Mapping Voids, Debonding, Delaminations, Moisture, and Other Defects Behind or Within Tunnel Linings (2013)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

47 a p p e N D I x G Introduction This appendix describes the progress of a particular non- destructive testing (NDT) technique known as acoustic sound- ing and outlines how this system will work within the framework of the SHRP 2 Renewal Project R06G. This system is in its final stages of development, and research thus far has shown it to be a promising technique capable of quickly determining the stage of tile debonding in tunnel lin- ings. Because the system remains under development, this appendix discusses how the system will be used in inspection procedures and provides an idea of what the end product will be. An evaluation of public tunnels and a series of test speci- mens will be conducted for this research and will be discussed in the final report. acoustic Sounding technique When debonding occurs on tiled surfaces, hammer sounding by ear or by microphone can readily differentiate bonded from debonded tile. Debonded areas have a characteristic lower- frequency pinging relative to fully bonded tiles. The goal here is to devise a less subjective method for inspectors to quickly and efficiently characterize the condition of tile bonding. Technical Needs In general, tile debonding can occur for two reasons: improper installation or external influences. Improper installation commonly includes the following: • Improper use of bonding agent (e.g., mixing ratios or the wrong type of agent); • Improper tile spacing; • Excessive open time; and • A low standard of workmanship (e.g., not “back buttering” the tile). External influences can include environmental conditions (e.g., thermal expansion) and/or excessive tunnel lining forces (e.g., damage from voids, cracks, delamination, or debonding). In either case, debonding of the tile does occur and can pose a danger to the public. This SHRP 2 project uses many NDT techniques to identify the onset of damage behind the tiled wall lining before debonding occurs and to quickly and efficiently identify regions that need immediate attention after debonding occurs. Research Approach The system under development is used with a laptop com- puter capable of recording audio signals and installed with a version of MATLAB (developed by MathWorks, http://www .mathworks.com/products/matlab/), along with an impact source (preferably a ball-peen hammer). As the operator lightly taps the center of each tile with the hammer, the lap- top’s internal microphone records the audio signal. MATLAB software performs a fast Fourier transform on the data set and uses pattern recognition techniques to monitor the fun- damental frequencies of flexural vibration for each individual tile. The modes of vibration frequencies in a voided tile can be predicted using acoustic theory for a rectangular plate with simply supported edges (Rossing and Fletcher 2003): 0.453 1 12 2 f c h m L n L mn L x y = +  + +        where cL = longitudinal wave speed, h = thickness of the tile, m and n = integers describing the current mode of excita- tion (m = n = 0 for the fundamental frequency of flexural vibration), and Lx and Ly = respective side lengths of the tile. Field Testing with Acoustic Sounding

48 The vibration frequencies increase as the voided sections of tile decrease (Liu et al. 2011). Therefore, it is theoretically possible to relate the fundamental frequency to the approxi- mate area of debonding. This technique can be incorporated into a program that assigns a color scale to the frequency spectrum of a tile wall under inspection. The research team envisions that the final program will be able to operate in two modes. The first is for near-real-time inspection. In this mode of operation, a threshold frequency from an expected frequency band repre- senting sound concrete is established and used to make a pass-fail decision, telling the operator whether a tile is most likely bonded or debonded. The second mode is intended for mapping a large region of tile, and the final result is a map of the tiles showing the degree of expected bond. As in the first mode, the operator selects a section of tile representing a fully bonded state for the program to determine the fundamental frequencies associated with bonded sections. The user then taps each tile in a predetermined order. For instance, the sec- tion might consist of an area 13 tiles high by 40 tiles wide. The program prompts the operator to select the layout desired, and after the operator taps each tile in the given order, the program displays a plot showing the frequency spectrum. Field application in the Washburn tunnel A rudimentary version of this technique was used for a proof- of-concept test in the Washburn Tunnel in Houston, Texas. The Washburn Tunnel is the only underwater vehicle tunnel in operation in Texas and was completed in 1950. It carries a federal road beneath the Houston Ship Channel, joining two Houston suburbs. The tunnel was constructed using the immersed tube method, with sections joined together in a prepared trench, 26 m (85 ft) below the water line. The entire inner wall is tiled with 110-mm by 110-mm (4.3-in. by 4.3-in.) ceramic tiles. Like many under- water tunnels with tiled walls, this one is experiencing debond- ing in various areas. Three sections of tile that contained debonded regions (as determined by an inspector performing hammer sounding by ear) were chosen. The regions, shown on the left side of Figure G.1, display the area under consideration outlined with blue painter’s tape. The debonded section (deter- mined by human ear) is indicated with a blue painter’s tape “x” on the debonded section. On the right side of Figure G.1, scans made via ultrasonic tomography (UST) are shown for each of the three regions. The depths of the C-scans (plan views) in Fig- ure G.1 range from 16 mm to 103 mm (0.63 in. to 4.1 in.). One of the areas investigated (Figure G.1, middle) was evaluated by using a rudimentary version of the acoustic sounding technique and is shown in Figure G.2. This example shows a strong cor- relation between hammer sounding by ear and the automated acoustic sounding technique. In Figure G.2, the bottom left plot depicts the tiles color coded in grayscale, with the higher frequencies (predicting a fully bonded state) as white and the lower frequencies (predict- ing a debonded state) as black. As previously discussed, the lower frequencies should theoretically correspond to larger voided areas behind the tile. The bottom right plot in Figure G.2 shows the output with a pass-fail algorithm denoting tiles that fall below the expected fully bonded state (red is the expected debonded state, and green is the expected fully bonded state). testing Criteria For the automated acoustic sounding device discussed here, no system is commercially available. The following testing criteria are given to estimate the usefulness in designing and implementing this technique. Precision, Accuracy, and Repeatability Precision and accuracy criteria will need to be determined on the basis of ground truth data (which were not available for the tunnel lining under inspection) on actual debonded tiles. Tech- nological difficulties prevented the research team from com- pleting a system for validation on test specimens within the time constraints of this project. The system previously described in Field Application in the Washburn Tunnel is only compared with hammer sounding (by ear) and UST, which should not be used in place of ground truth data. Because the detection of debonded tiles depends on the fre- quency band chosen to represent bonded tile, the threshold value for a pass-fail decision will vary. The researchers recom- mend rating the failures (debonded tiles) by color-coded signals based on the proximity of the fundamental frequency response to the chosen threshold. After this is experimentally tried, the precision and accuracy of the technique can be estimated. Repeatability will depend on the precise location of impact. A great deal of variance is possible depending on how far the point of impact is from the center of the tile. Calibration Procedures Calibration will have to be made on a section of tile evaluated by other NDT devices or otherwise assured to be sound. The researchers recommend determining a band from several sample locations of bonded tiles. After this frequency band is determined, it can be used as a threshold value for determin- ing debonded tiles. Testing Procedures The research team envisions that a fully developed automated sounding method will be able to operate in two modes. The

49 Figure G.1. Debonded regions of tile ( left ) paired with the associated UST C-scans (right). first is for near-real-time inspection. In this mode of opera- tion, the threshold frequency from an expected frequency band representing bonded tile; is established and used to make a pass-fail decision, telling the operator whether the tile is most likely bonded or debonded. The second mode is intended for mapping a large region of tile, and the final result is a map of the tiles showing the degree of expected bond. This pass-fail decision will be based on how close the fundamental frequency of the tile is to the threshold value. As in the first mode of operation, the operator will select a section of tile representing a fully bonded state for the pro- gram to determine the fundamental frequencies associated with bonded sections. The user will then tap each tile in a predetermined order.

50 The research team also recommends developing an appli- cation for a smart phone that would signal whether a tile is likely debonded or bonded. A threshold value could be cho- sen to represent bonded tile, and significant deviations from that threshold would result in a pass (green) or fail (red) screen. Cost The research team attempted to construct a viable proto- type, but it is in progress and not yet ready for field appli- cation. A final and proven technique is expected to be inexpensive. Limitations The limitations of this device are as follows: • Battery power. Any remote device will rely on battery- powered operation for long periods of analysis. • Consistent impact location. Repeatability of impact plays a huge role in precision and accuracy. The operator’s point of impact should not deviate significantly from the center of the tile. • Microphone quality. The research team is not certain at this time whether the microphone quality from a typical smart phone or laptop computer is sensitive enough to 2 4 6 8 10 12 1500 2000 2500 3000 2 4 6 8 10 12 Figure G.2. Debonded regions of tile (top) paired with the acoustic sounding results (bottom).

51 distinguish fundamental frequencies from the ambient noises present in a tunnel. The proof-of-concept method presented above used recordings from a smart phone video recorder and then processed the data with MATLAB code. When used in the field, the laptop computer had trouble recording usable data. Data Analysis and Interpretation The purpose of the automated acoustic sounding technique is to remove the subjective component of the process by allowing the software to make a pass-fail decision. Further analysis and decision making would involve other NDT techniques. Equipment and Systems Integration Requirements The research team recommends that devices use MATLAB soft- ware on any platform compatible with the version purchased. Conclusion This automated sounding technique is still under develop- ment. Many factors influence the peak frequencies observed in the frequency spectrum from a single tile tap, including the size of the void, whether or not the hammer tap was directly in the center of the tile, and multiple-mode interfer- ence. Preliminary results indicate that this technique, although basic in its approach, will offer the tunnel inspector a quick, efficient, inexpensive, and objective technique that provides sufficient information for repair procedures or fur- ther investigation. references Liu, S., F. Tong, B. Luk, and K. Liu. 2011. Fuzzy Pattern Recognition of Impact Acoustic Signals for Nondestructive Evaluation. Sensors and Actuators A: Physical, Vol. 167, No. 2, pp. 588–593. Rossing, T., and N. Fletcher. 2003. Principles of Vibration and Sound, 2nd ed. Springer, New York.