Reducing the numerical complexity of designs

Patent metadata

Field	Value
Publication number	US-11934756-B2
Application number	US-202017112417-A
Country	US
Kind code	B2
Filing date	Dec 4, 2020
Priority date	Dec 4, 2020
Publication date	Mar 19, 2024
Grant date	Mar 19, 2024

Abstract

A component library having a plurality of design components is received. Designs are predicted using the plurality of components using a machine learning model. The predicted designs comprise a subset of all possible designs using the plurality of components. A set of design criteria is received. At least one design solution is generated based on the set of design criteria and the predicted designs.

First claim

What is claimed is:

1. A method comprising:
training a machine learning model using training data;
transforming the training data into one or more different representations, the one or more different representations comprising a simplified molecular-input line-entry system (SMILES) representation;
receiving a component library having a plurality of design components;
predicting designs using the plurality of components using the machine learning model and the one or more different representations, the predicted designs comprising a subset of all possible designs using the plurality of components;
receiving a set of design criteria; and
generating at least one design solution based on the set of design criteria and the predicted designs.

2. The method of claim 1 , further comprising:
optimizing one or more components based on the predicted designs and the set of design criteria; and
generating at least one design solution based on optimized components.

3. The method of claim 1 , wherein the machine learning model is a generative model.

4. The method of claim 1 , further comprising training the machine learning model using a plurality of electrical circuits.

5. The method of claim 1 , further comprising training the machine learning model using training data from one or more of electrical, mechanical, and thermal domains.

6. The method of claim 5 , wherein transforming the training data into one or more different representations comprises:
transforming the training data to a graph representation; and
transforming the graph representation to the simplified molecular-input line-entry system (SMILES) representation.

7. The method of claim 1 , wherein training the machine learning model using examples from physical domains comprises training the machine learning model using at least one of mechanical, and thermal training data.

8. The method of claim 1 , wherein the training data comprises a plurality of components broken up into component types.

9. The method of claim 8 , wherein the component types comprise one or more of flow sources, effort sources, flow stores, effort stores, dissipators, transformers, and gyrators.

10. The method of claim 8 , further comprising tokenizing the component types.

11. A system, comprising:
a processor; and
a memory storing computer program instructions which when executed by the processor cause the processor to perform operations comprising:
train a machine learning model using training data;
transform the training data into one or more different representations, the one or more different representations comprising a simplified molecular-input line-entry system (SMILES) representation;
receive a component library having a plurality of components;
predict designs using the plurality of components using the machine learning model and the one or more different representations, the predicted designs comprising a subset of all possible designs using the plurality of components;
receive a set of design criteria; and
generate at least one design solution based on the set of design criteria and the predicted designs.

12. The system of claim 11 , wherein the processor is configured to:
optimize one or more components based on the predicted designs and the set of design criteria; and
generate at least one design solution based on optimized components.

13. The system of claim 11 , wherein the machine learning model is a generative model.

14. The system of claim 11 , further comprising training the machine learning model using training data from one or more of electrical, mechanical, and thermal domains.

15. The system of claim 14 , wherein transforming the training data into one or more different representations comprises:
transforming the training data to a graph representation; and
transforming the graph representation to the simplified molecular-input line-entry system (SMILES) representation.

16. A method comprising:
training a machine learning model using training data;
transforming the training data into one or more different representations, the one or more different representations comprising a simplified molecular-input line-entry system (SMILES) representation;
receiving an electrical component library having a plurality of electrical components;
predicting circuit designs using the plurality of electrical components using the machine learning model and the one or more different representations, the predicted circuit designs comprising a subset of all possible circuit designs using the plurality of electrical components;
receiving a set of design criteria; and
generating at least one design solution based on the set of design criteria and the predicted circuit designs.

Assignees

• Palo Alto Res Ct Inc
• Xerox Corp

Inventors

• Matei Ion
• Zhenirovskyy Maksym
• De Kleer Johan
• Feldman Aleksandar B

Classifications

• G06F30/27 Primary

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title

• G06F30/17

  Mechanical parametric or variational design · CPC title

• G06F2111/10

  Numerical modelling · CPC title

• G06F2111/12

  Symbolic schematics · CPC title

• G06F30/373

  Design optimisation · CPC title