Ultrasonics 62 (2015) 112–125

Contents lists available at ScienceDirect

Ultrasonics journal homepage: www.elsevier.com/locate/ultras

Defect detection around rebars in concrete using focused ultrasound and reverse time migration Surendra Beniwal, Abhijit Ganguli ⇑ Department of Civil Engineering, Indian Institute of Technology, Delhi 110016, India

a r t i c l e

i n f o

Article history: Received 23 November 2014 Received in revised form 28 April 2015 Accepted 16 May 2015 Available online 23 May 2015 Keywords: Focused ultrasound Reverse time migration Concrete Experimental investigation

a b s t r a c t Experimental and numerical investigations have been performed to assess the feasibility of damage detection around rebars in concrete using focused ultrasound and a Reverse Time Migration (RTM) based subsurface imaging algorithm. Since concrete is heterogeneous, an unfocused ultrasonic field will be randomly scattered by the aggregates, thereby masking information about damage(s). A focused ultrasonic field, on the other hand, increases the possibility of detection of an anomaly due to enhanced amplitude of the incident field in the focal region. Further, the RTM based reconstruction using scattered focused field data is capable of creating clear images of the inspected region of interest. Since scattering of a focused field by a damaged rebar differs qualitatively from that of an undamaged rebar, distinct images of damaged and undamaged situations are obtained in the RTM generated images. This is demonstrated with both numerical and experimental investigations. The total scattered field, acquired on the surface of the concrete medium, is used as input for the RTM algorithm to generate the subsurface image that helps to identify the damage. The proposed technique, therefore, has some advantage since knowledge about the undamaged scenario for the concrete medium is not necessary to assess its integrity. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Ultrasonic inspection is used for non-destructive evaluation of concrete, typically, as a complementary tool to Ground Penetrating Radar (GPR) [1,2]. The Ground Penetrating Radar is mainly used to detect rebars and metallic tendon ducts in concrete bridge decks. Shifts in the frequency spectrum, relative changes in the amplitude and arrival times of identifiable portions of the scattered GPR waveform are considered as indicators of the rebar integrity and the extent of damage [3–5]. If the scattered field emanating from a defect is less contaminated by clutter noise, the performance of the associated reconstruction algorithm is expected to be good. In the presence of a dense distribution of rebars in a concrete structure, there is strong reflection of the electromagnetic waves which may mask weaker scattering signatures from damage affected regions. Ultrasonic inspection may be useful in such situations since the waves are scattered relatively less by rebars but are extremely sensitive towards the presence of interfaces having strong mismatch in impedances, like cracks and air-voids. In this context, ultrasound can be used as a complementary tool to the GPR modality for assessment of damage in concrete. Besides GPR ⇑ Corresponding author. Tel.: +91 11 2659 1221; fax: +91 11 2658 1117. E-mail addresses: [email protected] (S. Beniwal), abhijit.ganguli@ civil.iitd.ac.in (A. Ganguli). http://dx.doi.org/10.1016/j.ultras.2015.05.008 0041-624X/Ó 2015 Elsevier B.V. All rights reserved.

there are methods based on half-cell potential measurements [6] or usage of embedded piezoelectric sensors for detection of delaminations or material damage in rebars [7,8]. The half-cell potential technique requires access to the exposed end of a rebar which may not be always possible in various structures. Structural health monitoring with embedded sensors may require expensive instrumentation for carrying out a survey. In such situations, surface based ultrasonic inspection provides an inexpensive and convenient alternative which can be applied on existing in-service structures. One of the common algorithms for imaging with scattered ultrasonic waves is the Synthetic Aperture Focusing Technique (SAFT) [9], also known as Synthetic Aperture Radar (SAR) [10]. Successful imaging of defects, using SAFT in materials like composites and concrete through numerical simulation and experimental investigation has been reported in [11–14]. Recent developments in the field of synthetic aperture imaging of concrete involve the application of 2D arrays [15–18] which employ shear waves for interrogation of the medium. Condition monitoring of tendon ducts using shear wave transducer arrays has been reported in [16,17], wherein the SAFT generated C-scan images of the scattered field are used to identify the embedded artificial defects as steel plate or air-voids. It is also demonstrated that the phase information corresponding to reflected signals can be used to distinguish

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

between different types of embedded objects [17]. Commercially available array systems that use the SAFT imaging modality have also been applied for monitoring the integrity of concrete pavements [19–22]. However, SAFT may have limitations with regard to detection of vertical interfaces or objects with steeply inclined boundaries. Moreover there may be artifacts or ghost images in the SAFT generated image due to surface waves, multiple reflections and mode conversions of the wave field originating at interfaces, resulting in wrong conclusions about the scatterer location. Despite these apparent limitations, SAFT still remains the most adopted algorithm in the field of ultrasonic non-destructive evaluation due to faster processing and low data storage requirements. There exists another subsurface imaging algorithm, popular in the field of geophysics, which is known as the Reverse Time Migration (RTM) technique [23–26]. The physics of the RTM algorithm is based on the principle of Time Reversal of waves investigated extensively by Fink et al. [27–33] and others [34–37]. In physical time reversal, the medium is first excited with ultrasound and the field scattered by an internal object is recorded by a receiver array. The transmitters are then introduced at the receiver locations and the recorded fields are time reversed and fed into the transmitters. The generated ultrasonic field focuses back to the location of the scatterer. In case of multiple scatterers, the field gets focused back on all of the scatterers inside the medium. However, in the presence of a dominant scatterer, the energy is pre-dominantly focused back towards that location. The RTM algorithm, for the purpose of imaging of a medium, requires high end computational resources with large memory and data storage capacity. These requirements restrict RTM from being used on the fly during inspection operations. Rather, RTM works as a post processing tool for image reconstruction. These reasons have possibly restricted the application of RTM mostly to the oil exploration industry. With advancements in parallel processing of computational resources, the technique has potential for being used in other fields. Application of RTM for imaging of defects with the scattered Lamb wave field has been reported in [38]. The application of this technique towards detection of acoustic emission sources due to crack initiation in concrete media has been reported in [39,40]. A recent application of RTM in the field of non-destructive evaluation is reported in [41], where the authors use synthetic data to image vertical interfaces and circular objects embedded in polyamide, and in concrete media in the presence of small distributed scatterers. The RTM technique utilizing a full elastic wave propagation algorithm does not have issues with multiple reflections [41] and the mode conversions (compressional to shear wave conversions and vice versa) arising from interaction of elastic waves with interfaces. For this reason, it is expected that the RTM generated image will contain less artifacts in comparison to the SAFT image of the same medium. SAFT performs back propagation using the compressional or the shear wave velocity information, which creates artifacts in case where the scattered field contains mode converted wave field data. This aspect therefore presents an inherent advantage of the RTM algorithm vis-à-vis SAFT. In this paper, we present results from experimental and numerical investigations on application of the RTM algorithm for evaluation of integrity of rebars embedded in concrete. Subsurface defects in concrete are always undesirable since they affect the strength of concrete and pose a risk of further degradation under adverse loading and environmental conditions. Defects around rebars include partial or complete delamination near the rebar surface due to the imperfect casting condition or cracking under fire or earthquake loads. These defects around the rebars are a matter of utmost concern since they may lead to dramatic decrease in the strength of the member due to increased possibility of slip at the steel–concrete interface. Damage of the material around the rebar

113

due to the rusting of steel under chloride or carbonation ingress may be another cause of the above mentioned defect. The term damage in this paper exclusively refers to the defects discussed in the above context. We use focused ultrasound to assess the condition of rebars since a focused field has a greater probability of detection of a defect existing in the chosen focal region of interest (ROI). The focused field directs more energy towards the chosen ROI and partially circumvents the scattering and attenuation of the ultrasonic field due to interactions with the aggregates and the cement mortar matrix. There have been attempts at focusing of Lamb waves in thin plates towards particular directions using beam forming techniques to magnify the scattered field from damaged regions [42,43]. Beam steering of ultrasound for concrete inspection using phased array technology has been reported in [44–47]. In this paper, we propose a similar concept of focusing compressional waves towards a known rebar location inside the concrete to enhance the possibility of detection of a defect in the vicinity of the rebar. The total field that is received at a receiver array aperture is used as input for the elastic wave based RTM algorithm for generating the subsurface image. To the best of the authors’ knowledge, the feasibility of such an approach involving focused ultrasound and a full elastic wave based RTM algorithm has not yet been investigated in the context of non-destructive evaluation of rebar integrity in concrete. The paper is organized as follows. The next section presents results from numerical simulation of focused ultrasonic wave propagation in concrete using the Finite Difference in the Time Domain (FDTD) technique. The section first provides background on the FDTD simulation scheme for generation of synthetic scattered field data. It is then followed by numerical investigations on the subsurface reconstruction using the RTM based algorithm and the widely applied Zero-Lag Cross-Correlation Imaging Condition [51]. In the subsequent sections, we assess the performance of the proposed technique by presenting results of ultrasonic experiments on an actual concrete medium. The paper ends with a conclusion section and description of future work.

2. Numerical simulations 2.1. FDTD simulation geometry The goal of FDTD simulations is to assess the proposed methodology using synthetic ultrasonic scattered field data. The simulation medium (Fig. 1) consists of five steel rebars of either 16 or 20 mm diameter in cement concrete. The medium is 400 mm wide and 250 mm deep. The centers of the rebars in the top layer (#1 and #2) are at a depth of 130 mm and the centers of bottom layer of rebars (#3 through #5) are at a depth of 170 mm from top surface. The coarse and fine aggregates have been modeled as a random spatial distribution of elliptical aggregate particles embedded in the cement paste (Fig. 1). The medium also contains 1% volume fraction of sub millimeter sized air voids. The coarse aggregates have a volume fraction 42% and their nominal maximum size is 20 mm, which is the common practice in concrete construction. The rebar on the top left (#1) and the bottom center (#4) have a delamination (in the form of air wrap) around them of 0.5 mm size, which separates the steel from the surrounding concrete. The other rebars (#2, #3 and #5) are perfectly bonded to the concrete and therefore simulate healthy rebars. The diameters of the rebars in millimeters are shown in Fig. 1, including the delamination around two of the rebars. The material properties of the constituents inside the simulation medium are presented in Table 1. For carrying out the calculations, an FDTD model has been developed by the authors using the MATLAB platform. The

114

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

Fig. 1. Numerical concrete medium having fine and coarse aggregates and air voids along with five steel rebars in two layers. Two of the rebars (#1 and #4) are having small sized (0.5 mm) delamination around them (shown in dark blue color). The medium is investigated using ultrasonic array of transducers at top surface. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1 Material properties for the simulation model. Material

Comp. wave velocity V c (m/s)

Density q (kg/m3)

Poisson’s ratio

Cement paste Fine aggregate Coarse aggregate Air Steel

3500 4000 5600 330 5800

2200 2500 2700 1.2 7800

0.21 0.21 0.21 – 0.30

program utilizes a Staggered Grid Stress-Velocity based FDTD scheme for simulating elastic wave propagation in accordance to the developments presented in [53,54]. Stress-free boundary conditions are used at all the four edges of the concrete medium in the simulation. In order to excite the medium, vertical velocity input is introduced at specific grid points representing transducer positions on the top surface. The grid points are spaced at 0.4 mm interval and the simulation time step is 0.025 ls. The simulations were run for 180 ls. Within this simulation time, there may be multiple reflections (multipathing) associated with the ‘source to sidewall to rebar to receiver’ trajectory.

2.2. The input signal The input signal in our simulations is taken from actual experimental data on concrete. The reflection of the ultrasonic field from the bottom is windowed out (Fig. 2) and used as the input signal for the numerical investigations. The Fourier spectrum of the input signal shows the 6 dB bandwidth to be between 44 and 156 kHz with 100 kHz as the peak frequency. The wavelengths of the longitudinal waves range from 27 to 96 mm corresponding to the frequency range of 44–156 kHz, used for the insonification of the medium. More details about the input signal are presented in Section 3.1. Since the medium is insonified with a focused ultrasonic field, the array has to be sampled at half wavelength of the compressional wave corresponding to the highest frequency in the signal. At a compressional wave velocity of 4200 m/s, the array sampling comes to about 13 mm at 156 kHz (highest dominant frequency). The array element spacing is kept as 10 mm, which is less than half

m

Volume fraction (%) 12 45 42 1 –

of the required wavelength and thus prevents formation of sidelobes during focusing [48].

2.3. The focusing action For purposes of this study, it is assumed that an average compressional wave velocity is known for the background medium. We also assume that the dispersion and attenuation properties of the medium are small in the chosen frequency band around 100 kHz. Referring to Fig. 2, it is observed that the dominant frequency in the windowed out backwall reflection is 100 kHz. This is attained after attenuation of higher frequencies while traversing the medium twice (0.5 m in the medium). The same has been observed in [55]. The signal is therefore representative of the field propagating inside the concrete medium. It has also been discussed in [55] that ultrasonic frequencies between 50 and 100 kHz are typically suitable for inspection of concrete due to less attenuation. It has also been reported in [56] that attenuation in concrete is the lowest in the frequency band of the windowed out signal in the current experiment. The compressional wave velocity is also constant in the chosen bandwidth indicating low dispersion. The focusing action is achieved by time delayed excitation of the array elements, using the compressional wave velocity of 4200 m/s. If Ri and Rmax are respectively the distances of the focal point (which coincides with the center of a particular rebar) from the ith array element and the array element located furthest from the focus, then the time delay between the input excitations at the two elements is computed as jRi RV max j. This causes a focusing of the compressional wave energy at the desired focal point. The focusing delays are calculated assuming that all the rebars are in a healthy

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

115

Fig. 2. Input signal (on left) as bottom reflection signature from actual experiment on concrete medium, and the frequency spectrum of the input signal (on right) showing narrow band centered around 100 kHz.

condition. Therefore, if a defect exists in the form of a void around the rebar, the scattering action will be produced from the interface between the damaged region and the concrete instead of the rebar. Further, the nature of the scattered field will be different in comparison to the field reflected by a healthy rebar. The focusing action is repeated for all the rebars and the scattered field for each focusing experiment is acquired at the surface and stored for imaging purposes.

scatterer. Referring to Eq. (1), for simplicity, assuming only perturbation of the density parameter at the scatterer location, the modified equation for the scattered field is given by:

2.4. Reverse time migration

urj ðt; RÞ ¼ Gmodel ðt; R; rÞ  uscat ðt; rÞ i ji

The details associated with the RTM algorithm are presented in this section. Assuming that a source, receiver and a scatterer are respectively located at positions r0, r and r0 with respect to the origin in an elastic medium, the analytical expression for the scattered displacement field in the ith direction, uscat ðt; rÞ at r is given i by [49]:

€ j ðt; r0 ; r0 Þ  ½dkðdjk dim Þ uscat ðt; rÞ ¼ ðdqÞGij ðt; r; r0 Þ  u i 0

0

þ dlðdjl dkm þ djm dkl ÞGij;k ðt; r; r Þ  ul;m ðt; r ; r0 Þ

ð1Þ

In Eq. (1), the scattered field uscat ðt; rÞ is either synthetically geni erated data acquired using the FDTD scheme or real experimental data. The parameters dq, dk and dl are the perturbations in the density and the elasticity properties at the scatterer location relative to the background. The term uj(t, r0 , r0) is the total field in the jth direction at r0 due to the source located at r0, which is the sum of the incident field and the scattered field. However, following the practice in existing RTM literature, the Born Approximation approach is followed here and the incident field usj ðt; r0 ; r0 Þ at the scatterer location (i.e., the field that would have reached when there is no scatterer located at that point) is considered in the formulation, with the superscript s indicating the source. The Green’s Function relating the source field in the jth direction at r0 with the received field at r in the ith direction is given by Gij(t, r, r0 ). The overdots in Eq. (1) correspond to differentiation with respect to time, the subscript after comma represents a spatial derivative and the ⁄ symbol implies convolution. The letters i, j, k, l and m vary between 1 and 2 in a 2D formulation. The Reverse Time Migration (RTM) algorithm involves two independent FDTD simulations using a model of the interrogated medium with the estimated values of the shear and the compressional wave velocities and the density of the medium. The medium is assumed to be isotropic and homogeneous and without any

€ sj ðt; r0 ; r0 Þ uscat ðt; rÞ ¼ ðdqÞGij ðt; r; r0 Þ  u i

ð2Þ

In the RTM algorithm, the time reversed version of uscat ðt; rÞ is i propagated back into the medium containing no scatterer. The corresponding analytical expression for the resulting field in the jth direction at any position R is given by:

ð3Þ

The superscript r represents the time reversed receiver-field ðt; R; rÞ is the Green’s Function of the simulated medium and Gmodel ji (approximated by the FDTD scheme) that relates the source in the ith direction at r to the received field in the jth direction at any other point R. Substituting for uscat ðt; rÞ from Eq. (2), the field at i R = r0 is:

€ sj ðt; r0 ; r0 Þ urj ðt; r0 Þ ¼ ðdqÞGmodel ðt; r0 ; rÞ  Gij ðt; r; r0 Þ  u ji

ð4Þ

Assuming that Gmodel ðt; r0 ; rÞ reasonably approximates the actual ji medium Green’s Function, by invoking reciprocity [50], it can be shown that Eq. (4) represents a zero lag cross-correlation between the two Green’s Function. This results in a high value of the convolution at the scatterer location representing temporal and spatial focusing. This result directly emerges from the physics of physical time reversal. The location and the time instant of the focal spot can be detected by recording the field in the medium at every step of the simulation. However, the approach is to apply an Imaging condition for which a second FDTD simulation is performed with the source excitation being fed at various transmitter locations and propagated into the same simulation medium without scatterers. The resulting field at any point r in the simulation medium  0 Þ, values of which constitutes the forward source field, i.e., usj ðt; r;r at every point inside the medium are stored. The Imaging Condition adopted in this work is given by the zero lag cross-correlation between the time reversed receiver field and the forward source field [51,52]. This is applied at every point of the simulated medium and a 2D map of the values is expected to demonstrate the location of the scatterer. The equation of the Imaging Condition is given by:

IðrÞ ¼

Z

 0 Þdt urj ðt; rÞusj ðt; r;r

ð5Þ

116

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

By observing Eqs. (4) and (5), IðrÞ is expected to have high values around the scatterer location, i.e., r ¼ r0 . In this work, instead of the displacement field, the imaging condition is developed using the velocity fields at every pixel location. 2.5. Numerical wavefield propagation A numerical experiment was performed to generate synthetic ultrasonic data at the receiver locations. The elastic concrete medium in Fig. 1 was excited on the top surface at 37 array locations using the windowed-out signal of Fig. 2 as vertical velocity input and with the predefined time delay sequence so as to achieve a focused compressional wavefield at the rebar locations. The snapshots of the focused field (towards rebar #5) propagating in the

elastic concrete medium of Fig. 1 are shown in Fig. 3a through Fig. 3d at various instants during the numerical wave propagation in our 2D FDTD calculations. These snapshots depict vertical component of the velocity field. The focused compressional wave field first converges on the undamaged bottom right rebar (#5) [Fig. 3a and b], interacts with the rebar [shown in Fig. 3c] and is partially reflected and partly transmitted [Fig. 3d]. The reflected wave field spreads around the rebar and a non-uniform spatial distribution due to the heterogeneous nature of the concrete medium is observed. The reflected field further undergoes multiple scattering by the aggregates before reaching the top surface. It may also be noted that the nominal size of the coarse aggregate (20 mm) is comparable to the size of the rebars (16–20 mm). This distortion of the propagating field

Fig. 3. Snapshots of the ultrasonic field propagating through the medium shown in Fig. 1 at various instants of time. The wavefields (vertical velocity) are depicted in (a) through (d) when focusing is performed on the bottom right rebar (#5); and in (e and f) when the focusing is performed respectively on the bottom left rebar (#3) and the bottom central rebar (#4). The color scales are in velocity units. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

due to the aggregates makes the task of detection of a damaged rebar more difficult in concrete. Further, the focused field converging at the rebar #3 is shown in Fig. 3e and f, which reveals that there is a significant interaction of the focused field with the delaminated rebar #1 during its advancement towards the focus. 2.6. Results of simulation The synthetic ultrasonic field received at the top surface was time reversed and input as vertical velocity excitations at the receiver locations and propagated into a simulated homogeneous medium with a background velocity of 4200 m/s, using the FDTD algorithm. The results of RTM based calculations with zero lag cross-correlation between forward and reverse migration fields are presented as 2D images as shown in Fig. 4a–e. All the images are normalized relative to their absolute maximum values, therefore the color scales are unitless. The image of the rebar appears as a bright focal spot which is emphasized by the dotted circle. It is observed in Fig. 4b, c and e that the focal spots which represent three healthy rebars, are having similar polarity of the bright spots.

117

In case of the two rebars having damage around them, the polarity pattern of the spots are opposite, as observed in Fig. 4 a and d. The details of the bright spots are also indicated by the arrows in Fig. 4a–e. The rebars with no damage around them have a negative (blue color) valued patch at the center of the bright spots and the rebars with damage around them have a positive (red color) valued patch at the center of the bright spots. The polarity pattern of the bright spots at rebar locations in the RTM generated images could therefore differentiate between the damaged and the undamaged scenarios. 3. Experiment and results 3.1. Input signal in RTM forward migration The RTM based reconstruction of the concrete subsurface is performed using experimental data obtained from measurements on a concrete block (Fig. 5) which has a geometry similar to that in Fig. 1. The RTM based algorithm comprises of two forward simulations using the FDTD methodology. The first step involves the

Fig. 4. The RTM generated images of individual rebars using a focused incident field on each rebar. The rebars #1 and #4 in (a and d) respectively have similar polarity of the bright spots which is opposite to that of healthy rebars of (b, c and e). Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

118

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

Fig. 5. Experimental setup showing (sub millimeter sized) styrofoam wraps around rebars, the cast specimen showing aperture lines (dotted black lines) corresponding to three vertical planes and the signal transmission-reception with a pair of transducers.

forward propagation of the input ultrasonic field into a numerical homogeneous and elastic medium (also called the forward field). The next step involves the simulation of the propagation of the time reversed field, received at various transducer locations at the top of the experimental block, into the same numerical model of the medium (which is the time reversed field). The first simulation requires excitation of the medium with the same input field that is transmitted by the transducers. The exact nature of this field is generally unknown under experimental conditions. In such a situation we propose utilization of the bottom reflection signature, which is windowed out from the received backscattered field, as the input field (please refer to Section 2.2 and Fig. 2). The bottom reflection signature is generally quite prominent and therefore identifiable in the received signal, due to strong impedance mismatch at the concrete–air interface. In the present work, several A-scans (received signals) at random locations on the top surface of the concrete block are collected, by using a bistatic transducer configuration. The transmitter–receiver pair has a constant distance of 70 mm between their centers which ensures that the bottom reflection occurs at approximately the same time instant in each A-scan. The bottom reflection signature is windowed out from each A-scan and an averaged template input signal is generated, which is shown in Fig. 2. The center frequency of this signal is around 100 kHz. This signal is used as vertical velocity input at the transducer positions for calculating the forward field in the RTM based reconstruction process. 3.2. Experimental setup and procedure The elastic properties of the simulated elastic medium, used in the numerical investigations presented in Section 2 and in the RTM based reconstruction, are derived from the experimental investigations conducted on a concrete block sample (refer to Fig. 5). The compressional wave velocity of the medium is found to be 4200 m/s and the Poisson’s Ratio is taken as 0.21. The coarse aggregates in the concrete mix have a maximum size of 20 mm. The size of concrete medium is 0.40 m  0.25 m  0.30 m. There are five rebars embedded in two layers. Two of these rebars have simulated

damage in the form of less than 1 mm thick styrofoam wrap at some portions of their length as shown in Fig 5. Experimental simulation of finite debonds on rebars using PVC wraps [57] and delaminations in concrete using foam have already been previously reported in [58,59]. The concrete medium is investigated along three parallel apertures, perpendicular to the rebar direction. Referring to Fig. 5 (top right), the vertical plane through aperture 1 contains a damaged rebar at the center of the bottom layer and four healthy rebars. The second plane contains five rebars that are all healthy and the third plane contains damage around the rebar at the top left and the remaining four rebars are healthy. The sizes of the rebars vary between 16 and 20 mm. The field reflected by the rebars contains both compressional and the shear wave components, which are received by the experimental array elements in units of Volts. Both of these wave fields (in time reversed form) are fed as vertical velocity excitations (at source locations) into the simulated homogeneous elastic medium to generate the time reversed field at every point in the medium. Propagating the wave field in the elastic medium results in the utilization of both of these modes to form the RTM images. This is not the case when an acoustic medium is used in the RTM based reconstruction. In this work, only one transmitter and one receiver with center frequency of 250 kHz (manufactured by Olympus) have been used in the absence of a commercially available transducer array system. The source transducer was excited by a 200 V square wave pulse generated from a pulser-receiver circuit (Olympus-5077PR). The diameter of the contact surface is 36 mm. The signals were acquired and digitized (128 averages) in a Tektronix 2004C oscilloscope, at a sampling frequency of 10 MHz for a time window of 250 ls. Grease is used as a coupling agent. The specific array line on which the ultrasonic data is taken is coated with a moderately thick layer of grease. The receiver is slid on the grease from one array point to another without taking it off the surface. Before recording the signal, the position of the receiver is carefully checked and the transducer is pressed against the concrete surface, so that the signal level is maximum. The collected waveforms were also normalized relative to their maximum before being fed into

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

119

Fig. 6. The RTM generated images of individual rebars in the plane #1 of Fig. 5. The healthy rebars in (a, b, c and e) have a similar polarity (see indicating arrows) but different from the polarity of unhealthy rebar in (d). Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

the RTM algorithm. This effectively takes care of issues arising out of variation in coupling between the transducers and the concrete surface. The array elements are placed at 37 equidistant positions along the aperture line, on a grid marked on the concrete surface. The inter-element grid spacing is 10 mm. The source transducer is first placed at a grid point and signals (A-scans) are received by manually placing the receiver at each of the remaining grid points. The process is repeated until the transmitter covers all the grid points on the aperture. In this manner, the data is collected in a full matrix capture modality on a particular aperture. In order to generate a B-scan corresponding to scattering of the field focused towards a particular point, a delay and sum operation is performed on the full matrix captured data. This is a post processing operation but it generates the same scattered field B-scan that would otherwise be generated through physical focusing with a real transducer array. The delay and sum operation for focusing is similar to the one described in [36].

3.3. Experimental results The RTM generated images using experimental data obtained on the aperture #1 are shown in Fig. 6. All these RTM images including the ones following later are normalized relative to their absolute maximum values. Therefore the color scales are unitless. The vertical plane through the aperture comprises of one rebar at the bottom center with damage around it. Fig. 6a and b shows that the images of the two healthy rebars in the top layer have identical polarity pattern (positive–negative–positive). The circles in black denote the exact size and location of corresponding rebar in the figure. The two healthy rebars, located on the extreme left and right of the bottom layer, have the same polarity pattern, as observed in Fig. 6c and e. A secondary spot is observable slightly above the actual rebar spot in Fig. 6c. This may be due to the field partially reflected by the rebar on the top left. In contrast, the bright spot corresponding to the rebar at the bottom center [Fig. 6d] shows a polarity pattern (negative–positive–negative)

120

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

Fig. 7. The RTM generated images of individual rebars in the plane #2 of Fig. 5. All the rebars are healthy and have the same polarity of bright spots indicated by arrows. Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

which is opposite of the rest of the four healthy rebars. The change in the polarity pattern of the bright spots in the RTM generated image helps to identify a damaged rebar condition. It is also worth mentioning that the condition of the rebar at the bottom center is correctly detected with a focused field, despite it being partially obstructed by the top two rebars (refer to Fig. 5) while being interrogated by the transducer array. The RTM generated images for all of the five rebars in the aperture #2 [Fig. 7a–e], show the same polarity pattern as observed in the healthy rebars in Fig. 6. This is expected since all the rebars in this plane have no damage around them. The investigations are further carried out in aperture #3, where the rebar on the top left has a simulated damage around it and the rest of the rebars are healthy. The top right, the bottom central and the bottom right rebars [Fig. 8b, d and e] have a polarity pattern of positive–negative–positive, which is consistent with the no damage condition. The image of the rebar on the top left in Fig. 8a has a polarity pattern of low negative-high positive-high negative, which is of opposite polarity in comparison to the three healthy rebars in Fig. 8b, d and e.

An alternative way of identifying a damaged or an undamaged rebar could be by observing the polarity of a particular portion of the bright spots lying just near the outer periphery of the concerned rebar. Referring to Figs. 6–8 (apertures #1, #2 and #3 respectively), the polarity of the image near the outer periphery is negative (blue colored patch) for rebars with no damage and positive (red colored patch) for the rebars surrounded by damage. These patches are indicated by horizontal arrows in Figs. 6–8. It is observed that the image of the rebar in Fig. 8c, is not clear and has noise surrounding it. Also the negatively-polarized amplitude in Fig. 8d is not as strong as compared to a similar situation in Fig. 7d. The possible causes behind these are discussed in the next subsection. 3.4. Discussions The causes of the inconsistencies in the images in the previous subsection, are discussed here. Repeatability issues due to variation in coupling of the transducers with the concrete surface resulting in inconsistent data, was investigated first. The

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

121

Fig. 8. The RTM generated images of individual rebars in the plane #3 of Fig. 5. The rebar in (a) has damage around while the others in (b–e) are healthy. Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

acquisition experiment was repeated on aperture #3, having inconsistencies in some of the RTM images. However, no significant difference could be observed between the RTM generated images of all the rebars obtained from the two exercises. So the adopted technique was found to be repeatable. The cause of the noise in Fig. 8c is now explored here. All the B-scans presented subsequently are comprised of raw experimental A-scans (for clarity) and the images are normalized to their maximum absolute values. The scattered field B-scans due to focusing on bottom left and bottom right rebars in aperture #2 [Fig. 9a and b] and the bottom left rebar in aperture #3 [Fig. 9c] were inspected. A partial hyperbola was observed corresponding to the scattering of the focused field by each of the two rebars in aperture #2 [Fig. 9a and b]. The position of the hyperbola was ascertained from time of flight calculations after incorporating the correct delays. These B-scans are also less noisy in comparison to the one in Fig. 9c, where the reflection by the bottom left rebar is not clear. Fig. 9c also shows scattering by secondary objects (apart from the top left rebar), possibly voids or honeycombs around the rebar.

Also, since the top left rebar is delaminated, the focused field may be partially obstructed resulting in less energy reaching the targeted rebar. This was observable in the focused field patterns of Fig. 3e and f obtained from the simulation exercise. The field reflected by the delaminated rebar may have reached the receivers along with the field scattered by the focused rebar, thereby contributing additional noise in Fig. 8c. It is also be noted that such a noisy image is not observable in Fig. 7, where all the rebars are in a healthy condition. The case of lower negative amplitude of the image of the bottom central rebar in Fig. 8d in comparison to Fig. 7d is further investigated by observing the B-scans of the scattered field when focusing is performed on the central rebars in apertures #2 and #3. The hyperbola corresponding to the central rebar reflection in Fig. 10a is more pronounced than that in Fig. 10b (i.e., wider due to the signal reaching more number of receivers). This leads to a clearer RTM image with higher intensity in Fig. 7d. The difference in the aggregate distribution and acoustical properties between the two apertures is likely to create differing scattered field signature in the receiver data. Moreover, as explained earlier

122

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

Fig. 9. Experimental B-scans corresponding to the ultrasonic focusing on (a) the bottom left, (b) the bottom right and (c) the bottom left rebars in the apertures labeled. Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 10. Experimental B-scans showing scattered field corresponding to the ultrasonic focusing on the bottom central rebar in the apertures labeled. Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

also, less amount of energy reaches the bottom central rebar (aperture #3) in the presence of delaminated top left rebar which is strong reflector of the ultrasonic field. The possibility of identifying a damaged situation by inspecting the scattered B-scan data in time and frequency domain was also explored. Fig. 11 shows the receiver B-scans from the three apertures when the field is focused on the top left rebar. These rebars have clean RTM images that are able to clearly distinguish between undamaged [Figs. 6a and Fig. 7a] and damaged states [Fig. 8a].

Referring to Fig. 11a and b, the time domain B-scans are qualitatively different and it is difficult to ascertain the fact that both of the rebars are healthy. Comparing Fig. 11c with Fig. 11a and b, not much significant deviation could be observed that would otherwise indicate a damaged rebar condition. A similar conclusion can be reached by observing the frequency domain B-scans in Fig. 11d–f. These images are, in general, very noisy. Fig. 11d and e do not show much similarity between themselves despite being images of the healthy rebar in apertures #1 and #2. Even though

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

123

Fig. 11. The time and frequency domain receiver B-scans from the three apertures for a focused ultrasonic field incident on the top left rebar. Color scales are unitless. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 11f is different from the other two images, no definitive indicator such as significant shift in frequency spectrum due to the delaminated state of the rebar is observed. Therefore no general inference could be drawn from the B-scans in time and frequency domains. On the other hand, the RTM based exercise has been able to reconstruct images with different polarities in several instances (Figs. 6–8) and has consistently been able to distinguish between delaminated and healthy rebars. This has been possible in spite of the noisy and difficult to interpret nature of the B-scans discussed above. Finally, it may be noted that the data collected in the experiments contains the information with regard to 3D wave propagation. Despite this fact, the images presented here, which are based on 2D RTM simulations, have been able to provide a good distinction between bonded and debonded rebar conditions. This can be ascribed to the focusing action that confines a major part of the ultrasonic field to a small focal area in front of the aperture. In this situation, reflections coming from off-axis scatterers would be reduced and their effect on the RTM reconstruction would not be significant. In summary, experiments performed on concrete using focused ultrasound and an elastic wave based RTM algorithm demonstrated potential to detect damage around rebars in concrete. The

image of a rebar with damage around it has a polarity pattern opposite to that of a healthy rebar. A limitation in the proposed technique is also identified where the rebar image is not clearly formed and is surrounded by some clutter. However, it can be concluded that any departure from a consistent pattern observed in various rebar images can be indicative of a possible variation in the medium properties in the vicinity of the rebar. 4. Conclusion The conclusions based on the investigations reported in this paper are: 1. An elastic wave propagation based RTM algorithm, in combination with a focused ultrasonic field, is a promising tool for the detection of damage around steel rebars embedded in a concrete medium. The focused field directs more energy towards a particular ROI, thereby, enhancing the possibility of detection of a potential defect. The RTM based algorithm is able to generate the image of a rebar with sub-millimeter sized damage around it in a medium containing aggregates as large as 20 mm in size.

124

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

2. A delamination type damage around a rebar is found to manifest as a rebar spot with opposite polarity in comparison to a healthy rebar. This fact has been verified through numerical and experimental investigations. The identification is made possible since most of the images generated by the proposed technique are generally clear and do not contain too many artifacts. 3. The condition of the rebars embedded in a moderately dense mesh of surrounding rebars can be selectively examined by using a focused incident field and the RTM algorithm. In this work we have been able to access the health condition of 16 and 20 mm sized rebars placed in an actual concrete medium in two layers having horizontal separation of 50 mm and vertical separation of 40 mm. 4. The proposed technique has been found to be efficient despite the presence of factors that may typically accompany an experimental environment e.g., lack of uniform coupling between the transducers and the test surface and complexities in the test media. The experimental and numerical results have been consistent with each other with respect to imaging of damaged or undamaged rebars. Future work will be oriented towards field-testing of the technique proposed in this paper. This would include ultrasonic investigation with a focused field and the RTM algorithm in more challenging situations like structural concrete joints containing dense distribution of main reinforcement bars surrounded by shear stirrups. Acknowledgements The authors gratefully acknowledge the support from the Indian Institute of Technology Delhi, India for the experimental facility that made this work possible. Financial support for the computational resources obtained from the Department of Science and Technology, India, under the project no. RP-02638 is also gratefully acknowledged. References [1] C. Maierhofer, G. Zacher, C. Kohl, J. Wöstmann, Evaluation of radar and complementary echo methods for NDT of concrete elements, J. NonDestructive Eval. 27 (2008) 47–57. [2] C. Kohl, D. Streicher, Results of reconstructed and fused NDT-data measured in the laboratory and on-site at bridges, Cement Concr. Compos. 28 (2006) 402– 413. [3] J. Hugenschmidt, R. Loser, Detection of chlorides and moisture in concrete structures with ground penetrating radar, Mater. Struct. 41 (2008) 785–792. [4] W.-L. Lai, T. Kind, M. Stoppel, H. Wiggenhauser, Measurement of accelerated steel corrosion in concrete using ground-penetrating radar and a modified half-cell potential method, ASCE J. Infrastruct. Syst. 19 (2013) 205–220. [5] S.-X. Hong, W.-L. Lai, R. Helmerich, Monitoring accelerated corrosion in chloride contaminated concrete using ground penetrating radar, Proceedings of the 14th International Conference on Ground Penetrating Radar, Sanghai, China, 2012. [6] B. Pradhan, B. Bhattacharjee, Half-cell potential as an indicator of chlorideinduced rebar corrosion initiation in RC, J. Mater. Civ. Eng. 21 (10) (2009) 543– 552. [7] X.Q. Zhu, H. Hao, K.Q. Fan, Detection of delamination between steel bars and concrete using embedded piezoelectric actuators sensors, J. Civ. Struct. Health Monit. 3 (2013) 105–115. [8] F. Wu, F.-K. Chang, Debond detection using embedded piezoelectric elements in reinforced concrete structures – Part I: experiment, Struct. Health Monit. 5 (5) (2006) 5–15. [9] M. Schickert, Progress in ultrasonic imaging of concrete, Mater. Struct. 38 (2005) 807–815. [10] M. Cheney, A mathematical tutorial on synthetic aperture radar, SIAM Rev. 43 (2) (2001) 301–312. [11] M. Krause, F. Mielentz, B. Milman, W. Muller, V. Schmitz, H. Wiggenhauser, Ultrasonic imaging of concrete members using an array system, NDT&E Int. 34 (2001) 403–408. [12] A. Shlivinski, K.J. Langenberg, Defect imaging with elastic waves in inhomogeneous-anisotropic materials with composite geometries, Ultrasonics 46 (2007) 89–104.

[13] M. Spies, W. Jager, Synthetic aperture focusing for defect reconstruction in anisotropic media, Ultrasonics 41 (2003) 125–131. [14] A. Ganguli, C.M. Rappaport, D. Abramo, S. Wadia-Fascetti, Synthetic aperture imaging for flaw detection in a concrete medium, NDT&E Int. 45 (2012) 79–90. [15] V.N. Kozlov, A.A. Samokrutov, V.G. Shevaldykin, Ultrasonic equipment for evaluation of concrete structures based on transducers with dry point contact, Proceedings of the NDE Conference on Civil Engineering, St. Louis, MO, USA, 2006. [16] M. Krause, B. Milmann, F. Mielentz, D. Streicher, B. Redmer, K. Mayer, K.J. Langenberg, M. Schickert, Ultrasonic imaging methods for investigation of post-tensioned concrete structures: a study of interfaces at artificial grouting faults and its verification, J. Non-Destructive Eval. 27 (2008) 67–82. [17] M. Krause, K. Mayer, M. Friese, B. Milmann, F. Mielentz, G. Ballier, Progress in ultrasonic tendon duct imaging, Proceedings of the Non-Destructive Testing in Civil Engineering (NDTCE 09), Nantes, France, 2009. [18] A.O.D.L. Haza, A.A. Samokrutov, P.A. Samokrutov, Assessment of concrete structures using the Mira and Eyecon ultrasonic shear wave devices and the SAFT-C image reconstruction technique, Constr. Build. Mater. 38 (2013) 1276– 1291. [19] K. Hoegh, L. Khazanovich, Correlation analysis of 2D tomographic images for flaw detection in pavements, ASTM Int. J. Test. Eval. 40 (2) (2012). [20] K. Hoegh, L. Khazanovich, Ultrasonic tomography technique for evaluation of concrete pavements, J. Transport. Res. Board 2232 (2011) 85–94. [21] K. Hoegh, B. Worel, T. Yu, L. Khazanovich, Subsurface joint deterioration detection: MnROAD blind test comparison of ultrasound array technology with conventional nondestructive methods, J. Transport. Res. Board 2367 (2013) 3–12. [22] K. Hoegh, L. Khazanovich, Concrete pavement joint diagnostics using ultrasonic tomography, J. Transport. Res. Board 2305 (2012) 54–61. [23] W.-F. Chang, G.A. McMehan, Reverse-time migration of offset vertical seismic profiling data using the excitation time imaging condition, Geophysics 51 (1986) 67–84. [24] W.-F. Chang, G.A. McMechan, Elastic reverse-time migration, Geophysics 52 (10) (1987) 1365–1375. [25] R. Sun, G.A. McMehan, Pre-stack reverse-time migration for elastic waves with application to synthetic offset vertical seismic profiles, Proc. IEEE 74 (3) (1986) 457–465. [26] W.-F. Chang, G.A. McMehan, 3-D elastic prestack, reverse-time depth migration, Geophysics 49 (4) (1994) 597–609. [27] M. Fink, Time reversal of ultrasonic fields-Part I: basic principles, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 39 (5) (1992) 555–566. [28] F. Wu, J.-L. Thomas, M. Fink, Time reversal of ultrasonic fields-Part II: experimental results, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 39 (5) (1992) 567–578. [29] M. Tanter, J.-L. Thomas, M. Fink, Time reversal and the inverse filter, J. Acoust. Soc. Am. 108 (1) (2000) 223–234. [30] M. Tanter, J.-L. Thomas, M. Fink, Focusing and steering through absorbing and aberration layers: application to ultrasonic propagation through the skull, J. Acoust. Soc. Am. 103 (5) (1998) 2403–2410. [31] J.-L. Thomas, F. Wu, M. Fink, Time reversal focusing applied to lithotripsy, Ultrason. Imaging 18 (2) (1996) 106–121. [32] C. Prada, E. Kerbrat, D. Cassereau, M. Fink, Time reversal techniques in ultrasonic nondestructive testing of scattering media, Inverse Prob. 18 (6) (2002) 1761–1773. [33] C. Draeger, D. Cassereau, M. Fink, Theory of the time-reversal process in solid, J. Acoust. Soc. Am. 102 (3) (1997) 1289–1295. [34] M. Griffa, B.E. Anderson, R.A. Guyer, T.J. Ulrich, P.A. Anderson, Investigation of the robustness of time reversal acoustics in solid media through reconstruction of temporally symmetric sources, J. Phys. D, Appl. Phys. 41 (8) (2008) 085415. [35] A.M. Sutin, J.A. Tencate, P.A. Johnson, Single-channel time reversal in elastic solids, J. Acoust. Soc. Am. 116 (5) (2004) 2779–2784. [36] A. Ganguli, R.X. Gao, K. Liang, J. Jundt, Experimental investigation of ultrasound wave focusing in attenuative solids, IEEE Trans. Instrum. Meas. 59 (9) (2010) 2261–2271. [37] H.W. Park, Numerical simulation and investigation of the spatial focusing of time reversal A0 Lamb wave mode using circular piezoelectric transducers collocated on a rectangular plate, J. Sound Vib. 332 (2013) 2672–2687. [38] X. Lin, F.G. Yuan, Detection of multiple damages by prestack reverse-time migration, AIAA J. 39 (11) (2001) 2206–2215. [39] E.H. Saenger, Time reverse characterization of sources in heterogeneous media, NDT&E Int. 44 (2011) 751–759. [40] E.H. Saenger, G.K. Kocur, R. Jud, M. Torrilhon, Application of time reverse modeling on ultrasonic non-destructive testing of concrete, Appl. Math. Model. 35 (2011) 807–816. [41] S. Müller, E. Niederleithinger, M. Krause, T. Bohlen, Reverse time migration: a seismic imaging technique applied to ultrasonic data, Int. J. Geophys. (2012) 7, http://dx.doi.org/10.1155/2012/128465. Article ID 128465. [42] V. Giurgiutiu, J.J. Bao, Embedded-ultrasonics structural radar for in situ structural health monitoring of thin-wall structures, Struct. Health Monitor. 3 (2) (2004) 121–140. [43] P. Malinowski, T. Wandowski, I. Trendafilova, W. Ostachowicz, A phased arraybased method for damage detection and localization in thin plates, Struct. Health Monitor. 8 (1) (2009). [44] O. Paris, P. Bredif, O. Roy, J.M. Rambach, G. Nahas, Study of phased array techniques for cracks characterization in concrete structures, Proceedings of

S. Beniwal, A. Ganguli / Ultrasonics 62 (2015) 112–125

[45] [46]

[47]

[48] [49] [50] [51] [52]

the International Symposium on Non-Destructive Testing in Civil Engineering, 2003. F. Mielentz, Phased arrays for ultrasonic investigations in concrete components, J. Non-Destructive Eval. 27 (2008) 23–33. L. Azar, S.C. Wooh, Experimental characterization of ultrasonic phased arrays for the non-destructive evaluation of concrete structures, Mater. Eval. 57 (2) (1999) 134–140. O. Paris, C. Poidevin, J.M. Rambach, G. Nahas, Study of phased array techniques for concrete inspection, Proceedings of the 8th International Conference of the Slovenian Society for Non-Destructive Testing, 2005. pp. 11–17. G.S. Kino, Acoustic Waves: Devices, Imaging and Analog Signal Processing, Prentice-Hall Series, 1987. P. Mora, Nonlinear two-dimensional elastic inversion of multioffset seismic data, Geophysics 52 (9) (1987) 1211–1228. K. Aki, P.G. Richards, Quantitative Seismology, University Science Books, 2002. J.F. Claerbout, Towards a unified theory of reflector mapping, Geophysics 36 (3) (1971) 467–481. S. Chattopadhyay, G.A. McMehan, Imaging conditions for prestack reversetime migration, Geophysics 73 (3) (2008) S81–S89.

125

[53] H.W. Chang, C.J. Randall, Finite-difference time-domain modeling of elastic wave propagation in the cylindrical coordinate system, Proc. IEEE Int. Ultrason. Symp. (1988) 397–402. [54] J. Virieux, P-SC wave propagation in heterogeneous media: velocity-stress finite-difference method, Geophysics 51 (4) (1986) 889–901. [55] M. Schickert, M. Krause, W. Muller, Ultrasonic imaging of concrete elements using reconstruction by synthetic aperture imaging technique, ASCE J. Mater. Civ. Eng. 15 (3) (2003) 235–246. [56] T.P. Philippidis, D.G. Aggelis, Experimental study of wave dispersion and attenuation in concrete, Ultrasonics 43 (2005) 584–595. [57] F. Wu, F.-K. Chang, Debond detection using embedded piezoelectric elements in reinforced concrete structures – Part I: experiment, Struct. Health Monitor. 5 (1) (2006) 5–11. [58] J. Zhu, J.S. Popovics, Imaging concrete structures using air-coupled impactecho, ASCE J. Eng. Mech. 133 (6) (2007) 628–640. [59] C. Cheng, M. Sansalone, The impact-echo response of concrete plates containing delaminations: numerical, experimental and field studies, Mater. Struct. 26 (1993) 274–285.

Defect detection around rebars in concrete using focused ultrasound and reverse time migration.

Experimental and numerical investigations have been performed to assess the feasibility of damage detection around rebars in concrete using focused ul...
8MB Sizes 0 Downloads 8 Views