Machine Learning Tool Detects Invisible Wood Coating Decay Before Visible Damage Appears

Recently, a team of researchers at Kyoto University and other Japanese institutions developed a novel machine learning (ML) diagnostic framework capable of identifying chemical deterioration in exterior wood coatings before any visible signs of failure emerge, according to a study published in Resources, Conservation, and Recycling.¹

Understanding how wooden artifacts and materials deteriorate is critical for evaluating their condition and creating effective conservation plans. Researchers are using advanced imaging technologies to investigate how chemical changes affect the wood’s structure and mechanical properties, providing insights that can support improved preservation efforts.² This study expands on this topic by investigating the utility of ML frameworks in this space, and how these frameworks can help support or enhance spectroscopic techniques.^1,2

What did the researchers do in their study?

The researchers tested wood coating decay by exploring how six commercial wood coatings, which included waterborne acrylic, polyurethane, and alkyd systems, held up to xenon-arc accelerated weathering. They then analyzed the resulting material changes using mid-infrared (MIR) spectroscopy paired with partial least squares (PLS) regression and a genetic algorithm for spectral wavenumber selection.¹ The models demonstrated high predictive accuracy and identified chemically meaningful degradation markers in each coating type.¹

What was the main finding of the study?

The main finding is that subvisible chemical changes, which is defined as those occurring well before cracking, flaking, or discoloration appears, can now be detected and quantified non-destructively.¹ Field validation under contrasting environmental conditions confirmed the method translates beyond laboratory conditions.¹

What are the main practical implications of this study?

Currently, visual inspections of building materials remain the most widely used method for determining the strength of these materials. Because of this common procedure, wood coatings end up being replaced reactively instead of proactively.¹ A diagnostic tool that flags chemical degradation at earlier stages could enable scheduled, targeted maintenance interventions, which reduces both material waste and labor costs.¹

By accelerating durability screening in the product development pipeline, coating formulators could use MIR–ML analysis to evaluate new formulations faster than conventional weathering trials allow.¹

References

1. Nishimura, K.; Isaji, S.; Ito, T.; Takano, T.; Ohki, H.; Teramoto, Y. Mid-infrared Spectroscopy and Machine Learning for Early Detection of Latent Deterioration in Exterior Wood Coatings Toward Extended Service Life. Res. Cons. Rec. 2026, 229, 108837. DOI: 10.1016/j.resconrec.2026.108837
2. Dal Fovo, A.; Cicchi, R.; Gagliardi, C.; Baria, E.; Fioravanti, M.; Fontana, R. Detecting Early Degradation of Wood Ultrastructure with Nonlinear Optical Imaging and Fluorescence Lifetime Analysis. Polymers (Basel). 2024, 16 (24), 3590. DOI: 10.3390/polym16243590