Inverse Design Acceleration via Resolution Mapping

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current methods for designing photonic devices, such as optical multiplexers and demultiplexers, are inefficient due to the high computational cost of full-resolution simulations, which hinders the optimization of structural parameters and increases the time required to achieve desired performance metrics.

Innovation Solution

An inverse design process utilizing reduced-resolution simulations and machine learning models to predict full-resolution performance results, where the system conducts operational and adjoint simulations at lower resolutions and updates the design based on predicted performance, thereby reducing computational time without compromising accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If full-resolution simulations are conducted to ensure accurate performance evaluation, then measurement precision is improved, but computational time increases significantly

Engineering Contradiction:

Improveperformance evaluation accuracyVSAvoidcomputational time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent creates a machine learning model that learns the mapping between reduced-resolution simulation results and full-resolution performance characteristics. The model is trained by comparing predictions against actual full-resolution simulation data, allowing it to copy the essential performance information without requiring expensive full-resolution simulations during the design optimization process.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent uses reduced-resolution simulations as a cheaper, faster alternative to full-resolution simulations. These lower-fidelity simulations are performed repeatedly during the iterative design process, and only full-resolution simulations are conducted periodically to retrain or validate the machine learning model, making the computational resources more efficiently utilized.

Inventive Principle:

Principle #27Cheap short-living objects (Disposable)

2Productivity

If reduced-resolution simulations are used to reduce computational time, then productivity is improved, but measurement precision deteriorates

Engineering Contradiction:

Improvedesign process speedVSAvoidperformance evaluation accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The machine learning model serves as an intermediary that translates reduced-resolution simulation results into accurate predictions of full-resolution performance characteristics. This mediator bridges the gap between computational efficiency and measurement precision, allowing the system to use fast low-resolution simulations while maintaining the accuracy of high-resolution evaluations through the learned mapping relationship.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent changes the resolution parameter of simulations from high to low, dramatically reducing computational cost. The machine learning model then compensates for this parameter change by learning the relationship between reduced-resolution inputs and full-resolution outputs, restoring the measurement precision that would otherwise be lost through the resolution reduction.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If iterative design optimization is performed to improve device performance, then productivity is improved, but computational time increases

Engineering Contradiction:

Improvedesign optimization efficiencyVSAvoidtotal computational time

Core Design Contradiction:

ProductivityVSLoss of time

Solution Approach 1:

The patent enables continuous iterative optimization by using the machine learning model to provide immediate feedback on design variations without requiring time-consuming full-resolution simulations. The model can be quickly retrained or updated with new simulation data, allowing the design process to continue smoothly through multiple iterations with minimal computational delays.

Inventive Principle:

Principle #20Continuity of useful action

Solution Approach 2:

The machine learning model is trained in advance using a subset of full-resolution simulation data, creating a predictive tool that can be applied to numerous design iterations without requiring repeated expensive simulations. This preliminary training action pays off during the iterative optimization process by providing fast, accurate performance predictions for each design variation.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS20230100128A1Accelerating an inverse design process using learned mappings between resolution levels

Publication Date: 2023.03.30 X DEVELOPMENT LLC

AI summary

A computer-implemented method of creating a design for a physical device using an inverse design process is provided. A computing system receives a proposed design. The computing system conducts an operational simulation based on the proposed design at a first resolution to generate a calculated performance result. The computing system provides the calculated performance result to a machine learning model to generate a predicted performance result of an operational simulation based on the proposed design at a second resolution, where the second resolution is higher than the first resolution. The computing system updates the proposed design based on the predicted performance result.