Topology Optimization Using Machine Learning for 3D Design

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current generative design methods for 3D objects face limitations due to high computational complexity in solving topology optimization problems, leading to reduced design space exploration and potentially sub-optimal design selections when time and computational resources are limited.

Innovation Solution

A computer-implemented method that converts a high-resolution shape into a coarse shape, using a trained machine learning model to generate a new shape with the original resolution based on coarse structural analysis data, thereby reducing computational complexity and enabling more comprehensive design space exploration.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If the number of structural analysis operations is increased to improve design quality, then the design space exploration is enhanced and design convergence with objectives is improved, but the computational complexity and resource consumption increase significantly

Engineering Contradiction:

Improvedesign qualityVSAvoidcomputational complexity

Core Design Contradiction:

Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent replaces traditional iterative mechanical optimization processes with a machine learning-based predictive system. The trained model learns from training data containing shape-parameter pairs and their structural analysis results, then directly predicts optimal parameters without performing iterative structural analyses during the optimization process itself. This substitution dramatically reduces computational complexity while maintaining design quality.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent performs preliminary training of the machine learning model using comprehensive structural analysis data before the actual optimization process. During the training phase, the system performs extensive structural analyses on training shapes to build the predictive model. Once trained, the model can quickly predict optimal parameters for new shapes without requiring iterative structural analyses, thus preparing the system in advance to handle production optimization efficiently.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If the number of topology optimization problems is reduced to decrease computational complexity, then the resource consumption is lowered, but the design space exploration is limited and sub-optimal designs may be selected

Engineering Contradiction:

Improveresource efficiencyVSAvoiddesign space exploration

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The machine learning model replaces the computationally intensive iterative optimization process, enabling the system to evaluate many more design problems within the same resource constraints. By substituting iterative structural analyses with direct predictive modeling, the system can explore a larger design space and solve more topology optimization problems efficiently.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Device complexity

If the number of iterations in topology optimization is limited to reduce computational complexity, then the resource consumption decreases, but the design convergence with objectives deteriorates and sub-optimal designs are produced

Engineering Contradiction:

Improvecomputational complexityVSAvoiddesign convergence

Core Design Contradiction:

Device complexityVSReliability

Solution Approach 1:

The patent replaces iterative structural analysis and optimization with a machine learning-based direct prediction approach. The trained model has learned the relationship between shape parameters and structural performance from extensive training data, allowing it to predict optimal parameters in a single step without iterative refinement. This substitution maintains design convergence reliability while dramatically reducing computational complexity.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system performs preliminary training with comprehensive iterative optimization data to capture convergence behavior. The training process includes shapes that have undergone full iterative optimization to convergence, allowing the model to learn optimal parameter configurations. This preliminary action ensures that the model's predictions are reliable and converge to optimal designs without requiring iterative refinement during production use.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11468634B2Machine learning techniques for generating designs for three-dimensional objects

Publication Date: 2022.10.11 AUTODESK INC

AI summary

In various embodiments, a topology optimization application solves a topology optimization problem associated with designing a three-dimensional (“3D”) object. The topology optimization application coverts a first shape having a first resolution and representing the 3D object to a coarse shape having a second resolution that is lower than the first resolution. Subsequently, the topology optimization application computes coarse structural analysis data based on the coarse shape. The topology optimization application then uses a trained machine learning model to generate a second shape having the first resolution and representing the 3D object based on the first shape and the coarse structural analysis data. The trained machine learning model modifies a portion of a shape having the first resolution based on structural analysis data having the second resolution. Advantageously, generating the second shape based on structural analysis data having a lower resolution reduces computational complexity relative to prior art techniques.