Text-driven high-precision involute spline design checking method

Abstract

The invention relates to the technical field of computer-aided engineering and mechanical transmission digital design checking, in particular to a text-driven high-precision involute spline design checking method. The method comprises the steps that an input text and a configuration set are obtained, and a semantic verification result is generated through format standardization, term normalization, lexical analysis, semantic association and other processing; on the basis, rule matching and expansion, numerical solution, inspection size calculation and geometric parameter reverse calculation are carried out, and inspection and reverse geometric parameters are obtained; then carrying out model construction, grid division, load application, solving, bearing capacity checking and wear life evaluation to form a bearing capacity and life calculation checking result; and finally, outputting a standardized report archiving structure through report template rendering, paragraph splicing, annotation proofreading, data arrangement, warning summarization and version packaging. According to the method, the automation of the whole process from text input to design checking is realized, the design precision and efficiency are improved, and the data traceability and term consistency are ensured.

Description

Technical Field

[0001] This invention relates to the field of computer-aided engineering and digital design verification of mechanical transmission, and in particular to a text-driven high-precision involute spline design verification method. Background Technology

[0002] Currently, most engineers use Excel and design software as their primary tools for spline design and verification. Using Excel typically involves setting calculation formulas in a spreadsheet according to relevant spline standards, with Excel automatically updating the results. The drawbacks of this method are that parameter input and results are usually on the same page, it cannot generate standardized calculation reports or complex logical judgments to indicate unreasonable design parameters, etc. Furthermore, calculating the verification dimensions of involute splines involves solving equations numerically, without a direct solution. Using design software generally involves the spline design function of mechanical or transmission design software. Its main drawback is limited support for spline standards, especially in load-bearing capacity verification. Supported load-bearing capacity verification methods are outdated and incomplete, and the geometric calculations of some software differ from the spline standards, resulting in inconsistencies between the calculation results and standard specifications. Additionally, design software is generally a black box; users cannot intervene in internal calculations or customize calculations and verification formulas. Most saved files are binary files, which are not directly readable by users, limiting their usability.

[0003] Furthermore, in the field of computer-aided engineering and digital design verification of mechanical transmission, existing solutions for design and verification scenarios that acquire input text blocks and constraint configuration sets and generate semantic verification result sets typically rely on general design software and table templates for parameter input, rule lookup, dimension conversion, and inspection item organization. These solutions suffer from limitations such as incomplete cross-step consistency checks and conflict detection, inconsistent standard rule base mapping and unclear version associations, and a lack of bidirectional association between the verification dimension domain and geometric parameter domain. Existing methods often rely on manual table lookups, fixed scripts, and forward solving, which are prone to matching ambiguities and rule expansion breakpoints under engineering context constraints, making it difficult to achieve stable verification and reverse structural implementation. For the joint processing of semantic verification result set structure, standard rule base and engineering context structure, existing technologies generally have shortcomings in the synchronization of matching and positioning, rule expansion, numerical solution and geometric inverse calculation, as well as in the judgment and recording stages. It is difficult to form a consistent process from the acquisition of input text blocks and constraint configuration sets to the inspection dimension table and report archiving structure in the application scenario of involute spline design verification. This results in the verification and inverse calculation structure being difficult to be directly referenced in model construction, mesh generation and load application. In the production and processing stages, problems such as inconsistent design annotation and delayed processing of inspection tool parameters are likely to occur. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a text-driven high-precision involute spline design verification method, comprising: The input text block and constraint configuration set are obtained, and format normalization and terminology normalization, lexical analysis and semantic association, consistency check and conflict detection are performed to generate a semantic verification result set structure. Obtain the semantic verification result set structure, perform matching and rule expansion processing, numerical solution, verification size calculation and geometric inverse processing, and generate the verification and inverse processing structure; The process involves obtaining the verification and reverse engineering structure, performing model building and mesh generation, load application and solution control, load verification and wear life assessment, and generating load and life criterion structures. The system acquires the load-bearing and lifespan criterion structure, performs report template rendering and paragraph splicing, annotation and proofreading, tool parameter organization, warning summary and version encapsulation, and generates a report archive structure.

[0005] Furthermore, the input text block and constraint configuration set include: The input text block includes a parameter assignment line, a standard selection line, a calculation instruction line, and a comment line; wherein, the parameter assignment line is used to declare the name-value pair and unit system, the standard selection line is used to specify the name and version identifier of the standard entry used, the calculation instruction line is used to trigger subsequent solving and retrieval, and the comment line is used to collect contextual hints and traceability information. The constraint configuration set includes a terminology dictionary, unit and tolerance standards, parameter boundaries, project metadata, and a load boundary configuration reference list. The terminology dictionary records the mapping of standard terms and synonyms; the unit and tolerance standards record the writing methods of length, angle, and dimension matching; the parameter boundaries record the upper and lower limits and enumeration sets; the project metadata records the part identifier, version identifier, and responsible person; and the load boundary configuration reference list only stores the name and version number for subsequent steps to look up.

[0006] Furthermore, the standardization of formatting and terminology includes: The format standardization includes unifying encoding, line breaks, whitespace, punctuation, and differences between full-width and half-width characters, converting units to the target format, removing illegal characters, and registering them in the processing log. The terminology normalization process includes merging synonyms according to the terminology dictionary, identifying line categories to label parameter assignment lines, standard selection lines, calculation instruction lines, and annotation lines, and performing template replacement and segmented indexing for unit and tolerance writing.

[0007] Furthermore, lexical analysis and semantic association processing include: The lexical analysis includes constructing a segmentation sequence of key, delimiter, value, unit, and comment segment for parameter assignment lines, extracting specification name, chapter number, and version identifier for standard selection lines, extracting verb phrases, object words, and parameter slots for calculation instruction lines, and introducing natural language processing capabilities for general word segmentation, phrase boundary recognition, and named entity recognition. The semantic association processing includes establishing three types of mappings for parameter assignment lines: parameter identifiers and units, parameter identifiers and boundaries, and parameter identifiers and comments; establishing three types of links for calculation instruction lines: verb phrases and object words, object words and parameter slots, and parameter slots and parameter identifiers; establishing binding relationships between specification entries and engineering context structure drafts for standard selection lines; and using a dependency relationship extractor to output dependency arcs and rewrite them as internal relation triples.

[0008] Furthermore, the matching, location, and rule expansion processes include: The matching and rule expansion process includes loading the semantic verification result set structure and standard rule base into the rule matching engine, locating candidate entries through the keyword index of the entry tree, forming a matching set through the cross-key convergence range of the index table, performing a two-stage screening based on the specification domain and unit domain in the engineering context structure to filter entries not in the selected version, removing entries lacking necessary input based on the available items of the parameter domain, and retaining unique entries by referring to the entry priority and upstream and downstream dependency records. The rule expansion process includes handling cross-entry references and internal references. Cross-entry references adopt a layer-by-layer dereferencing strategy, while internal references adopt an in-situ replacement strategy. Abstract placeholder terms in dimension and tolerance statements are replaced with parameter items in the engineering context structure or resolvable derived items in the mapping domain. Unit conversion is completed by the mapping table provided by the unit domain. Tolerance system conversion follows the system given in the specification domain. A rule control stack is constructed to store branch conditions, selected paths, and backtracking points.

[0009] Furthermore, the process of numerical solution, verification dimension calculation, and geometric inverse processing also includes: The numerical solution includes a dependency-based topology progression, starting with a subset of no-back edges to process basic entries without dependencies, then processing entries with single-layer and multi-layer dependencies, reading available initial values ​​and boundary ranges from the parameter domain of the engineering context structure, supplementing derived terms from the mapping domain, and performing interval shrinkage, step-size adaptive search, and tolerance sampling. The inspection dimension calculation includes mapping the geometric and tolerance parameter set into inspection dimensions of outer diameter, inner diameter, tooth thickness and reference ring, constructing a dimension chain sequence, connecting the basic geometric elements, derived geometric elements and tolerance items in a topological sequence, adding the gauge type, number of measuring points and reading method fields to each node, and sorting the dimension chain sequence according to gauge compatibility and item priority. The geometric reverse processing includes selecting the target dimension from the inspection dimension table, querying the rule mapping table to obtain the corresponding set of geometric elements and constraint relationships, extracting adjustable and non-adjustable items from the geometric and tolerance parameter set to form an adjustable item priority queue, eliminating invalid directions through monotonicity determination, approaching the target range through interval compression, verifying the boundary state through local recalculation, triggering multi-path back-substitution when a link conflict occurs, and reading the back-tracking point from the rule control stack through the back-substitution strategy.

[0010] Furthermore, the process of model building and mesh generation, load application and solution control also includes: The model construction and mesh generation process includes reading the part coordinate system, datum plane and tooth groove distribution, generating spline base, tooth profile, chamfer and transition fillet according to the geometric and tolerance parameter set, generating control sections locally according to the inspection dimension table and carrying out feature fusion, and establishing contact area sketch markings for inner and outer spline meshing surfaces, end face support surfaces and positioning datum penetration surfaces. The mesh generation process includes selecting the element type and integration strategy based on the material and contact parameter list, applying local mesh refinement based on the key geometric positions marked in the inspection dimension table, adopting a meshing strategy with a refinement level higher than the global level for the root transition fillet, tooth flank working surface and meshing start and stop zone, and using a multi-index joint judgment method for mesh quality control, evaluating the element twist, aspect ratio and minimum angle characteristics, and entering adaptive re-meshing when the threshold is not reached. The load application and solution control process includes mapping the torque, axial load, radial disturbance and angular misalignment defined in the load condition list to the corresponding geometric surfaces or feature lines, establishing a master-slave surface relationship through the contact list and writing friction parameters, applying constraint degrees of freedom at the positioning datum through the boundary constraint list, and generating load step schemes and contact activation schemes through the solution strategy suggestion items.

[0011] Furthermore, the process of load-bearing verification and wear life assessment also includes: The load-bearing verification and wear life assessment process includes extracting representative regions of the tooth flank working surface, root transition zone, and end face support from the finite element analysis result set; establishing an evaluation window for each region to record the spatial distribution statistics of maximum principal stress, equivalent stress, and contact pressure; performing threshold comparison, gradient continuity determination, and hot spot clustering; generating a load-bearing verification entry table; and extracting contact pressure time history, shear stress time history, and slip path time history for key regions. A load cycle spectrum is generated using rainflow counting and interval merging strategies, and damage accumulation and thickness attenuation accumulation are performed.

[0012] Furthermore, the process of rendering the report template and splicing paragraphs also includes: The report template rendering and paragraph splicing process includes reading the project name, part identifier, version identifier, selected specifications and unit system; reading the traceability key and paragraph anchor point configuration in the mapping domain; replacing placeholders using the template layout library and component list; generating cross-reference markers when encountering cross-segment references; automatically rearranging column widths and recalculating pagination; and performing paragraph splicing processing, including extracting dimension chain node trajectories, gauge types and unit writing specifications from the inspection dimension table to form dimension segment fragments; extracting threshold comparison results, gradient continuity judgment summaries and hotspot clustering summaries from the load verification item table to form load item fragments; extracting cycle spectrum summaries, cumulative records and status markers from the life assessment item table to form life item fragments; and extracting pending lists, anomaly tags and handling suggestions from the diagnostic domain summary to form warning segment fragments. The process also involves topic merging and sorting, prioritizing the chapter order given by the specification domain, and secondarily merging duplicate fields under the same topic according to the part identifier and key area index list numbering order. When fields conflict, the source priority and version timestamp in the mapping domain are queried to retain the higher-priority entry.

[0013] Furthermore, the process of labeling, verification, and processing of inspection tool parameters also includes: The annotation verification and tool parameter organization process includes scanning the dimension items in the dimension segment based on the bidirectional link between the report text block and the inspection dimension table, locating each item to the corresponding record in the inspection dimension table through the link key, verifying the consistency of unit writing specifications and tolerance system, verifying the closure relationship of the dimension chain, checking the consistency of terminology, generating an annotation verification result set, and selecting dimensions such as outer diameter, inner diameter, tooth thickness and reference circle from the inspection dimension table, and archiving the measuring tool type, number of measuring points and reading method to form a tool parameter sheet.

[0014] The key innovations of this invention include: (1) Obtain the semantic verification result set structure and standard rule base, perform matching and positioning, rule expansion and numerical solution, verification size calculation and geometric inverse processing, and generate verification and inverse processing structure.

[0015] (2) Obtain the set of configuration of the structure and load boundary for inspection and reverse calculation, perform model construction and mesh generation, load application and solution control, load verification and wear life assessment, and generate load and life criterion structure.

[0016] (3) Obtain the load-bearing and lifespan criterion structure and engineering context structure, perform report template rendering, paragraph splicing and annotation proofreading, tool parameter organization and warning summary and version encapsulation processing, and generate report archive structure.

[0017] The following are its main beneficial effects: (1) Around the semantic verification result set structure and standard rule base, through the continuous operation of matching and positioning, rule expansion, numerical solution, verification dimension calculation and geometric inverse calculation, a verification and inverse calculation structure is formed, which is directly called by model construction, mesh generation and load application and solution control, to achieve a verifiable input organization oriented to the engineering context structure.

[0018] (2) Around the verification and reverse calculation of the structure and load boundary configuration set, through the continuous operation of model construction, mesh generation, load application, solution control, load verification and wear life assessment, a load and life criterion structure is formed, which is rendered by the report template and referenced by paragraph splicing to complete the stable transfer from geometric parameters to criterion terms.

[0019] (3) Based on the load-bearing and life criterion structure and the engineering context structure, a report archiving structure is formed through the continuous operation of report template rendering, paragraph splicing, annotation proofreading, tool parameter organization, warning summary and version encapsulation, providing a traceable text and structural carrier for the subsequent organization and review of input text blocks and constraint configuration sets. Attached Figure Description

[0020] Figure 1 A text-driven human-computer interaction graphical interface layout diagram provided in this application embodiment; Figure 2 This is a flowchart illustrating a text-driven high-precision involute spline design verification method provided in an embodiment of this application. Detailed Implementation

[0021] The method described in this invention enables user interaction through a dual text editor interface. The left editor serves as the calculation input channel, supporting mixed editing and real-time modification of parameter assignment, standard selection, calculation instructions, and annotations. The right editor dynamically generates and renders calculation reports, supporting the viewing and revision of output content, making the interface user-friendly and easy to use. The system stores all input, output, and intermediate process files in plain text format, facilitating direct access to parameter values, calculation logic, and annotation information. Functionally, the system not only supports automatic design and verification based on standards and specifications but also provides fully customizable design and verification modes. A high-precision numerical solver ensures the accuracy of calculation results and possesses the ability to reverse engineer from inspection dimensions to geometric parameters. Regarding load-bearing capacity verification, the system integrates finite element analysis functionality, automatically performing structural solutions to provide quantitative references for design. Simultaneously, it expands the evaluation modules for spline wear and fatigue life, providing comprehensive life verification results.

[0022] As a preferred embodiment of this method, Figure 1 A text-driven human-computer interaction graphical interface layout diagram provided for embodiments of this application, such as... Figure 1As shown, the interface mainly includes a text input area on the left and a text output area on the right. Users can directly input original design text blocks containing parameter assignments, standard selection instructions, and design annotations in structured text format in the left editing area; the content is editable in real time. The bottom of the interface has function buttons for calculation, opening, saving, and report saving. When the user triggers a calculation, the system executes the method described in this invention, and the final standardized calculation report, fixture parameters, and verification conclusions will be dynamically rendered and displayed in the right output area. This implementation uses plain text files to store all input, output, and intermediate data, facilitating direct viewing and version management for users, and intuitively demonstrating the text-driven, user-friendly, and transparent characteristics of this invention.

[0023] In a preferred embodiment, text-driven interactive operation is implemented through a graphical user interface. The main layout of the interface includes two text editors: the left side is the input editing area, used to receive parameter assignments, calculation requirements, and design annotations input by the user in plain text form. All content can be freely edited, copied, and pasted, and can be saved as a general text file; the right side is the output display area, used to present the results generated by the system. The bottom of the interface has operation buttons such as "Calculate," "Open," and "Save." After the user completes the input, the calculation command is triggered, and the system will sequentially execute the following core processes: First, the input text on the left undergoes syntactic and semantic analysis to extract parameters and commands, and performs consistency checks and warning prompts; next, geometric parameter calculations and inverse calculations are performed, covering dimensional tolerance solutions, fixture recommendations, and inspection dimension calculations; then, the rationality of design parameters is verified, load-bearing capacity calculations based on multiple standards are performed, and wear and fatigue life assessments are conducted; finally, the system automatically generates a complete report containing all geometric parameters, inspection dimensions, tolerances, fixture parameters, and warning information, and can further drive geometric modeling and finite element analysis. The final output of this process, including a spline parameter table and an analysis report that can be directly used for engineering drawings, is displayed in the editor on the right and can be saved as a portable text file. This implementation intuitively demonstrates the technical features of this invention, which involve fully automatic generation of analysis reports from text input.

[0024] A preferred embodiment, referring to Figure 2 This is a flowchart illustrating a text-driven high-precision involute spline design verification method provided in an embodiment of the present invention. The process may include at least steps S100-S400: S100: Obtain the input text block and constraint configuration set, perform format normalization and terminology normalization, lexical analysis and semantic association, consistency check and conflict detection, and generate a semantic verification result set structure. S200: Obtain the semantic verification result set structure, perform matching and positioning and rule expansion processing, numerical solution, verification size calculation and geometric inverse processing, and generate the verification and inverse processing structure; S300, obtain the verification and reverse engineering structure, perform model construction and mesh generation, load application and solution control, load verification and wear life assessment, and generate load and life criterion structure; S400: Obtain the load-bearing and lifespan criterion structure, perform report template rendering and paragraph splicing, annotation proofreading and fixture parameter organization, warning summary and version encapsulation, and generate report archive structure.

[0025] S100: Obtain the input text block and constraint configuration set, perform format normalization and terminology normalization, lexical analysis and semantic association, consistency check and conflict detection, and generate a semantic verification result set structure. The input text blocks and constraint configuration sets originate from three channels: initialization interface input, file import, and version backfilling. The former involves operators inputting parameter items, rule lines, and comment lines in a structured area; the file import channel supports extracting text segments from historical archives and external documents; and the version backfilling channel originates from constraint configuration backfilling items in the report archive structure. An input text block is defined as a parsable text sequence organized by line, containing parameter assignment lines, standard selection lines, calculation instruction lines, and comment lines. Parameter assignment lines declare name-value pairs and unit systems; standard selection lines specify the name and version identifier of the adopted standard entry; calculation instruction lines trigger subsequent solving and retrieval; and comment lines collect contextual hints and tracing information. The constraint configuration set is defined as a set of rules and boundaries that runs throughout the entire process. It includes a terminology dictionary, unit and tolerance standards, parameter boundaries, project metadata, and a load boundary configuration reference list. The terminology dictionary records the mapping of standardized terms and synonyms; the unit and tolerance standards record the writing methods for length, angle, and dimensional fits; the parameter boundaries record upper and lower limits and enumeration sets; the project metadata records part identifiers, version identifiers, and responsible persons; and the load boundary configuration reference list only stores names and version numbers for later lookup. Specifically, the aforementioned input text block and constraint configuration set are sent together to the format standardization stage. Differences in encoding, line breaks, whitespace, punctuation, and full / half-width characters are uniformly processed; unit writing is converted to the target standard; illegal characters are removed and registered in the processing log; line breaks are merged without affecting semantic boundaries; and inline comments are retained as traceable fragments.

[0026] Furthermore, the terminology normalization step performs multiple rounds of mapping on the aforementioned text sequence. First, synonym merging is performed according to the terminology dictionary, converging different spellings of the same concept into standardized entries. Then, line category identification is carried out, marking parameter assignment lines, standard selection lines, calculation instruction lines, and annotation lines. Line category identification adopts a hybrid strategy of rule template matching and context verification. Next, template replacement and segmented indexing are performed on the writing of units and tolerances to form tags that can be referenced in subsequent parsing. Terminology normalization not only constrains word forms but also generates a draft of the engineering context structure to store project metadata, selected specifications, unit systems, and initial parameter tables. The draft is dynamically updated within this main step as the parsing progresses. All replacement actions and original fragments form a bidirectional index, and the index entries are entered into the processing log and backtracking list.

[0027] The lexical analysis stage segments and annotates the normalized text. For parameter assignment lines, a segmentation sequence is constructed for keys, delimiters, values, units, and comment segments; for standard selection lines, the specification name, chapter number, and version identifier are extracted; for calculation instruction lines, verb phrases, object words, and parameter slots are extracted. Natural Language Processing (NLP) capabilities are incorporated into the lexical analysis process for general word segmentation, phrase boundary recognition, and chunk merging; key nouns and part names are annotated using a Named Entity Recognition (NER) model; verb phrases are assigned grammatical roles using a Part-of-Speech (POS) model; and the model output is compared with the rule template and then incorporated into a unified parsing instruction set data structure. To reduce misclassification and omissions, boundary constraints are set at each segmentation node. If template conflicts or entity ambiguities are encountered, conflict candidates are recorded and written back to the alarm domain of the parsing instruction set. After parsing, the parsing instruction set and the engineering context structure draft enter the semantic association stage.

[0028] Semantic association is used to link lexical-level fragments into executable semantic units. First, three types of mappings are established for parameter assignment lines: parameter identifiers and units, parameter identifiers and boundaries, and parameter identifiers and comments. Then, three types of links are established for computation instruction lines: verb phrases and object words, object words and parameter slots, and parameter slots and parameter identifiers. Finally, for standard selection lines, a binding relationship is established between specification entries and the project context structure draft. During the linking process, a dependency extractor is used. The dependency arcs output by the extractor are rewritten as internal relation triples and stored in the association table within the parsing instruction set. The association table records the source line number, link strength, and source fragment. After association, the project context structure draft is upgraded to a project context structure, containing five parts: metadata field, specification field, unit field, parameter field, and mapping field. The mapping field points to the association table of the parsing instruction set, forming a stable reference. No evaluation is triggered here, and no standard rules are expanded; only structural construction and link solidification are completed.

[0029] Consistency checks involve static review and cross-checking of the project context structure and parsing instruction set. Static review includes line distribution verification, duplicate detection, null and illegal value identification, unit standardization verification, and parameter boundary existence confirmation. Cross-checking includes coupling checks between specification selection and parameter domains, constraint checks between parameter assignment and parameter boundaries, and slot filling completeness checks of the parsing instruction set. During the check, a verification entry table is constructed, recording the type, location, associated objects, and suggested actions. When a constraint conflict is detected, such as the same parameter having contradictory values ​​in different rows, the system triggers conflict detection. Conflict detection is completed collaboratively by two layers of strategies: the first layer determines the coverage relationship based on priority rules in the constraint configuration set, and the second layer determines the adoption order based on time sequence and source trust level. If the two layers of strategies still cannot resolve the issue, the entry is suspended, its suspension status is written to the conflict list, and a manual handling entry is opened in subsequent steps. All log entries, alarm entries, and conflict entries generated during the check process are uniformly numbered and bound to the original line number. The numbering strategy uses a project number prefix and an auto-incrementing sequence number for easy subsequent retrieval.

[0030] After conflict detection, a semantic verification result set structure is generated. This structure consists of three parts: the first part is a fixed snapshot of the engineering context structure, used to record the context state formed within this main step; the second part is a parsing instruction set, recording the executable calculation instruction sequence and semantic association information; the third part is a verification summary and a conflict list. The verification summary includes inspection statistics and a key issue overview, while the conflict list includes pending entries and adjudication trajectories. The semantic verification result set structure serves as the output field name of this main step, directly passed to rule solving and verification dimension inversion in subsequent main steps, and enters the input position "semantic verification result set structure" in S200. Simultaneously, the engineering context structure is accessed by the report template rendering and paragraph splicing in S400 when needed for subsequent text report generation and annotation proofreading. Version backfilling requirements form constraint configuration backfilling items when generating the report archive structure. These backfilling items enter the constraint configuration set in the next execution of this main step, forming a closed loop. The processing log and backtracking list remain searchable throughout the entire process, allowing downstream steps to review any original line and mapping trajectory by number.

[0031] The technical effects of this step can be summarized as follows: By serializing format standardization, terminology normalization, lexical parsing, semantic association, consistency checks, and conflict detection, a structured, executable, and traceable semantic verification result set is formed, and a stable input baseline is established for subsequent rule solving, geometric inverse solving, load assessment, and report generation.

[0032] S200: Obtain the semantic verification result set structure, perform matching and positioning and rule expansion processing, numerical solution, verification size calculation and geometric inverse processing, and generate the verification and inverse processing structure; The semantic verification result set structure originates from previous steps and includes a fixed snapshot of the engineering context structure, a parsing instruction set, a verification summary, and a conflict list. The standard rule base consists of multi-layered rule sets, including national standards, industry standards, enterprise standards, and project-defined rules, organized using a combination of entry trees and index tables. Each entry defines triggering conditions, involved geometric elements, dimensional and tolerance statements, exceptions, and priorities. Specifically, the semantic verification result set structure and the standard rule base are loaded into the rule matching engine. The rule matching engine reads verb phrases, object words, and parameter slots from the parsing instruction set, locates candidate entries using keyword indexes in the entry tree, and then converges the range using cross-keys in the index table to form a matching set. The matching set enters a two-stage filtering process. Stage one filters entries not included in the selected version based on the specification and unit fields in the engineering context structure. Stage two removes entries lacking necessary input based on available items in the parameter fields. If entries have mutual exclusion or inclusion relationships, a unique entry is retained by referring to the entry priority and upstream / downstream dependency records, and the source and reason for the replaced entry are recorded in the processing log.

[0033] After matching and location are completed, rule expansion begins. Rule expansion handles cross-entry and internal references for each retained entry. Cross-entry references employ a layer-by-layer dereferencing strategy, while internal references use an in-situ replacement strategy. During expansion, abstract placeholder terms in dimension and tolerance statements are replaced with parameter items in the engineering context structure or resolvable derived items in the mapping domain. Unit conversion is handled by a mapping table provided by the unit domain, and tolerance system conversion follows the system given in the specification domain, recording the conversion trajectory. To handle conditional branches in entries, rule expansion constructs a rule control stack. Stack frames store branch conditions, selected paths, and backtracking points. When a condition is unresolved, the alarm domain of the parsing instruction set is queried. If no criteria are found, the stack frame is marked as pending and added to the expansion pending list. After rule expansion, a rule mapping table is generated. The rule mapping table records the entry identifier, triggering conditions, input field name set, output field name set, dependency hierarchy, and conversion trajectory. Many-to-one relationships are aggregated into a single mapping record using a merging strategy. The merging strategy prioritizes entry priority, followed by the entry source hierarchy and specification domain preferences of the engineering context structure.

[0034] After the rule mapping table stabilizes, numerical solution is initiated. Numerical solution generates geometric and tolerance parameter sets for the mapping records. Specifically, the solver proceeds according to the dependency topology, starting with the subset without back edges and processing basic entries without dependencies, then processing entries with single-layer and multi-layer dependencies. Before each record enters the solution, available initial values ​​and boundary ranges are read from the parameter domain of the engineering context structure, derived terms are supplemented from the mapping domain, and constraint information related to the record is retrieved from the parsing instruction set. The solution process includes three stages: interval shrinkage, adaptive step-size search, and tolerance sampling. Interval shrinkage compresses feasible intervals based on boundary ranges; adaptive step-size search adjusts the step size based on historical iteration information; and tolerance sampling generates representative discrete points and calculates statistics under tolerance constraints. If an infeasible state occurs, backtracking is triggered. Backtracking returns to the previous feasible node according to the rule control stack record, marking this segment as an abandoned path and writing the entry identifier and timestamp in the diagnostic domain. After the numerical solution is completed, a set of geometric and tolerance parameters is generated. The parameter items are accompanied by source entries, conversion trajectories, boundary states and convergence indicators, and the derivation relationships are registered in the extended domain for direct reference by downstream applications.

[0035] After the geometric and tolerance parameter set stabilizes, the inspection dimension calculation begins. This calculation revolves around a readable dimensional system for field gauges and inspection tools, mapping the geometric and tolerance parameter set to inspection dimensions such as outer diameter, inner diameter, tooth thickness, and reference ring. The mapping rules are derived from the inspection section of the standard rule base. Specifically, a dimension chain sequence is first constructed, connecting basic geometric elements, derived geometric elements, and tolerance items topologically. Each node is appended with fields for gauge type, number of measuring points, and reading method. The unit domain format is then mapped to the dimension chain sequence for unified writing. Subsequently, the dimension chain sequence is sorted based on gauge compatibility and item priority to obtain a draft inspection dimension list. When gauge compatibility conflicts occur, the inspection dimension calculation queries the project metadata and load boundary configuration reference directory in the project context structure. If the project metadata provides a gauge preference, the preference path is adopted; otherwise, a candidate gauge option is generated in the diagnostic domain and marked for manual confirmation. After the inspection dimensions are calculated, an inspection dimension table is generated. Each inspection dimension includes source entries, dimension chain node trajectories, gauge types and unit writing specifications, and links to the corresponding geometric parameter items via links for subsequent traceability.

[0036] Subsequently, geometric reverse engineering is performed. Geometric reverse engineering constructs a reverse adjustment sequence based on the mapping between inspection dimensions and geometric parameters. First, the target dimension is selected from the inspection dimension table, and the corresponding set of geometric elements and constraint relationships are obtained by querying the rule mapping table. Then, adjustable and non-adjustable items are extracted from the geometric and tolerance parameter set, forming a priority queue for adjustable items. The reverse search eliminates invalid directions through monotonicity determination, approaches the target range through interval compression, and verifies the boundary state through local recalculation. When a link conflict occurs, multi-path back-substitution is triggered. The back-substitution strategy reads the backtracking points of the rule control stack and attempts to replace candidate items or change the gauge adaptation scheme. If unresolved constraints still exist, the dimension is marked as pending and written as a reverse engineering unresolved record in the diagnostic domain. After geometric reverse engineering is completed, a geometric reverse engineering set is formed. The geometric reverse engineering set records the geometric parameter revision amount, revision order, and back-substitution trajectory for each dimension, and is merged with the geometric and tolerance parameter set to generate an updated geometric and tolerance parameter set, maintaining the link with the engineering context structure, and only adding a revision snapshot in the extended domain.

[0037] Throughout the aforementioned process, anomaly handling and logging mechanisms are implemented. Discarded candidate entries, replaced reference entries, backtracked solution nodes, and marked pending items are all written to the diagnostic domain. The diagnostic domain is indexed by entry identifier and time sequence, supporting rapid retrieval across steps. Processing logs are persisted to disk at key nodes, containing input summaries, output summaries, and decision information for subsequent report template rendering and paragraph concatenation. Within this main step, the project context structure is updated only incrementally in the extended domain, without rewriting previous snapshots, facilitating review of the valid state at any point in subsequent steps. After all processing is completed, the product organization is an inspection and reverse structure, which consists of three parts: the first part is the updated set of geometric and tolerance parameters, the second part is the inspection dimension table, and the third part is the diagnostic domain. The output field is named Inspection and Reverse Structure. In the subsequent main steps, the inspection and reverse structure is directly read as the input position of S300 and used for model building, mesh generation, load application and solution control, load verification and wear life assessment. At the same time, the inspection dimension table and the diagnostic domain are referenced in the report template rendering and paragraph splicing in S400. The source trajectory and number are used to generate the dimension segment and tolerance segment and warning summary in the report text block, forming a field flow link that can be reviewed and traced.

[0038] The technical effects of this step can be summarized as follows: By linking matching and positioning, rule expansion, numerical solution, inspection dimension calculation and geometric inverse calculation in sequence, a structured and verifiable inspection and inverse calculation structure is formed, which provides stable input for subsequent load-bearing and life assessment and finite element analysis, and accumulates complete process information for easy cross-step calling and retrieval.

[0039] S300, obtain the verification and reverse engineering structure, perform model construction and mesh generation, load application and solution control, load verification and wear life assessment, and generate load and life criterion structure; The inspection and reverse engineering structure is derived from previous steps and includes an updated set of geometric and tolerance parameters, an inspection dimension table, and a diagnostic domain. The load boundary configuration set is maintained by the engineering context structure and includes a load case list, a boundary constraint list, a material and contact parameter list, and solution strategy suggestions. The inspection dimension table is used to mark key geometric locations readable by the gauges, the geometric and tolerance parameter set is used to drive 3D geometric reconstruction and local feature scale setting, and the diagnostic domain is used to handle alarms and pending items from the previous rule matching and reverse engineering stages. Specifically, the aforementioned inspection and reverse engineering structure and load boundary configuration set are fed into the model building pipeline. The pipeline first reads the part coordinate system, datum plane, and tooth groove distribution, and generates feature bodies such as spline matrix, tooth profile, chamfers, and transition fillets according to the geometric and tolerance parameter set. If there are non-standard tooth tip modification or local material filling requirements, a control section is generated locally based on the target dimension given in the inspection dimension table, and feature fusion is performed. During model building, a contact area sketch is created simultaneously to identify the internal and external spline meshing surfaces, end support surfaces, and positioning reference penetration surfaces. If there are geometric conflicts or dimensional chain break records in the diagnostic domain, isolation features are inserted at the corresponding locations and written to the construction log during the model building phase. The construction log maintains the traceable relationship between the geometry and the original entries.

[0040] Furthermore, mesh generation is triggered after model construction. The mesher selects element types and integration strategies based on the material and contact parameter list, and applies local mesh refinement based on key geometric locations marked in the inspection dimension table. Root transition fillets, tooth flank working surfaces, and meshing start and stop zones employ a refinement level higher than the global level. Mesh quality control uses a multi-index joint judgment method, batch evaluating element twist, aspect ratio, and minimum angular characteristics. Mesh blocks that do not meet the threshold are entered into an adaptive re-meshing queue. When re-meshing fails in geometrically extremely thin or small radius areas, the mesher calls an alternative strategy to generate local transition elements and records the reason for the alternative. After meshing is completed, a finite element model structure is generated. This structure includes element sets, node sets, material assignment tables, contact pair lists, mesh quality statistics, and geometry-to-element mapping keys. These mapping keys are linked back to the dimension chain nodes recorded in the inspection dimension table for subsequent result write-back and traceability.

[0041] Load application and solution control are based on a finite element model structure. The load case list defines load cases such as torque, axial load, radial disturbance, and angular misalignment; the boundary constraint list defines constraint items such as fixed supports, symmetry constraints, and flexible connections; and the material and contact parameter list defines attributes such as elastic parameters, surface roughness, normal and tangential contact stiffness, and friction coefficient. Specifically, load cases are mapped to corresponding geometric surfaces or feature lines. A master-slave relationship is established through the contact list, and friction parameters are written in. Constraint degrees of freedom are applied at the positioning datum through the boundary constraint list. Initial states are supplemented through additional items such as temperature or assembly preload. Solution control is driven by solution strategy suggestions, which include the nonlinearity level, initial step size, iteration upper limit, and convergence criterion selection. The controller generates load step schemes and contact activation schemes accordingly. If the diagnostic domain indicates the existence of unresolved dimensions or a choice between two gauges, a parameter placeholder is generated at the corresponding location. During the solution control phase, a catch-all value is used for this placeholder, and a prominent mark is added to the result domain. After the load is applied, a solution input snapshot is constructed, and the snapshot and the finite element model structure are entered into the solver together.

[0042] The solution phase invokes the Finite Element Analysis (FEA) algorithm kernel to perform structural solving. Understandably, the solver advances the calculation according to the load step scheme. Nonlinear contact and material nonlinearity are handled collaboratively by contact iteration and tangential stiffness updates. The convergence criterion uses a comprehensive judgment of three quantities: displacement, internal force imbalance, and contact penetration. When a divergence trend or contact flipping occurs, the controller reduces the step size and reverts to the previous stable state, while simultaneously writing the load step number and triggering reason into the diagnostic domain. For combined load cases involving angular misalignment and radial disturbance, the solver reconstructs the contact state in each substep to maintain the correctness of the meshing pair sequence. If insufficient activation or excessive adhesion of contact pairs occurs within a substep, the friction parameter correction template is called in that substep to update the tangential stiffness term and the iteration continues. After the solution is completed, the results of displacement field, stress field, contact pressure field and energy components are summarized to form a finite element analysis result set. The result set also records the working condition number, step size trajectory, iteration statistics and anomaly label. The result set is consistent with the geometry to element mapping key and is used to locate specific tooth surface areas or root transition zones in the future.

[0043] Load-bearing capacity verification is performed based on an itemized assessment of the finite element analysis result set. Specifically, representative areas of the tooth flank working surface, root transition zone, and end face support are extracted from the result set. An evaluation window is established for each area, recording the spatial distribution statistics of maximum principal stress, equivalent stress, and contact pressure. The evaluation windows are then mapped back to the items pointed to by the rule mapping table, generating a load-bearing capacity verification item table. The item table performs three procedures on each evaluation window: threshold comparison, gradient continuity determination, and hotspot clustering. Threshold comparison corresponds to the load-bearing capacity requirements in relevant standards, gradient continuity is used to identify local stress concentrations, and hotspot clustering is used to summarize spatially adjacent high-response areas. If there are unresolved dimensional or gauge records in the item table and diagnostic domain, the corresponding item is marked as pending verification, and this status is written to the extended domain. After completing the three procedures, the load-bearing capacity verification item table and a key area index list are output. The key area index is used to support the local load history sampling for subsequent life assessment.

[0044] Wear life assessment is performed based on a key region index list and a finite element analysis result set, using time-history-based cumulative calculations. Further, the cyclic load parameters in the load case list are expanded into a load history sequence. The life engine extracts the contact pressure time history, shear stress time history, and slip path time history for each key region, generating a load cycle spectrum using rainflow counting and interval merging strategies. Damage accumulation and thickness attenuation accumulation are then performed according to the material curves and wear coefficient library provided in the material and contact parameter list. The accumulation process writes the region-level damage increment and thickness reduction after each cycle block is completed. To handle multi-region coupling, the life engine performs load sharing redistribution and recursive accumulation for adjacent regions with shared boundaries. If a non-physical abrupt change occurs during the recursion, the corresponding load step's anomaly label in the solution phase is checked back, triggering a local recalculation. The recalculation uses the same mesh and an updated step size scheme, without altering the global iteration history. After all cycle blocks are processed, a life assessment entry table is generated. This table includes region number, cycle spectrum summary, cumulative records, and status markers, and is cross-indexed with the load verification entry table for easy summarization during the reporting phase.

[0045] Throughout the aforementioned process, anomaly logging, metadata storage, and traceability key maintenance are maintained. Construction logs, mesh quality statistics, solution input snapshots, iteration statistics, anomaly tags, key area indexes, and two types of entry tables are uniformly written into the extended domain. The extended domain shares a unified numbering system with the metadata domains in the project context structure. When events such as model construction failure, mesh quality consistently failing to meet thresholds, or multiple solution convergence rollbacks occur, the system generates remedial suggestion entries and writes them into the diagnostic domain. These remedial suggestions do not change the solution results but only provide indicative information for subsequent reporting stages. After all processing is complete, a load-bearing and lifespan criterion structure is generated. This structure consists of four parts: a load-bearing verification entry table, a lifespan assessment entry table, a key area index list, and a diagnostic domain summary. These serve as the output field names for this main step and enter downstream stages. In subsequent steps, the load-bearing and lifespan criterion structure fields are directly read for report template rendering and paragraph splicing. Simultaneously, corresponding numbers are written back to the size, tolerance, and warning summary sections of the report text block, forming a field flow link consistent with previous steps.

[0046] The technical effects of this step can be summarized as follows: Through the continuous operation of model building, mesh generation, load application and solution control, combined with itemized load verification and time-based wear life assessment, a load and life criterion structure is generated, which meets the requirements for cross-step reuse and traceability, and supports the content organization and numbering mapping in the subsequent report generation stage.

[0047] S400: Obtain the load-bearing and lifespan criterion structure, perform report template rendering and paragraph splicing, annotation and proofreading and fixture parameter organization, warning summary and version encapsulation, and generate report archive structure; The load-bearing and lifespan criterion structure originates from previous steps and includes a load-bearing verification item table, a lifespan assessment item table, a key area index list, and a diagnostic domain summary. The engineering context structure originates from the initial stage and includes a metadata domain, a specification domain, a unit domain, a parameter domain, and a mapping domain, and maintains a link with the inspection dimension table and diagnostic domain in the inspection and reverse engineering structure. Specifically, the load-bearing and lifespan criterion structure and the engineering context structure are jointly sent to the report generation pipeline. The pipeline first reads metadata such as project name, part identifier, version identifier, selected specification, and unit system, and then reads the traceability key and paragraph anchor point configuration in the mapping domain to establish an initial index for subsequent paragraph positioning and cross-referencing. If there are unresolved items or disposal suggestion items in the diagnostic domain summary, the pipeline reserves a placeholder in the corresponding topic position and writes the number and source line number in the record plane to maintain consistency with the mapping of upstream items.

[0048] The report template rendering is based on a template layout library and a component list. The template layout library defines the layout elements and pagination strategies for areas such as the cover, table of contents, parameter table, size section, tolerance section, load capacity section, lifespan section, warning section, and appendix. The component list defines the binding rules for paragraph-level placeholders, table cell placeholders, and anchor placeholders. Specifically, the item name, part identifier, version identifier, and date in the metadata field are bound to the cover and header / footer; the specification field and unit field are bound to the unit writing and tolerance system in the parameter table; and the traceability key in the mapping field is bound to the paragraph anchor. After loading the template, the renderer replaces the placeholders item by item. When encountering cross-paragraph references, it calls the anchor index library to generate cross-reference markers. When a field in the load capacity check item table or lifespan assessment item table exceeds the template column width, the renderer triggers automatic column width rearrangement and pagination recalculation, and writes back the paragraph number and page number lookup table after re-pagination to avoid reference failures caused by page number mismatches. After rendering, a draft of the report text block is obtained. This draft contains semantic anchors and index entries, which can be directly called by subsequent paragraph concatenation.

[0049] The paragraph splicing process integrates content fragments from various sources based on the initial draft of the report text block. Specifically, dimension segment fragments are formed by extracting dimension chain node trajectories, gauge types, and unit writing specifications from the inspection dimension table; load-bearing segment fragments are formed by extracting threshold comparison results, gradient continuity judgment summaries, and hotspot clustering summaries from the load-bearing verification item table; life-bearing segment fragments are formed by extracting cycle spectrum summaries, cumulative records, and status markers from the life-bearing assessment item table; and warning segment fragments are formed by extracting pending lists, anomaly tags, and handling suggestions from the diagnostic domain summary. The splicer performs topic merging and sorting on the fragments, prioritizing the chapter order given by the specification domain, and secondarily following the numbering order of part identifiers and key area index lists, merging duplicate fields under the same topic; if field conflicts exist between fragments, the splicer queries the source priority and version timestamp in the mapping domain, retaining high-priority entries and writing folding marks at low-priority entries. The folding marks indicate the source and reason for replacement in the warning segment. After completion, a report text block is formed, maintaining the paragraph anchor points and adding a source index at the end of the paragraph for subsequent annotation and proofreading.

[0050] The annotation verification is conducted based on a two-way link between the report text block and the inspection dimension table. The annotation verification rule set consists of three categories of rules: terminology consistency rules, unit and tolerance writing standard rules, and dimension chain closure and pointing correctness rules. Specifically, the verifier first scans the dimension items in the dimension segment, locates each item to the corresponding record in the inspection dimension table through the link key, and then maps the unit writing standard and tolerance system in the record back to the report text block. If there is an inconsistency between the two sides, a difference entry is generated and the source line number is marked. Then, the dimension chain closure relationship is checked by comparing the preceding and following references in the report text block with the node order in the inspection dimension table. If a break in the order or a pointing error is found, the break position and expected pointing are written in the difference entry. The terminology consistency is checked again by matching the key terms appearing in the report text block with the terminology entries in the engineering context structure item by item. If synonyms are mixed, a unified suggestion is output and a verification annotation mark is inserted in the segment. After the verification is completed, an annotation verification result set is generated, which includes a list of difference entries, a set of annotation locations, and a set of unified terminology suggestions for subsequent processing and encapsulation.

[0051] The tooling parameter compilation is a structured compilation designed to meet the readability requirements of the connection between inspection and tooling. Specifically, it selects dimensions such as outer diameter, inner diameter, tooth thickness, and reference ring from the inspection dimension table, and archives the tool type, number of measuring points, and reading method. This is combined with the unit and specification fields in the engineering context to form a tooling parameter sheet. When tool compatibility is marked as a choice between two options in a previous step, the parameter compiler marks the corresponding entry in the tooling parameter sheet as pending confirmation and includes two sets of parameters. The parameter source and entry identifier are written into the reference area of ​​the diagnostic domain summary for easy manual review later. After parameter compilation, the parameter sheet is transcribed into a table fragment and inserted into the tooling parameter section of the report text block. A downloadable copy of the parameter list is also generated in the appendix, sharing the same numbering system and anchor point relationship as the main text.

[0052] The warning summary is based on the event classification of the diagnostic domain summary. Specifically, events such as unresolved rule matching, solution rollback, contact reversal, gauge selection, local recalculation, and mesh replacement are summarized in a hierarchical manner. The hierarchical rules are provided by the metadata domain of the engineering context structure, which provides the level threshold. A summary paragraph is generated for each type of event, containing the event number, source entry, and time sequence, and mapped back to the relevant paragraph in the report text block. Jump anchors are added to the corresponding paragraphs to form a bidirectional link from the warning paragraph to the main text. If there are duplicate events across different operating conditions, the aggregator merges the numbers and retains the repeated page numbers in a paginated annotation manner to reduce redundancy while retaining traceability.

[0053] Version encapsulation packages and archives the report text blocks, annotation and verification result sets, and gauge parameter sheets. Specifically, the encapsulator generates two types of carriers: editable and non-editable versions. The editable version retains paragraph anchors, traceability keys, and annotation tags, while the non-editable version locks the format and page number mapping. Simultaneously, version metadata is generated, recording the template version, specification version, unit system, and mapping domain summary, and creating a snapshot list of references to the verification and life assessment item tables. To support the pre-configuration of the next round of tasks, the encapsulator generates constraint configuration backfill items based on the annotation and verification result set and warning summary. These constraint configuration backfill items correspond to constraint configuration set entries in the project context structure, including terminology unification suggestions, unit and tolerance writing settings, and the final state of gauge adaptation selection. After version encapsulation is complete, a report archive structure is output, containing editable carriers, non-editable carriers, version metadata, a snapshot list, and constraint configuration backfill items.

[0054] Understandably, the report archive structure serves as the output field name for this main step, while also acting as a link between steps. The report text blocks and their paragraph anchors are used for version comparison and difference tracking in subsequent iterations. The annotation and proofreading result set is reviewed in the next task to guide the revision of the input text blocks. Constraint configuration backfill items are written into the constraint configuration set of the next initial stage, serving as the pre-input for the corresponding step, forming a closed-loop link from result to configuration. Thus, the output field name is "report archive structure," the next input location is the constraint configuration backfill items in the constraint configuration set of the initial stage, and the numbering and mapping relationship remains consistent with the project context structure, supporting the location and traceability of any paragraph and any item in subsequent runs.

[0055] The technical effects of this step can be summarized as follows: Through the sequential processing of template rendering, paragraph splicing, annotation and proofreading, tool parameter organization, warning summary and version encapsulation, the transformation from analytical data to standardized text and structured archives is completed; the output has the ability to be continuously reviewed and traced back, and can be stably referenced by previous and subsequent steps; the constraint configuration backfill items formed in the version encapsulation provide direct input for the next round of tasks, maintaining process closure and terminology consistency.

Claims

1. A text-driven high-precision involute spline design checking method, characterized in that, Comprise: Obtain input text block and constraint configuration set, perform format normalization and term normalization processing, lexical analysis and semantic association processing, consistency check and conflict detection processing, generate semantic verification result set structure; Obtain semantic verification result set structure, perform matching positioning and rule expansion processing, numerical solution, geometric parameter calculation, inspection size calculation and geometric parameter back calculation processing, generate geometric parameters, inspection size parameters, gauge design parameters, inspection size back calculated geometric parameters, bearing capacity estimation, etc. Results; Through the geometric parameters and the semantic verification result set structure, according to the related formulas and empirical formulas of the standard specification, the preliminary formula calculation of the bearing capacity checking and the wear life evaluation are carried out, then the three-dimensional model construction and the mesh division processing, the load application and the solution, the bearing capacity checking and the wear life evaluation based on FEA are carried out, and finally the comprehensive bearing capacity checking and wear life evaluation results are generated; Through the geometric parameters, the inspection size parameters, the inspection size back calculated geometric parameters, the comprehensive bearing capacity checking and wear life evaluation results, the report template rendering and paragraph splicing processing is carried out, the calculation report is generated, the warning summary and version packaging processing is carried out, and the report archiving structure is generated.

2. The method of claim 1, wherein, The input text block and the constraint configuration set comprise: The input text block contains parameter assignment line, standard selection line, calculation instruction line, annotation description line; wherein, the parameter assignment line is used to declare name-value pair and unit system, the standard selection line is used to indicate the adopted specification item name and version identification, the calculation instruction line is used to trigger subsequent solution and retrieval, and the annotation description line is used to collect context prompt and traceability information; The constraint configuration set contains term dictionary, unit and tolerance system, parameter boundary, project metadata, load boundary configuration reference list; wherein, the term dictionary records the mapping of specification terms and synonyms, the unit and tolerance system records the writing method of length, angle and size fit, the parameter boundary records the upper and lower limits and enumeration set, the project metadata records the part identification, version identification and responsible person, and the load boundary configuration reference list only stores the name and version number for subsequent steps to find.

3. The method of claim 1, wherein, The format normalization and term normalization processing comprise: The format normalization comprises unified coding, line feed, space, punctuation and full and half difference, unit writing conversion to target system, illegal character elimination and registration in processing log; The term normalization processing comprises synonym merging according to the term dictionary, line type identification to mark parameter assignment line, standard selection line, calculation instruction line and annotation description line, and unit and tolerance writing template replacement and segmentation indexing.

4. The method of claim 1, wherein, The lexical analysis and semantic association processing comprise: The lexical analysis comprises constructing key, delimiter, value, unit, annotation segment split sequence for parameter assignment line, extracting specification name, chapter number and version identification for standard selection line, extracting verb phrase, object word and parameter slot for calculation instruction line, and introducing natural language processing capability for general word segmentation, phrase boundary identification and named entity recognition; The semantic association processing includes establishing three types of mappings of parameter identification and unit, parameter identification and boundary, and parameter identification and annotation for parameter assignment lines, establishing three types of links of verb phrase and object word, object word and parameter slot, and parameter slot and parameter identification for calculation instruction lines, establishing a binding relationship of specification item and engineering context structure draft for standard selection lines, and outputting dependency arcs by a dependency relationship extractor and rewriting into internal relationship triples.

5. The method of claim 1, wherein, The matching positioning and rule expansion processing includes: The matching positioning and rule expansion processing includes loading the semantic verification result set structure and the standard rule library into a rule matching engine, positioning a candidate item through a keyword index of an item tree, converging a range through cross keys of an index table to form a matching set, and performing two-stage screening to filter out items not in a selected version according to a specification domain and a unit domain in the engineering context structure, to eliminate items missing necessary inputs according to available items of a parameter domain, and to reserve unique items by referring to item priority and upstream and downstream dependency records; The rule expansion processing includes processing cross-item references and internal references, the cross-item references adopt a layer-by-layer dereferencing strategy, the internal references adopt an in-place replacement strategy, abstract placeholder terms in size and tolerance statements are replaced by parameter items in the engineering context structure or derivable items in the mapping domain, unit conversion is completed by a mapping table provided by the unit domain, and tolerance system conversion follows a system given in the specification domain, and a rule control stack is constructed to save branch conditions, selected paths, and backtracking points.

6. The method of claim 1, wherein, The numerical solution, verification size calculation, and geometric reverse processing further include: The numerical solution includes advancing according to a dependency topology, processing basic items without dependency from a loop-free subset, processing items with single-layer and multi-layer dependency, reading available initial values and boundary ranges from a parameter domain of the engineering context structure, supplementing derivable items from a mapping domain, and performing interval contraction, step adaptive search, and tolerance sampling; The verification size calculation includes mapping geometric and tolerance parameter sets to verification dimensions of an outer diameter type, an inner diameter type, a tooth thickness type, and a reference circle type, constructing a size chain sequence, concatenating basic geometric elements, derived geometric elements, and tolerance items according to a topology, adding a gauge type, a number of measurement points, and a reading mode field to each node, and sorting the size chain sequence according to gauge adaptation and item priority; The geometric reverse processing includes selecting a target dimension from a verification size table, querying a rule mapping table to obtain a corresponding geometric element set and constraint relationship, extracting adjustable items and non-adjustable items from geometric and tolerance parameter sets, forming an adjustable item priority queue, removing invalid directions through monotonicity determination, approaching a target range through interval compression, rechecking a boundary state through local recalculation, and triggering multi-path back substitution when a link conflict occurs, and a back substitution strategy reads backtracking points of a rule control stack.

7. The method of claim 1, wherein, The model construction and mesh division processing, load application and solution control processing further include: The model construction and meshing processing includes reading part coordinate system, reference surface and spline distribution, generating spline base, tooth profile, chamfer and transition fillet according to geometric and tolerance parameter set, generating control section locally according to inspection size table and carrying out feature fusion, establishing contact area sketch to identify internal and external spline meshing surface, end face support surface and positioning reference through surface; The meshing processing includes selecting element type and integration strategy according to material and contact parameter list, applying local mesh refinement according to key geometric position marked by inspection size table, adopting division strategy with higher refinement level than global for root transition fillet, tooth side working surface and meshing start-stop band, carrying out mesh quality control by using multi-index joint determination mode, evaluating element twist, aspect ratio and minimum angle, and entering adaptive re-meshing when not reaching threshold value; The load application and solving control processing includes defining torque, axial load, radial disturbance and angular misalignment in load case list to corresponding geometric surface or feature line, establishing master-slave surface relationship by contact pair list and writing friction parameter, applying constraint degree of freedom at positioning reference by boundary constraint list, and generating load step scheme and contact activation scheme by solving strategy suggestion.

8. The method of claim 1, wherein, The process of load carrying check and wear life evaluation processing further includes: The load carrying check and wear life evaluation processing includes extracting representative areas of tooth side working surface, root transition band and end face support part from finite element analysis result set, establishing evaluation window for each area to record spatial distribution statistics of maximum principal stress, equivalent stress and contact pressure, executing threshold value comparison, gradient continuity determination and hot spot clustering, generating load carrying check item table, and extracting contact pressure time history, shear stress time history and slip path time history for key areas, generating load cycle spectrum by using rain flow counting and interval merging strategy, and carrying out damage accumulation and thickness attenuation accumulation.

9. The method of claim 1, wherein, The process of report template rendering and paragraph splicing processing further includes: The report template rendering and paragraph splicing processing includes reading project name, part identification, version identification, selected specification and unit system, reading traceability key and paragraph anchor point configuration in mapping domain, using template layout library and component list to replace placeholders, generating cross-reference mark when encountering cross-paragraph reference, automatically rearranging column width and recalculating page, and carrying out paragraph splicing processing, including extracting dimension chain node track, gauge type and unit writing specification from inspection size table to form dimension paragraph segment, extracting threshold value comparison result, gradient continuity determination summary and hot spot clustering summary from load carrying check item table to form load carrying item segment, extracting cycle spectrum summary, accumulation record and state mark from life evaluation item table to form life item segment, and extracting pending list, abnormal label and disposal suggestion from diagnosis domain summary to form warning paragraph segment, carrying out theme merging and sorting, preferentially following chapter order given by specification domain, secondly following numbering order of part identification and key area index list, and merging repeated fields under the same theme, and querying source priority and version time stamp in mapping domain to retain high priority item when there is field conflict.

10. The method of claim 1, wherein, The process of label proofreading and gauge parameter arrangement processing further includes: The marking proofreading and gauge parameter sorting processing includes bidirectional link based on the report text block and the inspection size table, scanning the size item in the size section, positioning each item to the corresponding record of the inspection size table through the link key, checking the consistency of the unit writing specification and the tolerance system, checking the size chain closure relationship, checking the consistency of the terminology, generating the marking proofreading result set, and selecting the outer diameter class, the inner diameter class, the tooth thickness class and the reference circle class dimension from the inspection size table. The gauge type, the measurement point number and the reading mode are archived to form the gauge parameter sheet.

Images