|NDT.net Sep 2004 Vol. 9 No.09|
| 2nd MENDT Proceedings
Inspection of Steel Tendons in Concrete Using Guided WavesM. D. Beard, M. J. S. Lowe, and P. Cawley
Department of Mechanical Engineering, Imperial College, London, SW7 2AZ, UK
SynopsisThe use of ultrasonic guided waves to inspect concrete reinforcing tendons is complicated by energy leakage into the embedding material. Previous work has shown that high frequency, low-leakage guided waves can be used to minimise the leakage, but the maximum inspection range has not yet been determined. This paper presents the results of experiments designed to evaluate the effect of important factors such as leakage and defect geometry, enabling conclusions on the maximum inspection range to be reached.
IntroductionPost-tensioning is a common construction method used to increase the load bearing capacity of concrete structures. The technique is widely used in bridge construction, where steel tendons are run through ducts in the concrete after the concrete has hardened. These tendons are then tensioned to apply a compressive load to the concrete, hence increasing the maximum tensile load that it can carry. The tendon ducts are then filled with an alkaline grout, which protects against water ingress and corrosion. However, if the grouting is poor or incomplete, water may come into contact with the tendons, leading to corrosion and fracture. This process is often accelerated by the presence of road salts dissolved in the water. The tendons are critical to the strength of the structure, and failure can lead to a transfer of load to other tendons, leading to a chain reaction of failures that will eventually cause the structure to collapse. Following the collapse of a bridge at Ynys-y-Gwas , a major inspection program of all existing post-tensioned bridges in the UK was started. A moratorium was also placed on the construction of new post-tensioned bridges, which was only lifted after the introduction of new procedures designed to prevent poor grouting.
Although the new procedures should reduce the instances of poor grouting, the problem of how to reliably inspect tendons on existing structures remains. There are many inspection techniques available to bridge engineers, but the embedded nature of the tendons, combined with the fact that the tendons are located in ducts, makes the inspection of individual tendons extremely difficult. Standard methods, such as radiography, are capable of locating ducts, and also of detecting voids in the grout, but the resolution is not sufficiently good to determine tendon condition. The identification of corroded tendons usually relies on the detection of corrosion products in concrete samples, or visual examination of the tendons through bore holes. Clearly, such invasive methods cannot be used to examine a complete structure, and their use also carries a risk of further damage.
The concept of using guided waves for the inspection of individual tendons has been in existence for some years, and has recently been investigated further . The inspection method is a pulse-echo technique, with guided waves being excited at the free end of the tendon, which can be exposed in the anchorage region. These waves are then reflected from breaks or major defects, as shown in figure 1. The position of any defects can then be calculated from the reflection arrival time. The maximum inspection range is limited by the amount of attenuation the wave experiences as it propagates. The fundamental mechanisms of attenuation are energy leakage from the tendon into the grout material, and material damping effects, both of which are highly dependent on the mode properties at the frequency of excitation. High frequency, low-leakage modes which have minimum attenuation have recently been discovered , but although the total attenuation has been predicted, it has not been measured until now. There are also other significant factors that affect the inspection range, and these need to be taken into consideration.
The research presented in this paper aims to provide an objective understanding of the behaviour of these low-leakage guided wave modes, so that the feasibility of developing non-destructive testing techniques for different structures can be properly assessed. This is achieved primarily through the use of experimental techniques to measure the attenuation of the modes in grouted tendons, and the reflection coefficient of the modes from defects. This information is then used to determine the maximum possible inspection range for tendons of different diameter. Two types of reinforcing tendon are considered, single wires, typically 5mm or 7mm in diameter, and 15.9mm 7-wire strands, consisting of six 5.2mm diameter wires spirally wrapped around a central 5.5mm diameter wire. It has been found that the stranded tendons behave differently to single wires, and these differences are discussed along with the implications for tendon inspection.
Theory and ModelingPrevious work in this field  has shown that it is extremely beneficial to model guided wave behaviour analytically, in order to understand how the waves propagate. The development of the guided wave modeling software 'Disperse' [4,5] has greatly simplified the modeling of guided waves in leaky, multi-layer, cylindrical systems. Disperse was used to model the grouted tendon as a 2 layer axially symmetric system, consisting of a solid steel cylinder embedded in an infinite expanse of cement grout. The material properties that were used for the steel and grout respectively are: Young's Modulus, 216.9GPa and 11.2GPa, Poisson's ratio, 0.29 and 0.21, and density 7900kg/m3 and 1600kg/m3. At high frequency, material damping is also a significant cause of attenuation, and was therefore included in the model. The longitudinal and shear attenuation was taken to be 0.003 and 0.043 nepers/wavelength respectively for the steel, and 0.008 and 0.100 nepers/wavelength respectively for the grout.
Figure 2 shows the predicted attenuation dispersion curves for the axially symmetric 'longitudinal' modes in the tendon system. Each line represents a propagating mode, showing how the attenuation of that mode will vary with frequency. Only the longitudinal modes are shown, as they generally have lower attenuation than the non-axially symmetric 'flexural' modes. The axially symmetric nature of these modes also makes them relatively easy to excite with a single compression transducer. The modes are numbered sequentially in the usual series L(0,n), where n is a counter variable that increases with the mode order. The frequency and attenuation are expressed as products of the tendon radius, allowing the information to be used for any tendon diameter.
The most significant factor in selecting a suitable mode is minimising the attenuation, in order to maximise the inspection range. Figure 2 shows that each longitudinal mode reaches a minimum in attenuation, and that the higher order modes reach increasingly lower attenuation in the frequency range shown. These minima form the group of low-leakage modes identified by Pavlakovic  to exist in embedded structures. The fact that attenuation decreases with frequency is counter-intuitive, as the attenuation of ultrasound generally increases with frequency, due to material damping effects. However, these 'low-leakage' modes have specific properties that minimise the attenuation, which are discussed in detail in . It is also desirable to excite a mode at a point of maximum energy velocity (the speed at which the wave packet travels), to minimise the effects of dispersion. Fortunately, the attenuation minima shown in figure 2 all correspond with energy velocity maxima, and the energy velocity is also close to the steel longitudinal bulk velocity.
The selection of suitable modes is also governed by other practicalities. The use of low frequency modes such as L(0,1) is undesirable, because the long wavelength means that energy can travel in the beam or grouting duct as a whole, and not just within the tendon. At high frequency, it becomes more difficult to achieve repeatable results because of variation in the coupling conditions, and the attenuation also begins to increase because of material damping effects. In general, the L(0,8)-L(0,12) modes have been found to give the best compromise between ease of excitation and low attenuation. The behaviour of these modes is very similar, and although the results presented in this paper are for specific modes in this range, the results are typical of all the high frequency low-leakage modes.
Effect of Excitation SignalA highly damped immersion transducer with a centre frequency of 5MHz was used for all of the experiments reported in this paper. The transducer was driven by a custom-built pulse-echo amplifier with a maximum operating frequency of 10MHz. An external PC and Arbitrary Function Generator were used to generate the input signal, and a digital oscilloscope was used to capture the received signal, which could then be transferred to the PC for further signal processing. The transducer was mounted in a holder clamped onto the tendon ends, and coupled using an industrial coupling gel. A highly damped transducer was selected because it could be used over a wide range of frequencies, with minimal risk of ringing. The excitation signal consisted of a Gaussian windowed toneburst, with a centre frequency corresponding to the frequency of an attenuation minimum, allowing the ultrasonic energy to be targeted on a specific low-leakage mode. Increasing the number of cycles in the input signal reduces the frequency bandwidth of the signal, but also increases the length of the signal in time. This reduces the ability to inspect the near field, but since the wavelength is very small, the effect is relatively minor for moderate numbers of cycles.
The effect of the excitation signal bandwidth has been investigated by measuring the amplitude of the end reflection from a short length of embedded tendon, with different numbers of cycles in the input signal. Figure 3 shows the results for the L(0,8) mode. There is a steady improvement in amplitude with increasing cycles, as more energy is concentrated at the low-leakage frequency. However, with more than about 20 cycles, the rate of improvement increases. The inset in figure 3 shows the frequency spectrum of a 20-cycle excitation signal compared to the frequencies at which attenuation minima occur. It can be seen that at 20 cycles and above, all the energy in the input signal is concentrated on just one low-leakage mode. This is very beneficial, as the excitation of multiple modes can result in destructive interference between the modes, which all propagate at slightly different speeds. This explains the observed increase in the rate of improvement in end reflection amplitude in the single mode excitation region shown on figure 3. Testing at very high numbers of cycles was not possible, because of constraints on the length and maximum amplitude of the signal generated by the pulse-echo amplifier. The maximum test range will therefore be achieved with a very narrow bandwidth signal at the frequency of an attenuation minimum. Typically, input signals of 200 cycles have been used where near-field resolution is not important.
Measurement of AttenuationThe existence of the low-leakage points can be verified simply by examining the strength of a tendon end reflection at different frequencies . The actual value of the attenuation can be measured by comparing the signal amplitude at two points along an embedded length, but this is difficult to do because the displacement of the low-leakage modes is very small at the tendon surface. The attenuation measurements presented in this paper are based on the relative amplitude of multiple end reflections from a short length of embedded tendon, which is much easier to measure. Figure 3a shows the transducer and specimen, which was tested with a 100-cycle toneburst centred on an attenuation minimum. The experimental set up is shown schematically in figure 3b, and an example time trace showing 3 reflections from the specimen end is given in figure 3c. The attenuation can then be estimated from the reduction in signal amplitude between subsequent reflections.
This method of attenuation measurement assumes that the reflection coefficient of the low-leakage modes from the tendon ends is unity. Pavlakovic  showed by Finite Element methods that the reflection coefficient of the L(0,4) mode at a low-leakage point is around 90% for a complete, flat break in a free bar. The high reflection coefficient occurs because the axial stress is high relative to the shear stress at low-leakage points, hence the required stress-free boundary condition at the tendon end can be met almost entirely by the superposition of the incoming and outgoing low-leakage modes. This means that only a very small amount of mode conversion to other propagating and non-propagating modes occurs . The high number of elements needed for the simulation means that the reflection coefficients for higher order modes cannot currently be calculated using the author's finite element modeling facility. However, as the mode order increases, the relative amount of in-plane stress to shear stress increases, which will increase the reflection coefficient further, making a reflection coefficient of unity a reasonable assumption. The relative axial and shear stress mode shapes for the low-leakage point on the L(0,11) mode at 19.0MHz-mm are shown in figure 3d.
The experiment was repeated for the L(0,8) mode in 5mm and 7mm diameter single wires, and the inner and outer wires of a 15.9mm 7-wire strand. The measured attenuation is shown in Table 1, together with the predictions made by Disperse. The results show that the model generally overestimates the attenuation, although the behaviour of the individual specimens is very different. In the single wire specimens, the predicted attenuation is more accurate for the smaller wire diameters. One probable reason for the overestimation is that the model assumes that the grout and the steel are perfectly bonded, which is unlikely. Poor bonding has been shown to reduce the leakage in such systems . In addition, there may be error in the material damping coefficients, which are significant at high frequency, and can vary considerably between different steels. The strand specimen results are very different for the inner and outer wires, and are discussed in a later section of this paper.
Reflection from Breaks and DefectsIn order to determine the maximum inspection range, it is also necessary to know how much energy will be reflected from breaks, which may be aligned at some angle to the tendon axis
The effect of break angle has been measured experimentally, by cutting different end angles on a steel bar with the centre section embedded in mortar, shown schematically in figure 5. A series of pulse-echo tests was carried out on the specimen, with a new end angle cut between each test. The end angle was cut with a circular bench saw, and was varied from 0° to 55° to the normal to the bar axis. The transducer was left firmly clamped in place during all the cutting operations to ensure the contact conditions were varied as little as possible. The mortar section acts as a filter for the outgoing wave and the reflected wave, ensuring that only the low-leakage modes can be incident on the angled end, and be received by the transducer. The many other modes that may be generated by mode conversion at the angled end and by the transducer are highly attenuated in the embedded section, allowing just the behaviour of the low-leakage modes to be examined.
Figure 5 also shows the measured loss in end reflection amplitude for different end angles for three low-leakage points on the L(0,12) mode. There is a sharp and almost linear reduction in signal amplitude up to about 40dB loss at an angle of 10° . The loss then generally decreases to about 30dB between 10° and 55° . Following the experiment, a repeat test was made at 0° and the end reflection amplitude was within 2dB of the original 0° test, demonstrating that the transducer contact conditions had remained reasonably constant throughout the experiment. All three of the low-leakage modes are affected to about the same extent, irrespective of frequency. This is expected, because the displacement characteristics of each mode are similar at the low-leakage points. The poor reflection coefficient is primarily a result of the wavelength being only around 2mm, and so at 10° there is more than 1 whole wavelength difference in the length of the bar across the diameter. This is likely to lead to a large amount of mode conversion taking place to both flexural and other longitudinal modes; these mode-converted waves are not seen in the reflected signal because they are not able to propagate through the embedded section.
Effects in Stranded TendonsFigure 6 shows the time traces recorded from the inner and outer wires of a 15.2mm diameter stranded tendon. The L(0,8) mode was excited using a narrow bandwidth signal of 200 cycles.
The inner wire shows evidence of several modes propagating in addition to the low-leakage mode. This is because the centre wire is not well coupled to the grout, reducing the leakage of all the modes. This explains why the measured attenuation in the centre wire of a strand was lower than that of a single wire of similar diameter. The time trace recorded from the outer wire of a strand specimen also shows evidence of more modes, although in this case no second or third reflections can be seen, indicating high attenuation. The FFT of the end reflection from an outer wire is shown in the right hand side of figure 6, and it shows that very little energy propagates at the centre frequency, which corresponds to an attenuation minimum in a straight tendon. A similar effect has been observed in the testing of deformed epoxy bonded rock bolts , which suggests that the low-leakage points are lost when curvature is introduced to the system. It is therefore unlikely that low-leakage guided waves can be used to inspect the outer wires of stranded post-tensioning tendons effectively, because of the high attenuation experienced by the waves.
Discussion and ConclusionsThe measured attenuation and reflection coefficients can be used to determine the maximum length of tendon that can be inspected using the proposed method. The pulse-echo instrument amplifiers can compensate for about 100dB of attenuation before the received signal is lost in the noise floor. Loss through scattering at a complete break could account for 40dB of this attenuation, thus the total allowable attenuation due to material damping and leakage has to be less than 60dB. The inspection range for a complete break would therefore be limited to about 1.2m in 7mm diameter wires and 0.8m in 5mm diameter wires. In 15.9mm stranded tendons, about 1.5m of the centre wire could be inspected, but inspection of the six outer wires would be limited to about 0.5m, because of curvature effects. For partial depth defects, the inspection range could be reduced considerably, because the low-leakage modes are unlikely to be sensitive to defects near the tendon surface . A further problem is that an undamaged tendon would be indicated by a lack of any return signal, and it is therefore difficult to confirm the range that has been inspected.
The inspection of small diameter tendons using guided waves is therefore limited to short lengths of tendon in the anchorage region. This area is difficult to inspect using existing techniques, and therefore guided waves do offer some advantages. However, since the attenuation of the waves decreases with increasing bar radius, the opportunities for inspecting larger diameter grouted bolts and re-bars are much better. The predicted attenuation for a 25-30mm diameter bar reduces to around 15dB/m, which would give a 5m range if the end was flat. Tests on a number of 3m long epoxy-bonded rock bolts of 22mm diameter have shown that the method can be used to successfully determine the length of bolts, and to identify some defects .
AcknowlegementsThis work was funded by the Engineering and Physical Sciences Research Council and was carried out in collaboration with the Transport Research Laboratory Ltd., and Rock Mechanics Technology Ltd.