P‐113: Temperature Prediction and Optimization of LCD Modules Using a Stacked Machine Learning Algorithm

High temperatures represent a significant failure risk in the display industry, especially for high‐power LCD products, where thermal management plays a critical role in ensuring reliability and performance. The reliance on traditional FEA simulations, which generally take approximately one week during initial development stages, renders the thermal design and optimization of LCD modules a time‐intensive and resource‐intensive process, limiting efficiency and scalability. This study introduces a stacked machine learning algorithm designed to efficiently predict high‐temperature risks in critical components of LCD modules. By leveraging the ensemble learning approach, the algorithm effectively addresses challenges in generalization and stability inherent to few‐shot learning, achieving a MAE (mean absolute error) of 0.5°C in temperature prediction. Furthermore, it enhances simulation efficiency by more than 90% compared to traditional methods, providing a significant advancement in thermal analysis and design optimization. To enhance interpretability, SHAP‐based explainable AI technique is used to identify critical factors influencing module temperatures and quantify the relationships between design parameters and maximum component temperatures. This methodology enables rapid evaluation of diverse design configurations, facilitates optimization beyond traditional experience‐based approaches, and significantly mitigates the thermal failure risk in high‐power MNT products.