International Symposium (NDT-CE 2003)Non-Destructive Testing in Civil Engineering 2003
|Start > Contributions >Lectures > Ultrasonic 1:|
NON-CONTACT ULTRASONIC INTERROGATION OF COfig3NCRETEJames Berriman, T.H.Gan, D. A. Hutchins, P Purnell.
School of Engineering, University Of Warwick, Coventry CV4 7AL,UK.
There is already extensive use of ultrasonics in concrete diagnostics. These systems typically use narrow-bandwidth piezoelectric transducers, which must be in contact with the concrete surface, an example being the PUNDIT, (Portable Ultrasonic Non destructive Digital Indicating Tester). This is the most common approach to field testing. In this paper, it is compared to a new approach, using air-coupled ultrasound. This combines a pulse compression technique with electrostatic broadband transducers in air, so that contact to the sample is avoided. A comparison of results obtained using the PUNDIT and non-contact (NC) systems is presented here. There are some interesting differences in the data when the concrete material under test is non-homogeneous. They imply that the NC system is more sensitive to the paste properties, whereas the converse is true for the PUNDIT system, which seems to couple energy to the sample primarily via the aggregate.
The variation of the characteristics of concrete and the increased cost of infrastructure maintenance points towards the need for improved methods of NDE with good reliability to cost ratios. By developing improved NDE techniques, the reliability of our infrastructure will be improved while maintaining the tightest possible cost control .
The modern construction industry requires high strength, high performance concrete that is workable, self levelling or self compacting and sometimes retarded or accelerated. Quality requirements are high for durable concrete structures but this must be reconciled with a decrease in available on-site workmanship. Producers of concrete materials have a plethora of admixtures and additives available to them that can modify its properties in the fresh and/or hardened state. On-site use can lead to situations whereby the construction worker has used a variety of these products to achieve perceived maximum success without reckoning with the adverse effects of changing the mix.
Of principle interest is the use of ultrasonic NDE for the reliable assessment or detection of defects within concrete members. Current outlined aims are location of tendons, detection of voids, detection of honeycombing and information on geometric dimensions, although ultrasound is also known to be sensitive to the structure of the concrete, via changes in velocity and attenuation. For instance, a technique for estimating the level of porosity has been developed . This is a procedure to estimate the durability of concrete samples driven by the increasing number of structures with premature deterioration due to environmental action. Other investigators  have used high frequency ultrasound (0.5-1MHz), to quantify chemical damage in concrete. Using attenuation of surface waves it was shown that it was possible to detect and characterise cover degradation. Synthetic aperture focusing techniques (SAFT) have also received recent attention. This is a solution to the problem of flaw detection in concrete with single sided access. SAFT uses an ultrasonic echo method based on the application of multiple source and receiver locations. Data processing algorithms are then applied, which tend to reduce noise and increase image quality over that obtainable from a single scanned transducer.
The PUNDIT, developed in the 1970's, uses the pulse velocity method to determine material characteristics . It generates low frequency ultrasonic pulses and measures the time taken for them to pass from one transducer to the other. It has become part of many national standards for concrete testing  and is still used on site and in research . However, despite the PUNDIT's wide adoption by industry, it is desirable to develop an alternative system. The need for physical contact between the concrete surface and the transducers is inconvenient, as surface roughness is a derogatory factor in providing and maintaining sufficient acoustic coupling. Good contact with the sample is also a prerequisite of the application of SAFT, where positional scanning of source and receiver elements is often required.
The transducers used with the PUNDIT use resonant piezoelectric elements, with a centre frequency of 54kHz. They are driven with a large transient or delta function pulse that causes the transmitter to vibrate at the centre frequency of the piezoelectric elements. Thus, the sample is only exposed to a limited range of frequencies in a single test, and a range of transducers is required to examine responses to different frequencies. Efforts have been made to deliver 'chirp' pulses that contain a range of frequencies but detection was achieved using laser interferometry, owing to the limited sensitivity of piezoelectric transducers in broadband operation . This resulted in a complicated and expensive instrumentation system.
The work reported here takes a new approach to the problem of concrete testing, and uses broadband ultrasonic electrostatic transducers, designed to operate in air. These can accurately emit chirps defined by signal generators rather than being limited to resonant narrow band signals, as in the case of piezoelectric transducers. In these devices, a rigid conducting backplate and a thin metallised Mylar membrane are arranged to form a capacitor. Applying a modulated voltage causes the membrane to vibrate, generating ultrasound; changes in charge across the membrane can also be used for detection . Using air coupling has obvious advantages in terms of scanning, the lack of liquid couplant, and the ability to perform tests at a range of unprepared concrete surfaces. A possible disadvantage of this technique is the much reduced signal levels in the concrete, due to the large acoustic impedance mismatch between air (the coupling medium) and the concrete. However, this can be deal with by using a signal enhancement technique known as pulse compression.
Pulse compression uses the wide bandwidth of the air-coupled ultrasonic system to improve signal to noise ratios (SNR) significantly . In pulse compression, the received signal CT(t)is first bandpass filtered around the centre frequency of the original chirp so that all frequencies not included in the original chirp are removed. This filtered signal is then cross-correlated with the input signal C(t) to give the compressed pulse P(t). Cross-correlation involves incrementally time shifting CT(t), multiplying it with C(t) and plotting the result vs. the time shift. The value of P(t) will be approximately be zero except at a single time-shift where a sharp peak is evident; this represents the time-of-flight if the signal. The SNR improves dramatically with application of this technique, such that signals can be easily recovered from well below the noise floor of the detector. Time-of-flight resolution is improved since the signal will only correlate at a single well-defined time shift.
The use of pulse compression and broadband transducers for materials diagnostics is becoming an accepted technique in applications where contact is undesirable, such as in the testing of food . Recently, efforts have been made to apply these techniques to heterogeneous materials, and in particular to concrete . In the cited paper, it was shown that low power non-contact ultrasonics could be used for the NDE of concrete of up to 75 mm in thickness. It has been hypothesised that discrepancies between the air-coupled and more traditional (PUNDIT) systems were due to preferential coupling effects; the non-contact system being more sensitive to paste properties and the PUNDIT being more sensitive to aggregate .
This paper describes extensive experiments that confirm this preliminary conclusion. These will use the air-coupled pulse-compression system with capacitive transducers for concrete diagnosis, with a comparison to results from the PUNDIT as a comparative tool. The aim of this research is therefore to highlight and explain the differences in the results given by these two ultrasonic measurement systems.
Six different concrete mixes were prepared with target weights of 1.75 kg, using Portland cement, sand and 10mm aggregate in various proportions. The constituents of the concrete mixes were specified such that a range of different concrete samples was produced, all with varying amounts of aggregate. The water/cement ratio was kept constant at 0.5. Table 1 shows the target ratios for the six mixes.
The samples were mixed and set in purpose built wooden moulds to give lateral dimensions of 100mm x100mm, with a 30 mm thickness. The samples were then tested every day for seven days and five more sporadically up to a concrete age of 49 days using both ultrasonic systems. This was designed to demonstrate changes in the speed of sound through the thickness of the concrete with time as curing proceeded.
The experimental set-up for the air-coupled system is shown in figure 1. The nucleus of the system is the NCA1000 pulser/receiver unit from VN Instruments Limited. This contained a digital signal processor that generated tuned ultrasonic Hanning 'chirp' pulses that the capacitance transducers can accurately mimic.
The pulse generator contained a 200W broadband variable-gain power amplifier and the receiver contained a variable-gain low-noise amplifier (maximum gain of 90dB), followed by an A/D converter. This was then fed to the source transducer. The emitted chirp signal travelled through air to the sample, through the sample thickness, and out into the air on the far side. Here, a receiver was positioned co-axially with the source, so that a through-transmission signal could be obtained. A Cooknell charge amplifier model CA6/C, which had a gain of 250-mV pC-1, amplified the output from the receiver transducer before being fed into the NCA1000 input. The transducers were of a novel metallised polymer film/micromachined silicon backplate capacitance design described previously . Upon signal reception, the embedded software within the digital signal processor performed on-line pulse compression processing. The output of the NCA1000 was then superimposed upon a +200V dc bias voltage.
Figure 2 illustrates how the pulse compression was used with the broad bandwidth of the transducers to get good signals. Here, waveforms are presented for signals transmitted across an air gap (i.e. without a sample being present). Figure 2(a) is an example of a typical unfiltered hanning modulated broadband chirp pulse that is transmitted across the air gap. It has a chirp centre frequency of 450 kHz, a bandwidth of 300 kHz with a pulse duration of 200 ms. Fig. 2(b) is the time waveform after application of bandpass filtering. This is utilised to reduce low frequency oscillations that occur either as a result of small traverse dimensions, (allowing a low frequency airwave to bypass the sample), or as a consequence of using a thin plate like sample in which low frequency flexural modes can be stimulated.
Fig. 2(c) is the FFT of the filtered time waveform. This waveform depicts the wideband frequency response of the capacitance transducer air-coupled system limited by the 300 kHz bandwidth of the applied chirp. Finally the waveform of fig. 2(d) shows the pulse compression output from the NCA1000 pulser/receiver unit. The single peak is expected and is a direct result of the cross correlation. As can be seen, the time resolution is now excellent. Note that the pulse compression is improved for both wider bandwidths (i.e.) a greater range of frequency sweep and/or a longer chirp pulse in time.
The pulse compression system was now used on real concrete samples, and the arrival time of the peak signal from the pulse compression output measured for the range of concrete samples listed in Table 1 at various times. Knowing the sample thickness, the longitudinal sound velocity could then be calculated as a function of time. The measurements were repeated in every case using the contacting PUNDIT system, and a comparison made of the results. The results indicated that the air-coupled system was sufficiently sensitive to be able to test concrete samples containing a range of aggregate percentages, for a sample thickness of 30 mm.
Results and discussion
The results that were obtained for the samples of Table 1, using both air-coupled and PUNDIT systems, are shown in Figure 3 for all six samples. The first, Figure 3(a), was just paste with no aggregate or sand. As can be seen, both transducer types measured the expected increase in longitudinal velocity with time, as the material hardened. It is interesting to note that for some of the measurements there is reasonably good agreement between the air-coupled and PUNDIT systems, whereas at intermediate times, the PUNDIT seems to produce a lower value for the velocity. In samples 2 and 3, a greater proportion of aggregate was present, and in these cases it was observed that the PUNDIT now measured a higher velocity than the air-coupled system, and that this discrepancy increased with the greater proportion of aggregate. The addition of sand in samples 4-6 (Figures 3(d)-(f)) caused a similar phenomenon.
The above results may be summarised in Figure 4, where the difference between the velocities measured by the air-coupled and PUNDIT systems are compared. It will be seen that the addition of aggregate in particular has led to a difference in velocity being measured by the two methods, with the air-coupled approach routinely measuring a lower velocity than the PUNDIT.
The reasons for this difference are thought to be due to the nature of the coupling between the transducers themselves and the concrete samples. The air-coupled systems have to get energy into and out of the samples via the air/sample interface. The amount of energy transmitted across the interfaces is strongly dependent on the acoustic impedance Z of the sample, noting that Z = rc , where r is density and c acoustic velocity. Air has a very low value of Z compared to concrete. More energy is transmitted as the impedance mismatch to air gets smaller or as the impedance of the concrete decreases. Paste has lower values of both r and c, and hence for a sample containing aggregate, it would be expected for the energy to be initially transmitted from the air into the paste primarily, and not into the aggregate. The converse would be true for the PUNDIT system, which uses a piezoelectric element in a metallic case. Here the impedance is higher than the paste, and more closely matched to that of the aggregate; hence, we would expect energy to better coupled to the aggregate initially.
The data of Figures 3 and 4 can be explained by these changes in coupling. PUNDIT would couple energy into the aggregate primarily, and signals would pass through the concrete sample with a bias towards measuring the aggregate properties. Conversely, the air-coupled system couples preferentially into the paste, and hence would be expected to be more sensitive to changes in the paste as it cures. This is indeed what is observed, with a greater range of acoustic velocities, at lower values, being observed in the air-coupled data.
The important feature of the above is that the PUNDIT may be underestimating the changes in acoustic properties (and hence the elastic moduli) of the concrete samples during the curing cycle. Conversely, it may be measuring more about the aggregate properties. In both cases, it is recognised that a complicated wave propagation path is followed. Current work is looking at this phenomenon, with a view to quantifying this effect.
It has been demonstrated that air-coupled systems can be used to test concrete samples, and that the longitudinal acoustic velocity can be measured. It has also been shown that the results show a difference to that from a contacting PUNDIT system. It is thought that these differences arise from the coupling to the sample in each case. It is suspected, however, that the air-coupled system may be more sensitive to changes in properties of the paste as it cures, and this may be of interest to researchers in this area.