Combined machine learning–wave propagation approach for monitoring timber mechanical properties under UV aging

This study proposes a combined machine learning–wave propagation approach for nondestructive prediction of the modulus of elasticity (MOE) and rupture (MOR) of timber subjected to ultraviolet (UV) radiation. Fir, poplar, alder, and oak wood specimens were subjected to artificial UV aging and assessed using the Lamb wave propagation. Different features including the wave characteristics and the viscoelastic properties of the specimens were obtained from the Lamb wave propagation tests. The extracted features trained a decision tree model for MOE and MOR prediction. The UV radiation caused a decrease in the elastic properties of wood but increased its viscoelasticity. The results also showed a decrease in the wave velocity and an increase in the wave amplitude decay with the UV exposure time. It was revealed that compared with the wave velocity, the wave amplitude decay was better correlated to the MOE of MOR of UV-degraded wood. The MOE and MOR of UV-degraded wood were accurately predicted by the machine learning models fed by the features extracted from the Lamb wave propagation tests, where the shear storage modulus was found as the most important feature for training the models. It was concluded that the proposed approach offers a great tool for in-situ monitoring of wood structures under weathering and photodegradation conditions.