Inspection of Enamel Removal Using Infrared Thermal Imaging and Machine Learning Techniques

Area of Science:

• Manufacturing Engineering
• Materials Science
• Artificial Intelligence

Background:

• Current quality assurance in aerospace and automotive manufacturing relies heavily on post-process inspection, missing opportunities for in-process defect detection.
• A significant gap exists in research concerning the inspection of termination manufacturing processes, particularly enamel removal on Litz wire.

Purpose of the Study:

• To develop and evaluate an automated inspection system for the enamel removal process on Litz wire using infrared thermal imaging and machine learning.
• To assess the feasibility and performance of various machine learning classifier models for detecting residual enamel.

Main Methods:

• Utilized infrared thermal imaging to capture temperature profiles of Litz wire bundles with and without enamel.
• Applied machine learning techniques, including Gaussian Mixture Models and Support Vector Classification, for automated enamel removal inspection.
• Evaluated and compared the classification accuracy and evaluation time of different machine learning models.

Main Results:

• The Gaussian Mixture Model (GMM) with expectation maximization achieved a 100% enamel classification accuracy and a rapid evaluation time of 1.05 seconds.
• Support Vector Classification (SVC) models demonstrated high training and classification accuracy (over 82%) but had significantly longer evaluation times (134 seconds).
• GMM proved to be the most effective model for accurate and efficient automated inspection of enamel removal.

Conclusions:

• Infrared thermal imaging combined with machine learning offers a viable solution for automated, in-process inspection of enamel removal on Litz wire.
• The Gaussian Mixture Model is a highly effective tool for this application, ensuring consistent product quality and reducing manufacturing scrap in critical industries.