US 12,379,716 B2
Methods and systems for verification of machine learning-based varnish analysis
Robert Schroeter, Livonia, MI (US); and Ife Siffre, Detroit, MI (US)
Assigned to Ford Global Technologies, LLC, Dearborn, MI (US)
Filed by Ford Global Technologies, LLC, Dearborn, MI (US)
Filed on Oct. 3, 2022, as Appl. No. 17/937,664.
Prior Publication US 2024/0111276 A1, Apr. 4, 2024
Int. Cl. G05B 19/418 (2006.01); G06T 7/00 (2017.01)
CPC G05B 19/41875 (2013.01) [G05B 19/4187 (2013.01); G06T 7/0004 (2013.01)] 7 Claims
OG exemplary drawing
 
1. A system for evaluating a deep learning tool for estimating varnish fill percentages of a stator, comprising:
a housing enclosing a UV light source and digital imaging equipment;
one or more replicas of a section of the stator having varnish configured to fluoresce when irradiated by the UV light source, the one or more replicas having amounts of varnish corresponding to predetermined varnish fill percentages; and
a processor configured with the deep learning tool and instructions stored on non-transitory memory that, when executed, cause the processor to:
receive images of the one or more replicas from the digital imaging equipment;
process and analyze the images using the deep learning tool by cropping the images and analyzing a fluorescence signature of the images, the deep learning tool trained to identify and quantify the varnish using deep learning algorithms;
output estimated varnish fill percentages from the deep learning tool;
compare the estimated varnish fill percentages to the predetermined varnish fill percentages; and
display a notification to a user at a display device in response to a difference between the estimated varnish fill percentages and the predetermined varnish fill percentages being greater than a threshold difference.