Abstract— In this paper we compare the effectiveness of the time, wavelet and Hilbert Huang domain analysis of waveforms from a Non Destructive Test (NDT) equipment. The analysis of the signals from NDT systems is one of the most important activities since it allows to locate the actual defects. Very often the echoes traveling back to the sensors are superimposed with the signal from the transducer. We analyze the waveforms obtained by couples of sensors in order to extract the echoes and to recognize the points were they originate. A preliminary numerical analysis has been used to study the main features of the elastic waves propagations in inhomogeneous materials Keywords— Non Destructive Test, Hilbert Huang transform, Magnetostrictive transducers and sensors, Ultrasonic waves, Wavelet transform. I. INTRODUCTION LTRASOUNDS based techniques have been introduced for long range NDT in different engineering fields [1], [2]. In particular, in the case of plants with systems of pipes, ultrasonic torsional waves, guided by the walls of the pipes themselves, have been used to inspect portions of pipes up to 20 meters long. The presence of a defect in the structure produces an echo detected by properly positioned sensors. Although the position of the defects can be evaluated with sufficient precision, very little can be said about the extension and the shape that are, if necessary, investigated with other techniques thus reducing the time needed to complete the investigation. We propose the same approach for the NDT analysis of concrete or masonry walls; the first step would consist of the inspection of a relatively large portion of the structure with the aim of detecting and locate possible defects. The extension and the shape of the defect can be subsequently investigated by the use of the signal processing techniques on a narrow zone. The feasibility of a long distance inspection technique for concrete structure based on the propagation of the elastic ultrasonic waves has been investigated by the authors in Manuscript received May 21, 2007; Revised Received November 21, 2007. All authors are with Department of Electric Systems and Automation, University of Pisa, Italy, via Diotisalvi 2, 56126 Pisa, Italy (phone number +39 050 2217300, email {barmada, raugi, musolino, tucci, turcu}@dsea.unipi.it). previous papers [3], [4]. We will now focus on the analysis of the signals detected by the sensors that is crucial to the practical feasibility of the proposed approach. The presence of a defect produces an echo traveling back to the transducer. We can also consider the defect as a secondary source that activates when the wavefront from the transducer impacts the defect itself. The presence of superior modes across the thickness of the wall (when its thickness is comparable with the wavelength of the excitation) and the multiple reflections between the boundary surfaces makes it difficult to identify the echo in the waveform by sensor. However, if the sensors are placed at two points close to each other, it is possible to recognize the presence of a backward traveling wave by performing a proper comparison of the two waveforms. The knowledge of the arrival instants of the backward waves on an array of couples of sensors makes it possible to estimate the position of the defect. A correct estimate of the arrival instant is then crucial for a good performance of the triangulation algorithm that is usually used. The accuracy of the results produced by time domain analysis of the couple of waveforms from the sensors may be improved by performing further analysis in the wavelet and Hilbert Huang domains. II. THEORETICAL ANALYSIS In order to asses the validity of the proposed long range inspection technique several numerical analysis of the propagation of elastic waves have been performed both on concrete and brick masonry walls. We simulated the action of the transducer by imposing a tangential force on a small portion of the surface of the walls. Magnetostrictive transducers driven by known currents have been used in the experimental campaign. The waveform shown in fig. 1 has been used in all the simulations. The force is zero for instants after 0.12 ms. We considered a concrete wall 4 meters high (y direction), 1.5 meters long (x direction) with a thickness of 15 centimeters. The transducer is located in the centre of the upper surface. It is 2 centimeters high and 10 centimeters long and is able to transmit an x-directed force to the underlying portion of the surface. Table I shows the physical constant assumed for the concrete. Hilbert Huang and Wavelet Processing of Time Domain Signals from Ultrasoinc Guided Waves Magnetostrictive Sensors Arrays Sami Barmada, Antonino Musolino, Marco Raugi, Mauro Tucci, and Florin Turcu U INTERNATIONAL JOURNAL OF CIRCUITS, SYSTEMS AND SIGNAL PROCESSING Issue 4, Volume 1, 2007 327