Multimodal ML for 3D Design Exploration

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional computer-aided design techniques for 3D objects require significant knowledge of physics and engineering to define design spaces, leading to omission of optimized design alternatives and reduced quality of final products due to the need for quantitative user input.

Innovation Solution

A computer-implemented method using machine learning models to generate images representing design alternatives for 3D objects based on user-intended qualitative input, involving keyword prompts, multimodal similarity models, and prompt-to-image models to automate the design exploration process without requiring quantitative user knowledge.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If conventional generative design applications are used to automatically synthesize design alternatives, then productivity is improved, but device complexity increases due to the need for significant knowledge of physics and engineering to define design spaces

Engineering Contradiction:

Improvedesign alternative generation speedVSAvoiduser knowledge requirement

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces an intermediary system comprising a natural language processing module and a generative design module that translates qualitative user input into quantitative design parameters. This intermediary layer mediates between the user's simple textual description and the complex generative design algorithms, eliminating the need for users to directly specify complex physical constraints and engineering parameters.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system enables self-service by allowing users to define design spaces through simple qualitative descriptions without requiring expertise in physics or engineering. The automated parameter extraction and constraint generation processes perform the complex analytical work that would otherwise require specialized knowledge, making the tool accessible to non-experts.

Inventive Principle:

Principle #25Self-service

2Manufacturing precision

If quantitative goals and physical constraints are specified to define design space, then manufacturing precision is improved, but ease of operation deteriorates due to the complexity of defining proper design spaces

Engineering Contradiction:

Improvedesign alternative qualityVSAvoiduser input simplicity

Core Design Contradiction:

Manufacturing precisionVSEase of operation

Solution Approach 1:

The patent replaces the traditional mechanical approach of manually specifying quantitative parameters and constraints with an intelligent system that automatically extracts and translates qualitative descriptions into precise design parameters. The natural language processing and automated parameter generation substitute for the complex manual configuration process.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system dynamically transforms parameters from qualitative user input into quantitative design parameters automatically. Instead of requiring users to specify precise numerical values for physical constraints, the system converts descriptive text into the appropriate quantitative parameters and constraint values needed for high-quality design generation.

Inventive Principle:

Principle #35Parameter changes

3Reliability

If design space is over-constrained with unnecessarily high minimum load constraints, then reliability is improved, but weight of moving object increases due to bulky structural reinforcements

Engineering Contradiction:

Improvedesign constraint satisfactionVSAvoidobject mass

Core Design Contradiction:

ReliabilityVSWeight of moving object

Solution Approach 1:

The system incorporates feedback mechanisms where the AI model learns from user preferences and design outcomes to automatically adjust constraint specifications. The system analyzes the relationship between specified constraints and generated design quality, providing feedback that helps refine constraint definitions to avoid over-constraint while maintaining reliability.

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The patent applies partial action by allowing the system to selectively apply constraints based on their actual necessity rather than requiring all possible constraints to be explicitly specified. The automated parameter extraction identifies which constraints are truly needed versus which are redundant or overly restrictive, applying only the appropriate level of constraint to achieve reliable designs without excessive weight.

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20240104275A1Techniques for using multimodal machine learning models to generate design alternatives for three-dimensional objects

Publication Date: 2024.03.28 AUTODESK INC

AI summary

In various embodiments, a design exploration application generates images that represent design alternatives for three-dimensional (3D) objects. The design exploration application generates a keyword prompt based on design intent text that describes a 3D object. The design exploration application executes a first machine learning model on the keyword prompt to generate a first set of keywords. The design exploration application generates a rephrase prompt based on a second set of keywords that includes at least one keyword from the first set of keywords. The design exploration application executes the first machine learning model on the rephrase prompt to generate a final text prompt. The design exploration application executes a second machine learning model on the final text prompt to generate a set of images.