Development of an Early Degradation Diagnosis Technology for Wood Coatings Using Mid-Infrared Spectroscopy and Machine Learning

Accurately predicting the degradation state of wood coatings is challenging, and as the coating deteriorates, it becomes a major factor accelerating the degradation of the wood itself. In this study, we aimed to develop a diagnostic technology that combines mid-infrared spectroscopy and machine learning to assess latent coating degradation, which cannot be detected visually, and to enable appropriate maintenance before visible degradation occurs. This approach is expected to reduce recoating costs, lower environmental impact, and contribute to the expanded use of wood materials. In this study, we obtained mid-infrared spectroscopy data from model coatings artificially degraded in a short period using an accelerated weathering test device and constructed a predictive model for degradation time through regression analysis. Furthermore, field tests were conducted to verify the applicability of this diagnostic technology to actual buildings.