Analysis of adhesive degradation mechanisms and strength prediction based on principal component analysis

Understanding the degradation mechanisms of adhesives is essential for improving their durability. Although the chemical state variations of adhesives can be evaluated through infrared (IR) spectroscopy, quantitative analysis has proven challenging owing to various influencing factors. This study demonstrates the effectiveness of principal component analysis (PCA), a dimensionality reduction technique employed in adhesive research. By utilizing this method, the chemical variations were primarily represented by three major factors: moisture, oxidation, and filler. Incorporating these extracted factors as features in a machine learning model enabled high-accuracy prediction of adhesive tensile strength, even for samples exposed to unknown degradation environments, thereby extending the applicability of conventional methods. Notably, the principal components associated with moisture demonstrated a strong correlation with the tensile strength of cured epoxy adhesive, providing quantitative support for initial hypotheses. This study demonstrates the robustness and versatility of PCA for diagnosing adhesive degradation and assessing adhesive performance, validating the long-term durability and enhancing quality control, positioning PCA as a basis for broader practical implementation.