CLASSIFICATION OF WOOD SPECIES BY NEURAL-NETWORK ANALYSIS OF ULTRASONIC SIGNALS

The passage of ultrasonic waves through an anisotropic inhomogeneous m aterial such as wood involves complex interactions between the physica l vibrations of the ultrasound and the elastic response of the wood. T he initial ultrasound signal is modified by the transmission medium in a way characteristic of the elastic anisotropy of the medium. The man y species of wood have subtly different elastic responses. In this wor k the characteristic signals formed by these responses is examined. A neural network system is used to classify these signals in terms of sp ecies. The neural network is shown to have a high success rate in iden tifying wood species from the ultrasonic trace. It is established that this identification is not possible using wave velocity or received s ignal amplitudes. The most appropriate propagation direction for speci es identification is also considered. (C) 1998 Elsevier Science B.V.