Methods and systems for predicting stator insulation condition from stator sections

Patent metadata

Field	Value
Publication number	US-12333735-B2
Application number	US-202217937649-A
Country	US
Kind code	B2
Filing date	Oct 3, 2022
Priority date	Oct 3, 2022
Publication date	Jun 17, 2025
Grant date	Jun 17, 2025

Abstract

Methods and systems are provided for an insulation system of a stator. In one example, a method may include receiving images of the stator at a processor of a computing system and feeding the images to a deep learning tool to generate processed images by segmenting and cropping the images according to slots identified in the images. Further, the varnish in the processed images may be quantified based on fluorescence of the varnish, converted into estimated varnish fill percentages, based on an output from analysis of the processed images, and the estimated varnish fill percentages may be displayed in a report.

First claim

The invention claimed is:

1. A method for automatically analyzing images of a stator, comprising:
receiving images of the stator at a processor of a computing system, the images depicting varnish deposited in slots of the stator;
feeding the images to a deep learning tool implemented at the processor to generate processed images by segmenting and cropping the images according to slots identified in the images;
extracting and quantifying the varnish in the processed images, via the deep learning tool, based on fluorescence of the varnish, the deep learning tool trained to identify and analyze the fluorescence using results from machine learning-based color distribution analysis;
converting quantification of the varnish into estimated varnish fill percentages, via the deep learning tool, based on an output from analysis of the processed images; and
displaying the estimated varnish fill percentages in a report at a display device.

2. The method of claim 1 , wherein the images of the stator are images of cross-sections of the stator, and wherein the cross-sections are one or more of transverse cross-sections and axial cross-sections.

3. The method of claim 2 , wherein the transverse cross-sections are obtained by slicing the stator along one or more planes perpendicular to a central axis of rotation of the stator, and wherein the axial cross-sections are obtained by slicing the stator along one or more planes parallel to the central axis of rotation of the stator.

4. The method of claim 1 , wherein receiving the images of the stator includes receiving the images from at least one fluorescence detector configured to acquired fluorescence images of the stator.

5. The method of claim 1 , wherein segmenting the images includes locating the slots in the images and dividing the images into segments, each of the segments depicting one of the slots.

6. The method of claim 5 , wherein cropping the images includes trimming the segments to borders of the slots and removing regions of the images outside of the borders.

7. The method of claim 1 , wherein extracting and quantifying the varnish includes identifying regions of the varnish in the images that are not visually discernable by a user.

8. The method of claim 1 , wherein the deep learning tool includes a convolutional neural network configured to learn to locate and quantify the varnish in the images.

9. The method of claim 1 , wherein displaying the estimated varnish fill percentages in the report includes displaying the report within less than 5 seconds of receiving the images at the deep learning tool.

10. A system for evaluating a varnish condition of a stator, comprising:
a housing enclosing a UV light source and a fluorescence detector; and
a processor configured with executable instructions stored in non-transitory memory that, when executed, cause the processor to:
receive images of cross-sections of the stator from the fluorescence detector;
input the images to a deep learning tool to segment and crop the images into processed images according to slots of the stator identified in the images;
identify and quantify varnish, via the deep learning tool, in the processed images, the deep learning tool trained to analyze the processed images based on outputs from machine learning models utilizing clustering and color distribution analysis;
convert quantification of the varnish into varnish fill percentages via the deep learning tool, the varnish fill percentages determined based on analysis of the processed images; and
display the varnish fill percentages as a report at a display device.

11. The system of claim 10 , wherein the images are of axial cross-sections of the stator, obtained by slicing the stator along a plane parallel with a central axis of rotation of the stator through at least one of the slots, and wherein the images show a side surface of the at least one of the slots.

12. The system of claim 11 , wherein the outputs from the machine learning models include cluster-only images and binary masks to enable identification and quantification of the varnish in the cluster-only images and the binary masks.

13. The system of claim 10 , wherein the images are of transverse cross-sections of the stator obtained by slicing the stator along plane perpendicular to a central axis of rotation of the stator, and wherein the images depict cross-sectional areas of conductors of the stator and insulating paper surrounding the conductors.

14. The system of claim 10 , wherein the outputs from the machine learning models include binary masks enabling identification and quantifications of voids in the varnish.

15. The system of claim 10 , wherein the varnish fill percentages are displayed in the report according to the slots of the stator identified in the images.

16. The system of claim 15 , wherein the varnish fill percentages are further displayed according to a twist end, a crown end, and a central region of the stator for each of the slots.

17. A method for evaluating a varnish condition of a stator, comprising:
illuminating a cross-section of the stator with light from a UV light source;
obtaining a fluorescence image of the cross-section via a fluorescence detector and transmitting the fluorescence image to a deep learning tool implemented at a processor;
segmenting the fluorescence image, via the deep learning tool, into segments corresponding to slots of the stator;
cropping the segments, via the deep learning tool, to areas in the segments depicting the slots;
identifying and analyzing varnish in the segments using deep learning algorithms of the deep learning tool trained to identify and quantify varnish based on a fluorescence signature of the varnish; and
estimating varnish fill percentages of the slots and displaying the varnish fill percentages at a display device as a report.

18. The method of claim 17 , wherein illuminating the cross-section of the stator includes irradiating the cross-section with UV light at a wavelength inducing fluorescence at the varnish deposited in gaps between conductors of the stator and surfaces of the slots.

19. The method of claim 17 , wherein obtaining the fluorescence image includes generating the fluorescence image using pre-set, uniform image acquisition settings of the fluorescence detector.

20. The method of claim 17 , wherein the deep learning tool is trained based on training data generated by machine learning models utilizing cluster analysis and color-based thresholding to identify and quantify voids in the varnish and a relative amount of the varnish in the slots.

Assignees

• Ford Global Tech Llc

Inventors

• Schroeter Robert
• Avery Seth
• Wolf Chris
• Lenz Jackson
• Tan Boratha

Classifications

• G06V10/26 Primary

  Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion · CPC title

• G06T2207/10064

  Fluorescence image · CPC title

• G06T2207/10024

  Color image · CPC title

• G06T2207/30164

  Workpiece; Machine component · CPC title

• G06T2207/20084

  Artificial neural networks [ANN] · CPC title