Prediction of mechanical properties of weathered injection-molded polycarbonate/acrylonitrile-butadiene-styrene blends using Fourier-transform infrared spectroscopy-based 1D convolutional neural network

In this study, accelerated degradation tests under various temperature–humidity conditions and a 12-month outdoor weathering test were conducted on a polycarbonate/acrylonitrile–butadiene–styrene (50:50) blend to comprehensively evaluate the degradation mechanisms and behaviors of the material through mechanical, chemical, and surface analyses. Furthermore, a predictive model was developed to estimate the mechanical properties based on the degradation-induced variations in the material's Fourier-transform infrared (FTIR) spectral characteristics, even without prior knowledge of environmental factors, such as temperature, humidity, or exposure duration. A 1D convolutional neural network architecture, which is highly effective in learning spectral features, was employed to develop the prediction model. Using FTIR data obtained from the accelerated degradation tests and from specimens exposed outdoors for six months, the model successfully predicted the properties of specimens subjected to 12 months of outdoor exposure. The proposed model can thus perform reliable non-destructive assessment without requiring explicit information on the target's degradation environment. The findings of this study are expected to facilitate lifetime prediction and quality control of materials in various industrial applications.