SAFT-Reconstruction in ultrasonic immersion technique using phased array transducers J. Kitze, J. Prager, R. Boehm, U. Völz, H.-J. Montag Federal Institute for Materials Research and Testing Berlin, Division 8.4, Unter den Eichen 87, 12205 Berlin ABSTRACT. The two main preconditions for the application of the Synthetic Aperture Focusing Technique (SAFT) are: (i) a large divergence of the sound beam of the transducer and (ii) an exact knowledge about the sound propagation path. These requirements are easily fulfilled for point sources directly mounted on the surface of the specimen. In many cases, however, the transducer is wedge mounted and/or coupled using a water delay line, e.g. in immersion technique. These delay lines change the beam index and the propagation path has to be evaluated for each pixel separately considering Fermat s principle. Using phased array transducers, a sector scan can improve the divergence of the sound beam. The introduced method combines the advantages of using a phased array transducer in immersion technique to improve SAFT reconstruction. An algorithm is presented accounting the influence of the delay line on the reconstruction method. The applicability of the algorithm is shown by validation with simulated echo responses and with experimental results collected from a specimen with artificial flaws. Keywords: Ultrasonic, phased array, SAFT, immersion technique PACS: 43.58.+z, 43.60.+d INTRODUCTION SAFT has been used successfully as a reconstruction procedure of ultrasonic inspections in contact technique in order to improve the Signal to Noise Ratio (SNR), the resolution and the sizing capability [1]. To guarantee a constant acoustic coupling and a high resolution, a practical implementation of the ultrasonic testing is easier to realize in immersion technique, i.e., coupled using a constant water delay line, than in contact technique. Thus, we investigated which preconditions must be fulfilled so that the well known SAFT algorithm can be applied also to the ultrasonic measuring data in immersion technique. Furthermore, we examined the advantages of phased array transducers on the SAFT reconstruction in immersion technique. An existing software was enhanced and ultrasonic tests were performed to verify the applicability of the modified SAFT algorithm. THE SAFT PRINCIPLE SAFT is a travel time based analysis method with subsequent image reconstruction. Even if the method is also available in 3-D, for sake of simplicity, only the 2-D reconstruction will be described in the following. A detailed formulation of the method is presented e.g. in [2].
During the scanning process the test volume - Region Of Interest (RIO) - is insonified by ultrasound waves as sketched in Figure 1. The reconstruction algorithm implies high frequency (RF), digital measurement data Φ( x, dt, ) from a wide angular range of the test specimen. Accordingly, the test specimen has to be scanned using a divergence angle as large as possible in a linear path for a 2D-reconstruction. The large divergence can be achieved using a conventional transducer with a small transducer area or using phased array transducers with a swiveled sound beam performing a sector scan. FIGURE 1. SAFT principle, Isochrones for SAFT reconstruction at different scanning positions For the reconstruction in 2D the ROI given as ( x, z d) is discretized into a pixel grid where each pixel is considered successively as a reflector and gets assigned o x, z. Therefore, the SAFT-method arranges the echo signals a reflection amplitude ( ) ( x, dt, ) From each scanning position (, ) Φ into this grid according to their time of flight and divergence angle (Fig. 1). x d with x SM, where S M is the scan path, isochrones can be constructed as ct = x x 2 + d z 2. (1) ( ) ( ) Note, in pulse-echo technique this distance is travelled twice. Summing up the contributions from all scanning positions along S M yields the amplitudes in each pixel as 2 o x, z x, d, t x x d z dx. ( ) = Φ = ( ) + ( ) 2 2 S M c (2) In this numerical superposition of the Time Of Flight (TOF) dependent echoes the phase positions are taken into account. For the actual reflection spot o( x 0, z 0), constructive interference provides increasing amplitude in the reconstruction with the number of echoes from different transducer positions. For all other pixels the superposition yields only small or negligible echoes due to destructive interferences. This superposition requires knowledge of the TOF for potential echoes which might come from the pixels [3]. This is trivial as long as using point sources directly mounted on the surface of the specimen. Thus, simple geometrical relations are valid for calculating the sound propagation path and therewith the TOF from the transducer to each pixel in the specimens ROI (see Fig. 2, left). Due to the refraction the determination of the beam index and the propagation path is not trivial for transducers with conventional dimensions and additional delay lines. A relative shift of the beam index in space (Δx) and time (Δt) must be considered for each pixel separately (see Fig. 2, right).
In the existing SAFT algorithms this shifts are not considered. For a small transducer only the solid marked paths along the sound beam axis are taken into account during computation (Fig. 2, right). The real path of the ultrasound according to Fermat`s principle and Snell`s law, however, is the dashed marked path (Fig. 2, right). This results in a relative shift of the beam index (Δx) and the TOF (Δt). This shift must be considered for the calculation of the propagation path and the correct allocation of the received echo signals towards every pixel in the SAFT reconstruction (see Fig. 3a). In fact, the refraction law has to be fulfilled for each pixel separately [4]. FIGURE 2. SAFT in contact technique (left) vs. SAFT in immersion technique (right) While using a phased array probe the shift of the beam index and the delay time are already determined for every angle of incidence when calculating the delay times for the sector scan (see Fig. 3b). Reusing these values, the time consuming correction of the variable delay line for each pixel can be omitted, which reduces the computation time drastically. Strictly speaking, these values are only valid at the axis of each sound beam. As the divergence of the sound beam is quite small when using a phased array transducer, the values are adequate for all pixels inside the divergence of each single beam. Thus, only these approximate correction values are used in the SAFT algorithm. It is possible to apply this principle also to the small transducer to save computation time by discretizing the whole divergence angle of the sound beam into sufficiently small sectors. Therewith, the relative shift of the beam index and the time of flight are calculated only for each sector and not for each pixel (see Fig. 3c). FIGURE 3. Alternatives for the determination of the TOF in the SAFT algorithm for, a) single element, for each pixel, b) phased array, for each swivel angle, c) single element, for each sector In the case of curved surfaces additionally it is necessary to consider the tilt (β) of the normal surface at the beam index to determine the sound propagation path accurately (Fig. 4). The tilting of the surface normal increases with the increase of the curvature of the surface.
These three factors - the relative shift of the beam index (Δx), the TOF (Δt) and the tilt of the surface normal (β) in case of curved surfaces - have to be considered for the successful application of SAFT in ultrasonic immersion technique. The SAFT software, developed at the Federal Institute for Materials Research and Testing, is enhanced by these three features. FIGURE 4. Tilting of the surface normal depending on the curvature of the reference block MEASUREMENT SET-UP Ultrasonic investigations to verify the modified SAFT algorithm are carried out on a planar and a cylindrical test specimen using a 10 MHz linear array. The water delay line is set to 20 mm. The specimens contain three side drilled holes with a diameter of 2 mm (Fig. 5). The aim of our investigations is to detect and localize the test reflectors and to examine the applicability of SAFT for inspections using ultrasonic immersion technique. All tests are performed using a manipulator scanning the surfaces of the specimens. A Compas XL phased array equipment [5] with an integrated SAFT software module, modified by implementing the additional features explained above, is used to collect and process data. FIGURE 5. Measurement set-up using a 10 MHz linear array in ultrasonic immersion technique, test specimen with three side drilled holes planar (left) and cylindrical (right) MEASUREMENT RESULTS Planar Test Specimen First we have carried out investigations using one single element of the phased array transducer simulating a small conventional transducer with a large divergence of the sound beam, to fulfill the general precondition for SAFT. Figure 6 shows the rectified SAFT reconstruction of the measurements using one single element and illustrates the error that occurs without compensation of the variable delay line for each pixel. The shown echo indications are greatly broadened because the assumed TOF are inaccurate without compensating the variable delay line for each pixel.
FIGURE 6. SAFT reconstruction 10 MHz linear array, single element, divergence 60, planar specimen, without compensation of the variable delay line for each pixel As expected, the SAFT reconstruction with additional compensation of the variable delay line for each pixel shows small indications of the three side drilled holes (Fig. 7). Due to this compensation it is now possible to localize the received echoes accurately. However, the calculations for each pixel increase the computational effort drastically. FIGURE 7. SAFT reconstruction 10 MHz linear array, single element, divergence 60, planar specimen, with compensation of the variable delay line for each pixel As stated above, the compensation of the variable delay line for each pixel is not required using phased array transducers performing a sector scan. Here the approximate values for the beam axis calculated during determination of the delay times are used. Again, we have used a 10 MHz linear array, but with 8 active elements. The sound beam is steered between -30 and 30 with an angular step size of 5. Figure 8 shows the result of the SAFT reconstruction of the measurements using a 10 MHz linear array with compensation of the variable delay line only for every swivel angle, not for each pixel. The result differs only slightly from the reconstruction with a single element with compensation of the variable delay line for each pixel.
FIGURE 8. SAFT reconstruction 10 MHz linear array, phased array (8 elements), sound beam swiveled from -30 to 30, divergence 5, planar specimen, with compensation of the variable delay line for every swivel angle, not for each pixel The advantages of the investigations using phased array transducers are: on the one hand, an increase of the SNR of about 6 db, which is observed while comparing the background noise level from Figure 7 and 8 and on the other hand, a significant reduction of computation time. As discussed before, the discretization of the large divergence angle of a single element transducer into sufficiently small sectors produces similar results as those using a phased array transducer with sector scan (see Fig. 9). However, less computation time is required compared to the exact results presented in Figure 7 as the relative shift of the beam index and the time of flight are calculated only once for each sector and not for each pixel. The disadvantage is the loss of SNR and the occurrence of artefacts. FIGURE 9: SAFT reconstruction 10 MHz linear array, single element with discretization of the angular range of 5 (sectors), planar specimen, with compensation of the variable delay line for every sector, not for each pixel Cylindrical Test Specimen The ultrasonic inspections at the cylindrical test specimen are performed in the same manner as for the planar test specimen. The findings are unrestrictedly valid for both cases. Figure 10 shows the SAFT reconstruction of a 180 scan of the cylindrical
test specimen using a phased array transducer. The three side drilled holes are clearly visible in the reconstruction. The size of the indications near the surface is in the range of the wavelength. However, the reconstruction does not show such clear indications as for the planar test specimen. There are two possible reasons for this difference, (i) a not exactly circular test specimen and (ii) inaccuracies in the adjustment. The longer the sound path, the greater is the influence. That s why the side drilled hole in the middle of the test specimen is more blurred than those near the surface. FIGURE 10. SAFT reconstruction 10 MHz linear array, phased array (8 elements), sound beam swiveled from -30 to 30, divergence 5, cylindrical specimen, scan of 180, with compensation of the variable delay line for every swivel angle, not for each pixel COMPARISON WITH SIMULATION RESULTS We have carried out comparative simulations to verify our experimental results. Figure 11 shows the comparison of the experimental and simulated SAFT results for a phased array transducer with compensation. FIGURE 11. SAFT reconstruction 10 MHz linear array measurement vs. simulation, phased array 8 elements, with compensation of variable delay line for every swivel angle, not for each pixel The results of the simulation are in good agreement with the experimental data. Minor deviations are observed in shape and echo height due to the reason that the simulations are calculated for the ideal case, i.e., for the ideal reflector, the ideal sound field and the ideal adjustment. Minor inaccuracies when the determining the sound velocity, the sound path or the curvature of the curved test specimen have a strong influence on the results of the SAFT reconstruction.
CONCLUSIONS AND OUTLOOK The SAFT principle is extended for ultrasonic immersion technique for planar and cylindrical components. For those applications it is necessary to calculate the variable TOF caused by the refraction between the water delay line and the specimen. In case of small transducers the variable TOF must be considered for each pixel in the reconstruction. When using phased array transducers the necessary correction of the variable delay line can be simplified by reusing the information on the shift of beam index and delay time. These values are provided by the phased array system. This reduces computation time. The error occurring within the small divergence of the sound beam was shown to be negligible. Additionally the use of phased array probes increases the SNR. Using a small transducer it is feasible to save computation time by discretization of the angular range given by the divergence. REFERENCES 1. T. Hauser, H.-J. Montag, R. Boehm, U. Voelz: Vergleich von Rekonstruktionsverfahren auf der Basis von Gruppenstrahler-Ultraschalldaten, (in German) in Deutsche Gesellschaft für Zerstörungsfreie Prüfung Jahrestagung, edited by DGZfP Berlin,1998, Bamberg, 7-9 September 1998,, pp. 561-69. 2. K.-J., Langenberg, R. Marklein, K. Mayer: Theoretische Grundlagen der zerstörungsfreien Materialprüfung mit Ultraschall (in German), in Oldenbourg Wissenschaftsverlag GmbH, München, 2009. 3. A. Chahbaz, R. Sicard: Comparative Evaluation Between Ultrasonic Phased Array And Synthetic Aperture Focusing Techniques, in Review of Quantitative Nondestructive Evaluation, edited by Thompson and Chimenti, Vol. 22, American Institute of Physics (AIP) Conference Proceedings, Vol. 657, 2003, pp. 769-776. 4. T. Stepinski, F. Lingvall: Synthetic aperture focusing techniques for ultrasonic imaging of solid objects, 8th European Conf. on Synthetic Aperture Radar, Aachen, Germany, 7-10 June 2010. 5. G. Schenk, U. Völz, E. Dohse, BAM Berlin, L. Bauer, ASIC Design Werder: COMPASXL - Outstanding number of channels with a new phased array system, in Proceedings of ECNDT, Vol. 11, Berlin, 25-29 September 2006.