WO1992008128A1

WO1992008128A1 - Detecting defects in concrete

Info

Publication number: WO1992008128A1
Application number: PCT/GB1991/001905
Authority: WO
Inventors: Patrick Andrew Gaydecki; Frederick Michael Burdekin
Original assignee: Capcis Limited.
Priority date: 1990-10-30
Filing date: 1991-10-30
Publication date: 1992-05-14
Also published as: GB9023555D0; EP0555298A1

Abstract

A method and apparatus for detecting the presence and condition of components or defects embedded in a non-homogeneous body, in which the body is first insonated with an acoustic signal including a predetermined band of frequencies, for example from 0 to 400 kHz. A receiver transducer is positioned in contact with the body, to detect acoustic signals. Detected signals are transformed to the frequency domain to produce a frequency spectrum. The signals are then filtered to produce output signals representing the magnitude of the acoustic signals in each of at least two discrete frequency bands. The magnitudes of the output signals are then compared to derive frequency dependent data relating to defects in the body. The invention enables ultra-sonic techniques to be applied to the detection of small faults in non-homogeneous materials such as concrete.

Description

DETECTING DEFECTS IN CONCRETE

The present invention relates to a method and apparatus for detecting the presence and condition of components or defects embedded in a non-homogeneous body, for example a structural member fabricated from reinforced concrete.

Ultra-sonic fault detection has been used for many years to inspect the internal composition of metal components and castings.

In the case of ultra-sonic waves with a frequency of less than 10 MHz many metals are broadly homogeneous in structure and isotropic and accordingly the waves will travel with equal speeds in all directions through the material. In contrast, concrete is highly inhomogeneous. If acoustic waves are transmitted through an inhomogeneous material such as concrete the large number of inclusions gives rise to such a variety of scattering and absorption mechanisms that signals which can be picked up from the surface of the concrete are very difficult to interpret in a meaningful way.

The paper "Flaw Detection in Concrete by Frequency Spectrum

Analysis of Impact-Echo Wave Forms" by N. J. Carino, M. Sansalone and

N. N. Hsu "International Advances in Non-destructive Testing", 1986,

Volume 12, pages 117-146 describes the application of frequency spectrum analysis to determine the thickness of plate structures and to locate planar internal flaws in such structures. The Fast Fourier

Transform Technique is used to compute the frequency content of received pulses. This study shows that the frequency spectrum includes dominant frequencies corresponding to for example boundaries of the .plate and planar faults the presence of which within the plates are known. A low frequency peak was noted which corresponded to reflections from the bottom surface of the plate and a higher frequency peak was noted corresponding to reflections from a small diameter disc incorporated in the plate. This technique therefore makes use of wavelengths produced by thickness-mode vibrations of the sample, that is, the depth of the planar defects determines the wavelengths generated.

It is well known that the frequency content of acoustic waves transmitted through a body determine the size of the flaw that can be detected by a monitoring the amplitude of those acoustic signals. For a flaw to be detected it must be of the same order of, or larger than, the component wavelengths in the acoustic wave. Higher frequency components have shorter wave lengths and it is only higher frequency components which will be reflected by smaller faults. Unfortunately, higher frequency pulses have less penetrating ability as they are attenuated more quickly than low frequency pulses in a non- homogenous material such as concrete.

Although the above paper describes techniques which can be used to detect relatively simple faults, the frequencies used are relatively low. Typical instruments generate frequencies of up to 40 kHz. Since the propagation velocity in concrete is around 4000 metres per second depending upon the age, mix and strength, this gives a wavelength of 100 mm. Such a wavelength is far too long to provide any detailed information regarding small scale flaws such as breaks in reinforcing cable or small voids in ducts. Much greater frequencies are required to detect small flaws and in particular frequencies of the order of 400 kHz are required to detect flaws associated with reinforcing cables in concrete. The use of such high frequencies has not heretobefore been suggested because of poor propagation. Essentially it has been considered impractical to work at higher frequencies because concrete gives rise to such a confusion of echo noise that echoes arising from the non-homogeneous structure of the concrete are of much larger magnitude than echoes arising from faults to be detected. Thus ultrasonic techniques have not found wide application for the detection of embedded components or defects in concrete " or similar non-homogeneous material in contrast to the application of such devices to the inspection of homogeneous materials such as most metals.

Ultrasonic testing of concrete has generally been restricted to use of transit time measurement at one nominal (low) frequency, to determine pulse velocity, from which physical properties of the concrete itself have been inferred by empirical correlations.

It is an object of the present invention to obviate or mitigate the problems outlined above.

According to the present invention, there is provided a method for detecting the presence^' and condition of components or defects embedded in a non-homogeneous body, wherein the body is insonated with an acoustic signal including a predetermined band of frequencies, a receiver transducer is positioned in contact with the body, acoustic signals detected by the transducer are transformed to the frequency domain to produce a frequency spectrum, the detected signals are filtered to produce output signals representing the magnitude of the acoustic signals in each of at least two discrete frequency bands, and the magnitudes of the output signals are compared to derive frequency dependent data relating to defects in the body.

Thus the analysis technique of the present invention is entirely different to that described in the above paper by Carino and ^'Sansalone. In that case, an analysis is made of wavelengths resulting from thickness-mode vibrations, i.e. the frequencies generated are geometry dependent. In the case of the present invention, the transmitter itself generates all frequencies that form part of the analysis, irrespective of sample size or geometry.

Preferably the body is insonated with acoustic waves in the band from 0 to 400 kHz. This 40Q kHz wide band is then divided into eight bands each 50 kHz in width by filtering to produce eight sets of data which are then inverse transformed. The energies in the eight bands are then compared.

The method of comparison depends on the kind of defect which is being sought. For example, comparison can be in the form of cross- multiplication, in which all the energy bands are simply multiplied together, an operation similar in some ways to cross-correlation. This approach is particularly successful in locating moderate to large size voids when increased reflection of all wavelengths is to be expected. In the case of smaller faults, for example small regions of delamination or complete breakage or separation of reinforcing cables, the higher frequency components display changes in energy beyond the limits normally attributable to random variability. Lower frequencies, because of their longer wavelengths, remain relatively unaffected by such small faults. Thus in such circumstances a ratio is taken of the high frequency energy in for example the 350-400 kHz band to the low frequency energy in for example the 0 to 50 kHz, using the low frequency energy to check for background variability. Thus filtering in accordance with the present invention enables data to be extracted from signals despite the fact that those signals are dominated by frequency components which do not carry any information of interest. The invention thus makes ultra-sonic techniques applicable to the detection of small faults in concrete in a manner not suggested by the prior art. The use of low frequencies to correct for background variability is not suggested by the prior art.

The present invention also provides an apparatus for detecting the presence and condition of components or defects embedded in a non-homogeneous body, comprising means for insonating the body with a wide band acoustic wave, a receiver transducer which is positionable in contact with the body to detect acoustic signals, means for transforming the detected acoustic signals to the frequency domain to produce a frequency spectrum, means for filtering the detected signals to produce output signals representing the magnitude of the acoustic signal in each of at least two discrete frequency bands and means for comparing the magnitudes of the output signals to derive frequency dependent data related to defects in the body.

An embodiment of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which:

Figure 1 is a schematic illustration of an embodiment of the present invention; and

Figures 2 and 3 illustrate typical results obtained using the apparatus illustrated in Figure 1.

Referring to Figure 1, the system comprises a pair of roller probes, the rollers 1, 2 of which are in use placed in contact with the surface 3 of a concrete body to be investigated. The roller 1 couples signals to the surface 3 and thus acts as a transmitter whereas the roller 2 detects acoustic signals from the surface 3 and acts as a receiver. The rollers are C.N.S rolling probes which are 80 mm in diameter and 40 mm in width with a nominal resonance of 360 kHz. In use, good acoustical couplings to the concrete surface 3 is assured by the soft outer neoprene tyres of the transducers which roll across the surface 3. The surface 3 is also sprayed with water so that a continuous water film is built up between the surface and the probe tyres. The probes are matched symmetrical devices, in that they may act as both transmitters or receivers. Thus the probes are rolled in parallel across a surface to be scanned, the spacing between the rollers in one example being 65 mm as measured from the tyre mid points.

The transmitting probe 1 is energized by a Steinkamp pulser unit 4 which generates a 600 volt spike lasting for a few microseconds once every second. A pulse of this nature shocks the transducer into resonance in much the same way that a strike from a clapper excites a bell into ringing at its fundamental frequency. The resulting ultrasonic wave contains the fundamental resonance frequency and in addition many harmonics so that a wide band of frequencies make up the ultrasonic wave which travels through the concrete. The concrete body is thus insonated. The ultrasonic wave is not focused and thus spherical waves are propagated through the body underlying the surface 3, the spherical waves being centred on the point of contact between the transmitting transducer 1 and the surface 3. Acoustic waves are then detected by the probe 2, amplified by an amplifier 5 and applied to a waveform analyzer 6.

At the same instant that an excitation pulse is generated by the pulser unit 4, a synchronization signal is also generated by the unit 4 and applied to the analyzer 6. The synchronization signal instructs the analyzer to record the data then present at its input, that is the signal being picked up by the receiving probe 2. The wave form analyzer 6 may be obtained from the Analogic Corporation, comprising a model 630 analogue to digital converter and a model 610 processing unit.

The A to D converter is a twin channel device, featuring independent time bases, a sampling rate ranging from 1.7mHz to 35MHz, a maximum record buffer length of 8192 points, and an in-built anti^¬ aliasing filter set at 5MHz. Standard sampling resolution is 9 bits (512 levels), with 12 bits in signal compression mode (4096 levels). The device can be triggered in a wide variety of ways, either internally or externally, or clocked for each sample point. Pre- and post-triggering is available as standard.

The 6100 processing unit is built around a 16 bit 68000 microprocessor, and has a very comprehensive repertoire of. operations, with more than 50 waveform analysis, manipulation, signal processing and display functions. These include forward and inverse FFTs (typically 400ms for a record length of 512 points), chirp-Z transforms, convolution, correlation and all standard scalar calculations.

The analyzer 6 is controlled by a host computer 7 (IBM or compatible) via an IEEE communications interface. A "supervisor" software system instructs the analyzer to perform various signal processing operations on acquired data (described below), and also stores the data onto its hard disc and floppy disc. Using the same program, the data may be read back into the analyzer at a later stage for off-line processing.

The supervisor software is a menu based program written in Turbo-Pascal revision 4.0. Its purpose it to provide the user of the system with a simple, powerful and rapid means of control of the analyzer. For instance, many of the operations performed on the data, such as digital filtering of up to 100 records with subsequent energy calculation, would' require a great number of key-presses (not to mention time) if they were executed using the analyzer's front panel controls. By merely providing the supervisor software with the basic information, i.e. which records are to be processed and in what manner, the program instructs the analyzer to automatically sequence through the data, processing and storing the results of the analysis.

The roller type transmitting transducer 1 is connected to the 600V output of the pulser/synchronisation unit 4, and the synchronisation output of this device is connected to the trigger input of the A to D converter. The roller type receiver 2 is then connected to one input of the 4-channel amplifier 5, with the gain set to 40dB (X100). The output of the amplifier is in turn fed to one channel input of the A to D converter. The analyzer itself is connected to the host computer by means of a standard IEEE cable. The analyzer, amplifier and computer are then switched on. The supervisor software held in the computer is activated and the program automatically configures the analyzer to the set up normally required for a standard scanning operation.

Once the system has been configured, the concrete surface 3 to be scanned is sprayed with water and a transducer bogie supporting the other transducer is positioned at the first scan point. The pulser is then turned on, and a reading is taken every 10 mm. The time records are then stored in the analyzer's memory after each successive reading. After the specimen has been scanned, the data are transferred to the computer's hard disc by using a file-handling sub^¬ menu.

After the data has been stored, they may be analyzed in a variety of ways by simply accessing a data analysis sub-menu for the program. This sub-menu may instruct the analyzer to - calculate simple scalar measurements on each record and present them as a single trend trace; alternatively it can instruct the analyzer to perform "brick- wall" filtering on the data and then make the above scalar measurements on the filtered data. Once the trends have been calculated, the supervisor software instructs the analyzer to plot the traces on a Hewlett Packard plotter (not shown).

With the described system, a wide-band signal ranging from near d.c. to several hundred KiloHertz, travels through the concrete. The concrete contains inclusions such as small voids and steel cables, with interface regions between the steel and the concrete. The wide-band signal rapidly loses high frequency energy due to the heterogeneous nature of the medium. Where these frequencies encounter boundary conditions, reflection will occur, the strength being dependent upon the size of the defect in relation to the wavelength. The total signal that arrives at the receiver transducer 2 will include low frequency components of large amplitude and relatively little information content, and high frequency components that do change in strength according to the internal conditions but whose energy may be of the order of a thousand times weaker than the whole signal.

For this reason, analysis of the total signal will yield little of value. Analysis requires filtering to isolate specific bands of interest. An ideal filter passes, without attenuation or phase distortion, those frequencies and only those frequencies for which it has been designed. In practice, such a filter is impossible to realise as totally sharp cut-off points are never achieved since attenuation and phase change always occur.

By contrast, it is possible to implement "ideal" filters in the Fourier domain by setting unwanted harmonics to zero and performing an inverse transform to generate a time record composed only of the frequencies of interest. This has proven very successful in identifying areas containing voids, major cable breaks and areas of delamination.

The supervisor software instructs the analyzer to convert each ultrasonic record to the frequency domain through the use of its FFT function, and to produce eight new time records of different bandwidths by filtering and inverse transforming the data. These bands are 50 kHz in width, ranging from 0-50 kHz to 350-400 kHz. The analyzer then compares the energies in these bands. The manner in which the bands are compared or combined depends on the kind of defect that is being sought; it can take the form of cross- multiplication, in which all energy bands are simply multiplied together (an operation not dissimilar to cross-correlation), and this has been particularly successful in locating moderate to large sized voids when increased reflectance of all wavelengths is to be expected. In other cases involving corrosion, small regions of delamination or complete breakage and separation of the cable, the higher frequency components only tend to display changes in energy beyond the limits normally attributable to random variability. The low frequencies, because of their longer wavelength, remain relatively unaffected. In these instances, the ratio is taken of the high frequency energy to the low frequency energy, using the latter to correct for background variability.

Figures 2 and 3 show typical results obtained using the system described above from two regions of a test beam in which paper towels, saturated in salt solution, were wrapped around the reinforcing cables prior to pouring the concrete. The scans were conducted 14 months after casting. Each scan comprised 30 readings, in this case with a scan step of 20 mm. The scans started and finished 300 mm either side of the mid-point of the defects. The vertical axis represents high frequency energy between 350 and 400 kHz, enhanced by normalizing against low frequency variability. The peaks clearly indicate the defective regions. These traces were obtained by using a method of signal processing that ratioed the high frequency energy content against the low frequency energy, thus attempting to neutralize the variations in background noise induced by the random nature of the aggregate profile, and, to a lesser extent, the variations in signal strength resulting from changes in coupling consistency.

The described system is capable of detecting, through non¬ destructive means, many defective conditions inherent to prestressed concrete structures that have hithertofore only been identifiable using invasive or partially destructive methods. It is considered prudent however for the described system to be used in conjunction with other existing systems to obtain the fullest information about the condition of steel components buried in concrete.

Tests of the described equipment have shown that the equipment is capable of identifying voids of the order of 30 mm diameter at depth of the order of 100 mm from the surface and major breaks in cables involving a total separation of not less than 30 mm at a depth of approximately 70 mm. The system is also sensitive to regions affected significantly by chloride contamination over similar depths. -Detection of single breaks in a multi-wire cable is however more difficult but this is not surprising given the frequencies used in the described equipment and the relationship between frequency and the size of defects which can be detected. It should be possible however to make the equipment more sensitive to smaller defects by extending the upper limit of the frequency range.

It is possible that the system could be implemented with a plurality of receiving transducers in order to enable tracking and comparison of signals from adjacent regions to be carried out. Thus a central transmitting transducer could be provided with four receiving transducers evenly distributed therearound.

In the described embodiment of the invention, the analysis is based mainly on calculations of the energy content of filtered signals. However, without any further modification to either the hardware or software, the system is capable of measuring many other scalar quantities relating to the filtered signals, such as amplitude, rates of decay, duty cycle and point-in-time of signal peak. By incorporating these measurements into the analysis, the accuracy, resolution and diagnostic properties of the system can be enhanced. Moreover, by including a multiple probe arrangement, the spatial accuracy of the system may be_. improved.

Claims

CLAIMS:

1. A method for detecting the presence and condition of components or defects embedded in a non-homogeneous body, wherein the body is insonated with an acoustic signal including a predetermined band of frequencies, a receiver transducer is positioned in contact with the body, acoustic signals detected by the transducer are transformed to the frequency domain to produce a frequency spectrum, the detected signals are filtered to produce output signals representing the magnitude of the acoustic signals in each of at least two discrete frequency bands, and the magnitudes of the output signals are compared to derive frequency dependent data relating to defects in the body.

2. A method as claimed in claim 1, wherein the body is insonated with acoustic waves in the band from 0 to 400 kHz.

3. A method as claimed in claim 2 wherein the 400 kHz wide band is divided into eight bands each 50 kHz in width by filtering so as to produce eight sets of data which are then inverse transformed, the energies in the eight bands then being compared to derive the said frequency dependent data.

4. A method as claimed in any preceding claim, wherein the said comparison is in the form of cross-multiplication in which all the energy bands are multiplied together.

5. A method as claimed in any preceding claim, wherein a ratio is taken of high frequency energy in a first frequency band to low frequency energy in a second frequency band, the low frequency energy being used to check for background variability.

6. A method as claimed in claim 5, wherein the first frequency band is 350-400 kHz^' and the sec nd frequency band Is 0 to 50 kHz.

7. An apparatus for detecting the presence and condition of components or defects embedded in a non-homogeneous body, comprising means for insonating the body with a wide band acoustic wave, a receiver transducer which is positionable in contact with the body to detect acoustic signals, means for transforming the detected acoustic signals to the frequency domain to produce a frequency spectrum, means for filtering the detected signals to produce output signals representing the magnitude of the acoustic signal in each of at least two discrete frequency bands and means for comparing the magnitudes of the output signals to derive frequency dependent data related to defects in the body.