Development of Photo-Polymerization-Type 3D Printer for High-Viscosity Ceramic Resin Using CNN-Based Surface Defect...

Ceramic materials are known for their hardness and brittleness, which make traditional machining methods inefficient and prone to defects. While additive manufacturing has been explored for ceramics, it faces significant challenges in handling high-viscosity resins and detecting surface flaws during printing. Prior research has shown that conventional cutting techniques often result in poor-quality outputs and machining difficulties. It was already known that ceramic additive manufacturing struggles with maintaining structural integrity due to defects like pores and cracks. No prior work had resolved how to monitor and adjust for these defects in real time. That uncertainty drove the need for a system that integrates image processing and machine learning to detect and classify surface flaws. This gap motivated the development of a CNN-based approach for real-time monitoring. The study builds on existing knowledge of ceramic material behavior and image classification techniques.

The aim of this study was to develop a system for real-time monitoring of surface defects during the printing of high-viscosity ceramic resin using convolutional neural networks. The specific problem addressed is the poor quality and machining challenges caused by defects like pores and cracks in ceramic additive manufacturing. The motivation for this work stems from the limitations of conventional cutting and additive methods in handling ceramic materials. The study sought to improve classification accuracy and efficiency of defect detection by leveraging image processing and CNN algorithms. The researchers proposed a system that not only detects defects but also provides feedback for process adjustments. The goal was to enhance the quality of printed ceramic structures by identifying and mitigating defects during the printing process. The study focused on classifying defects into four categories and testing their impact on the final product.

The study utilized a CNN-based system for real-time monitoring of surface defects during ceramic resin printing. To classify defects, the researchers categorized them into four types: pore, minor, critical, and error. Image preprocessing steps included cropping, dimensionality reduction, and RGB pixel standardization to improve classification accuracy. The preprocessed images were then trained and tested using the DenseNet algorithm. The effect of each defect type on the printed structure was evaluated through experiments. A reblading process was tested to improve surfaces with pore and minor defects. The study combined image processing techniques with machine learning to develop a feedback system for process modifications. The results were validated through classification accuracy and experimental improvements in defect surfaces.

The study achieved a classification accuracy of 98% using the DenseNet algorithm after preprocessing images with cropping, dimensionality reduction, and RGB pixel standardization. Defects were successfully classified into four types: pore, minor, critical, and error. Experiments confirmed that the reblading process improved surfaces with pore and minor defects. The CNN-based system demonstrated high efficiency in distinguishing between normal and defective states. The feedback system allowed for real-time process adjustments based on defect classification. The results suggest that the proposed system can effectively monitor and mitigate surface defects during ceramic resin printing. The high accuracy of the classification system indicates its potential for practical application in additive manufacturing. The study provides evidence that integrating image processing and machine learning can enhance the quality of printed ceramic structures.

The study concluded that a CNN-based system can effectively monitor surface defects during the printing of high-viscosity ceramic resin. The proposed system achieved high classification accuracy and enabled real-time process adjustments. The researchers proposed that the feedback system based on defect classification can improve the quality of printed ceramic structures. The study demonstrated that pore and minor defects can be mitigated through the reblading process. The results suggest that the system can be applied to enhance the reliability of ceramic additive manufacturing. The authors highlighted the importance of integrating image processing and machine learning for defect detection. The study provides a framework for monitoring and adjusting printing processes in real time. The findings support the potential of the proposed system for practical use in ceramic additive manufacturing.