repair manual, and historical fault records, a fault knowledge constraint information system for oil and gas drilling pumps is established. This fault knowledge constraint information includes the topological relationships of drilling pump components, fault propagation rules, and handling association rules, expressed as follows: ,in, Indicates the topological relationships of drilling pump components; Indicates the rules for fault propagation; Indicates the rules for handling related matters; The topological relationship of drilling pump components uses hydraulic end components, power end components, and transmission connection components as component nodes, and establishes component connection edges according to assembly connection relationships, power transmission relationships, medium flow relationships, or inclusion relationships, as shown below: ; in, Represents a set of component nodes. This represents the set of connecting edges between component nodes; For any two component nodes and The connecting edge between the two is represented as: ; in, Representing component nodes With component nodes The connecting edges between them This indicates the connection type between component nodes. Connection types include assembly connection, power transmission connection, media flow connection, or inclusion connection. For example, an assembly connection can be established between a valve seat and a valve box; a mating connection can be established between a piston and a cylinder liner; a power transmission connection can be established between a crankshaft and a connecting rod; and a media flow connection can be established between hydraulic components.

[0020] Fault propagation rules Used to define state change terms, influence characterization terms, and permissible propagation relationships between component nodes. State change terms include punctures, leaks, wear, and fractures; influence characterization terms include pump pressure drop, displacement fluctuations, temperature rise, and abnormal vibrations. Rules for handling related matters This is used to define the permissible combinations between component nodes, state changes, and action items. Action items include replacement, repair, cleaning, tightening, and adjustment. For example, a leaking valve seat can be addressed by replacing the valve seat or repairing the valve box; a failed seal can be addressed by replacing the seal; insufficient lubrication can be addressed by adding lubricating oil or checking the lubrication channels; and bearing wear can be addressed by replacing the bearing or adjusting the bearing clearance. Figure 2 As shown, the component topology Fault propagation rules and handling related rules Together they constitute fault knowledge constraint information This is used for subsequent verification of candidate faulty chain segments.

[0021] S2, Component anchoring event unit formation For the description of the drilling pump failure to be parsed, a component anchor term list is first established. The component anchor term list includes the standard name, alias and abbreviation of each component node. For example, "discharge valve seat" and "discharge valve seat" can be mapped to the same discharge valve seat component node; "crosshead pin" and "pin shaft" can be mapped to the corresponding component node according to the preset term list; "cylinder liner seal" and "cylinder liner seal" can be attributed to the corresponding cylinder liner assembly according to semantic rules.

[0022] Component anchor words appearing in the fault description and The component nodes are matched, and the state change fragments, influence fragments, and disposal fragments associated with the same component anchor word are collected to form a component anchor event unit. i The unit for anchoring an individual component is represented as follows: ; in, Indicates a state change term; Indicates the influence of the characterization term; Indicates the action to be performed; This indicates the location information of each item in the original fault description; When a component anchoring event unit does not contain any impact characterization items or handling action items, the corresponding item can be set to empty, while retaining the identified component nodes, state change items and their positioning information; The location information for any state change item, impact characteristic item, or action item is represented as follows: ; in, Indicates the original fault description number; Indicates the segment number; Indicates the starting position of the character; Indicates the end position of the character.

[0023] The status change items can be determined based on status change markers such as "puncture" and "leakage"; the impact characterization items can be determined based on causal connection markers such as "cause" and "lead to" and the subsequent operating status description; the handling action items can be determined based on handling action markers such as "replace" and "maintain".

[0024] S3, Candidate Fault Chain Generation Based on the connection markers in the component anchoring event unit and the topological associations between component nodes, candidate fault chain segments are generated; Based on the causal connection markers between state change items and influence characterization items in component anchored event units, a candidate transmission relationship is established from state change items to influence characterization items; for different component anchored event units, when there is a topological association between corresponding component nodes, the corresponding candidate fault chain segment is marked as a connectable candidate fault chain segment. Candidate faulty chain segments are represented as follows: ; in, Indicates the first i One candidate faulty chain segment; Indicates candidate transmission relationship; Indicates the impact characteristics or actions taken; This indicates the text location information corresponding to the candidate faulty chain segment; when To influence the characterization term When, candidate fault chain segments are used to characterize the fault propagation process of "component node - state change item - impact characterization item"; when For handling action items At that time, the candidate fault chain segment is used to characterize the maintenance and handling process of "component node - state change item - handling action item".

[0025] For the fault description "the discharge valve seat leak causes a drop in pump pressure, requiring replacement of the valve seat assembly", the following candidate fault chain segments can be generated: ; in, Corresponding discharge valve seat, Corresponding to puncture and leakage, Corresponding to a drop in pump pressure, Replace the valve seat assembly accordingly.

[0026] S4, Candidate Faulty Chain Segment Verification For each candidate faulty chain segment, query the component nodes within it. For the corresponding nodes in the process, when the candidate fault chain involves two or more component nodes, further determine the component reachable paths between the relevant component nodes; From component nodes To component node The reachable path of a component is represented as: ; in, Indicates the component connection edges in the path; This represents intermediate component nodes in the path.

[0027] When a candidate fault chain segment targets only one component node, the path corresponding to that component node itself is considered a valid path; when a candidate fault chain segment involves multiple component nodes, it is determined whether there is an assembly connection path, power transmission path, medium flow path, or inclusion path among the relevant component nodes that matches the content of the candidate fault chain segment.

[0028] The component topology verification results of the candidate fault chain segment are denoted as When the component nodes in the candidate fault chain are the same, or when there is a component reachable path between the relevant component nodes that matches the candidate fault chain, ;otherwise, ; The failure propagation verification result of the candidate faulty chain segment is denoted as When the state change term, the connection relationship, and the influence representation term meet the requirements... And when the state change term precedes the influence characterization term, ;otherwise, ; The validity verification results of the candidate faulty chain segment are expressed as follows: ; when When the corresponding candidate faulty chain segment is determined as a valid faulty chain segment, it is represented as: ; S5. Fault Knowledge Chain Construction and Handling Action Integration like Figure 3 As shown, the state change items or influence characterization items in the effective fault chain segment are used as chain segment continuation anchor points. When two effective fault chain segments have the same continuation anchor point, or the influence characterization item of the previous effective fault chain segment is the same as the state change item of the subsequent effective fault chain segment or satisfies the preset correspondence, it is determined whether the two satisfy the continuity of the component reachable path and the continuity of the text positioning order.

[0029] When two adjacent valid fault chain segments have the same component node in their corresponding component reachable paths, or the termination node of the component reachable path of the preceding valid fault chain segment is the same as the starting node of the component reachable path of the following valid fault chain segment, or there is a direct connecting edge between them, the component reachable paths of the two valid fault chain segments are determined to be continuous. When the component reachable paths are continuous and the text positioning order is continuous, multiple valid fault chain segments are constructed into a fault knowledge chain, represented as follows: ; in, to This indicates a valid faulty chain segment that can be connected end-to-end based on the chain segment connection anchor point, the continuity of the component reachable path, and the text positioning order.

[0030] For the handling action items in the component anchored event units that constitute the fault knowledge chain, query the handling association rules. Determine whether the action item forms an allowed combination with the corresponding component node and state change item, and record the action association verification result as follows. ,when hour, ;when hour, ; when At that time, the action item will be handled. Link it to the corresponding fault knowledge chain; when the corresponding component anchor event unit does not contain a handling action item, set the handling association verification result corresponding to the component anchor event unit to 1, which does not affect the retention and construction of the fault knowledge chain. For example, for the fault description "discharge valve seat puncture caused pump pressure drop, valve seat assembly needs to be replaced", "discharge valve seat - puncture - pump pressure drop" can be used as the fault propagation chain segment; when "discharge valve seat - puncture - replace valve seat assembly" satisfies When this happens, "Replace valve seat assembly" will be linked to the corresponding fault knowledge chain.

[0031] S6. Link verification record generation and fault knowledge chain output For each fault knowledge chain, its component nodes, component reachable paths, original text location information, fault propagation rule matching results, and handling association rule matching results are associated to form a link verification record: ; in, This represents the set of reachable paths for components involved in the fault knowledge chain; This represents the set of original text location information corresponding to each valid fault chain segment. This indicates the result of the fault propagation rule matching; This indicates the result of handling the matching of related rules.

[0032] For a fault knowledge chain consisting of multiple valid fault chain segments, the fault propagation rule matching result can be expressed as: ; in, to They respectively represent the components of the first m Fault propagation verification results for each valid fault chain segment of the fault knowledge chain; The matching result of its association rule can be expressed as: ; Among them, for valid fault chain segments that do not contain handling action items, the corresponding The value is 1.

[0033] For multiple fault knowledge chains with the same starting component node, state change item, and impact representation item, determine whether their component reachable paths match the chain segment content, and determine whether the location order of each valid fault chain segment in the original fault description is continuous. m The text location sequence verification result of the fault knowledge chain is denoted as When the valid fault chain segments that make up the fault knowledge chain are arranged sequentially according to the starting position in the original fault description, ;otherwise, ; For duplicate fault knowledge chains with the same starting component node, state change item, impact representation item, and handling action item, their original text location information is merged; for fault knowledge chains where the component reachable path does not match the chain segment content, violates the fault propagation rules, the handling action item does not meet the permitted combination, or the text location order is not continuous, they are not used as the final output result.

[0034] The final output set of fault knowledge chains is represented as follows: ; in, This represents the final output set of structured fault knowledge chains. The final output includes at least component nodes, state change items, impact characterization items, handling action items, fault propagation direction, component reachable path, original text location information, and link verification records. The structured fault knowledge chain can be written into a fault knowledge graph and can also be used for fault tracing, maintenance plan query, fault diagnosis, and operation and maintenance decision support.

[0035] In a specific application, for the fault description "The discharge valve seat is leaking, causing a drop in pump pressure, and the valve seat assembly needs to be replaced," first map "discharge valve seat" to... The component nodes are defined; "leakage" is identified as a state change item, "pump pressure drop" as an impact characteristic item, and "replace valve seat assembly" as a handling action item; then, corresponding candidate fault chains are generated, and based on... The propagation direction of "leakage leading to pump pressure drop" was checked; then based on... Perform a handling association verification for "discharge valve seat - leak - replace valve seat assembly"; when the corresponding component path, propagation direction, handling association and text location sequence all meet the requirements, output a structured fault knowledge chain containing the fault propagation process, handling action items and link verification records.

[0036] The method of the present invention has been described above. Those skilled in the art can implement the method of the present invention based on this description. Other embodiments obtained by those skilled in the art based on the above description of the present invention without inventive effort should all fall within the scope of protection of the present invention.

Claims

1. A method for generating a drilling pump knowledge chain based on component topology and fault propagation, characterized in that, Includes the following steps: S1: Establish knowledge constraint information for oil and gas drilling pump failures, including the topological relationships of drilling pump components, failure propagation rules, and handling association rules; S2: Using the component nodes in the topological relationship of oil and gas drilling pump components as anchoring objects, the component association fragments of the fault description to be analyzed are collected to form component anchoring event units containing state change items, impact characterization items and handling action items, and their original positioning information is retained. S3: Generate candidate fault chains based on the causal connection markers between the state change items and the influence characterization items within the component anchored event unit; determine the connectable candidate fault chains based on the topological associations between component nodes of different component anchored event units. S4: Based on the topological relationship of drilling pump components and the fault propagation rules, the reachable paths between component nodes involved in the candidate fault chain segment and the propagation direction from the state change item to the influence characterization item are checked to obtain the effective fault chain segment; S5: Based on the connection anchor points between valid fault chain segments, the continuity of the reachable path of the component, and the text positioning order, construct the valid fault chain segments into a fault knowledge chain, and attach the handling action items in the component anchor event unit to the corresponding fault knowledge chain according to the handling association rules. S6: For fault knowledge chains, associate the reachable paths of components, text location information, and rule matching results. Retain fault knowledge chains where the reachable paths of components match the content of the chain segments, the text location order is continuous, and the rule matching results meet the requirements, and output the structured fault knowledge chain.

2. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S1 includes the following steps: S11: The hydraulic end components, power end components, and transmission connection components of the oil and gas drilling pump are used as component nodes, and the topological relationship of the drilling pump components is established according to the assembly connection relationship, power transmission relationship, medium flow relationship, or inclusion relationship between the components. S12: For different types of component connection relationships, set the allowed propagation direction between corresponding component nodes when state change items cause impact characterization items, as well as the allowed consecutive state change items and impact characterization items, to form fault propagation rules; S13: For each component node and its corresponding state change item, set the allowed handling action items to form handling association rules.

3. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S2 includes the following steps: S21: Use the standard name, alias or abbreviation of each component node in the component topology as component anchor words, and locate the component association fragment containing the component anchor words in the fault description to be parsed. S22: Collect the text content associated with the component anchor word and containing state change markers, causal connection markers, or action markers into component anchor event units; S23: In the component anchoring event unit, determine the component node, state change item, impact characterization item and handling action item, and record the location information of each item in the original fault description.

4. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S3 includes the following steps: S31: Determine the state change item corresponding to the component node based on the state change flag in the component anchoring event unit; S32: Based on the causal connection marker between the state change item and the influence characterization item, establish the candidate transmission relationship between the state change item and the influence characterization item, and retain the disposal action item and its positioning information in the component anchoring event unit; S33: Combine component nodes, state change items, impact characterization items, and candidate transmission relationships into candidate fault chain segments; S34: For candidate fault chains in different component anchor event units, when there is a topological association between their component nodes, and the influence characterization term of the previous candidate fault chain is the same as the state change term of the next candidate fault chain after terminology normalization, or when the two constitute a combination that is allowed to be continued in the fault propagation rules, the candidate fault chain is marked as a continueable candidate fault chain.

5. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S4 includes the following steps: S41: Query the corresponding node of the component node in the candidate fault chain segment in the drilling pump component topology, and determine the component reachable path between the component nodes when multiple component nodes are involved; S42: According to the fault propagation rules, determine whether the connection direction between the state change item and the influence characterization item in the candidate fault chain segment is consistent with the preset propagation direction, and determine whether the positioning order of the state change item in the original fault description is before the influence characterization item. S43: When the component nodes in the candidate fault chain segment are the same, or there is a component reachable path between multiple component nodes that corresponds to the preset propagation direction, and the candidate fault chain segment satisfies the fault propagation rule, the candidate fault chain segment is determined as a valid fault chain segment.

6. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S5 includes the following steps: S51: Use the state change items or influence characterization items in the effective fault chain segment as the chain segment continuation anchor point. S52: When two valid fault chains have the same chain connection anchor point, or the influence characterization item of the previous valid fault chain is the same as the state change item of the subsequent valid fault chain after terminology normalization, or the two constitute a combination that is allowed to be connected in the fault propagation rules, the two valid fault chains are connected in series to form a fault knowledge chain by combining the continuity of the reachable path of the component and the positioning order in the original fault description. S53: For the action items in the component anchored event units that constitute the fault knowledge chain, determine whether the action item forms an allowed combination with the corresponding component node and state change item according to the action association rules; when an allowed combination is formed, attach the action item to the corresponding fault knowledge chain.

7. The drilling pump knowledge chain generation method based on component topology and fault propagation according to claim 1, characterized in that, S6 includes the following steps: S61: Form a link verification record by associating the fault knowledge chain with the component node identifier, component reachable path, location information in the original fault description, fault propagation rule matching result, and handling association rule matching result; S62: For multiple fault knowledge chains with the same starting component node, state change item, and impact representation item, determine whether the component reachable path in each fault knowledge chain matches the chain segment content, and whether the location order of each valid fault chain segment is continuous. S63: The fault knowledge chain is such that the reachable path of the deleted component does not match the content of the chain segment, the fault propagation rules are violated, the handling action items do not meet the handling association rules, or the location order is not continuous. S64: Output a structured fault knowledge chain containing link verification records.

Images