Flexible electronics are becoming the next generation of devices due to their advantages, such as mechanical flexibility, eco-friendliness, large-area applicability, and scalability for mass production. However, solution-based manufacturing processes are prone to defects like discontinuities and local smudging, which can significantly degrade both device quality and yield. To tackle these challenges, rapid and accurate defect classification is crucial for real-time diagnosis during manufacturing. This study investigates the impact of data scale and key training hyperparameters on the performance of deep learning–based defect diagnosis models, using a dataset of conductive pattern defects in flexible electronics. We specifically examine how the number of training images affects model accuracy and generalization, and we analyze how adjustments to hyperparameters—such as L2 regularization and dropout—influence model performance in data-limited scenarios. Our findings offer insights into optimal training strategies tailored to different data scales and learning constraints, providing practical guidelines for designing and developing AI-based defect diagnosis models for flexible electronic devices.

Printed electronics is a manufacturing technology that fabricates electronic devices using printing techniques. Due to its characteristics of low cost and simple process, a roll-to-roll printing technique has been used to achieve the large area and mass production of flexible electronic devices such as a thin film transistor. In the roll-to-roll printing process, a fidelity of the engraved pattern position is one of the most important techniques to fabricate high resolution multi-layer electronic devices. In this study, an engraved register mark position measurement system was developed to numerically evaluate the position accuracy of engraved mark in printing roll. The proposed system is based on a high-precision encoder based position control system and a high-resolution machine vision system. The measurement error of the developed system is within the camera resolution ±2.1 μm, verifying the superiority of the system. Using the developed system, we measured the position errors of the engraved register marks for six industrial scale printing rolls. This study suggests that the position error of the engraved mark should be considered to achieve a high precision register control below ±10 μm and necessity of the developed system.

Among the various next-generation solar cells, a perovskite solar cell can solve the economic problem because it can perform the low temperature solution process and the material is inexpensive. Photovoltaic conversion efficiency is comparable to silicon solar cells and thin-film solar cells. However, to commercialize the perovskite solar cells, there are many problems to be resolved, such as stability, upscaling, and efficiency. Thus, in this study, perovskite crystallization experiments were conducted according to the coating conditions such as the coating speed of the meniscus solution sheared coating process, and large-area perovskite solar cells with p-i-n structures were fabricated. Perovskite crystallization is one of the crucial factors that determine the efficiency of solar cells, and it is an integral process condition for manufacturing large-area perovskite solar cells.