P‐158: <i>Distinguished Poster:</i> Early Detection of Crack Vulnerability in Foldable Displays Through Critical‐Angle Curvature Analysis

Foldable displays have become increasingly popular due to their portability and the provision of larger screens when unfolded. However, the use of flexible materials makes them more susceptible to mechanical failures, particularly crack formation in the hinge area where stress concentration is highest during folding operations. This study presents a novel approach to predict crack formation in foldable displays by measuring the curvature of the hinge area at a specific critical angle—where the mechanical stress peaks—and analyzing this data using Principal Component Analysis (PCA). The curvature measurements were initially reduced using PCA, retaining two principal components that captured around 60% of the dataset's variance. While this facilitated visualization, it was insufficient for clear classification between defective and non‐defective samples due to overlapping data points and information loss. To enhance predictive accuracy, we introduced an uncertainty quantification method based on a ‐ nearest neighbors approach with distance‐weighted contributions. This method calculates an uncertainty score for each sample, reflecting the confidence in classifying it as susceptible or resistant to cracking. By integrating the uncertainty analysis into an artificial intelligence (AI) model, we assigned a numerical score ranging from +1 (clearly non‐defective) to ‒1 (high risk of cracking) to each sample. Applying this model to a larger dataset of 200 samples, we successfully quantified the risk of crack occurrence—a task previously unfeasible with conventional methods. The model demonstrated strong discriminative power, effectively distinguishing between at‐risk and safe samples. This integrated methodology combining precise curvature measurement, uncertainty analysis, and AI modeling provides a robust framework for early detection and prediction of crack formation in foldable displays. The approach enhances product reliability and contributes valuable insights into material behavior under stress, supporting improved quality control measures in manufacturing. Future work will focus on expanding the dataset and incorporating additional features to further refine the model's predictive capabilities.