Search

CN-122020856-A - Automobile wire harness design intention analysis and CAD graph generation system based on large model

CN122020856ACN 122020856 ACN122020856 ACN 122020856ACN-122020856-A

Abstract

The invention discloses an automobile wire harness design intention analysis and CAD graph generation system based on a large model, which comprises a design intention acquisition module, a semantic analysis module, a knowledge analysis module, a geometric modeling module, an engineering verification module, an adaptive optimization module and a label generation module, wherein the design intention acquisition module is used for acquiring and preprocessing automobile wire harness design input data, the semantic analysis module is used for carrying out semantic analysis and constraint generation through the large language model, the knowledge analysis module is used for carrying out semantic disambiguation, parameter complementation and consistency verification, the geometric modeling module is used for converting normalized design intention expression into CAD drawing instructions and generating a wire harness initial geometric model, the engineering verification module is used for carrying out multidimensional constraint verification, the adaptive optimization module is used for iteratively optimizing the wire harness initial geometric model, and the label generation module is used for carrying out view projection and label drawing and generating target CAD graphs and engineering label data. The invention integrates a large language model and a CAD generation technology, realizes intelligent analysis and constraint optimization of the design intention of the automobile wire harness, and has the advantages of automation, high precision and strong compliance.

Inventors

  • LIANG XINGANG
  • HUANG ZIRUI

Assignees

  • 深圳市爱智慧科技有限公司

Dates

Publication Date
20260512
Application Date
20260206

Claims (9)

  1. 1. Automobile wire harness design intention analysis and CAD graph generation system based on large model, which is characterized by comprising: The design intention acquisition module is used for acquiring and preprocessing the design input data of the automobile wire harness and generating a normalized design instruction text; The semantic analysis module is used for inputting the normalized design instruction text into a large language model, carrying out semantic analysis and constraint generation by combining a wire harness engineering knowledge base, and outputting a structural design intention text; The knowledge analysis module is used for carrying out semantic disambiguation, parameter complementation and consistency verification on the structured design intention text to generate normalized design intention expression; The geometric modeling module is used for converting the normalized design intention expression into a CAD drawing instruction and generating a wire harness initial geometric model according to the mapping rule and the parameterized template; The engineering verification module is used for carrying out multidimensional constraint verification on the initial geometric model of the wire harness based on engineering constraint parameters to generate an engineering verification result; The self-adaptive optimization module is used for constructing an objective function based on the engineering verification result, iteratively optimizing the initial geometric model of the wire harness and outputting the objective geometric model; and the annotation generation module is used for executing view projection and annotation drawing based on the target geometric model and generating target CAD graph and engineering annotation data.
  2. 2. The large model-based automotive harness design intent analysis and CAD graphics generation system of claim 1, wherein the modules are implemented by the following method: obtaining design input data of an automobile wire harness, performing semantic preprocessing, and generating a normalized design instruction text; inputting the normalized design instruction text into a large language model, performing multi-layer semantic analysis and constraint perception generation processing, and outputting a structural design intention text; analyzing the structural design intention text based on the wire harness engineering knowledge base, disambiguating and checking consistency to generate normalized design intention expression; Converting the normalized design intent expression into a CAD drawing instruction sequence according to a mapping rule from semantics to CAD commands and a parameterized template, and executing the CAD drawing instruction sequence in a CAD engine to generate a wire harness initial geometric model; Performing multidimensional constraint verification and compliance analysis on the initial geometric model of the wire harness based on engineering constraint parameters to generate an engineering verification result; Constructing an objective function based on the engineering verification result and the engineering constraint parameters, generating a candidate adjustment data set, iteratively optimizing an initial geometric model of the wire harness, and outputting a target geometric model according to a convergence judgment result; And performing view projection, parameter mapping and annotation drawing processing based on the target geometric model to generate target CAD graph and engineering annotation data.
  3. 3. The large model based automotive harness design intent resolution and CAD graphics generation system of claim 2, wherein the design input data includes designer's voice or text instructions, harness structure descriptions, component parameters and connection constraint descriptions, and the semantic preprocessing includes voice recognition, text merging, term normalization, named entity recognition and dependency analysis.
  4. 4. The large model based automotive harness intent analysis and CAD graphics generation system of claim 2, wherein the outputting of the structured intent text comprises the specific steps of: Inputting a normalized design instruction text into a large language model, performing tokenization and position coding on the text content, and converting the text content into a semantic representation sequence, wherein the large language model is improved based on Vicuna-13B architecture, and specifically comprises the steps of adding a graph semantic solution terminal on the basis of a single text decoding head of an original model, adding a program synthesis head constrained by context-free grammar in parallel in an output layer structure, and introducing a constraint evaluation submodule and a constraint fusion submodule in an autoregressive decoding cycle; in the graph semantic solution wharf, based on a semantic representation sequence, generating probability weights of node candidates, edge candidates and attribute slot candidates according to a decoding step according to a node type set and a relation type set of a wire harness engineering knowledge base, and outputting a graph semantic probability weight set; In the program synthesis head, on the basis of a semantic representation sequence, grammar shielding is carried out on candidate instructions according to a generation set of a context-free grammar, instruction word elements meeting grammar are reserved, and a program instruction probability weight set is generated; In the constraint evaluation submodule, calculating the deviation amount under each constraint dimension according to the project constraint parameter pair graph semantic probability weight set and the program instruction probability weight set, constructing constraint cost according to the deviation amount, and mapping the constraint cost into the correction amount of the probability weight according to the penalty coefficient; In each decoding step, fusing the graph semantic probability weight set and the program instruction probability weight set through a constraint fusion sub-module, performing subtractive correction and renormalization according to the correction quantity to obtain a constraint-perceived joint probability weight set, and selecting a probability maximum term as output in the current decoding step; After all decoding steps are completed, summarizing output results of all decoding steps, extracting a node set, an edge set, an attribute slot set and an instruction sequence corresponding to a program synthesis head corresponding to a graph semantic solution terminal, and performing hierarchical mapping and unified coding processing to generate an intermediate representation of the structural design intent; And carrying out semantic alignment and consistency arrangement on the intermediate representation of the structural design intention according to semantic association rules, establishing a corresponding relation among the target object, the operation type, the parameter set, the engineering constraint parameter and the context association, and outputting a structural design intention text.
  5. 5. The large model-based automotive harness design intent parsing and CAD graphics generation system of claim 2, wherein the generation of the normalized design intent representation comprises the specific steps of: Analyzing a target object, an operation type, a parameter set and engineering constraint parameters in the structural design intention text to be associated with the context, and extracting semantic elements in an analysis result to generate a candidate entity set; performing name normalization and reference digestion processing on the candidate entity set according to the term list and the alias list of the wire harness engineering knowledge base to generate a standard candidate name set; Performing retrieval on the standard candidate name set in the component library index and the geometric feature index to obtain a component item related to candidate name semantically and a corresponding geometric feature parameter, and generating a candidate component set; Performing multi-strategy matching on the candidate component set according to node types, connection types and characteristic parameters defined in the wire harness engineering knowledge base, calculating semantic similarity, mapping an item with a matching score meeting a similarity threshold condition into a component library code and a geometric characteristic identifier corresponding to a target object, and generating a matching result set by combining characteristic contribution priority weights; performing consistency verification on the matched result set, and according to a vehicle model three-dimensional geometric model, design specifications of a host factory and an enterprise and engineering constraint parameters, verifying interface type compatibility, connection end position correspondence, conductor section and current capacity matching, minimum bending radius and minimum clearance constraint and space accessibility to obtain a verification result set; On the basis of the verification result set, combining the historical project data with engineering constraint parameters, performing completion processing on the missing or incomplete parameter items, and generating a completion parameter set according to rule priority and statistical confidence; And based on the verification result set, carrying out field mapping and data merging on the complement parameter set, updating the missing attribute, unifying the format and the coding, and generating the normalized design intent expression.
  6. 6. The large model-based automotive harness design intent resolution and CAD graphics generation system of claim 2, wherein the generation of the harness initial geometric model comprises the specific steps of: analyzing a target object, an operation type, a parameter set and engineering constraint parameters in the normalized design intent expression, establishing a mapping relation between the object and a parameterized template, and calibrating a template slot position and a corresponding parameter type for each target object; Based on the mapping relation, according to the mapping rule from the semantics to the CAD command, disassembling the object-level action into a drawing atomic instruction to form an instruction sequence framework comprising coordinate, unit, layer, line type and color coding definition; Taking an instruction sequence skeleton as input, and executing parameter instantiation and constraint embedding on each atomic instruction by combining a parameter set and engineering constraint parameters to generate a CAD drawing instruction sequence with instantiation parameters; according to node and fixed point parameters defined in a CAD drawing instruction sequence, establishing a corresponding relation between a three-dimensional assembly coordinate system and a two-dimensional projection coordinate system, determining the spatial positions of a start node and a stop node and a path control point, and generating a geometric primitive of the node and the fixed point; In the established space coordinate system, carrying out path solving and optimizing on a trunk line and branch lines based on the geometric primitives of nodes and fixed points, wherein a path cost function consists of a weighted sum of path length, corner number, gap penalty term and adherence deviation, each weight is set according to engineering constraint parameters, and a path result meeting geometric and constraint conditions is output; based on the path result, calculating the diameter and bending radius of the wire harness segment, inserting arc transition at the turning position, generating nodes and attribute slots at the branching position, marking geometric characteristic identifiers, and assembling the wire harness segment geometry into a wire harness primary geometric model according to the topological sequence; Calling sheath structure and shielding structure definition in a parameterized template on a primary geometric model of the wire harness, superposing to generate outer geometry and assembly constraint, and generating assembly hole sites and constraint symbols for fixed points and buckle positions to form structural wire harness geometric data; And inputting the structured harness geometric data into a CAD engine to execute drawing instructions, and generating a harness initial geometric model.
  7. 7. The large model-based automotive harness design intent resolution and CAD graphics generation system of claim 2, wherein the generation of the engineering verification result comprises the specific steps of: Extracting geometric and attribute information of a trunk line, branch lines, nodes, fixed points, a sheath and a shielding structure from an initial geometric model of the wire harness, establishing a spatial adjacent relation and a check unit index, loading engineering constraint parameters, and outputting a check input data set; performing electromagnetic compatibility risk calculation on the space layout information of the wire harness in the verification input data set and electromagnetic compatibility constraint in engineering constraint parameters, identifying parallel sections of the sensitive object, calculating an electromagnetic compatibility risk score and marking the risk sections; Conducting conductor current carrying and temperature rise assessment on conductor section attribute and electrical parameter information in the verification input data set, calculating current density and temperature rise estimated values, respectively comparing the current density and temperature rise estimated values with conductor current carrying and temperature rise constraint in engineering constraint parameters, and marking an irregular wire bundle section; performing minimum bending radius coincidence verification on the extracted geometric information of the central path of the wire harness in the verification input data set, performing discrete sampling along the central path of the wire harness, calculating the local curvature radius of each sampling position, comparing the local curvature radius with the minimum bending radius constraint in engineering constraint parameters, and marking the non-compliance bending position; performing minimum clearance compliance verification on the space adjacent relation in the verification input data set and the assembly layout information, calculating the nearest distance between the wire harness and the adjacent part and between the wire harness, comparing the nearest distance with the minimum clearance constraint in engineering constraint parameters, and marking an unconformity clearance pair; Performing space collision detection on the geometric topological structure in the verification input data set, the surface information of the assembly body and the space collision constraint in the engineering constraint parameters, identifying collision pairs by adopting a crossing mode of the hierarchical bounding body and the three-dimensional triangular grid, recording a contact surface, a penetration depth and a contact position, and marking collision elements; and collecting the risk section, the non-compliance wire harness section, the non-compliance bending position and the non-compliance gap pair with collision elements to generate an engineering verification result.
  8. 8. The large model-based automotive harness design intent resolution and CAD graphics generation system of claim 2, wherein the outputting of the target geometric model comprises the specific steps of: Reading an engineering verification result and a wire harness initial geometric model, extracting engineering constraint parameters, setting an iteration upper limit and a convergence threshold, and outputting an optimized input data set; Constructing an objective function based on the optimized input data set, wherein the objective function is a weighted sum of each constraint deviation and corresponding weight, and the constraint deviation is defined as a non-negative part of a difference value between an actual measurement index and an engineering constraint allowable range; based on the objective function definition and the optimized variable set, performing local disturbance on the path control point coordinates, performing translation on the fixed points and the buckle positions, adjusting the conductor section specification in the allowable gear, switching the shielding layer structure parameters in feasible combinations, switching the laying mode codes in candidate sequences, and outputting a candidate adjustment data set; recalculating each index corresponding to the engineering constraint parameters and calculating each constraint deviation amount for each candidate adjustment scheme in the candidate adjustment data set to obtain a corresponding objective function value and constraint satisfaction mark; Selecting an improved solution according to the objective function value and the constraint satisfaction mark, adopting a candidate adjustment scheme with the minimum objective function value and updating geometry and parameters when all constraint satisfaction candidate adjustment schemes exist; according to the continuous iterative objective function change trend, self-adaptively adjusting step length coefficients and constraint weights, reducing step length and increasing the weight corresponding to the violation of the constraint when the objective function is not reduced for a plurality of times, amplifying the step length and keeping the weight unchanged when the objective function is reduced, and returning to the candidate adjustment scheme generation step to continue iteration if the convergence threshold or the iteration upper limit is not reached; and integrating the geometric and parameter data obtained by final iteration according to the convergence judgment result, updating the wire harness geometric data, recording an optimal adjustment instruction sequence, optimizing convergence information and the deviation of the residual constraint, and outputting a target geometric model.
  9. 9. The large model-based automotive harness design intent resolution and CAD drawing generation system of claim 2, wherein the generation of the target CAD drawing and engineering annotation data comprises the specific steps of: loading a target geometric model, establishing an orthographic projection and axonometric projection coordinate system according to drawing standards and scale factors, and determining a view set to be generated; Performing projection conversion on the target geometric model based on the view set, mapping the three-dimensional coordinates into two-dimensional coordinates, extracting visible boundaries and generating view geometric primitives; establishing a size base line and a reference on the view geometric primitive, and calculating a linear size, an arc radius and a bending angle to form a size calculation result; According to the attribute data of the target geometric model, mapping the terminal number, the line number, the color code, the section specification and the component code number to geometric elements corresponding to the dimension calculation result to generate a parameter mapping result; performing annotation drawing based on the parameter mapping result, arranging size annotation, symbol annotation and extraction annotation, and performing layout and collision avoidance according to the minimum spacing and alignment rule to form an annotation layout result; setting a layer, a line type, a line width and a character style according to the drawing specification on the labeling layout result, generating a title bar and a proportion mark, and forming structured graphic data; Performing annotation consistency check on the structured graph data, comparing and correcting the annotation value with the geometric measurement result, and updating the checked annotation data; and integrating the geometric primitives of each view and the checked annotation data to generate a target CAD graph and engineering annotation data.

Description

Automobile wire harness design intention analysis and CAD graph generation system based on large model Technical Field The invention relates to the technical field of computer aided drawing, in particular to an automobile wire harness design intention analysis and CAD graph generation system based on a large model. Background The existing automobile wiring harness design mainly depends on a CAD drawing system driven by manual operation and rules, a designer needs to manually construct geometric structures such as a trunk line, branch lines, nodes, fixed points and the like through complex graphic commands, and the process is complex and is easily influenced by subjective differences. A part of the system introduces a semi-automatic design mode based on a template or a knowledge base, but only can call a part model within a preset structural range, and is difficult to cope with design requirements of diversified vehicle types and complex constraints. In addition, the existing technology for converting natural language into CAD instructions stays in a general scene, and lacks deep understanding capability for semantics and process specifications of automobile wire harness engineering. The existing harness CAD system mostly adopts a post-checking and manual correction mode in engineering constraint processing such as electromagnetic compatibility, conductor current carrying and temperature rise, minimum bending radius, space collision and the like, and lacks dynamic constraint sensing and self-adaptive optimizing mechanisms in a generation stage, so that the design period is long, the modification cost is high, and the compliance rate is low. The lack of integrated mechanisms for semantic parsing, engineering constraints, and geometry generation makes it difficult for the prior art to achieve intelligent closed-loop from design intent to compliance graphics. Therefore, how to provide a large model-based automotive harness design intent resolution and CAD graphics generation system is a problem that needs to be solved by those skilled in the art. Disclosure of Invention The invention aims to provide an automobile wire harness design intention analysis and CAD graph generation system based on a large model, integrates a large language model and a CAD generation technology, realizes intelligent analysis and constraint optimization of the automobile wire harness design intention, and has the advantages of automation, high precision and strong compliance. According to the embodiment of the invention, the automobile wire harness design intention analysis and CAD graph generation system based on the large model comprises the following steps: The design intention acquisition module is used for acquiring and preprocessing the design input data of the automobile wire harness and generating a normalized design instruction text; The semantic analysis module is used for inputting the normalized design instruction text into a large language model, carrying out semantic analysis and constraint generation by combining a wire harness engineering knowledge base, and outputting a structural design intention text; The knowledge analysis module is used for carrying out semantic disambiguation, parameter complementation and consistency verification on the structured design intention text to generate normalized design intention expression; The geometric modeling module is used for converting the normalized design intention expression into a CAD drawing instruction and generating a wire harness initial geometric model according to the mapping rule and the parameterized template; The engineering verification module is used for carrying out multidimensional constraint verification on the initial geometric model of the wire harness based on engineering constraint parameters to generate an engineering verification result; The self-adaptive optimization module is used for constructing an objective function based on the engineering verification result, iteratively optimizing the initial geometric model of the wire harness and outputting the objective geometric model; and the annotation generation module is used for executing view projection and annotation drawing based on the target geometric model and generating target CAD graph and engineering annotation data. Optionally, the modules are realized by the following method: obtaining design input data of an automobile wire harness, performing semantic preprocessing, and generating a normalized design instruction text; inputting the normalized design instruction text into a large language model, performing multi-layer semantic analysis and constraint perception generation processing, and outputting a structural design intention text; analyzing the structural design intention text based on the wire harness engineering knowledge base, disambiguating and checking consistency to generate normalized design intention expression; Converting the normalized design intent expression into a CAD drawing