Volume based prediction of physical properties in sports articles

Patent metadata

Field	Value
Publication number	US-11763144-B2
Application number	US-201916684700-A
Country	US
Kind code	B2
Filing date	Nov 15, 2019
Priority date	Nov 27, 2018
Publication date	Sep 19, 2023
Grant date	Sep 19, 2023

Abstract

Described are methods of estimating a physical property associated with a region of a sports article. The method includes determining a plurality of structural features within the region, determining, for each structural feature, a feature value, mapping each feature value to a physical property, wherein the mapping is based on a machine learning algorithm from a limited plurality of samples, and wherein each sample associates a feature value with a value of the physical property, and using the mapping to estimate the physical property for the region.

First claim

That which is claimed is:

1. A method of estimating a physical property associated with a region of sports article comprising, the steps:
determining a plurality of structural features within the region, wherein each structural feature of the plurality of structural features comprises a distribution of a material within the region;
determining, for each structural feature, a feature value of the distribution of the material within the region;
mapping each feature value to a physical property, wherein the mapping is based on a machine learning algorithm from a limited plurality of samples, wherein each sample associates a feature value with a value of the physical property, and wherein the physical property comprises printability of the structural feature; and
using the mapping to estimate the physical property for the region.

2. The method of claim 1 , wherein the region is divided into sub-regions.

3. The method of claim 1 , wherein the region is bounded by a hexahedron.

4. The method of claim 1 , wherein the region is bounded by a quadrilateral.

5. The method of claim 2 , wherein the structural features are associated with the sub-regions.

6. The method of claim 2 , wherein the sub-regions are voxels or pixels which subdivide the region.

7. The method of claim 2 , wherein each sub-region is associated with a feature value for a structural feature within the sub-region.

8. The method of claim 7 , wherein the feature value associated with the sub-region depends on an amount of material present in the sub-region.

9. The method of claim 1 , wherein the mapping is based on a machine learning algorithm that was trained using the plurality of samples.

10. The method of claim 1 , wherein the mapping is based on an artificial neural network that was trained using the plurality of samples.

11. The method of claim 10 , wherein the artificial neural network is a convolutional neural network.

12. The method of claim 1 , wherein an association between the feature value with a value that is representative of the physical property is obtained for each sample using a finite element method.

13. The method of claim 1 , wherein the sports article is a shoe and the region is located in a midsole of the shoe.

14. The method of claim 13 , further comprising a step of subdividing the region into voxels, wherein at least one of the plurality of structural features is associated with each voxel.

15. The method of claim 1 , wherein the physical property further comprises any one of elasticity, stiffness, and shearing resistance of the structural feature within the region.

16. The method of claim 15 , wherein the structural feature is a beam that connects nodes.

17. The method of claim 1 , wherein the sports article is a shoe and the region is located in an upper of the shoe.

18. The method of claim 17 , further comprising a step of subdividing the region into pixels, wherein at least one of the plurality of structural features is associated with each pixel.

19. The method of claim 17 , wherein the structural feature is a patch of material to be attached to the upper of the shoe.

20. A method of manufacturing at least a part of a sports article, comprising the steps:
a. estimating the physical property associated with a plurality of regions of a part of the sports article according to claim 1 ;
b. optimizing the physical property using the estimate and considering at least one structural constraint for the part of the sports article; and
c. manufacturing the part of the sports article.

21. The method of claim 20 , wherein the step of manufacturing comprises additive manufacturing.

22. The method of claim 20 , wherein the step of manufacturing comprises:
providing a blank; and
placing at least one patch on the blank.

23. A part of a sports article which is manufactured according to a method of claim 20 .

24. A Sports article comprising a part according to claim 23 .

25. A method of manufacturing at least a part of a sports article, comprising:
a. estimating a physical property associated with each of a plurality of regions of a part of the sports article according to a process that includes, for each region of the plurality of regions:
(i) determining a plurality of structural features within the region, wherein each structural feature of the plurality of structural features comprises a distribution of a material within the region;
(ii) determining, for each structural feature, a feature value of the distribution of the material within the region;
(iii) mapping each feature value to a physical property, wherein the mapping is based on a machine learning algorithm from a limited plurality of samples, wherein the mapping is based on a machine learning algorithm that was trained using the plurality of samples, wherein each sample associates a feature value with a value of the physical property, and wherein the physical property comprises printability of the structural feature; and
(iv) using the mapping to estimate the physical property for the region, wherein the sports article is a shoe and the region is located in a midsole of the shoe;

b. optimizing the physical property using the estimate and considering at least one structural constraint for the part of the sports article; and
c. manufacturing the part of the sports article.

Assignees

• Adidas Ag

Inventors

• Siegl Christian
• Suessmuth Jochen
• Perrault Jacques
• Luther Derek
• Schneider Andrew
• Kendrick Dustin
• Poerner Mario
• Henderson Mark

Classifications

• G06F30/27 Primary

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

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/09

  Supervised learning · CPC title

• B33Y50/00 Primary

  Data acquisition or data processing for additive manufacturing · CPC title

• G06N3/08 Primary

  Learning methods · CPC title