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.

In a roll-to-roll continuous system, winding is one of the most important processes since it determines the quality of the final manufactured products such as flexible film and printed electronic devices. Since an adequate winding tension can reduce the incidence of the defects that are derived from the inner stress of the wound roll such as starring and telescoping, it is necessary to determine the optimal taper-tension profile. In this study, an algorithm for the setting of an optimal taper-tension profile in consideration of the residual stress in the wound roll is suggested; furthermore, the algorithm was adjusted for the determination of an optimal tapertension profile regarding the winding process of 10 ㎛ polypropylene (PP) film. As a result of the algorithm-generated, optimal taper-tension profile, the residual stress and radial stress in a PP wound roll were decreased to 27.37 % and 40.05 % (mean value), respectively.

Roll-to-Roll printing process has become a great issue as a breakthrough for low cost and mass production of electronic devices such as organic thin film transistor, and etc. To print the electronic devices, multi-layer printing is essential, and high precision register control is required for this process. Unlike stop-and-repeat printing process, it is impossible to control the register in a static state since the roll-to-roll process is a continuous system. Therefore, the behavior of web such as polyethylene terephthalate (PET) and polyimide (PI) by the tensile and thermal stress generated in the roll-to-roll process as well as motor control of driven rolls has to be considered for a high precision register control. In this study, the correlation between curing temperature and thermal deformation of PET web is analyzed. Finally, it is verified experimentally that the temperature disturbance generates the more serious register error under the higher curing temperature.