Systems and methods for non-destructively testing stator weld quality and epoxy thickness

Patent metadata

Field	Value
Publication number	US-12427606-B2
Application number	US-202318172643-A
Country	US
Kind code	B2
Filing date	Feb 22, 2023
Priority date	Feb 22, 2023
Publication date	Sep 30, 2025
Grant date	Sep 30, 2025

Abstract

A method includes obtaining pre-weld image data of the electric motor stator from one or more image sensors; performing a welding process in response to obtaining the pre-weld image data, obtaining post-weld image data of the electric motor stator from the one or more image sensors in response to performing the welding process, obtaining epoxy image data of the electric motor stator from the one or more image sensors in response to obtaining the post-weld image data, performing a difference-based image processing routine based on the post-weld image data and the epoxy image data to generate a digital twin of the electric motor stator, and determining one or more epoxy characteristics of the electric motor stator based on the digital twin.

First claim

What is claimed is:

1. A method for non-destructively testing an electric motor stator, the method comprising:
obtaining pre-weld image data of the electric motor stator from one or more image sensors;
performing a welding process in response to obtaining the pre-weld image data, wherein the welding process includes welding hairpin wires and connectors of the electric motor stator;
obtaining post-weld image data of the electric motor stator from the one or more image sensors in response to performing the welding process;
obtaining epoxy image data of the electric motor stator from the one or more image sensors in response to obtaining the post-weld image data;
performing a difference-based image processing routine based on the post-weld image data and the epoxy image data to generate a digital twin of the electric motor stator; and
determining one or more epoxy characteristics of the electric motor stator based on the digital twin.

2. The method of claim 1 further comprising:
determining one or more pre-weld characteristics of the electric motor stator based on the pre-weld image data; and
setting one or more parameters of the welding process based on the one or more pre-weld characteristics.

3. The method of claim 2 , wherein the one or more pre-weld characteristics include a weld gap, a lateral offset, a vertical offset, a degradation characteristic, a trimming characteristic, or a combination thereof.

4. The method of claim 1 further comprising:
performing a post-weld verification routine to determine one or more post-weld characteristics of the electric motor stator based on the post-weld image data; and
selectively performing a post-weld corrective action based on the one or more post-weld characteristics.

5. The method of claim 4 , wherein the one or more post-weld characteristics include a weld size, a weld shape, a weld alignment, a pore visibility, a discoloration, or a combination thereof.

6. The method of claim 1 , wherein the difference-based image processing routine is a computer vision routine.

7. The method of claim 6 , wherein performing the computer vision routine based on the post-weld image data and the epoxy image data to generate the digital twin of the electric motor stator further comprises:
determining, for each pixel from among a plurality of pixels of the post-weld image data, a post-weld intensity value;
determining, for each pixel from among a plurality of pixels of the epoxy image data, an epoxy intensity value; and
determining, for each pixel from among the plurality of pixels of the epoxy image data, an intensity difference between the epoxy intensity value and a corresponding post-weld intensity value, wherein the digital twin is further based on the intensity difference associated with each pixel from among the plurality of pixels of the epoxy image data.

8. The method of claim 1 , wherein the difference-based image processing routine is a convolutional neural network routine.

9. The method of claim 8 , wherein performing the convolutional neural network routine based on the post-weld image data and the epoxy image data to generate the digital twin of the electric motor stator further comprises:
aligning the post-weld image data and the epoxy image data based on a reference object of the electric motor stator;
determining, in response to aligning the image data and the epoxy image data and for each pixel from among a plurality of pixels of the post-weld image data, a post-weld intensity value;
determining, for each pixel from among a plurality of pixels of the epoxy image data, an epoxy intensity value;
determining, for each pixel from among the plurality of pixels of the epoxy image data, an intensity difference between the epoxy intensity value and a corresponding post-weld intensity value; and
generating the digital twin based on the intensity difference associated with each pixel from among the plurality of pixels of the epoxy image data.

10. The method of claim 9 further comprising:
providing the digital twin to one or more convolutional layers to generate one or more feature maps; and
providing the one or more feature maps to one or more pooling layers to generate a reduced image, wherein the one or more epoxy characteristics are further based on the reduced image.

11. The method of claim 1 , wherein the one or more epoxy characteristics include an epoxy thickness, an epoxy distribution, or a combination thereof.

12. A method for non-destructively testing an electric motor stator, the method comprising:
obtaining pre-weld image data of the electric motor stator from one or more image sensors;
determining one or more pre-weld characteristics of the electric motor stator based on the pre-weld image data;
performing a welding process based on the one or more pre-weld characteristics, wherein the welding process includes welding hairpin wires and connectors of the electric motor stator;
obtaining post-weld image data of the electric motor stator from the one or more image sensors in response to performing the welding process;
obtaining epoxy image data of the electric motor stator from the one or more image sensors in response to obtaining the post-weld image data;
performing a difference-based image processing routine based on the post-weld image data and the epoxy image data to generate a digital twin of the electric motor stator; and
determining one or more epoxy characteristics of the electric motor stator based on the digital twin.

13. The method of claim 12 , wherein the one or more pre-weld characteristics include a weld gap, a lateral offset, a vertical offset, a degradation characteristic, a trimming characteristic, or a combination thereof.

14. The method of claim 12 further comprising:
performing a post-weld verification routine to determine one or more post-weld characteristics of the electric motor stator based on the post-weld image data; and
selectively performing a post-weld corrective action based on the one or more post-weld characteristics, wherein the one or more post-weld characteristics include a weld size, a weld shape, a weld alignment, a pore visibility, a discoloration, or a combination thereof.

15. The method of claim 12 , wherein performing the difference-based image processing routine based on the post-weld image data and the epoxy image data to generate the digital twin of the electric motor stator further comprises:
determining, for each pixel from among a plurality of pixels of the post-weld image data, a post-weld intensity value;
determining, for each pixel from among a plurality of pixels of the epoxy image data, an epoxy intensity value; and
determining, for each pixel from among the plurality of pixels of the epoxy image data, an intensity difference between the epoxy intensity value and a corresponding post-weld intensity value, wherein the digital twin is further based on the intensity difference associated with each pixel from among the plurality of pixels of the epoxy image data.

16. The method of claim 15 , wherein the difference-based image processing routine is a computer vision routine.

17. The method of claim 15 , wherein the difference-based image processing routine is a convolutional neural network routine, and wherein the method further comprises:
aligning the post-weld image data and the epoxy image data based on a reference object of the electric motor stator;
determining, in response to aligning the image data and the epoxy image data and for each pixel from among the plurality of pixels of the post-weld image data, the post-weld intensity va

Assignees

• Ford Global Tech Llc

Inventors

• Schroeter Robert
• Oelscher Michael

Classifications

• H02K15/106

  Insulating between conductors and cores · CPC title

• H02K15/33

  Connecting winding sections; Forming leads; Connecting leads to terminals · CPC title

• B23K31/006

  relating to using of neural networks · CPC title

• B23K2101/36

  Electric or electronic devices · CPC title

• G01N33/442 Primary

  Resins; Plastics · CPC title