ML-Based CAD Object Labeling for Assembly Design

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional CAD applications lack tools for automatically labeling and providing additional information for design objects in assembly designs, making it difficult for subsequent designers without expertise to understand and contribute to the design.

Innovation Solution

A computer-implemented method that displays a design space with design objects, generates a prompt for object identifiers and labels, transmits it to trained ML models, and receives responses to display object labels and additional information within the design space.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If conventional CAD applications are used without automated labeling tools, then the initial designer can efficiently create assembly designs using their expertise, but subsequent designers without expertise cannot understand or contribute to the design

Engineering Contradiction:

Improveease of understanding design objectsVSAvoidloss of design object information

Core Design Contradiction:

Ease of operationVSLoss of information

Solution Approach 1:

The patent introduces an intermediary system consisting of ML models that automatically generate labels and descriptions for design objects. This intermediary bridges the gap between the initial designer's expertise and the subsequent designer's lack of expertise by converting implicit design knowledge into explicit, accessible information without requiring direct human intervention from either party.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system enables self-service by allowing the design objects to automatically generate their own labels and descriptions through ML models. Instead of requiring the initial designer to manually label each object or the subsequent designer to manually study the design, the system autonomously generates the necessary information, making the design self-explanatory.

Inventive Principle:

Principle #25Self-service

2Reliability

If manual study and research are performed to understand design objects, then subsequent designers can gain knowledge about the assembly, but significant effort and time are required

Engineering Contradiction:

Improveunderstanding of design objectsVSAvoidtime required for design review

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs preliminary action by automatically generating labels, descriptions, and design history information before the subsequent designer needs to review the assembly. This advance preparation of information eliminates the need for time-consuming manual research and study, as all necessary design object information is already available in an accessible format.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent replaces the mechanical system of manual study and research with an automated information generation system using ML models. Instead of the subsequent designer physically studying and analyzing design objects, the system automatically generates and presents the necessary information, substituting human cognitive effort with automated computational processes.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Productivity

If automated ML models are used to generate object information, then subsequent designers can quickly understand design objects, but additional system complexity is introduced

Engineering Contradiction:

Improvespeed of design reviewVSAvoidcomplexity of CAD application

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent applies universality by designing the ML model integration to serve multiple functions: generating labels, creating descriptions, retrieving design history, and providing contextual information. This multi-functional approach consolidates what would otherwise require multiple separate tools or processes into a single integrated system, reducing the perceived complexity while enhancing productivity.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Loss of information

If no automated labeling tools are provided, then the CAD application remains simple, but design objects lack descriptive information for understanding

Engineering Contradiction:

Improveinformation about design objectsVSAvoidautomation of labeling process

Core Design Contradiction:

Loss of informationVSExtent of automation

Solution Approach 1:

The system enables self-service by allowing design objects to automatically generate their own labels and descriptions through ML models. Instead of requiring the initial designer to manually label each object or the subsequent designer to manually study the design, the system autonomously generates the necessary information, making the design self-explanatory.

Inventive Principle:

Principle #25Self-service

Data Source

PatentUS20250117527A1Automated design object labeling via machine learning models

Publication Date: 2025.04.10 AUTODESK INC

AI summary

In various embodiments, a computer-implemented method for displaying object information associated with a computer-aided design, the method comprising displaying a design space that includes a plurality of design objects, generating a prompt that includes a set of object identifiers corresponding to a first set of design objects included in the plurality of design objects and a first query for a set of object labels corresponding to the first set of design objects, transmitting the prompt to at least one trained machine learning (ML) model for processing, receiving, from the at least one trained ML model, a first ML response that includes the set of object labels corresponding to the first set of design objects, and displaying the set of object labels within the design space.