You are on page 1of 10
0D — OL SS ——. METECHNICAL PAPER @ pail Defect Imaging by Improved Ultrasonic Synthetic Aperture Focus Techniques ne Sternini*, Albert Y. Liang", and Francesco Lanza di Scaleat apsTRACT his paper presents a prototype system for the ultra- soricimaging of defects in rails that uses an improved ersion of the synthetic aperture focus technique (GAFD. Aral flaw imaging system is needed to quanti- tatvely assess the size and the shape of a defect in a manner that goes beyond the operator's judgment. ‘improvements in rail maintenance practice on several eyes can be expected if the rail defect verification step is truly quantitative, The SAFT was chosen over a traditional phased array imaging system due to the reduced hardware complexity, improved focus, and speed. The system being developed implements sarious steps to further improve the performance of the SAFT including: (a) compounding various wave propagation modes to reduce artifacts and increase aray gain; (b) including a wedge in the beamforming algorithm for optimum detection of transverse defects; (Qutiizing the parallel processing structure of the sgephical processing unit (GPU) architecture for increased imaging rates; and (¢) stitching together 2D sices to reconstruct 3D volumetric images. Results are shown on rail sections with simulated and natural transverse defects borrowed from the Federal Railroad Administration (FRA) Rail Defect Library. While the ‘ocus ofthe current research is on manual handheld faw imaging, several of the proposed algorithmic steps can be useful for in-motion rai inspections. Kerworos: ultrasonic imaging, synthetic aperture focus technique (SAFT), delay-and-sum (DAS), Compounding, transducer wedge, GPU processing, ary, rail inspection, nondestructive testing "TOE al ath Nonna Deptt Sra Figen ay exons 085054 iSeries fet yaar ent ge nnd Maee sr nay a Cnc ola Sons RSS a gator ana Introduction entation of defects in aod tracks res) OF and risk management. The cot of rail fiure can be catast and substantial efforts required to edace the sk Proper alloc tin of resources for al defect management requires not only detection, but als precise localization and characterization ? detection and have difficulty Current techniques are limited to i identifying thesia andloatingthe orientation ofthe ahead defects due to operators judgment. ransverse defects (TDs) an detail facturesare some ofthe most common types of defects. ‘TDs can originate as manufacturing discontinuities such as hydrogen flakes and brittle subsurface inclusions (Jeong 2003). Improved manufacturing techniques over the years have reduced discontinuities arising from manufacturing; however, TDs and detail factures initiated from surface discontinuities asa result of rolling contact fatigue (RCF) are widespread and continue to grow (Cannon etal-2003) Ultrasonic inspection is the most common technique used for defect detection in rails (Anon 1990; Lanza di Sealea 2007). Typical ultrasonic testing for rails uses A-scans, which cannot directly provide information on the defect size. For unbiased sng, ultrasonic arrays are needed ina beamforming technique. The most common beamforming method uses the phased array ultrasonic testing (PAUT) technique. Early imple- ‘mentations of phased arrays in ultrasonic wheels for rail inspec- tions were studied in the 1970s to better detect subsurface and surface breaking defects (Pagano 1979). More recent applica- tions of PAUT for al defect imaging were examined by researchersat the Transportation Technology Center in Pueblo, Colorado, for both manual verification (Garcia and Zhang 2006) and insmotion inspection (Witte and Poudel 2016). ‘The objective ofthis paper isto demonstrate the ability to age rail defect using the ultrasonic synthetic aperture focus technique (SAFT) (Flaherty et al, 1967; Schmerr 2016), which is variation of phased array imaging where each transducer can be treated individually and the focusing is done synthetically rather than physically. These features bring some key advantages in terms of image quality image speed, and harchare simplification, Experimental tests were performed to validate the algorithm for various defect environments, The ultrasonic imaging al developed utes the allowing main components: cea mage compounding to increase aray gain without adding to the physical aperture ofthe aray, wedge imaging through ray tacing for optimal identification of near vertical defects (such as TDs), ‘maim reflectivity normalization to highlight the presence of ial for ral safety phic, WWLY 2019 + MATERIALS EVALUATION 931 ————— ee Z ME TECHNICAL PAPER © ULTRASONICSYNTHETIC APERTURE FOCUS FOR RAILS ey rail defects, and real-time imaging using a graphical processing PU) based computation compared toa slower central processing unit (CPU) architecture Imaging results (both 2D and 3D) are presented from AIRE ral sections with simulated defects (flat-bottom holes drilled in the railhead to simulate ’'Ds) and from a 136RE rail sample with a naturally occurring TD. ‘Synthetic Aperture Focus Technique 'SAET for ultrasonic imaging has been around since the late 1960s and bas found numerous applications in the medical maging and structural health monitoring fields (Lanza di Scalea cet al. 2017; Sternini et al. 2017). typical SAFT approach uses an array of piezoelectric transducers that can act as both transmit- ters and receivers of ultrasonic waves. The image is constructed by extracting features from the received ultrasonic waveforms that are appropriately backpropagated in time to account for delay due to the spatial position of the transmitter, receiver, and focus point. ‘This approach of temporal backpropagation is also commonly known as the delay-and-sum (DAS) algorithm, Another key aspect of SAFT beamforming is the selection of weights atrib- uted to each collected waveform (Halland Michaels 2010). Basic SAFT algorithms use unity weights, orno weights applied, to the backpropagated and summed waveforms. Static apodization weights, such as hanning or kaiser-bessel windows, are also widely utilized to decrease artificial ringing effects in the image, but at the expense of smearing the image results and decreasing spatial resolution. Hanning or kaser-bessel windows are considered static ince they ae applied independently of the focus point. In this work, unity weights have been selected as apodization weights in the SAFT algorithm, even though weights based on the physics of the propagating waves could be applied to farther improve the imaging results (Lanza di Scalea etal 2017). ‘Theoretical Framework Consider an ultrasonic transducer array with M transmitters and N receivers. Let the spatial coordinates of each Co Focus P on Figure 1. Schematic ofthe imaging setup. 932 MATERIALS EVALUATION + JULY 2019 teansmiter i= 1, 3 Mbe (3); andthe spa of ech eceiverj= I, Nalsobe (1,9), asinyay ote A standard DAS algorithm constructs an iment EL summing, at each pixel P(xy), the amplitudes, of 59) by ignals, A, appropriately backpropagated, foreach st tion of transmitter and receiver). Inthe medio backpropagated DAS algorithm is written ay the 9 PSe = BE ay Ail yn} Aa where Wij tepresents the apodization weights previouly ‘The backpropagation time, jay >vesD0nds othe tay time of the wave from the transmitter ito the focus point P(xjy) and back to the receiver j, and is fi, + esa) + “Ls 4 OD Sixy~ where the denominators can be the longitudinal wave speed cz or the shear wave speed cs inthe solid. ‘The received signal, A, in Equation 1 can be computed directly from the raw waveforms from an enveloped version of the raw waveforms, of from the analytical signal representation of the raw waveforms. Inthe latter case, each waveform is decomposed into its in-phase and phase-quadratare compo nents through the hilbert transform. Equation 1 would then be applied to each of the hilbert-transformed components separately (Frazier and O'Brien 1998). The final image would then be constructed by computing the modulus ofthe two contributions at each pixel (x,y). This technique is wile to generate the results shown in this work. Compounding of Multiple Wave Modes ‘The possibility of combining different wave mode combina- tions comes from the realization that, in general, a defectina solid can reflect one or both ofa longitudinal L-wave anda shear S-wave, through either same mode reflection or mode conversion (Lanza di Scalea 2017). In the most general 3s since the excitation from a transducer with a small footprint (width of 0.6 mm for the sensors used in this work) can produce both L-waves and S-waves, thus acting as point source, there can exist up to four combinations of wave modes available for imaging in a bulk solid: L-wave transmitted and reflected (LL), L-wave transmitted and S-wave reflected (LS), S-wave transmitted and reflected ($8), and S-wave trans- ‘mitted and L-wave reflected (SL). For the case where a transducer wedge is used, only the L-wave is considered inside the wedge, and it can be refracted as both an L-wave and an S-wave in the test medium. Both ‘wave modes can be reflected by the discontinuity within the test medium both as same-mode reflections and as mode- converted reflections, and refract back to the L-mode in the

You might also like