Prediction of the mechanical properties of wood using guided wave propagation and machine learning

Abstract

Accurate nondestructive prediction of the mechanical properties of wood is a critical quality control task for timber grading. Currently, the ultrasonic wave method is widely used for this purpose, which overestimates the static mechanical properties of wood. To resolve this challenge, this paper proposes a robust machine learning-based model for predicting the modulus of elasticity (MOE) and modulus of rupture (MOR) of wood with varying moisture content (MC) using the guided Lamb wave propagation method. It combines the Lamb wave velocity with the “group method of data handling” to predict the mechanical properties of wood. The effect of MC and density on the Lamb wave velocity and the mechanical properties of seventy green poplar wood specimens were analyzed and discussed. It was shown that the Lamb wave velocity is highly sensitive to the MC, density, and mechanical properties of wood. It enables the guided Lamb wave propagation to be used as a reliable method for nondestructive timber characterization. The MOE and MOR of wood specimens predicted by the developed machine learning model were compared with those from the three-point bending tests, and the obtained accuracy was revealed to be higher than that obtained in the literature using the conventional ultrasonic method. It was concluded that the proposed combined Lamb wave propagation and machine learning method offers a strong tool for nondestructive prediction of the mechanical properties of wood.