Automobile wire harness process rule automatic matching method based on knowledge graph

Abstract

The invention discloses an automobile wire harness process rule automatic matching method based on a knowledge graph, and the method comprises the following steps: collecting wire harness design data, and carrying out the standardization processing; analyzing the process rule base, extracting key attribute fields and generating a process rule metadata set; semantic modeling and structured fusion are carried out, and a process knowledge graph is constructed; performing semantic association analysis, causal constraint fusion and feasibility judgment processing by utilizing a semantic retrieval enhancement module; carrying out provable retrieval, risk assessment and conflict resolution based on the candidate process rule set; converting the target process rule set into a process instruction, and driving a design system to perform synchronous updating and rule labeling; and updating the process knowledge graph based on system feedback data, and outputting an optimized process verification report and updating a design version. The method is based on the knowledge graph and the semantic causal fusion technology, intelligent matching of the wire harness process rules is achieved, and the method has the advantages of being high in matching precision, high in interpretability and capable of achieving self-adaptive optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent design technology for automotive wiring harnesses, and in particular to an automatic matching method for automotive wiring harness process rules based on knowledge graphs. Background Technology

[0002] Current automotive wiring harness design and process formulation typically rely on manual rule matching and experience-based judgment. Design engineers must manually search for applicable rules in a vast database of process rules and compare and verify them item by item according to design parameters. This results in low rule retrieval efficiency, unstable matching results, and inconsistencies in rule application across different design stages. While some systems have introduced knowledge graphs and semantic modeling techniques to achieve a structured association between wiring harness design and process knowledge, these often remain at the static knowledge level, lacking effective modeling of rule constraints, conflicts, and dynamic applicability.

[0003] Meanwhile, existing methods struggle to balance semantic relevance and causal logic consistency during automatic matching of process rules, making it impossible to quantify the feasibility and risks of the rules. They also lack an incremental update mechanism based on system feedback, resulting in insufficient rule adaptability and verifiability, and the process optimization process relies on manual adjustments.

[0004] Therefore, how to provide an automatic matching method for automotive wiring harness process rules based on knowledge graphs is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an automatic matching method for automotive wiring harness process rules based on knowledge graphs. This invention achieves intelligent matching of wiring harness process rules based on knowledge graph and semantic causal fusion technology, and has high matching accuracy, strong interpretability and adaptive optimization capabilities.

[0006] An automatic matching method for automotive wiring harness process rules based on knowledge graphs according to an embodiment of the present invention includes the following steps:

[0007] Collect wire harness design data, perform standardization processing, and form a standardized design dataset;

[0008] Parse the process rule base, extract the applicable objects, triggering conditions, parameter thresholds, priorities, conflict relationships, execution actions and applicable scope, and generate a process rule metadata set;

[0009] Semantic modeling and structured fusion of standardized design datasets and process rule metadata sets are performed to construct a unified semantically related process knowledge graph;

[0010] Using the semantic retrieval enhancement module, semantic association analysis, causal constraint fusion and feasibility judgment are performed on process rule nodes and design parameter nodes on the process knowledge graph, and a set of candidate process rules that meet the causal orientation and constraint feasibility is output.

[0011] Based on the candidate process rule set, provable retrieval, risk assessment and conflict resolution are performed to generate a target process rule set that conforms to causal orientation and risk constraints.

[0012] The target set of process rules is converted into process instructions, which drive the design system to perform synchronous updates and rule annotations, and output process verification reports and updated design versions.

[0013] Based on system feedback data, the process knowledge graph is compared and embedded for updates, the semantic association and constraint consistency of rules and design parameters are adjusted, and an optimized process verification report and an updated design version are output.

[0014] Optionally, the harness design data includes harness component identification, wire diameter, insulation class, shielding layer type, path geometry parameters, minimum bending radius, fixing point location and spacing, connector model, materials and distances of adjacent components, list and intensity of electromagnetic interference sources, location and power of heat sources, and applicable standard number. The standardization process includes unified data structure, unified parameter units, standardized coding, data consistency verification, and parameter semantic mapping.

[0015] Optionally, the generation of the process rule metadata set specifically includes:

[0016] The text rules, logical expressions, and parameter constraints related to wire harness design in the process rule library are parsed. The rule content is organized hierarchically according to the hierarchical structure and scope of application of the rules, and the original rule entries related to wire harness components, path structure, electrical performance, and thermal management are extracted.

[0017] The original rule entries are semantically segmented and parameter mapped, and the limiting conditions, operational constraints and environmental constraints involved in the rules are transformed into structured fields, and the semantic fields are normalized.

[0018] Based on normalized structured fields, the applicable objects, triggering conditions, parameter thresholds, priorities, conflict relationships, execution actions, and applicable scope are extracted to generate an attribute vector for each rule;

[0019] Hierarchical annotation is performed on the conflict relationships between rules based on attribute vectors. The conflict level is determined according to the rule priority coefficient and constraint strength weight, and conflict annotation results are generated.

[0020] Based on the conflict annotation results, the execution actions are structured and encoded, and the operation instructions, parameter adjustment methods and applicable conditions are bound to form a rule execution template, and a unique identifier code is assigned to each process rule;

[0021] The rule data, after semantic parsing, attribute extraction, conflict annotation, and execution encoding, are merged to generate a set of process rule metadata.

[0022] Optionally, the construction of the process knowledge graph specifically includes:

[0023] Based on the parameter structure of the standardized design dataset and the semantic attributes of the process rule metadata set, the node types and relationship types of the knowledge graph are defined;

[0024] Based on node type and relationship type, the parameter information in the standardized design dataset is instantiated into node attribute data, and semantic connections between nodes are established according to the rule objects corresponding to each field in the process rule metadata set.

[0025] Perform normalization mapping on continuous parameters in node attribute data, calculate the ratio according to the difference between the maximum and minimum values ​​of the parameters, unify the numerical range of the parameters, and output normalized node attribute data.

[0026] Based on normalized node attribute data, a semantic similarity comparison method is used to perform semantic comparison and mapping on parameter names, codes and units in the standardized design dataset, and to uniformly assign synonymous or homologous parameters to the same knowledge entity node, generating entity alignment mapping data.

[0027] The relationship weights between nodes are adjusted based on the entity alignment mapping data, and the relationship strength is calculated based on the product of parameter correlation and rule importance to obtain weighted node relationship data.

[0028] Based on normalized node attribute data, entity alignment mapping data, and weighted node relationship data, node fusion and relationship integration processing are performed to construct a unified semantic association process knowledge graph.

[0029] Optionally, the output of the candidate process rule set specifically includes:

[0030] Based on the node and relationship structure in the process knowledge graph, semantic representations of design parameter nodes and process rule nodes are extracted to construct a semantic retrieval vector set;

[0031] Using the semantic retrieval enhancement module, semantic association retrieval of process rule node embedding vectors and design parameter node embedding vectors is performed on the semantic retrieval vector set, generating a similarity score matrix between rule nodes and design parameter nodes. The semantic retrieval enhancement module is based on the LightRAG framework and improves upon it by introducing causal constraint association modeling and differentiability feasibility fusion mechanism.

[0032] Based on the similarity score matrix, causal constraint modeling is performed on the data relationship between rule nodes and design parameter nodes. Causal weights are marked for rule nodes with constraint conflicts, and a causal dependency graph is generated.

[0033] Based on the similarity score matrix and causal dependency graph, semantic scores and causal weights are jointly calculated using semantic causality. A weighted fusion function is then used to perform causal constraint weighted fusion processing on the joint calculation results to generate a comprehensive score.

[0034] The weighted fusion function takes the joint calculation result data as input, dynamically adjusts the fusion weight according to the constraint relationship in the causal dependency graph, obtains the comprehensive influence value of each rule node by calculating the difference between the expected value of the result of the rule node under the intervention condition and the expected value of the result under the non-intervention condition, and uses the comprehensive influence value as the comprehensive score result of the rule node.

[0035] Based on the comprehensive scoring results, the constraint satisfaction of each rule node is differentiable and optimized to generate a feasibility correction matrix.

[0036] By combining the comprehensive scoring results with the feasibility correction matrix, a joint screening calculation is performed to output a set of candidate process rules that meet the requirements of causal orientation and constraint feasibility.

[0037] Optionally, the generation of the target process rule set specifically includes:

[0038] Based on the node relationship structure in the candidate process rule set and process knowledge graph, rule nodes, design parameter nodes and corresponding causal relationship paths are extracted to construct provable retrieval graph data.

[0039] In the provable retrieval graph data, causal inference path search is performed on the associated paths of candidate rule nodes. The provability score of the path is calculated based on the path coverage and node association weight, and the set of paths with the highest provability score is selected.

[0040] Based on the set of paths with the highest provability scores, a minimum proof subgraph is extracted, which includes target rule nodes, design parameter nodes, and corresponding constraint relationships. The minimum proof subgraph is formed by jointly minimizing the relationship weights between nodes and semantic similarity.

[0041] Risk factors are extracted and weights are calculated for the rule nodes included in the minimum proof subgraph. A rule risk function is constructed based on process complexity, constraint conflict frequency and historical failure probability. The risk distribution of the rule nodes is weighted and the risk score of each rule node is output. The rule risk function is optimized with the expected value of non-compliance deviation at the confidence level.

[0042] Candidate rule nodes are sorted in ascending order based on rule risk scores to construct a rule risk ranking list. Conflicting rule nodes in the ranking list are then processed to resolve conflicts, retaining the set of rule nodes with the lowest risk scores and logical consistency.

[0043] After conflict resolution, the set of rule nodes is subjected to consistency verification and structural integrity verification to generate a set of target process rules that conform to causal orientation and risk constraints.

[0044] Optionally, the output of the process verification report and the updated design version specifically includes:

[0045] Map each rule node in the target process rule set to the design system parameter field, and generate a structured process instruction set based on the constraint type and control objective corresponding to the rule;

[0046] Standardized coding is performed on the operating parameters in the structured process instruction set, converting harness spacing limits, minimum bending radius limits, fixed point spacing limits, shielding level configurations, and path adjustment actions into control instruction formats that the system can recognize;

[0047] The structured process instruction set is written into the rule control module of the design system, and the design model parameters are updated synchronously and the corresponding rule sources are marked.

[0048] Perform compliance verification calculations in the updated design model, compare design parameters against the constraints of the target process rule set, and generate verification result data;

[0049] The verification results data are organized into a report structure to form a process verification report that records the process verification status, constraint matching status, and correction parameters;

[0050] Based on the verification results, the design system is updated and the rule annotation is synchronized, and a process verification report and an updated design version are output.

[0051] Optionally, the optimized process verification report and the updated design version output specifically include:

[0052] Based on system feedback data, the difference comparison of rule nodes, design parameter nodes and constraint relationships recorded in the process rule base is performed to extract rule trigger deviation and verification deviation information.

[0053] The extracted deviation information is normalized, and the rule update weights are calculated based on the deviation magnitude and rule triggering frequency.

[0054] Adjust the scope of application, threshold and priority of process rule nodes according to the rule update weight, and remove invalid or conflicting rule nodes and mark the update status.

[0055] Based on the updated state of the rule nodes and design parameter nodes, incremental updates are performed on the applicable relationships, constraint relationships, conflict relationships and priority relationships recorded in the process knowledge graph, the relationship weights are adjusted and the entity alignment mapping data of the parameters and rule nodes are updated;

[0056] The updated nodes and relationships are recalculated by embedding, and new node embedding representations are generated and written into the process knowledge graph based on changes in node attributes and adjustments to relationship weights.

[0057] Based on the updated process knowledge graph, the semantic association and constraint consistency between process rule nodes and design parameter nodes are comprehensively recalculated, and an optimized process verification report and an updated design version are output.

[0058] The beneficial effects of this invention are:

[0059] This invention achieves unified semantic fusion of wiring harness design data and process rule knowledge by constructing an automatic matching system for automotive wiring harness process rules based on a knowledge graph. Through semantic modeling and entity alignment of standardized design datasets and structured process rule metadata sets, this invention establishes a process knowledge graph covering wiring harness components, design parameters, process rules, constraint relationships, and priority relationships, forming a computable semantic network between wiring harness design information and process constraints. Building upon this, a semantic retrieval enhancement module is introduced. Using an improved LightRAG framework, semantic association retrieval between design parameter nodes and process rule nodes is achieved. Combined with causal constraint association modeling and a differentiable feasibility fusion mechanism, the system can maintain semantic relevance while considering constraint consistency and logical dependencies, thereby improving the accuracy and robustness of rule matching.

[0060] This invention constructs a comprehensive scoring system that considers semantic matching, causal weights, and feasibility constraints through semantic-causal joint computation and weighted fusion function design, achieving causal-guided matching and dynamic filtering of rule nodes. This mechanism quantifies the complex semantic relationships between design parameters and process rules while adaptively adjusting matching weights based on causal dependency graphs, significantly improving the intelligence and interpretability of process rule matching. Furthermore, the feasibility correction matrix generated through differentiable optimization ensures the matching results have mathematically continuous and adjustable characteristics, guaranteeing the smoothness and verifiability of rule filtering and fundamentally avoiding the accuracy loss and mismatch problems caused by traditional static threshold-based filtering methods.

[0061] Furthermore, this invention introduces a provable retrieval and risk assessment mechanism based on the candidate process rule set, achieving a balance between rule verifiability and risk controllability using a minimum proof subgraph and rule risk function. The system can resolve conflicts and reorder priorities based on causal path coverage and risk distribution, ensuring that the output target process rule set possesses logical consistency and safety boundaries. By transforming the target process rule set into structured process instructions and writing them back to the design system, automatic verification of design parameters and rule annotation are achieved, resulting in a design version with compliance verification. Combined with the incremental update mechanism of system feedback data, the process knowledge graph and rule base can adaptively evolve, continuously optimizing the semantic associations and constraint consistency between rule nodes. Therefore, this invention not only improves the intelligence level and matching accuracy of automatic matching of wiring harness process rules but also achieves interpretable reasoning and continuous optimization of process rules, significantly enhancing the automation and compliance reliability of automotive wiring harness design. Attached Figure Description

[0062] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0063] Figure 1 This is a flowchart of an automatic matching method for automotive wiring harness process rules based on knowledge graphs, as proposed in this invention.

[0064] Figure 2 This is a schematic diagram of the node types and relationship structure of the process knowledge graph for the automatic matching method of automotive wiring harness process rules based on knowledge graph proposed in this invention. Detailed Implementation

[0065] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0066] refer to Figure 1-2 An automatic matching method for automotive wiring harness process rules based on knowledge graphs includes the following steps:

[0067] Collect wire harness design data, perform standardization processing, and form a standardized design dataset;

[0068] Parse the process rule base, extract the applicable objects, triggering conditions, parameter thresholds, priorities, conflict relationships, execution actions and applicable scope, and generate a process rule metadata set;

[0069] Semantic modeling and structured fusion of standardized design datasets and process rule metadata sets are performed to construct a unified semantically related process knowledge graph;

[0070] Using the semantic retrieval enhancement module, semantic association analysis, causal constraint fusion and feasibility judgment are performed on process rule nodes and design parameter nodes on the process knowledge graph, and a set of candidate process rules that meet the causal orientation and constraint feasibility is output.

[0071] Based on the candidate process rule set, provable retrieval, risk assessment and conflict resolution are performed to generate a target process rule set that conforms to causal orientation and risk constraints.

[0072] The target set of process rules is converted into process instructions, which drive the design system to perform synchronous updates and rule annotations, and output process verification reports and updated design versions.

[0073] Based on system feedback data, the process knowledge graph is compared and embedded for updates, the semantic association and constraint consistency of rules and design parameters are adjusted, and an optimized process verification report and an updated design version are output.

[0074] In this embodiment, the wire harness design data includes wire harness component identification, wire diameter, insulation class, shielding layer type, path geometric parameters, minimum bending radius, fixing point location and spacing, connector model, materials and distances of adjacent components, list and intensity of electromagnetic interference sources, location and power of heat sources, and applicable standard number. The standardization process includes unified data structure, unified parameter units, standardized coding, data consistency verification, and parameter semantic mapping.

[0075] In this embodiment, the generation of the process rule metadata set specifically includes:

[0076] The text rules, logical expressions, and parameter constraints related to wire harness design in the process rule library are parsed. The rule content is organized hierarchically according to the hierarchical structure and scope of application of the rules, and the original rule entries related to wire harness components, path structure, electrical performance, and thermal management are extracted.

[0077] The original rule entries are semantically segmented and parameter mapped, and the limiting conditions, operational constraints and environmental constraints involved in the rules are transformed into structured fields, and the semantic fields are normalized.

[0078] Based on the normalized structured fields, the applicable objects, triggering conditions, parameter thresholds, priorities, conflict relationships, execution actions and applicable scope are extracted to generate the attribute vector of each rule. The parameter thresholds are composed of each constraint variable in the parameter set. Each constraint variable corresponds to the allowable value range of the harness design parameters within the limited range. The value range is obtained by calculating the difference between its upper and lower bounds and combining it with the mean square deviation of historical compliant samples.

[0079] Based on attribute vectors, hierarchical labeling is performed on the conflict relationships between rules. The conflict level is determined according to the rule priority coefficient and constraint strength weight, and the conflict labeling result is generated. The priority coefficient is the product of the average triggering frequency of the rule in historical design and the process risk weight. The constraint strength weight is the logarithmic function result of the probability of non-compliance caused by parameter over-limit, which is used to measure the influence strength between rules.

[0080] Based on the conflict annotation results, the execution actions are structured and encoded, and the operation instructions, parameter adjustment methods and applicable conditions are bound to form a rule execution template, and a unique identifier code is assigned to each process rule;

[0081] The rule data, after semantic parsing, attribute extraction, conflict annotation, and execution encoding, are merged to generate a set of process rule metadata, including fields such as applicable objects, triggering conditions, parameter thresholds, priority, conflict relationships, execution actions, and applicable scope.

[0082] In this embodiment, the construction of the process knowledge graph specifically includes:

[0083] Based on the parameter structure of the standardized design dataset and the semantic attributes of the process rule metadata set, the node types and relationship types of the knowledge graph are defined. The node types include wire harness component nodes, design parameter nodes, process rule nodes, constraint nodes, environment nodes, and standard nodes. The relationship types include applicability relationships, constraint relationships, conflict relationships, and priority relationships.

[0084] Based on node type and relationship type, the parameter information in the standardized design dataset is instantiated into node attribute data, and semantic connections between nodes are established according to the rule objects corresponding to each field in the process rule metadata set.

[0085] Perform normalization mapping on continuous parameters in node attribute data, calculate the ratio according to the difference between the maximum and minimum values ​​of the parameters, unify the numerical range of the parameters, and output normalized node attribute data.

[0086] Based on normalized node attribute data, a semantic similarity comparison method is used to perform semantic comparison and mapping on parameter names, codes and units in the standardized design dataset, and to uniformly assign synonymous or homologous parameters to the same knowledge entity node, generating entity alignment mapping data.

[0087] The relationship weights between nodes are adjusted based on the entity alignment mapping data, and the relationship strength is calculated based on the product of parameter correlation and rule importance to obtain weighted node relationship data.

[0088] Based on normalized node attribute data, entity alignment mapping data, and weighted node relationship data, node fusion and relationship integration processing are performed to construct a unified semantic association process knowledge graph. The process knowledge graph records the applicable relationships, constraint relationships, conflict relationships, and priority relationships between wire harness components, design parameters, process rules, constraints, environment, and standards.

[0089] In this embodiment, the output of the candidate process rule set specifically includes:

[0090] Based on the node and relationship structure in the process knowledge graph, semantic representations of design parameter nodes and process rule nodes are extracted to construct a semantic retrieval vector set;

[0091] The construction of the semantic retrieval vector set includes: performing word segmentation, denoising, and word form merging on the text attributes of nodes in the process knowledge graph to extract semantic keywords and establish a set of node text features; based on the extracted text features, vectorizing the design parameter nodes and process rule nodes respectively, converting parameter names, units, value ranges, and rule description fields into embedding vectors of a unified dimension; calculating the semantic relevance between design parameter nodes and process rule nodes, and performing normalization and alignment processing on the embedding vectors of each node to ensure that the semantic features are in the same vector space; within the unified space, performing hierarchical aggregation and feature compression on the node embedding vectors to form a semantic retrieval vector set including the embedding representations of design parameter nodes and process rule nodes;

[0092] Using the semantic retrieval enhancement module, semantic association retrieval of process rule node embedding vectors and design parameter node embedding vectors is performed on the semantic retrieval vector set, generating a similarity score matrix between rule nodes and design parameter nodes. The semantic retrieval enhancement module is based on the LightRAG framework and improves upon it by introducing causal constraint association modeling and differentiability feasibility fusion mechanism.

[0093] The generation of the similarity score matrix includes: performing feature normalization on each node vector based on the embedding vectors of design parameter nodes and process rule nodes in the semantic retrieval vector set to ensure that the semantic representations are comparable in a unified feature space; calculating the semantic similarity between nodes on the normalized embedding vectors, and generating an initial similarity score based on the inner product or cosine similarity measurement method in the embedding space; performing weighted smoothing and confidence correction on the initial similarity score, and adjusting the score distribution by combining node semantic weights and structural proximity factors to reduce the impact of semantic noise and node sparsity on the results; sequentially traversing the process rule nodes and design parameter nodes to form a two-dimensional score matrix that records the degree of semantic matching between nodes, and outputting the similarity score matrix between the rule nodes and the design parameter nodes;

[0094] Based on the similarity score matrix, causal constraint modeling is performed on the data relationship between rule nodes and design parameter nodes. Causal weights are marked for rule nodes with constraint conflicts, and a causal dependency graph is generated.

[0095] The generation of the causal dependency graph includes: constructing an initial set of associated edges between nodes based on the semantic association strength between rule nodes and design parameter nodes in the similarity score matrix; performing constraint consistency analysis on the initial set of associated edges to identify node pairs with multiple condition conflicts or unmet constraints, and calculating the causal correlation coefficient based on the degree of constraint conflict; determining the causal connection direction based on the causal correlation coefficient and the semantic adjacency relationship of nodes to form causal relationship edges between rule nodes and design parameter nodes; performing weight normalization processing on the causal relationship edges to comprehensively calculate the causal weight between nodes by combining the conflict strength and semantic influence; and generating a causal dependency graph that records the causal dependency relationship between rule nodes and design parameter nodes and their weight distribution.

[0096] Based on the similarity score matrix and causal dependency graph, semantic scores and causal weights are jointly calculated using semantic causality. A weighted fusion function is then used to perform causal constraint weighted fusion processing on the joint calculation results to generate a comprehensive score.

[0097] The semantic-causal joint calculation includes: based on the similarity score matrix and the node weights recorded in the causal dependency graph, performing data pairing and index alignment on the semantic scores and causal weights of rule nodes and design parameter nodes to establish a semantic-causal relationship table; performing weight balancing processing on each node pair in the semantic-causal relationship table, calculating the linear combination value of semantic score and causal weight to form the initial joint score between nodes; normalizing and correcting the confidence level of the initial joint score, and adjusting the score distribution by combining the node semantic density factor and the causal association strength to improve the matching consistency of semantic and causal information; and aggregating and calculating the joint scores of all rule nodes and design parameter nodes in the normalized space to generate a joint calculation result of the comprehensive relationship between the semantic matching and causal association of the recorded nodes.

[0098] The weighted fusion function takes the joint calculation result data as input, dynamically adjusts the fusion weight according to the constraint relationship in the causal dependency graph, obtains the comprehensive influence value of each rule node by calculating the difference between the expected value of the result of the rule node under the intervention condition and the expected value of the result under the non-intervention condition, and uses the comprehensive influence value as the comprehensive score result of the rule node.

[0099] Based on the comprehensive scoring results, the constraint satisfaction of each rule node is differentiable and optimized to generate a feasibility correction matrix.

[0100] The generation of the feasibility correction matrix includes: aligning the semantic similarity scores and constraint thresholds of rule nodes and design parameter nodes based on comprehensive scoring data; calculating the constraint deviation value of each rule node based on the difference between the actual values ​​of the design parameters and the corresponding rule constraint thresholds; constructing a feasibility optimization function with semantic matching terms and feasibility loss terms as optimization objectives based on the constraint deviation values ​​and semantic similarity scores; weighting the semantic similarity scores and feasibility loss terms using a balance coefficient to comprehensively reflect the balance between semantic matching accuracy and constraint satisfaction; performing gradient-differentiable iterative optimization on the feasibility optimization function, dynamically adjusting the ratio of semantic matching weights to feasibility loss weights according to the gradient change direction to minimize constraint deviation values ​​and improve the constraint feasibility consistency between rule nodes and design parameter nodes; and normalizing the optimized constraint satisfaction results, mapping them to an interval between zero and one to form a feasibility correction matrix representing the feasibility ratio of rule nodes and design parameter nodes within the constraint space.

[0101] By combining the comprehensive scoring results with the feasibility correction matrix, a joint screening calculation is performed to output a set of candidate process rules that meet the requirements of causal orientation and constraint feasibility.

[0102] The joint screening calculation includes: based on the comprehensive scoring results and the feasibility correction matrix, index matching and matrix alignment are performed on the comprehensive scoring value and feasibility ratio of each rule node to establish a joint evaluation matrix of rule nodes and design parameter nodes; a dual-threshold screening process is performed on each element in the joint evaluation matrix, the first threshold is determined based on the comprehensive scoring value to determine the semantic matching lower limit, and the second threshold is determined based on the feasibility ratio to determine the constraint satisfaction lower limit; rule nodes that simultaneously satisfy the semantic matching threshold and constraint feasibility threshold are marked and their corresponding design parameter node indices are recorded to generate a preliminary screening set; multidimensional normalization and confidence correction calculations are performed on the preliminary screening set, and the confidence weight of the screening results is adjusted according to the causal dependency strength and conflict relationship between nodes; the confidence weighted results are hierarchically aggregated and sorted, and conflicting rule nodes and redundant node connections are removed to generate a candidate process rule set that conforms to causal orientation and constraint feasibility.

[0103] In this embodiment, the generation of the target process rule set specifically includes:

[0104] Based on the node relationship structure in the candidate process rule set and process knowledge graph, rule nodes, design parameter nodes and corresponding causal relationship paths are extracted to construct provable retrieval graph data.

[0105] In the provable retrieval graph data, causal inference path search is performed on the associated paths of candidate rule nodes. The provability score of the path is calculated based on the path coverage and node association weight, and the set of paths with the highest provability score is selected.

[0106] Based on the set of paths with the highest provability scores, a minimum proof subgraph is extracted, which includes target rule nodes, design parameter nodes, and corresponding constraint relationships. The minimum proof subgraph is formed by jointly minimizing the relationship weights between nodes and semantic similarity.

[0107] Risk factors are extracted and weights are calculated for the rule nodes included in the minimum proof subgraph. A rule risk function is constructed based on process complexity, constraint conflict frequency and historical failure probability. The risk distribution of the rule nodes is weighted and the risk score of each rule node is output. The rule risk function is optimized with the expected value of non-compliance deviation at the confidence level.

[0108] Candidate rule nodes are sorted in ascending order based on rule risk scores to construct a rule risk ranking list. Conflicting rule nodes in the ranking list are then processed to resolve conflicts, retaining the set of rule nodes with the lowest risk scores and logical consistency.

[0109] After conflict resolution, the set of rule nodes is subjected to consistency verification and structural integrity verification to generate a set of target process rules that conform to causal orientation and risk constraints.

[0110] In this embodiment, the output of the process verification report and the updated design version specifically includes:

[0111] Map each rule node in the target process rule set to the design system parameter field, and generate a structured process instruction set based on the constraint type and control objective corresponding to the rule;

[0112] Standardized coding is performed on the operating parameters in the structured process instruction set, converting harness spacing limits, minimum bending radius limits, fixed point spacing limits, shielding level configurations, and path adjustment actions into control instruction formats that the system can recognize;

[0113] The structured process instruction set is written into the rule control module of the design system, and the design model parameters are updated synchronously and the corresponding rule sources are marked.

[0114] Perform compliance verification calculations in the updated design model, compare design parameters against the constraints of the target process rule set, and generate verification result data;

[0115] The verification results data are organized into a report structure to form a process verification report that records the process verification status, constraint matching status, and correction parameters;

[0116] Based on the verification results, the design system is updated and the rule annotation is synchronized, and a process verification report and an updated design version are output.

[0117] In this embodiment, the output of the optimized process verification report and the updated design version specifically includes:

[0118] Based on system feedback data, the difference comparison of rule nodes, design parameter nodes and constraint relationships recorded in the process rule base is performed to extract rule trigger deviation and verification deviation information.

[0119] The extracted deviation information is normalized, and the rule update weights are calculated based on the deviation magnitude and rule triggering frequency.

[0120] Adjust the scope of application, threshold and priority of process rule nodes according to the rule update weight, and remove invalid or conflicting rule nodes and mark the update status.

[0121] Based on the updated state of the rule nodes and design parameter nodes, incremental updates are performed on the applicable relationships, constraint relationships, conflict relationships and priority relationships recorded in the process knowledge graph, the relationship weights are adjusted and the entity alignment mapping data of the parameters and rule nodes are updated;

[0122] The updated nodes and relationships are recalculated by embedding, and new node embedding representations are generated and written into the process knowledge graph based on changes in node attributes and adjustments to relationship weights.

[0123] Based on the updated process knowledge graph, the semantic association and constraint consistency between process rule nodes and design parameter nodes are comprehensively recalculated, and an optimized process verification report and an updated design version are output.

[0124] Example 1:

[0125] To verify the feasibility of this invention in practice, it was applied to the wiring harness process design stage of an automobile manufacturing company. In the vehicle design process, the wiring harness system contains numerous branches, connectors, and path constraints. Traditionally, designers need to manually retrieve process rules based on experience and compare parameters. The numerous sources and frequent updates of these rules make it difficult to detect constraint conflicts in a timely manner, resulting in long design verification cycles and frequent version inconsistencies and rework. This invention aims to solve the problems of low efficiency in process rule matching, delayed rule updates, and reliance on manual judgment for design verification. By constructing a knowledge graph and semantic retrieval mechanism, it achieves automated association and dynamic updating of process rules.

[0126] In this implementation, wiring harness design data was first collected, including wire diameter, path geometry parameters, fixing point locations, insulation class, shielding configuration, electromagnetic interference source distribution, and heat source information. The data was then standardized, its units unified, and semantically mapped to generate a standardized design dataset. Simultaneously, the process rule base was analyzed, extracting applicable objects, parameter thresholds, execution actions, and priority information, which were then structured and encoded to generate process rule metadata. Subsequently, a process knowledge graph was constructed based on these two types of data, unifying the modeling of wiring harness components, design parameters, process rules, and constraint relationships to achieve global semantic-level association.

[0127] After the knowledge graph is constructed, this invention utilizes a semantic retrieval enhancement module to perform semantic vector encoding and similarity calculation on design parameter nodes and process rule nodes. Combined with causal constraint modeling and feasibility correction mechanisms, it performs semantic causal joint screening on rule nodes, outputting a set of candidate process rules that meet the constraints. During the matching process, the system can automatically identify conflicting relationships, eliminate invalid rules, and reallocate priorities, enabling complex process judgments that originally required manual comparison to be completed automatically in a short time, significantly improving matching accuracy and computational efficiency.

[0128] After outputting candidate rules, the system further performs provable retrieval and risk assessment, verifies the logical feasibility of each rule through causal reasoning paths, and estimates the risk weight of each rule based on historical failure data, thereby forming the final target process rule set. This set, after structured transformation, is written into the design system control module, automatically generating corresponding process instructions and updating model parameters. Designers can view rule triggering results, constraint satisfaction status, and verification status in real time on the system interface. The system simultaneously outputs a process verification report and updates the design version, achieving closed-loop verification of process control and traceable rule management.

[0129] During long-term system operation, this invention continuously collects verification reports and design system feedback data to perform discrepancy comparisons and embedding updates on the knowledge graph. The system automatically adjusts relationship weights and recalculates node embedding representations based on rule-triggered deviations, verification deviations, and changes in node attributes, achieving adaptive evolution of process knowledge. When new rules are added or old rules conflict, the system can automatically adjust within the update cycle, ensuring that subsequent design tasks directly apply the latest rule system.

[0130] Through application in multiple rounds of wiring harness design tasks by enterprises, this invention has achieved semantic association between design parameters and process rules, automatic feasibility determination, and dynamic rule optimization. This significantly reduces manual retrieval and comparison steps, improving the accuracy and consistency of rule matching. Through an incremental update mechanism of the knowledge graph, the system maintains synchronous evolution between rules and design data, ensuring the stability and compliance of design results. This embodiment demonstrates that this invention can effectively improve the intelligence level and process verification efficiency in the wiring harness design stage, providing reliable technical support for the digitalization of automotive wiring harness processes and knowledge-driven design.

[0131] Table 1. Performance Comparison of the Invention and the Traditional Manual Matching Method for Automotive Wiring Harness Process Rules

[0132] Indicator Categories Traditional methods Method of the present invention Average process rule matching time (minutes) 185 42 Match accuracy (%) 83.6 96.2 Rule conflict detection rate (%) 68.4 94.7 Process validation cycle (hours) 21.5 6.3 Design version write-back consistency rate (%) 79.2 95.5 Rule update response time (hours) 10.2 2.1 Percentage of human involvement (%) 64.8 18.7 System stable operation cycle (days) 30 90 Process validation error rate (%) 9.3 2.4

[0133] As can be clearly seen from Table 1, the method of the present invention is superior to the traditional method in many indicators.

[0134] The average process rule matching time was reduced from 185 minutes to 42 minutes, a reduction of approximately 77.3%, mainly due to the knowledge graph semantic modeling mechanism and semantic retrieval enhancement module introduced in this invention. This module enables the system to quickly locate candidate rules in a structured semantic space through semantic embedding and similarity comparison of design parameter nodes and process rule nodes, thereby significantly reducing the time spent on manual retrieval.

[0135] The matching accuracy improved from 83.6% to 96.2%, an increase of 15.1%. Traditional methods rely on experience-based judgment, and rule triggering conditions are easily misunderstood or overlooked. This invention, however, integrates rule applicability with contextual constraints through causal constraint modeling and a semantic-causal joint computation mechanism, avoiding mismatches. Furthermore, this invention introduces a weight adaptive update mechanism into the process knowledge graph, making the semantic connections between nodes more consistent with actual design logic, thereby further improving matching accuracy.

[0136] In terms of rule conflict detection rate, this invention achieves 94.7%, a 38.4% improvement over traditional methods. This is because the system can automatically identify logical conflicts and constraint contradictions between multiple rules through dynamic comparison and priority recalculation of conflict relationship nodes, providing early warnings to designers and thus avoiding frequent rework in the later stages of design. Simultaneously, the process verification cycle is shortened from 21.5 hours to 6.3 hours, a reduction of 70.7%, indicating that the verification process has shifted from manual serial inspection to automatic parallel verification by the system, achieving full-process digital verification.

[0137] The consistency rate of design version write-back increased from 79.2% to 95.5%, effectively ensuring the synchronization between process updates and design versions. This improvement is mainly due to the two-way synchronization mechanism adopted by the system when executing rule annotation and version write-back. It can verify and record the one-to-one correspondence between design parameters and rule sources during the update phase, ensuring that the updated version fully inherits the constraints and semantic tags.

[0138] The reduction in rule update response time is particularly significant, decreasing from 10.2 hours to 2.1 hours, an improvement of 79.4%. This invention automates rule updates through difference comparison and embedded recalculation. The system can instantly adjust node attributes and relationship weights upon detecting deviation information, achieving real-time self-learning and incremental updates. This significantly accelerates the adaptation speed to new standards or process changes during the design phase.

[0139] The proportion of human intervention decreased from 64.8% to 18.7%, a reduction of 71.2%, demonstrating the significant advantages of process knowledge graphs in design automation and intelligence. Through semantic parsing, causal determination, and feasibility correction, the system can complete most of the rule screening and verification processes without human intervention, requiring engineers to confirm only at necessary stages.

[0140] Furthermore, the system's stable operation cycle was extended from 30 days to 90 days, an improvement of 200%, indicating that the method maintains high reliability and consistency under continuous operation and version iteration; the process verification error rate decreased from 9.3% to 2.4%, a decrease of 74.2%, proving that the present invention has significant advantages in anomaly detection and semantic consistency verification.

[0141] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A knowledge graph-based automatic matching method for automobile wiring harness process rules, characterized in that, The method comprises the following steps: Collecting wire harness design data, performing standardization processing, and forming a standardized design data set; Analyzing a process rule library, extracting applicable objects, trigger conditions, parameter thresholds, priority levels, conflict relationships, execution actions, and applicable ranges, and generating a process rule metadata set; Performing semantic modeling and structured fusion on the standardized design data set and the process rule metadata set, and constructing a process knowledge graph with unified semantic association; Using a semantic retrieval enhancement module, performing semantic association analysis, causal constraint fusion, and feasibility determination processing on process rule nodes and design parameter nodes on the process knowledge graph, and outputting a candidate process rule set that meets the causal orientation and constraint feasibility; Based on the candidate process rule set, performing provable retrieval, risk assessment, and conflict resolution to generate a target process rule set that meets the causal orientation and risk constraints; Converting the target process rule set into process instructions to drive the design system to perform synchronous updating and rule labeling, and outputting a process verification report and an updated design version; Based on system feedback data, performing difference comparison and embedded updating on the process knowledge graph, adjusting the semantic association and constraint consistency of rules and design parameters, and outputting an optimized process verification report and an updated design version.

2. The method of claim 1, wherein the method is based on a knowledge graph of an automotive wiring harness process rule automatic matching method. The wire harness design data includes wire harness component identification, wire diameter, insulation level, shielding layer type, path geometry parameter, minimum bending radius, fixed point position and spacing, connector model, adjacent component material and distance, electromagnetic interference source list and intensity, heat source position and power, and applicable standard number. The standardization processing includes data structure unification, parameter unit normalization, coding standardization, data consistency checking, and parameter semantic mapping.

3. The method of claim 1, wherein the method comprises: The generation of the process rule metadata set specifically includes: Analyzing the text rules, logical expressions, and parameter constraint conditions in the process rule library related to wire harness design, hierarchically organizing the rule content according to the hierarchical structure and applicable range of the rules, and extracting original rule items related to wire harness components, path structure, electrical performance, and thermal management; Performing semantic segmentation and parameter mapping on the original rule items, converting the limiting conditions, operation constraints, and environmental constraints involved in the rules into structured fields, and performing normalization processing on the semantic fields; Based on the normalized structured fields, extracting applicable objects, trigger conditions, parameter thresholds, priority levels, conflict relationships, execution actions, and applicable ranges to generate an attribute vector for each rule; Based on the attribute vector, performing hierarchical labeling on the conflict relationships between rules, determining the conflict level according to the rule priority coefficient and constraint intensity weight, and generating a conflict labeling result; Based on the conflict labeling result, structurally encoding the execution actions, binding the operation instructions, parameter adjustment methods, and applicable conditions to form a rule execution template, and assigning a unique identification code to each process rule; Fusing the rule data after semantic analysis, attribute extraction, conflict labeling, and execution coding processing to generate a process rule metadata set.

4. The method of claim 1, wherein the method is based on a knowledge graph of an automotive harness process rule automatic matching method. The construction of the process knowledge graph specifically includes: Defining the node types and relationship types of the knowledge graph according to the parameter structure of the standardized design data set and the semantic attributes of the process rule metadata set; Based on the node type and the relationship type, the parameter information in the standardized design dataset is instantiated as node attribute data, and the semantic connection between nodes is established according to the rule objects corresponding to the fields in the process rule metadata set; Normalization mapping is performed on the continuous parameters in the node attribute data, the ratio is calculated according to the difference between the maximum value and the minimum value of the parameters, the numerical interval of the design parameters is unified, and the normalized node attribute data is output; Based on the normalized node attribute data, a semantic similarity comparison method is used to compare and map the parameter names, codes and units in the standardized design dataset, and the same meaning or homologous parameters are uniformly attributed to the same knowledge entity node, and entity alignment mapping data is generated; According to the entity alignment mapping data, the relationship weight between nodes is adjusted, the relationship strength is calculated according to the product of the parameter correlation degree and the rule importance, and the weighted node relationship data is obtained; Based on the normalized node attribute data, the entity alignment mapping data and the weighted node relationship data, node fusion and relationship integration processing are performed to construct a process knowledge graph with unified semantic association.

5. The method of claim 1, wherein the method is based on a knowledge graph of an automotive harness process rule automatic matching method. The output of the candidate process rule set specifically includes: Based on the node and relationship structure in the process knowledge graph, the semantic representation of the design parameter node and the process rule node is extracted, and a semantic retrieval vector set is constructed; Using a semantic retrieval enhancement module, the semantic association retrieval of the process rule node embedding vector and the design parameter node embedding vector is performed on the semantic retrieval vector set to generate a similarity score matrix of the rule node and the design parameter node, and the semantic retrieval enhancement module is improved based on the LightRAG framework and introduces a causal constraint association modeling and differentiability feasibility fusion mechanism; Based on the similarity score matrix, the data relationship between the rule node and the design parameter node is modeled with causal constraints, and the rule nodes with constraint conflicts are marked with causal weights to generate a causal dependency graph; Based on the similarity score matrix and the causal dependency graph, the semantic score and the causal weight are jointly calculated, a weighted fusion function is used to perform causal constraint weighted fusion processing on the joint calculation result to generate a comprehensive scoring result; The weighted fusion function takes the joint calculation result data as input, dynamically adjusts the fusion weight according to the constraint relationship in the causal dependency graph, obtains the comprehensive influence value of each rule node by calculating the difference between the result expected value under the intervention condition and the result expected value under the non-intervention condition, and takes the comprehensive influence value as the comprehensive scoring result of the rule node; Based on the comprehensive scoring result, the constraint satisfaction degree of each rule node is optimized to generate a feasibility correction matrix; Combined with the comprehensive scoring result and the feasibility correction matrix, joint screening calculation is performed to output a candidate process rule set that meets the causal orientation and constraint feasibility.

6. The method of claim 1, wherein the method is based on a knowledge graph of automotive harness process rules. The generation of the target process rule set specifically includes: Based on the candidate process rule set and the node relationship structure in the process knowledge graph, the rule node, the design parameter node and the corresponding causal association path are extracted, and a provable retrieval graph data is constructed; In the provable retrieval graph data, a causal reasoning path search is performed on an associated path of a candidate rule node, a path provability score is calculated according to path coverage and node association weight, and a path set with the highest provability score is selected; Based on the path set with the highest provability score, a minimum proof subgraph including a target rule node, a design parameter node and corresponding constraint relationships is extracted, and the minimum proof subgraph is formed by joint minimization of relationship weight and semantic similarity between nodes; Risk factor extraction and weight calculation are performed on the rule nodes included in the minimum proof subgraph, a rule risk function is constructed according to process complexity, constraint conflict frequency and historical failure probability, a weighted calculation is performed on the risk distribution of the rule nodes, and risk scores of the rule nodes are output, and the rule risk function takes the expected value of non-compliance deviation at a confidence level as the optimization target; According to the rule risk score, an ascending order sorting is performed on the candidate rule node, a rule risk sorting list is constructed, and a conflict resolution processing is performed on the conflict rule node in the sorting list, and a rule node set with the lowest risk score and logical consistency is reserved; Consistency checking and structural integrity verification are performed on the rule node set after conflict resolution, and a target process rule set meeting the causal orientation and risk constraint is generated.

7. The method of claim 1, wherein the method is based on a knowledge graph of automotive harness process rules. The output of the process verification report and the updated design version specifically includes: Map each rule node in the target process rule set to the design system parameter field, generate a structured process instruction set according to the constraint type and control target corresponding to the rule; Standardize the operation parameters in the structured process instruction set, convert the wire harness spacing limit, minimum bending radius limit, fixed point spacing limit, shielding level configuration and path adjustment action into a system recognizable control instruction format; Write the structured process instruction set into the rule control module of the design system, synchronously update the design model parameters and mark the corresponding rule source; Perform compliance verification calculation in the updated design model, compare the design parameters according to the constraint conditions of the target process rule set, and generate verification result data; Organize the verification result data into a report structure to form a process verification report recording the process verification state, constraint matching situation and modified parameters; According to the verification result data, perform version update and rule annotation synchronization on the design system, and output the process verification report and the updated design version.

8. The method of claim 1, wherein the method is based on a knowledge graph of automotive harness process rules. The output of the optimized process verification report and the updated design version specifically includes: Based on system feedback data, perform difference comparison on the rule nodes, design parameter nodes and constraint relationships recorded in the process rule library, extract rule trigger deviation and verification deviation information; Perform parameter normalization processing on the extracted deviation information, calculate rule update weight according to deviation amplitude and rule trigger frequency; According to the rule update weight, adjust the applicable range, threshold and priority of the process rule node, perform elimination processing on the failed or conflicting rule node and mark the update state; Based on the updated state of the rule node and the design parameter node, the applicable relationship, constraint relationship, conflict relationship and priority relationship recorded in the process knowledge graph are incrementally updated, the relationship weight is adjusted, and the entity alignment mapping data of the parameter and rule node are updated; The updated nodes and relationships are executed for embedding recalculation processing, and new node embedding representations are generated based on node attribute changes and relationship weight adjustments and written into the process knowledge graph; Based on the updated process knowledge graph, the semantic association and constraint consistency of the process rule node and the design parameter node are comprehensively recalculated, and the optimized process verification report and the updated design version are output.

Images