ULTRASONIC IMAGING of COPPER MATERIAL USING HARMONIC COMPONENTS

ULTRASONIC IMAGING of COPPER MATERIAL USING HARMONIC COMPONENTS T. Stepinski P. Wu Uppsala University Signals and Systems P.O. Box 528, SE- 75 2 Uppsala Sweden

ULTRASONIC IMAGING of COPPER MATERIAL USING HARMONIC COMPONENTS T. Stepinski P. Wu Uppsala University Signals and Systems P.O. Box 528, SE- 75 2 Uppsala Sweden Abstract In our experiments with inspection of copper specimens it appeared that when transmitting ultrasound pulses in one frequency band in the immersed solids the received echoes had a broad frequency spectrum containing both the fundamental and higher harmonic frequency bands. The setup of the experiments consisted of two ultrasonic transducers with different center frequencies and bandwidths. A narrowband transmitting transducer had a lower center frequency and was excited with a pulse, while a broad band transducer with higher center frequency was used as a receiver. The higher resonances of the transmitter excited by the broadband pulse resulted in the higher harmonics. It is known from medical ultrasonic imaging that the high harmonics due to the nonlinearitiy of wave propagation can be combined to enhance the signal to noise ratio and the resolution in B-scan images. Using the similar idea the paper presents an attempt to apply such technique to the NDE for improving the quality of ultrasound B-scans acquired in immersion inspection of copper. To enhance the signal to noise ratio, a novel method for processing signals is proposed in the paper. The method implements 2-D matched filters for filtering B-scans. The algorithm uses a prototype signal, a B-scan region selected from the acquired ultrasonic data in a selected frequency band as the prototype signal for designing a matched filter. The resulting filter was used to process individual B-scans and then a C-scan was extracted from the processed B-scans. The ultrasonic data used were acquired in the inspection of electron beam welds in copper canisters for spent nuclear fuel, and processed by the proposed algorithm with the aim of enhancing the echoes from voids and porosity masked by backscattering from the weld structure.

Introduction Over the recent years harmonic imaging technology has been rapidly developed and widely applied to medical ultrasound imaging that has enabled a substantial improvement in the image quality. This technology was initially developed for contrast harmonic imaging (also called contrast agent harmonic imaging) [] and later for tissue harmonic imaging (also called native tissue harmonic imaging) [2]. The contrast harmonic imaging is based on the fact that small contrast microbubbles radiate significant energy at multiples of the interrogation frequency. The tissue harmonic imaging, on the other hand, exploits the nonlinearities of ultrasound propagation that gradually change the shape of the wave as it passes through tissue, and thus result in the harmonic components [3]. In addition to the above two sources, harmonic components can also be produced by transducers. In this paper, we will present an attempt to apply such technique to the NDE for improving the quality of ultrasound B- scans acquired in immersion inspection of copper. Our idea is similar to that in the study by Ward, Baker, and Humphrey [4] in relation to native harmonic imaging, but motivated by our experience with copper specimens. Inspection of the copper specimens was made in pitch-catch mode by two ultrasonic transducers with different center frequencies and bandwidths. The transmitter had a lower center frequency and a narrow band, and was excited with a broad band pulse. A broadband transducer with a higher center frequency was used as a receiver. In such an experimental setup, the higher resonances of the transmitter were excited due to the broadband pulse and resulted in the higher harmonics. Because of the broadband frequency property the receiver captured echoes containing both the fundamental and harmonic components. All the fundamental and harmonic components were extracted and then used separately or in combination. Our detailed ultrasonic investigation was made on copper specimens welded by means of electron beam (EB) technique. The interest for EB welded copper originates from our project concerned with ultrasonic inspection of copper canisters for spent nuclear fuel (5). One of our main goals during ultrasonic assessment of EB welds is detecting small voids and porosity present in the weld zone. Unfortunately, the weld zone is characterized by a coarse internal material structure, which results in a great deal of backscattering from this volume that limits the detectability of voids and porosity (6). We have already tested, without great success, various methods known to be capable of suppressing material noise, such as split spectrum processing (SSP) or non-coherent detection (NCD) (7). In this situation we are investigating all possible physical phenomena that could be used for improving the imaging of defects in the EB weld. Making use of higher harmonic components in the ultrasonic signals opened a new direction in our research. Although harmonic components may be created by the nonlinearity of wave propagation, they are small compared to those created by the high resonances of transducers in the present case. Thus, we limit ourselves to the higher harmonic components from transducers and investigating their usefulness for the

imaging. The presented results are preliminary but we hope, interesting for the NDE community. Experimental method The aim of the experiments was to explore the higher harmonics mainly resulting from higher resonant modes of transmitters. A transducer pair (transmitter and receiver) aiming at the same volume in the metal specimen immersed in water was used for this purpose. By scanning the transducer pair a number of B-scans were acquired and used for extracting C-scans. In the experimental setup we used ALLIN Ultrasonic Array System manufactured by R/D Tech, France (Fig. ). The ALLIN system consisted of a PC controlled electronic system and a New Port scanner. DC motors moved the transducers mounted on the vertical axis in the x- and y-directions in the horizontal plane (see Fig.). A spherically focused transducer (PANAMETRICS V392) with a 38-mm diameter and a 9-mm focal length was used as a transmitter. It had a.85-mhz center frequency (measured), a 87%, 6-dB bandwidth (although relative bandwidth is broad, the absolute bandwidth is not because of the low center frequency). The receivers used were two broadband planar transducers (PANAMETRICS V326 & V327) that had center frequencies of 4.7 and 7 MHz (measured), respectively. For each transducer pair, a 3-D RF data set had been acquired and used to create the A-, B- and C- scans of specimen. Figure. Experimental setup with two transducers, lower frequency focused transmitter (left) and higher frequency receiver (right).

The inspected copper specimen was a section of copper canister with an electron beam weld; it had a set of sidedrilled holes (SDH) and bottom- drilled holes (BH) (see Fig. 2). Two identical sets of holes were manufactured in the specimen, the holes # to #9 and # to #8 (upper and lower holes in Fig. 2, respectively). The only difference was that the upper holes were drilled in the welded zone while the lower holes were in solid copper. A substantial scattering from the weld structure was expected and the lower holes were made as a reference. The specimen was placed in the tank and inspected from the top in pitchcatch mode. The reflecting surfaces of the holes were approximately at a 6-mm depth beneath the top surface. Figure 2. Geometry of the inspected copper canister section with drilled holes. Matched 2-D filter A simple 2-D matched filter is proposed for the filtering ultrasonic data in both temporal and spatial frequency domains simultaneously. This means that the filter acts both as a band-pass filter extracting the respective harmonic component and also as a spatial filter suppressing the ultrasonic responses in directions other than those of the ones defined in the prototype. The matched 2-D filter requires a prototype, a small matrix that is used as a finite impulse response (FIR) 2-D filter. In our case this matrix is chosen by manual selecting a part of B-scan that is representing the interesting echoes. This matrix is first inverted with 8 and than convoluted with the processed B-scan using a 2-D convolution operation. To avoid artifacts the prototype can be windowed by an appropriate windowing function. The result of filtering is illustrated in Figure 5 below, using a B-scan with well-pronounced vertical lines that have been suppressed by the filter. Experimental results The experiments with the above described transducer pairs had two particular aims: To detect and evaluate the presence of higher harmonics in an ultrasonic signal scattered from a copper specimen immersed in water.

To evaluate the information carried by the higher harmonics and its usefulness for imaging the internal specimen structure. To detect the presence of the higher harmonic components in the reflected ultrasonic signal a number of B-scan has been acquired and spectra of the individual scans were estimated. This has been done for both transducer pairs using broadband spike pulse. The higher harmonic components were detected in the signal. 5 2 4 4 Transducer position in mm 6 8 2 4 6 Power spectrum in db 3 2 8 2 22 34 35 36 37 38 39 4 4 42 43 Depth in copper in mm - -2 2 3 4 5 6 7 8 9 frequency in MHz Figure 3a. B-scan of the lower holes (left) and its power spectrum (right). Signal from hole # corresponds to the lowest echo in the B-scan. Power spectrum was calculated for the A-scan at 5-mm (response of hole #2). 5 Transducer position in mm 2 4 6 8 2 4 6 Power spectrum in db 4 3 2-8 2 22 34 35 36 37 38 39 4 4 42 43 Depth in copper in mm -2-3 2 3 4 5 6 7 8 9 frequency in MHz Figure 3b. B-scan of the upper holes (left) and its power spectrum (right). Signal from hole # corresponds to the lowest echo in the B-scan. Power spectrum was calculated for the A-scan at 5 mm (response of hole #3).

An example of this result is shown in Figure 3 where two B-scans of the specimen are shown together with power spectra of selected A-scans. The B-scans were acquired along the specimen, so that they contained the responses of all drilled holes. The SDH s are the strongest reflectors (in the bottom part of Figure 3). The B-scan in Figure 3a corresponds to the area without the weld (lower holes) while the scan in Figure 3b was measured in the welded part of the specimen. The backscattering from the weld structure is clearly visible. In both cases the power spectra of A-scan corresponding to 5 mm were calculated using Welch method with Hanning window, and clear pronounced maxima are seen at frequencies of approx. MHz, 3MHz and 5.5 MHz (the measurement was performed using the transducer pair V392 and V327). A-scan before filtering.5 5 Transducer position in mm 5 -.5-5 52 54 56 58 6 62 64 66 68 7.5 A-scan after filtering 2 -.5 34 35 36 37 38 39 4 4 42 43 Depth in copper in mm - 5 52 54 56 58 6 62 64 66 68 7 Figure 4a. B-scan of the lower holes after low pass filtering (left). A-scan 5 mm before, and after LP filtering (right). A-scan before filtering.5 5 Transducer position in mm 5 -.5-5 52 54 56 58 6 62 64 66 68 7.5 A-scan after filtering 2 -.5 34 35 36 37 38 39 4 4 42 43 Depth in copper in mm - 5 52 54 56 58 6 62 64 66 68 7 Figure 4b. B-scan of the lower holes after high pass filtering (left). A-scan 5 mm before, and after HP filtering (right).

Now we will try to evaluate usefulness of the harmonic components for the imaging of discontinuities inside the specimen. The first step in this evaluation consisted in isolating individual frequency bands in the ultrasonic signals and studying the filtered B-scans. The results are shown in Figure 4, where the B-scan from Figure 3a is shown after a low-pass and a high pass filtering (Figure 4a and 4b, respectively). The LP and HP filters were realized in frequency domain by windowing the A-scans FFT with a rectangular window. The cut off frequency of these windows was at the first dip of the power spectrum, that is approx. 2.2 MHz. The right panels of Figure 4 show the A-scan 5 mm before and after filtering, while the left panel presents B-scans consisting filtered A-scans. It is apparent that the HP filtered B-scan contains information about the holes drilled in the specimen. B-scan before filtering B-scan after filtering 5 5 5 5 2 2 34 36 38 4 42 34 36 38 4 42 Figure 5. B-scan of the upper holes before (left panel) and after (right panel) 2-D matched filtering. The next step of our evaluation was comparing the C-scans obtained from the filtered B- scans. However, before extracting C-scans the B-scans were filtered by the above described 2-D matched filters. Their aim was extracting the respective harmonic component and also vertical lines in the B-scan (cf. Figure 5). A section of B-scan corresponding to the response of the hole # (SDH) was used as a prototype for the filter. It is easy to see that the filter has suppressed the vertical lines and limited the spectrum (corresponding to the first harmonic in this case). The C-scans extracted from the processed A-scans (BP filters) and B-scans (2-D matched filetrs) are presented in Figure 6. All the C-scans were extrated from the A-scan envelope calculated using Hilbert transform. Since side-drilled holes result in very strong reflections only the left part of the C-scans is shown in Figure 6. The B-scan lines

corespond to the horizontal direction in the C-scans. Backscattering from the welded area of the specimen is visible in the upper part of each C-scan. The C-scans in the left column of Figure 6 were obtained from the A-scans processed by the described above band-pass filters operating in frequency domain, while the C-scans in the right column were extracted using the 2-D matched filters. The C-scans in the upper row of Figure 6 were obtained using the whole frequency spectrum of the signals, in other words, the left C-scan was extracted from the unfiltered data and the right one from the B-scan data filtered by a 2-D filter constructed for the unfiltered B-scan. The matched filter has averaged the weld scattering without affecting the spatial resolution (cf. scan lines at right column of Figure 7). It has to be remembered that the spatial processing of the 2-D matched filter results in processing along the horizontal lines in Figure 6. The lower BHs can easily be distinguished while the upper ones are masked by backscattering from the EB weld. The C-scans obtained from the information contained in the first, second, and third harmonic component, are presented in the consecutive lower rows of Figure 6, respectively. The profile lines corresponding to 62 mm are presented in the same way in Figure 7. Fom the analysis of Figures 6 and 7 it appears that the spatial resolution increases with the number of the harmonic component. This can be observed in the C- scans in Figure 6 and is also pronounced by the shape of peaks corresponding to the SDHs in Figure 7. The peaks corresponding to higher harmonic C-scans are narrower and the distance between them, measured in the bottom, increases. This can be observed both for the unprocessed and the filtered C-cans. The observed resolution improvement agrees with the theory since the higher harmonic components correspond to the shorter wavelengths. The C-scan obtained from the second harmonic, shown in Figure 6, seems to have optimal resolution and signal to noise ratio measured for the bottom-drilled holes. This can be seen even better in Figure 8, where a profile line crossing the upper holes at mm is shown. All three holes # 4, 5, and 6 can be distinguished at the processed C-scan extracted from the second harmonic component. It is very interesting that both the resolution and the signal to noise ratio is better for the second harmonic component than for the whole signal spectrum. Based on the above C-scans we can draw conclusion that all frequency bands, referred above as harmonic components, contain similar information about discontinuities in the inspected specimen. The matched filter performs nice smoothing of the C-scans and essentially improves detectability of the holes masked by backscattering from the EB weld. Generally, the resolution improves considerably for higher harmonic components. However the third harmonic component has a very low energy and is corrupted by discretization noise. It should be also noted that this component is affected by the strongest material attenuation.

2 3 4 5 6 7 8 2 3 4 5 6 7 8 2 4 6 8 2 2 4 6 8 2 2 3 4 5 6 7 8 2 3 4 5 6 7 8 2 4 6 8 2 2 4 6 8 2 2 3 4 5 6 7 8 2 3 4 5 6 7 8 2 4 6 8 2 2 4 6 8 2 2 3 4 5 6 7 8 2 3 4 5 6 7 8 2 4 6 8 2 2 4 6 8 2 Figure 6. C-scans of the right part of the specimen (bottom holes only). Left extracted from the unprocessed B-scan data. Right extracted from B-scans processed by 2-D matched filter. Sequential rows from the top show C-scans extracted from raw data, first, second, and third harmonic, respectively.

.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25.8.6.4.2 5 5 2 25 Figure 7. Profiles of line 62 mm in the C-scans of the whole specimen (the largest peaks correspond to the side-drilled holes). Left extracted from the unprocessed B-scan data. Right extracted from B-scans processed by 2-D matched filter. Sequential rows from the top show lines extracted from raw data, first, second, and third harmonic, respectively.

.8.8.6.4.6.4.2.2 5 5 5 5.8.8.6.4.6.4.2.2 5 5 5 5.8.6.4.2 5 5.8.6.4.2 5 5.8.6.4.2 5 5.8.6.4.2 5 5 Figure 8. Profiles of line mm in the C-scans of the right side of specimen (from the right peaks corresponding to the holes # 4, 5, and 6, respectively). Left extracted from the unprocessed B-scan data. Right extracted from B-scans processed by 2-D matched filter. Sequential rows from the top show lines extracted from raw data, first, second, and third harmonic, respectively.

Conclusions The presence of high harmonic components in ultrasonic signals obtained in pitch-catch inspection of copper specimens in immersion was experimentally confirmed. These components mainly originated from the higher resonant modes of transmitters that are under broadband excitations. Although energy contained in those frequency bands was considerably lower than that in the fundamental one, a useful information about the specimen discontinuities was detected in the high harmonic bands. The study reveals that the signals sent out by a transducer as transmitter (usually with limited bandwidth) contain much more information than when they are received in pulse-echo mode by the same transducer. In other words, much information (in higher harmonic components) contained in the signals sent out by a transmitter is lost in pulse-echo mode but can be captured in pitch-catch mode using another transducer that has broader frequency bandwidth. Therefore, the present study demonstrates an effective method of exploiting high frequency information contained in ultrasonic signals. C-scan images obtained by filtering A-scans in frequency domain (cutting out all frequencies except the interesting one) were compared to those obtained using a novel 2-D matched filter, performing filtering both temporal and spatial frequencies. The matched filter provided an efficient smoothing of C-scans without loss of their resolution. The filter was also capable of suppressing backscatter from the EB weld in the inspected copper specimen. Further research is performed aiming at developing algorithms for suppressing this backscatter using the information contained in the upper frequency bands. Acknowledgments This research was supported by the Swedish Nuclear Fuel and Waste Management Co. (SKB). The authors wish to thank Eider Martinez who assisted in the measurements of the copper specimens. References. F. Forsberg, D. A. Mertom, J. B. Liu, L. Needleman and B. B. Goldberg, "Clinical applications of ultrasonic contrast agents," Ultrasonics, vol. 36, pp. 695-7, 998. 2. F. Tranquart, N. Grenier, V. Eder, and L. Pourcelot, "Clinical use of ultrasound tissue harmonic imaging," Ultrasound in Med. & Biol., vol. 25, no.6, pp. 889-894, 999. 3. S. Makarov, and M. Ochman, Nonilnear and Thermoviscous Phenomena in Acoustics, Part I III, Acustica, vol. 82 (996), p. 579.

4. B. Ward, A. C. Baker, and V. F. Humphrey, "Nonlinear propagation applied to the improvement of resolution in diagnostic medical ultrasound," J. Acoust. Soc. Amer., vol., no., pp. 43-54, 997. 5. T. Stepinski, and P Wu, Ultrasonic Inspection of Nuclear Copper Canisters, SKB Projektrapport 97-6, August 997. 6. T. Stepinski, P. Wu, M.G. Gustafsson, and L. Ericsson, Ultrasonic array technique for the inspection of copper lined canisters for nuclear waste fuel, Proc of the 7th ECNDT, Copenhagen, May 998, pp. 377-383. 7. L. Ericsson, T. Stepinski, and M. Gustafsson, Suppressing Ultrasonic Grain Noise Using Non-linear Filtering Techniques, Presented at the First International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurized Components, Amsterdam, October 2-22, 998.