Adaptive Voronoi Diagrams for Leakage Detection in Composite Manufacturing

Abstract

The aim of this study is to develop and compare custom machine learning model against multiple classical machine learning model for detecting potential leakage region in the vacuum film used in the production of Carbon Fiber Reinforced Polymer (CFRP) components. Ensuring the airtightness of the vacuum film is crucial in CFRP manufacturing to prevent defects in the final product. In practice, it is possible that the vacuum film may contain leaks that are not visible to the human eye. Therefore, to predict the leakage region, we implement a custom machine learning framework using TensorFlow’s GradientTape, while classical machine learning approaches, including K-Nearest Neighbors, Decision Trees, and Classification Network (MLP-Classifier), are utilized for comparison. Comparative analysis reveals that custom machine learning models perform well overall and does better generalization as compared to the classical machine learning algorithms, indicating its potential effectiveness in practical applications within CFRP manufacturing processes. This study highlights the effectiveness of the custom machine learning approach, particularly TensorFlow’s GradientTape, in accurately detecting leakage regions in vacuum films for CFRP component manufacturing. These findings underscore the importance of leveraging advanced machine learning techniques to enhance quality control and ensure the integrity of composite materials production.