NOVEL CABLE SENSOR FOR CRACK DETECTION OF RC STRUCTURES

Proceedings of the Structural Materials Technology (SMT): NDE/NDT for Highway and Bridges 24 14-17 24, Buffalo, New York NOVEL CABLE SENSOR FOR CRACK DETECTION OF RC STRUCTURES Genda Chen, Ph.D., P.E. Associate Professor of Civil Engineering, University of Missouri-Rolla 328 Butler-Carlton Hall, 187 Miner Circle, Rolla, MO 6541, Tel: (573) 341-4462, Fax: (573) 341-4729, Email: gchen@umr.edu In this paper, the development and validation of a fundamentally new, topology-based cable sensor design concept is summarized for crack detection in reinforced concrete (RC) structures. The sensitivity, spatial resolution, and signal loss of sensors are investigated both numerically and experimentally. Two sensors were fabricated and validated with small- and large-scale laboratory tests under different loads. Both were proven sensitive to crack of various sizes from visually undetectable to excessive, giving the location and severity of damage simultaneously. One sensor has been installed on a three-span bridge for its long-term monitoring. It has a unique memory feature, capable of recording damage that has occurred during a recent event. INTRODUCTION Crack in RC members may lead to structural degradation due to reinforcement corrosion associated with the water leakage and chloride invasion, particularly in maritime facilities. The maximum crack width that has structural implications is approximately.33 mm for interior exposure or.41 mm for exterior exposure. In the case of nuclear reactors or other waste solid treatment plants, however, this limit would be much smaller in order to prevent any leakage of hazardous materials. On the other hand, a cracked structure can still support significantly more loads before it becomes unstable. Therefore, the crack width of engineering significance covers a wide range, making it challenging to detect cracks with embedded sensors. Lin et al [1] designed a strain sensor with coaxial cables and claimed, based on their calibration tests, that the sensor was more sensitive than commercial products under applied loads. However, both were designed based on the change in geometry, which can only achieve incremental improvement in sensitivity unless a special design of cable cross section is introduced [2]. More importantly, sensors were subjected to strain effects when embedded in concrete; previous studies thus need to be validated for practical applications. MEASUREMENT PRINCIPLE AND NEW CRACK SENSOR Electrical Time Domain Reflectometry (ETDR) ETDR is a remote sensing technology based on the propagation of electromagnetic waves in an electrical cable or a transmission line, which functions both as a signal carrier and a sensor. It uses a digital sampling oscilloscope with an ETDR sampling head. The sampling instrument launches a series of low-amplitude and fast-rising step pulses onto the transmission line and samples the reflected signal caused by an electrical property change, i.e., a discontinuity in Fig. 1, along the cable. The arrival time of the reflected signal represents the distance from the point of monitoring to the discontinuity while the intensity of the signal represents the degree of the discontinuity. A cable sensor embedded in concrete can thus detect both the 1

location and width of a crack. The amplitude of the reflected wave, V, normalized to that, V, of the incident voltage wave is known as the reflection coefficient, Γ + V Γ = + V Z Z = Z + Z, (1) in which Z and Z are the characteristic impedances of the coaxial cable sensor and the measurement system, respectively. Incident voltage step Reflected voltage step Digital sampling oscilloscope with a SD-24 TDR sampling head Coaxial cable Distance between points of monitoring and discontinuity Fig. 1 Principle of Cable Sensor Measurement The Concept of Crack Sensors Based on Topology Change To significantly improve the sensitivity of commercial cables for structural health monitoring, two innovative designs of cable sensors were developed based on the change in topology, or electrical structure, of coaxial cables. They are schematically shown in Fig. 2. The first sensor, Prototype I, was designed with a dielectric rubber tube around which a copper tape with adhesives is spirally wrapped as the outer conductor of the cable to facilitate the change of the electrical structure under strain conditions [2]. The second sensor, Prototype II, was designed with a Teflon dielectric layer and a steel spiral that can slide along the Teflon surface under strain conditions [3]. A key factor in their fabrications is to ensure that any two adjacent spirals are electrically in contact, but separate easily under loading. Solder cover Commercial stainless steel spiral Teflon ( ψ = 5.6mm) Silver plated, branded copper wires ( ψ = 4.6mm ) (a) Prototype I (b) Prototype II Fig. 2 Schematic View of Two Crack Sensors In Fig.2(b), the parameter ψ represents the outer diameter of the various layers. The characteristic impedance of the Prototype II is approximately 19 ohms, and the width of the spiral is 4.3 mm. Although the exact dimensions of the cable were finalized from the available materials in the market, the process of leading to the final selection was based on the optimal design of Sensor-III from the numerical simulations in SENSITIVITY ANALYSIS. The presence of a partial or complete separation between adjacent spirals, which act as the outer conductor of the cable sensor, will force the return current on 2

Proceedings of the Structural Materials Technology (SMT): NDE/NDT for Highway and Bridges 24 14-17 24, Buffalo, New York the transmission line outer shield to change its flow path as shown in Fig. 3. This effect introduces an added inductance, L-gap, according to the transmission line theory [4]. A portion of the incident wave will therefore be reflected when it encounters this discontinuity. The equivalent transmission line model of a coaxial cable sensor with a separation between spirals is shown in Fig. 4. The relation between the reflection coefficient in Eq. (1) and the added series equivalent inductance can be derived from the transmission line theory. Current flow path L-gap Partial separation of spirals Fig. 3 Change of Current Flow Path Z + V V Fig. 4 Equivalent Transmission Line Model Z SENSITIVITY ANALYSIS Numerical Simulations Four cable sensors were modelled and simulated with the software, FIDELITY. Their dimensions and characteristic impedances are presented in Table 1. These sensors can be divided into two groups by size; Sensor-I and Sensor-II have a larger diameter of the outer and inner conductor than that of Sensor-III and Sensor- IV. Each group has two cables with different copper tape widths. The dimension of a cable determines its characteristic impedance, and the value of the inductance introduced when turns of the spiral outer conductor separated completely or partially. The copper tape width determines the density of the separation turns, which leads to the different values of the added inductance. Table 1 Spiral Wrapped Coaxial Cable Sensors Sensor Diameter of the inner conductor Diameter of the outer conductor Spiral wrapping copper tape width Characteristic impedance I.8 mm 7.9 mm 3.2 mm 97 Ω II.8 mm 7.9 mm 6.4 mm 97 Ω III.6 mm 3.2 mm 3.2 mm 69 Ω IV.6 mm 3.2 mm 6.4 mm 69 Ω Fig. 5(a) relates the reflection coefficient to the length between separations. It is observed from the figure that Sensor-I and Sensor-III provide a larger value of the reflection coefficient because of their higher turn density. The higher turn density corresponds to a larger value of the added inductance for a certain length between separations, and results in a larger value of the reflection coefficient. Therefore, the two sensors made of a 3.2mm-wide copper tape are expected to perform satisfactorily. The added inductance resulted from a separation between two adjacent spirals can be determined from the ETDR waveform [4]: Z A L _ gap = 4, (2) V o 3

2 2 Reflection coefficient (milli rho) 18 16 14 12 1 8 6 4 Sensor I Sensor II Sensor III Sensor IV Reflection coefficient (milli rho) 18 16 14 12 1 8 6 4 Sensor I Sensor II Sensor III Sensor IV 2 2.2.4.6.8 1 1.2 1.4 Length between separations (cm) 2 4 6 8 1 12 14 16 Added inductance (nh) (a) (b) Fig. 5 Reflection Coefficient vs. Length between Separations or Added Inductance in which A is the area of a reflected pulse and V o is the peak voltage of a step pulse launched by the TDR. Fig. 5(b) presents the peak reflection coefficient as a function of the added inductance. The smaller the characteristic impedance of sensors (Sensor- III and Sensor-IV), the higher the sensitivity of sensors is. Combining the observations from Figs.5(a, b), it can be concluded that within the same length between separations, Sensor-III is considered as the best design among the four sensors due to its smaller geometry, smaller characteristic impedance and higher sensitivity. Static Tests with Small-Scale Beams Two Prototype-I sensors were fabricated and mounted near the tension surface of two RC beams of.914 m long. Each beam was tested monotonically with three point loads. The crack pattern and its corresponding results are presented in Fig. 6 for Beam 2a with Sensor-IV and in Fig. 7 for Beam 1c with Sensor-I. The reflection waveform along the entire length of Beam 2a reaches the maxima at two locations, one more significant than the other. They coincide well with the locations of two cracks. Outside the crack zone, the reflection coefficient is within 2~3 milli rho, indicating small elongation in the non-cracking area. Similarly, Fig. 7 indicates that the multiple-crack pattern has been successfully identified by the embedded sensor. The reflection coefficient corresponding to the peak of the waveform, reaches 8 milli rho for Sensor-I in Beam 1c and 15 milli rho for Sensor-IV in Beam 2a. Note that each reflection waveform, shown in all figures, represents the difference of measurement at a certain load from the baseline measurement when the beam is unloaded. 4

Proceedings of the Structural Materials Technology (SMT): NDE/NDT for Highway and Bridges 24 14-17 24, Buffalo, New York Crack:.155 mm.79 mm.53 mm.23 mm.15 mm 5 mm 2nd 1st Fig. 6 Crack Pattern and Measured Reflection Coefficient of Beam 2a 8th Crack: 1.6 mm.488 mm.457 mm.175 mm.119 mm.84 mm.5 mm 5 mm 3rd 6th 1st crack 2nd crack 4th crack 5th 7th Fig. 7 Crack Pattern and Measured Reflection Coefficient of Beam 1c VALIDATION TESTS Full-Scale RC Girder under Cyclic Loading A square, hollow RC girder of approximately 15 m long was loaded under a progressively increasing cyclic twist until its ultimate strength. The girder started cracking at the end of 1.5 cycles, and experienced yielding in the reinforcing steel during the 23rd cycle. The crack pattern is illustrated in Fig. 8 after 4.5 cycles. Sensor Starts Crack 1 Crack 2 Crack 3 Crack 4 Crack 5 Sensor Crack 7 Crack 6 Sensor Ends Fig. 8 Crack Pattern at the End of 9 Half Cycles (Peak Negative Twist) A Prototype-II sensor was embedded to one of the side faces, 76 mm below the top edge of the girder, in a precast groove that was 12.7 mm deep into the concrete and covered with cement grout. The sensor was subjected to tension and compression alternately within each loading cycle. During the tests, ETDR measurements were taken from both the near and far ends of the sensing cable. The ETDR reflection waveforms corresponding to the loading half cycles of No. 4, 6 and 8 that theoretically causes Cracks 1, 2, and 5 in Fig. 8, are presented in 5

Fig. 9(a, b) when data were taken from the near and far ends of the sensor, respectively. The sensor starts at.55 m and ends at 3.7 m; both were measured from the near end of the cable. In comparison with Fig. 8, Fig. 9(a) indicates that the sensor has successfully identified the location of Crack 1 and Crack 5, but likely missed Crack 2. Three spikes in the waveforms represent the location and relative magnitude of the cracks. The maximum signal response measured from the near end is 62 milli rho at Crack 1, corresponding to a twist of.2 rad/m. When measured from the far end, however, the maximum signal response to the same level of twist becomes 11 milli rho, due mainly to signal loss. Fig. 9 also shows that, as the number of cycles of an applied twist increases, the reflection coefficient at the location of cracks in general increases appreciably. These results indicate that the cracks are widening due to cyclic loading. Note that the peak at approximately 2 m in Fig. 9 means that Crack 3 associated with negative twist effects, separation of spirals of the sensor, is not fully closed when the girder is loaded in an opposite direction. The reflection waveforms recorded from the far end were measured by looking backward through the sensor, and therefore would be the mirror image of those measured from the near end in a perfect lossless cable. The waveforms in Fig. 9(a) are tapering off to the right side while those in Fig. 9(b) are tapering off to the left side due to signal loss. Reflection coefficient (milli rho) 7 6 5 4 3 2 1 Crack1 Crack2 Crack3 Crack5 No.4 No.6 No.8 Reflection coefficient (milli rho) 14 12 1 8 6 4 2 Crack1 Crack2 Crack3 Crack5 No.4 No.6 No.8-1 1 2 3 4 Distance (meter) -2 1 2 3 4 Distance (meter) (a) Data recorded from near end (b) Data recorded from far end Fig. 9 Measured TDR Waveforms at Positive Loading Cycles The peak reflection coefficient of the sensor, corresponding to Crack 4 under negative twist in Fig.8, was correlated in Fig.1 with the average strain measured at four reinforcing steel bars close to Crack 4. There seems a linear relationship between the measured strain, which is related to crack width, and the peak reflection coefficient from the crack sensor. The half cycle numbers shown in Fig. 1 correspond to the TDR data taken at that time. The discontinuity at the strain level of.25 is likely due to the fact that between the first half-cycle of loading at that level and the next peak, one or more of the steel reinforcing bars yielded, affecting the width of Crack 4 from that point forward. 6

Proceedings of the Structural Materials Technology (SMT): NDE/NDT for Highway and Bridges 24 14-17 24, Buffalo, New York Reflection coefficient (milli rho) 5 4 3 2 1 Half Cycle Number 4, 6, 8-1 1 2 3 Milli Strain 22 12, 14 16, 18, 2 28 3 Yielding Fig.1 Reflection Coefficient versus Strain 34 Reflection coefficient (mrho) Exposed cracks Hidden cracks After a series of tests During one test Fig. 11 Waveforms from a RC Column Small-Scale RC Column under Dynamic Loading An 1143-mm (45-inch) tall RC square column of 23mm 23mm (8" 8") was tested on the shake table. Its lower 635mm (25") was strengthened with fiber reinforced polymers (FRP). The column was subjected to a sinusoidal displacement input. At a stroke of 1.78mm and an excitation frequency of 4 Hz, the reflection coefficient waveform and the corresponding crack pattern during one test are shown in Fig. 11. It is clearly seen that the peaks of the waveform capture all cracks, exposed or hidden. More importantly, the fact that a similar waveform was recorded after the completion of tests indicates a unique memory feature of the crack sensor. Marked for clarification, the exposed cracks in Fig.11 were actually closed and unidentifiable under gravity load. The sensor can thus record damage that has occurred during a recent event, such as an earthquake, even though damage is visually undetectable. SIGNAL LOSS As observed from the girder tests, significant loss of a traveling signal took place either in the crack sensor or the extension cables that were used to connect the sensor to a TDR. To understand the signal loss mechanism, two 1-meter-long sensors (Prototype II) were fabricated and tested together with a 1-meter-long commercial cable as a reference. Actually, the two sensors were constructed by replacing the outer conductor (double shielded, braided silver plated copper wires) of additional two commercial cables with a tin plated stainless steel spiral and a gold plated steel spiral, respectively. The thickness of the gold plating on the spiral was.25mm. Signal often losses in three different forms: skin effect, dielectric absorption, and multiple reflections. Their contributions to the total peak loss are summarized in Table 2 [5]. Since their dielectric layers are identical, the two sensors and the reference cable have the same dielectric loss. Several observations can be made from the table. First, the skin effect is dominant in signal loss. Using a high conductivity metal for the inner conductor and the outer shield of a cable sensor significantly reduces the signal loss. Second, multiple signal reflection depends highly on the fabrication quality of a cable sensor. A controlled fabrication process of cable sensors is necessary to minimize this effect. Finally, by selecting a low loss dielectric material as the insulation layer of a cable sensor, e.g., polyethylene and Teflon, the dielectric loss can be well controlled within a wide frequency range from DC to 6 GHz. 7

Table 2. Skin Effect, Dielectric Absorption, and Multiple Reflections Specimen Skin effect Dielectric loss Reflection loss Total loss Gold plated spiral sensor 9 % 3% 5% 17% Steel spiral sensor 17 % 3% 1% 21% Reference cable 6 % 3% % 9% The signal loss can be compensated to a certain degree by de-convoluting a signal attenuation function with the measured signal from a crack sensor [5]. In the event that clear pulses were observed in a reflection coefficient waveform measured from a crack sensor, the area of each pulse instead of its peak value can be used to recover the potentially lost information since it is directly related to the added inductance by Eq. (2) and further to the crack width. Fig. 12 shows the change of the peak and the area of a pulse with the distance that the pulse travels. It is clearly observed that the peak loss increases exponentially with the length of the sensor while the area of the pulse remains nearly constant as the length of the sensor increases. Therefore, the added inductance of a crack can be accurately estimated with Eq. (2) even though a long, lossy cable sensor has been employed. 8 1 Peak loss [%] 7 6 5 4 3 2 1 steel spiral sensor gold plated spiral sensor reference cable Normalized Integral of pulse [V*S].99.98.97.96.95 steel spiral sensor gold plated spiral sensor reference cable 2 4 6 8 1 Distance [meter].94 2 4 6 8 1 Distance [meter] (a) Signal loss effect on peak response (b) Significantly reduced loss effect on waveform area Fig. 12 Effect and Compensation of Signal Loss BRIDGE APPLICATION The Dallas County Bridge is a three-span structure of a total length of 38.7m (127 ft), Fig. 13(a). The superstructure has three RC beams that were cast integrally with the deck. Built in 1956, the Bridge is currently posted for 18-ton trucks at a speed of 24 km/h. Two Prototype-II sensors have been installed at the bottom face of the deck, perpendicular to the traffic direction as illustrated in Fig. 13(b). Two load tests were conducted in the past six months. The sensors showed no sign of cracks on the deck. The readings from the two tests were consistent. 8

Proceedings of the Structural Materials Technology (SMT): NDE/NDT for Highway and Bridges 24 14-17 24, Buffalo, New York crack sensors (a) Overview (b) Installed sensors Fig. 13 Sensor Implementation on Dallas County Bridge CONCLUDING REMARKS Based on the crack-induced change of their topology, two prototype cable sensors were designed and fabricated. Prototype I can be used to detect cracks induced by static or cyclic loading while Prototype II is applicable for all type of loadings with a special memory feature under dynamic effects. Both sensors have been demonstrated in the laboratory sensitive to cracking, from visually undetectable to excessive crack, and able to accurately identify the location of cracks. Future research will be directed to further enhance the uniformity of sensor performance by using a spray-up type of plasma coating or automatic manufacturing of sensors. In addition, crack sensors must be validated in field conditions and demonstrated for their superior performance in full-scale structures. Potential applications of the developed crack sensors include: Monitoring the behavior of RC structures that are inaccessible, such as pile and shaft foundations, Monitoring the behavior of massive concrete structures such as dams, Monitoring hidden cracks in RC columns retrofitted with steel, concrete, or FRP jacketing, and Recording damage that has occurred during a recent disaster event such as earthquake, explosion, and wind gust. This application is particularly attractive for a rapid post-event assessment of the structural condition of critical buildings or bridges to facilitate emergency responses. ACKNOWLEDGEMENT This study was supported in part by the U.S. National Science Foundation under Award Nos.CMS-9733123 and CMS-2381, and University Transportation Center on campus. The results, findings, and opinions expressed in this paper are those of the author only and do not necessarily represent those of the sponsors. Thanks are extended to Drs. D. Pommerenke and J.L. Drewniak as well as graduate students for their contributions. REFERENCES 1 Lin, M.W., Abatan, A.O., and Zhou, Y. 2. Transverse shear response monitoring of concrete cylinder using embedded high-sensitivity ETDR sensors. In S.C. Liu (ed.), Smart Structures and Materials 2: Smart Systems for Bridges, Structures, and Highways; Proc. of 7 th SPIE. Newport Beach, CA. 9

2 Chen, G.D., Mu, H.M., Pommerenke, D. and Drewniak, J.L. 24. Damage detection of reinforced concrete beams with novel distributed crack/strain sensors. J Struct. Health Monitoring (in print). 3 McDaniel, R., Chen, G.D., Sun, S.S., D. Pommerenke, Huang, X., and Courtright, C. 24. Performance of coaxial cable sensors for crack detection under dynamic loads. Proc. 17 th ASCE Engineering Mechanics Conference, Newark, DE. 4 Poazr, D.M., 1998. MicroWave Engineering, 2 nd Ed. New York: John Wiley & Sons, inc. 5 Sun, S.S., D. Pommerenke, J.L. Drewniak, and Chen, G.D. 24. Signal loss, spatial resolution, sensitivity of long coaxial crack sensors. In S.C. Liu (ed.), Smart Structures and Materials 24: Sensors and Smart Structures Technologies for Civil, Mechanical and Aerospace Systems; Proc. of 11 th SPIE. San Diego, CA. 1