Improving acoustic tomography resolution by optimizing the wave travel time detection

Different techniques are available to assess the presence and. extent of decay without causing damage to the trees. Among them, classical acoustic methods are based on the determination of the propagation velocity of a sound wave when measured on the radial axis of the trunk. Commercial tools include the Impulse Hammer (acoustic) and the use of low frequency ultrasounds. An extension of the classical acoustic methods has been the development of acoustic tomography. Acoustic (therefore ultrasonic) tomography is a safe testing technique that has been used for in situ inspection of trees, logs, lumber, poles and wood based-composites, as reported in publications since 1986 (Tomikawa et al. 1986; Johnstone et al. 2010). The resolution of tomographic images depends on the signal wave-length, on the accuracy of the time of flight (TOF) determination, on the number of sensors and on the inversion algorithm. In the particular case of the TOF determination, the commercial devices use the same principle of velocity measurement than the classical acoustic methods in transmission mode. In non-destructive testing with ultrasound, the TOF measurement is usually determined from the instantaneous power by taking the first point above the noise threshold (Beall 2002). Another approach is to determine the threshold proportional to the maximum amplitude of the temporal signal (Kanda et al. 1998). In general, for the case of acoustic and ultrasonic tomography, the literature does not deal with this subject. However, the Akaike Information Criterion has been used as the travel time picker (Schubertet al. 2009).The aim of this work is to improve the accuracy of the TOF determination. In a first step, five algorithms were tested with numerical signals in which the signal to noise ratio (SNR) varied. In the case of a bad SNR, due for example to a mismatch between the acoustic impedances of the transducers .and the object, the TOF identification on raw signals can be ambiguous and there solution of the tomographic image is very low. In the second step of the study, the conclusions of the numerical simulations were compared with experimental TOF detections on an isotropic material (academic or reference object) and on wood samples. (Résumé d'auteur