Automated Acoustic Scanning of Concrete Bridge Decks

Automated Acoustic Scanning of Concrete Bridge Decks HONGBIN SUN, SUNYUN HAM and JINYING ZHU

ABSTRACT Chain drag testing is commonly used in current practice for bridge deck evaluation due to its low cost and ease of use. However, this method is subjective, and highly depends on the experience of the operators. An automated acoustic scanning system is developed to detect delaminations in concrete bridge decks. The system consists of an array of chains or impactors, a noncontact MEMS microphone sensor array, multichannel data acquisition device, RTK GPS positioning system, and signal processing schemes. Acoustic signal collected by the microphone is processed by short time Fourier transform (STFT). STFT amplitudes in the frequency range from 0.5 kHz to 5 kHz is summed up to form one dimensional data set vs. scanning time for each channel, and then data sets for each channel are stacked together to generate a two-dimensional map. A delamination identification algorithm is then developed to identify the center positons, dimensions of delaminations and recognize false positive delaminations. This system is validated on a 300 feet long bridge near Lincoln NE. INTRODUCTION 1

Concrete bridge deck deterioration is a major concern to highway agencies. Accurate and efficient evaluation of bridge deck will help highway agencies make proper maintenance decisions and reduce repair cost. The Impact-echo (IE) is a proven NDE method that had been used in the last 25 years for concrete components evaluation [1], [2]. The IE test not only identifies delaminations but also gives depth information of delamination. However, the IE test requires contact between sensor/source and concrete surface, which makes it time-consuming and labor intensive. Zhu and Popovics [3] proposed air-coupled IE test by using a microphone to replace the contact sensor. Other researchers have attempted to develop automated impact source and air-coupled sensing to increase the test efficiency [4]–[7]. However, these systems need complicated electrical and mechanical control for consistent impacts and the contact impact source remains a challenge for rapid scanning. Chain drag is another commonly used method Hongbin Sun, Jinying Zhu, Dept. of Civil Engineering, University of Nebraska-Lincoln, 1110 S 67th St. Omaha, NE 68182 Sunyun Ham, Dept. of Civil Engineering, University of Texas at Arlington, 416 Yates St., 433 Nedderman Hall, Arlington, TX 76019

2655

for delamination evaluation due to its low cost and ease of use. A major drawback of chain drag is that it relies on subjective interpretation of the inspector. The advantage of chain drag is that the steel chain used is a simple acoustic source for continuous excitation. Costley et al. [8], [9] developed an automated chain dragging system that includes dragging chains, a microphone, and signal acquisition and processing components. If working properly, the automated system will be more efficient and consistent than the manual chain drag test. However, when the authors investigated the chain drag acoustic evaluation system, we found that the chains generated broadband noises that may lead to misjudgment in some cases. In this study, a multi-channel acoustic scanning system is presented which integrates continuous impact excitation with the newly developed ball chains, MEMs microphones, and a high accuracy RTK GPS positioning system. The system enables automated imaging of structures for rapid mapping delaminations in bridge decks with high spatial resolution. A major improvement from the chain drag test is the ball-chain impactors, which allow impact-type acoustic excitations in a continual manner. A post processing algorithm is developed to identify and characterize delaminations. ACOUSTIC TESTING SYSTEM Acoustic sensor and data acquisition The acoustic scanning system includes excitation sources (chains), acoustic sensors (microphones), and positioning devices. Low cost MEMs (Micro-electro-mechanical systems) microphones (Adafruit SPW2430) with a frequency range of 100 Hz to 10 kHz were used as acoustic sensors. Microphones and chains were installed on a scanning frame. The testing frame includes 8 channels, and each channel includes one MEMs microphone and one chain installed below each microphone. The channel spacing varies between 0.1 m to 0.15 m (4 to 6 inches) which determines the lateral spatial resolution of scanning. Acoustic signals received by the microphone array are digitalized by an oscilloscope (PicoScope 4824) with a sampling rate of 100 kHz and transferred to a computer using a LabVIEW program. The positioning data was also synchronized and recorded. A laser distance sensor was used to record positions for the laboratory test, and a GPS was used for field testing on bridge decks. A schematic diagram of the acoustic testing system is shown in Figure 1.

Figure 1. Schematic diagram of the automated scanning system.

2656

Excitation with steel link-chains and ball-chains To differentiate the conventional chains used in the chain drag test and the newly developed ball- chain impactor, we denote the conventional chain as “link-chain”, and the new source as “ball-chain”. Figure 2 shows two different types of steel link-chains (C1 and C2) and the newly developed ball-chain. Both steel link-chains C1 and C2 have a 6.35 mm (1/4 inch) diameter, while C2 has galvanized finish. The ball-chain is made of 12.7 mm (1/2-inch) and 15.9 mm (5/8-inch) diameter brass balls with a spacing of 2 cm. Signals of 1.0 second duration were collected when each chain was dragged on a solid concrete surface. Each signal was processed by short time Fourier transform (STFT) and the spectrogram images are shown in Figure 3.

Figure 2. Two steel link-chains and a ball-chain.

Figure 3. STFT spectrograms (0.5-20 kHz) of (a) C1 on solid; (b) C2 on solid; (c) Ball-chain on solid. (all images have the same color scale).

Both steel link-chains generated high level noises around 15 kHz (see Figure 3a and Figure 3b) when dragging on solid area. Experimental study showed that the 15 kHz noises were caused by clapping between chain links. In addition, the link-chains also caused random broadband noises due to friction with rough concrete surface. The ballchain gave clean signals when tested on solid concrete surface, which is desired for

2657

concrete delamination detection [see Figure 3c]. According to a semi-analytical analysis for square concrete delaminations by Kee and Gucunski [5], for delaminations with a depth of 20 mm~80 mm and width of 0.2 m ~ 1 m, the resonance frequency of flexural vibration modes ranges from 0.5 to 5 kHz. Therefore, the major criterion for selecting proper excitation sources is to ensure low noise level in 0.5~5 kHz frequency range in solid concrete regions. In this frequency range, chains C1 and C2 show higher noise levels than the ball-chain. Other link chains with different sizes were also investigated, and they all showed various levels of noises in this frequency range. These noises will overlap with the acoustic signals of delamination responses and may give false positive indication of delamination. The ball-chain was used for field testing on a concrete bridge deck. FIELD TEST ON CONCRETE BRIDGE DECK Scanning system and results A scanning cart was designed with 8 channels of microphones for field testing on bridges, as shown in Figure 4. A final scanning image with positioning information was generated when the scanning was finished. A Real Time Kinematic (RTK) GPS (Piksi GPS, Swift Navigation, Inc.) provided real time position during the scanning. The field testing was conducted on a bridge at Warlick Boulevard in south Lincoln Nebraska. Only the southwest bound lane was scanned as indicated by the yellow area (90 m long) in Figure 5. A coordinate system was built in the figure while x-axis represents the bridge length direction and y-axis in the transverse direction. The GPS base was placed at the origin and the (x, y) coordinates were measured using the relative positions between the GPS rover and base units. Since the scanning frame is 1.2 m (4 feet) wide, three runs were conducted to cover the 3.6 m (12 feet) wide lane. Each run took about 70 seconds. Each time domain signal was first processed by STFT to generate a STFT spectrogram. The STFT amplitudes in the frequency range from 0.5 kHz to 5 kHz were summed up to form one dimensional data set vs. scanning time for each channel, and then the data sets from all channels were stacked together to generate a two-dimensional map, with two axes representing dimensions in longitudinal (scanning) and transverse directions. This process was repeated for each run and the images of three runs were then combined to form a map for the full lane.

Figure 4. Acoustic scanning cart with RTK GPS.

2658

Figure 5. Bridge deck for field testing [10].

Figure 6. Scanning images for bridge deck: (a) 40-60 m, (b) 60-80 m.

Figure 6 shows a 40-m long segment (40-80 m) of the scanned bridge deck. In order to clearly show locations of delamination, the image is split into two parts (40-60 m and 60-80 m). Before the automated acoustic scanning, manual chain drag test was performed by the Nebraska Department of Road (NDOR) and five delaminations (#1, #2, #3, #5, #6) were identified and marked on the bridge deck. They are shown in the scanning image as black rectangles. The acoustic scanning confirmed four delaminations (#1, #2, #3, #5). The positions of these four delaminations match well with the manual chain-drag results. The acoustic scanning also indicates a small piece of asphalt patch (confirmed using the GoPro video) in Figure 6(a) and a new delamination #4 in Figure 6(b), but did not detect delamination #6. It is found that delamination #6 is very narrow, which could be missed because no ball impacted on

2659

that location. This problem can be addressed by using ball-chains with multiple parallel balls to improve longitudinal resolution. Traffic noises (circled areas) can be observed in the images, which does not affect identification of the delaminations. The amplitude of traffic noises is much lower than the amplitude of delamination responses. Delamination identification algorithm and results A delamination identification algorithm was developed to identify the number and size of the delaminations and recognize false positive delaminations. Two-dimensional matrix 𝑃 represents the pixel amplitudes of the final scanning image (Figure 6). The algorithm can be described as the following steps: 1) Matrix 𝑃 is converted to a logical matrix using threshold 𝑃" : 0 𝑖𝑓 𝑃 𝑖, 𝑗 < 𝑃" 𝑃 𝑖, 𝑗 = 𝑃" = 400 (1) 1 𝑖𝑓 𝑃 𝑖, 𝑗 ≥ 𝑃" 2) Moore-Neighbor tracing algorithm modified by Jacob's stopping criteria is used to trace the boundary of each possible delamination; 3) A rectangular boundary is acquired based on the same length and width as the real boundary of each delamination. 4) This rectangle is treated as delamination. Figure 7 shows the boundary tracing results of Delamination #1 compared with the manual chain drag results. Real boundary was acquired using the tracing algorithm and plotted in black dashed curve. The rectangular boundary (red rectangle) has the same length (transverse direction) and width (longitudinal direction) as the real boundary. The rectangle’s width is smaller than the manual chain drag results, while the length is similar to the manual results. All possible delaminations were traced and the center position, width and length of each delamination are summarized in Table I including whether confirmed by manual chain drag results. We checked time domain signals for all delaminations and found some false positive delaminations which are generated by broadband noise. These may be caused by rough surface or small rocks. The remaining delaminations are confirmed by the manual results.

Figure 7. Delamination boundary tracing of Delamination #1.

2660

TABLE I. DELAMINATION IDENTIFICATION RESULTS X(m)

Y(m)

Width (m)

Length (m)

Manual results confirmed?

25.61

2.39

0.07

0.20

False positive

45.71

2.23

0.05

0.10

Asphalt patch

1

52.10

1.36

0.18

0.71

Yes

2

53.90

3.12

0.14

0.34

2

53.90

2.78

0.04

0.10

3

57.77

1.56

0.13

0.68

3

58.17

1.50

0.23

0.68

4

70.83

2.16

0.12

0.38

No

5

73.12

1.96

0.13

0.63

Yes

76.50

1.13

0.07

0.22

False positive

82.52

1.11

0.05

0.23

False positive

90.29

0.85

0.06

0.20

False positive

95.66

1.12

0.01

0.04

False positive

95.69

1.10

0.02

0.07

False positive

Del. #

Yes Yes

CONCLUSIONS In this study, a ball-chain source was developed as a new scanning acoustic source for bridge deck evaluation. The ball-chain was first tested on a solid concrete surface and showed higher signal-noise ratio, sensitivity and repeatability than the conventional steel link-chains. Finally, an automated acoustic system was developed and validated on a concrete bridge deck. Based on the results in this paper, the following conclusions are drawn: 1. The developed ball-chain has higher S/N than the traditional steel link-chain when dragging on concrete surface due to ball-chain’s periodical jumping and impact. 2. MEMs microphone is an effective acoustic sensor for delamination detection due to its high sensitivity, low cost, ease of use. 3. RTK GPS provides high precision positioning data (1 cm accuracy), which can be synchronized with acoustic signals to provide real time test positions. 4. The results of automated scanning on a bridge deck agree well with manual chain drag test. The automated system may miss small delaminations if no ball impacts on the delamination. This problem can be addressed by decreasing test speed or increase the number of ball impactors to improve spatial resolution in scanning direction. 5. A delamination identification algorithm was developed to locate delaminations and estimate the delamination size.

2661

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

C. Cheng and M. Sansalone, 1993. “The impact-echo response of concrete plates containing delaminations: numerical, experimental and field studies,” Mater. Struct., 26(5): 274–285. M. J. Sansalone and W. B. Streett, Impact-echo. Nondestructive evaluation of concrete and masonry. Bullbrier Press, Jersey Shore, PA. J. Zhu and J. S. Popovics, 2007. “Imaging concrete structures using air-coupled impact-echo,” J. Eng. Mech., 133(6): 628–640. N. Gucunski, M. Yan, Z. Wang, T. Fang, and A. Maher, 2012. “Rapid bridge deck condition assessment using three-dimensional visualization of impact echo data,” J. Infrastruct. Syst., 18(1): 12–24. S. Kee and N. Gucunski, 2016. “Interpretation of flexural vibration modes from impact-echo testing,” J. Infrastruct. Syst., 22(3): 1–10. G. Zhang, R. S. Harichandran, and P. Ramuhalli, 2012. “An automatic impact-based delamination detection system for concrete bridge decks,” NDT E Int., 45(1): 120–127. B. A. Mazzeo, J. Larsen, J. McElderry, and W. S. Guthrie, 2016. “Rapid multichannel impactecho scanning of concrete bridge decks from a continuously moving platform,” in Review of Progress in Quantitative Nondestructive Evaluation 2016 (QNDE 2016). R. D. Costley, M. E. Henderson, and G. N. Dion, “Acoustic inspection of structures,” US Patent No. 6,581,466, 2003. R. D. Costley and G. M. Boudreaux, 2003, “Finding delaminations in concrete bridge decks,” in 146th ASA Meeting, Austin, TX. “Google Map.” [Online]. Available: https://www.google.com/maps/@40.7505475,96.7122382,672a,20y,1.4t/data=!3m1!1e3?hl=en&authuser=0.

2662