There is considerable interest at the moment in inspecting pipes using ultrasound. For example water mains and the risers used in oil-wells. For the water supply industry the extent of water lost from water mains has become a politically sensitive subject in the UK. This issue of Innovation News has two articles on this subject. In a conventional test a transducer launches a plane wave into, say, water and collects an echo from the inside wall. From the signal it is desirable to measure the inside diameter of the pipe (comparing the value with what is known to be there and hence infering deterioration), to detect regions of corrosion directly, to find perforations and to assess the soil or rock conditions on the outside of the pipe. The list is ordered in terms of technical difficulty.

Cambridge Ultrasonics has the skill and equipment that can help in the design and development of a more advanced inspection system. This article and the article in our Hot Hints on the rear page illustrate the approach we would recommend. Our visualization equipment is particularly useful in developing this kind of system.

The curved surface of the interior of a pipe causes plane waves to be focused. Earlier work several years ago looking at focusing of ultrasound was motivated by wanting to know what happens when a finite amount of ultrasonic energy is focused into a small space. This field of research is still of interest in cavitation studies and sonochemistry. Focused ultrasound can make water cavitate, which is the formation of voids in the water that collapse asymmetrically, often with small jets of water penetrating them. Very high localised temperatures can be created when water cavitates, sufficient to produce scintillations of light.

However, we were interested in the caustic cusp that forms when pulsed ultrasound is focused under cylindrical aberration. Caustic cusps can be predicted by ray-tracing and the then-popular Catastrophe theory. One of the more interesting diffraction effects found around the focus is the formation of dislocations in the wave fronts. These are similar to the dislocations found in crystal planes where a closed (Burger's) circuit around the tip of the dislocation leaves a residue of one cycle or period of the wave, instead of 0 when a dislocation is not present.

How has this research been of practical benefit? What happens if plane waves are used to probe a pipe wall and the transducer is placed at the focal point of the pipe? One effect is immediately apparent when we look at the photograph shown here from a visualization experiment. Although the acoustic energy is high at the focus the phases of the waves around the focus are varying very rapidly with distance. This causes, firstly, the signal from the receiver to be very small (degraded signal-to-noise ratio), assuming the transmitter is used as a receiver, because the rapidly varying phase causes destructive interference over the surface of the transducer. Secondly, the received signal will be sensitive to small variations in geometry; in this case there will be sensitivity to vibrations of the order of 1/10th of the wavelength. The (small) signal will fluctuate greatly for geometrical variations of the order of the wavelength.

Conditions of disrupted phase are poor for making a measurement but changing the transducer design can improve performance substantially. However, it is worth noting first that conditions of disrupted phase are the norm for most applications of ultrasound and should not be considered abnormal. Phase disruption is rather like entropy in thermodynamics it usually gets worse. Textbooks often state that random scattering, which causes phase disruption, is a form of attenuation in materials like absorption; this is slightly misleading because it assumes that large aperture transducers are used exclusively for testing. Acoustic energy is not lost by elastic scattering instead destructive cancellation occurs over the receiver making it only appear that the energy is lost.

A few examples should help illustrate when phase disruption can occur:

• Gas and liquid flow-metering using X-shaped propagation paths - vortices in the flow cause random scattering and phase disruption, also diffraction around the port protecting a receiver causes phase disruption over the receiver.
• Reflection from rough surfaces - if the roughness is greater than about 1/10th of a wavelength then phase disruption becomes significant.
• Reflection of plane waves from non-planar surfaces - such as the cylindrical focusing example given above, but any surface that causes the reflected waves to be non-planar can lead to phase disruption.

How does one design a transducer to be more robust against phase disruption? The answer is simple: don't use a large aperture receiver although large aperture transmitters can still be used. What precisely is a large aperture transducer? It is a transducer with a surface dimension greater than one wavelength. To put it another way the receiver must be an omni-directional source. Point-sources have intrinsically low sensitivity and to overcome this it is usually necessary to build an array of point-source receivers in an application. However, a large number of point-sources is not needed, sometimes just a few, say six, is sufficient and it is seldom necessary to use as many as the 128-elements commonly used to steer beams in medical scanners.

It is essential to combine the output signals to remove polarity, say by taking the magnitude of the analytic signal (optimum energy detection algorithm) or more simply by rectifying the signal. The result of combining the array signals in this way is a phase-insensitive energy detector. It is possible to transmit from all of the point-sources together, making a coherent transmitter that approximates to a large aperture transmitter. If a large aperture transducer is replaced by a point-source array then the replacement may be very similar in size and overall shape. So it should be possible in many cases to up-grade existing systems. The electronic drive and receiving circuits are usually more complicated.

The benefits of using a point-source array are: they allow a more sensitive measurement to be made; they make the system more robust against vibrations; they are less sensitive to geometrical misalignments; they give better signal-to-noise ratios.

Cambridge Software Development Service is a part of Instrumentation Innovation Ltd, an independent consulting company based close to Cambridge. We have been building instrumentation and writing software for PCs since 1987. At the request of a customer we have offered pure software development service recently and as a new initiative we are offering our in-house software development to clients to assist with their projects.

The personal computer is probably the most cost-effective computer upon which to base new applications. Performance of PCs is now as good as workstations and costs are much lower.

With advanced compilers such as Visual Basic, Visual C++ and Watcom and reusable libraries such as MFC and OCX controls it is possible to reduce development times and costs significantly. Software that would have taken several man-years to write only a few years ago can be prepared by one person in a few months - that means cost savings.

Instrumentation Innovation Ltd also trades as Cambridge Ultrasonics. The company is known for developing novel ultrasonic inspection systems for concrete. It was commissioned to work on a project sponsored by the German Institute of Standards and the German Concrete Association that resulted in a PC and supporting hardware that could probe concrete to a depth of 1 m with a resolution of about 3 cm. It is hoped that a portable commercial system will be released soon based upon this prototype.
In one of our recent applications software had two tasks: to control a lock-in amplifier for collection of ultrasonic resonance spectra and to perform signal processing of the spectra. An interface to an embedded artificial neural network was also included.

Another example is a program to test industrial components on a manufacturing line. Again the program had to control plug-in boards in the host PC, process the signals collected and control the test operation. The test system was developed for a major automotive manufacturer.

At the time of writing the most popular platform for new software systems is MS Windows 95/98. Out of all the compilers we use we think Visual Basic 5.0 offers the greatest productivity and hence the lowest cost for clients. Unfortunately, it does not support hardware access, apart from standard device drivers such as filing, serial communications, graphics, keyboard, mouse. Custom device drivers are needed for control and measurement applications, using custom plug-in boards. There are many such devices available, usually in the form of DLL files that have to be included in the software. We are happy to recommend or investigate suitable hardware for specific applications. We can also develop custom interfacing circuits when necessary in support of our software service but it is nearly always cheaper to use existing, commercial products.

In issue number three of Innovation News we announced that Cambridge Ultrasonics and the Institut für Massivbau of the Technical University of Darmstadt (formerly Technische Hochschule Darmstadt) were promoting a project to develop a second generation of ultrasonic inspection equipment for concrete structures. Several other organisations expressed interest, three meetings have been held in Darmstadt and four organisations generously volunteered to support the work financially. Since then one of the sponsors has withdrawn and we unfortunately have not been able to attract sufficient additional sponsors for the project to start.

The partners would like the project to be given Eureka status and we have approached our national governments for project guidance and financial assistance. In the UK three proposals have been made to various arms of the UK government including: DETRA, Highways Agency and DTI. So far the first two have turned-down the proposals although DETRA assessed the proposal as technically very good. We are awaiting the conclusions from the DTI.

The project is known as the second generation project because it will build upon the results of two previous projects in which Cambridge Ultrasonics has taken a leading role. The best aspects of each of these successful projects will be developed to a stage where prototypes are ready for commercialization as instruments and as services in the hands of inspection service providers. The instruments resulting from the project should find application in a wide range of structures including: offshore oil production, nuclear power stations, dams, bridges, roads, telegraph poles, cat cracker linings and liquid petroleum gas confinement slabs.

We hope to announce better progress in the future and all existing partners hope the project will start later this year. For details contact: Dr David R Andrews at Cambridge Ultrasonics.

Pulsed ultrasound is generally used for measuring the elastic tensor values of crystals. But continuous ultrasonic sine wave excitation can also be used. Standing wave patterns are created in in the test sample. The method is known as ultrasonic resonance spectroscopy. The standing waves are closely related to the eigen-modes of vibration and these depend upon the elastic properties of the sample under test.

Resonance methods have a number of advantages over pulse-echo for measuring crystal elastic tensor values:

• It is less important to try to maximise the acoustic energy entering the test sample in a resonance test, simplifying experimental procedures.
• A lock-in amplifier or phase sensitive detector is used to measure the received signal and these are theoretically known to be the optimum methods for measuring the amplitude and phase of a continuous sine wave in terms of the signal-to-noise ratio.

The elastic tensor values, as derivatives of the free energy, are related to the thermodynamic properties of the sample such as the specific heat, Debye temperature, and Gruneisen parameter. Measurement of the elastic tensor allows a check to be made on a theoretical model of the sample under test. Damping of the ultrasound also provides information on coupling to relaxation mechanisms, electron mobility and anharmonicity. Variation of the elastic tensor values with temperature allows phase transitions to be explored such as superconductivity.

Samples can be tested that are as small as 100 microgrammes and as large as several kilogrammes. There are three important elements in the measurement method:

• Precise control of the test conditions such as: frequency and amplitude of the ultrasound, noise in the received signal, the effect of transducer coupling, the influence of transducer resonances, temperature, environment (any surrounding gas can damp resonances).
• Calculation of the eigen-modes in terms of the elastic tensor values, sample geometry and mass.
• Inversion of the previous calculation to obtain the elastic tensor values from the experimental data.
  

In this way it is possible to achieve accuracies better than 0.1%. A good overview of the method is given by Maynard, J,D., Physics Today 49:26-31 (1996).

The environment of the sample can present experimental difficulties. Two transducers are needed, one to inject the ultrasound the other to receive the response of the sample. Transducers have their own mechanical resonances. Every spectrum from a sample will have the spectrum of the transducers convolved with it. This can present a difficulty if peaks due to the transducer are interpreted as eigen-modes of the sample. Mechanical coupling between the transducers can also create a signal greater than the signal from the sample.

When operating at low or high temperatures one is usually forced to use coupling rods to carry the ultrasound from the transducers to the sample. A coupling rod provides, say, room temperature at the end coupled to the transducer and the temperature of the test environment at the end coupled to the sample. The spectrum of the coupling rods is an added complication but an interesting approach is to use long, thin alumina rods for high temperature experiments. These have many damped resonances, forming a smooth spectrum above which the resonances of the test sample appear. This approach relies on the sample having high Q (quality factor) values at its resonances (Q>1000), which is usually true for carefully made samples with parallelepiped shapes, low intrinsic damping and low porosity.

Other interesting approaches have been to use polarised PVdF film as the transducers. PVdF is a ferro-electric polymer that can be polarised to become piezoelectric. It is a highly damped material and because it can be made into thin films its natural resonances can be confined to a single, high frequency thickness-mode peak with, possibly, a harmonic. This resonance can be designed to be at frequencies above the experimental range for the sample. The working temperature range of PVdF is rather narrow, having a Curie temperature of about 500 K, above which spontaneous de-polarisation occurs, so PVdF is restricted to a range around room temperature.

A parallelepiped sample is usually held at two opposing corners. It is simpler to find the eigen-modes of vibration by rotating the sample slightly between tests, moving the transducers laterally to one another. It is the frequencies of the eigen-modes that are used in subsequent calculations to find the elastic tensor and not the other mixed modes also found in experimental spectra. A slight rotation of the sample affects the relative amplitudes of modes as observed in the experimental spectrum, generally leaving eigen-modes less affected than mixed modes.

Normal modes of vibration of an anisotropic sample are shown here. Copyright permission of American Institute of Physics.

Are you suffering from PDW? The symptoms are mysterious loss of signal-to-noise ratio, unexpectedly large sensitivity to mechanical vibration and small dancing or disappearing received signals! You know there is lots of energy there but where has it all gone? These are tell-tale symptoms of phase disrupted wave fronts or PDW (see first article in this issue for more details).

We are all familiar with the effect of PDW in setting the beam width of a piston transducer. The first zero happens when PDW leads to the total destructive interference of the acoustic energy over the surface of the transducer. Its not that there is no energy its just that the transducer fails to register a signal at that angle.

There is a remedy for unwanted PDW and that is to make your receiver phase-insensitive by turning it into an array. You don't need many elements although the more you have the better your signal-to-noise ratio will become. Six to ten is usually sufficient. Make sure that the size of each element is not greater than the wavelength of the ultrasonic waves and add the signals from the array together after removing the phase information (magnitude of the analytic signal is best but simple rectification can also work quite well).

Another trick if you intend to make an array is to make it a spatial matched-filter to the received waves. Matched filtering is the optimum linear filter for detecting a known signal in noise and matched filtering in the time domain is well-known. But it is possible to do the same thing spatially, where the receiver array is constructed to be more sensitive to the spatial wave profile of the waves you want to receive. For example, in the visualization sequence the wave fronts emerging from the focus form a radially expanding, circular profile. A transducer with the same profile as the waves at the same position will be spatially matched.

Thought of in this way we can see immediately that a large aperture, coherent receiver is nothing more than a coherent, matched, spatial filter for plane waves. That is why it only works well with plane, coherent waves and why PDW causes problems with large aperture, coherent transducers.

SAW filters are an example of matched filtering but in one-dimension, in which the inter-digital electrode profile is a physical representation of the time-domain signal. A SAW filter is therefore a fast, traditional matched filter, providing pulse compression. Another interesting development over the last few years is the time-reversal mirror array. Here the array generally has a flat surface but time delays are used to create a similar effect to a profiled transducer surface. In a time-reversal mirror array the flexibility given by storing the signal allows the system to adapt to any reflecting target but the price is in the size and expense of providing parallel fast acquisition, signal memory, analogue to digital conversion and amplification - each element in the array needs sophisticated electronics, so a system is expensive and bulky.

Receiver profiles or shapes matched to the waves of interest can transform performance. If you intend to make a spatially matched filter then the problem is what is the profile of the received wave? Working with pulsed wavefields creates another complication for theoretical predictions of wave profiles because pulsing introduces Green’s functions into the solution - it quickly becomes messy. Finite element solutions need very small time steps to predict profiles accurately and errors tend to get magnified; computation times can be very long indeed. At the end of any finite element calculation one always wonders, “How accurate and representative are the results?” Experimental verification is highly desirable. Experimental visualization is ideal for this application.

The accompanying figures show visualization photographs of ultrasonic waves around a focus in water. The centre frequency of the transducer was about 1 MHz. The waves were traveling from bottom to top and are shown passing through a caustic cusp focus (white dotted lines). Shown in schematic form in grey with white outlines are three hypothetical transducers. In the top figure is a traditional, coherent, large aperture transducer. In the middle figure is an array of eight point-source receivers; in the bottom figure is an array of eight point receivers curved into a profile to match the waves diverging from the focus. Which receiver will produce the best signal?

If by best we mean the sharpest pulse then the bottom transducer will be best. Sharp pulses give the best results for accurately measuring distances and in resolving between two adjacent reflectors on the acoustic path The bottom transducer should also have the greatest signal-to-noise ratio and best fidelity to resonances coming from the thickness of the pipe wall. It is these resonances that can be used to infer the thickness and condition of the pipe wall (degree of corrosion) and even the soil condition outside the pipe.

The bottom transducer does have some disadvantages. Do you know what they are?

Figure captions:

• A large aperture, coherent, traditional ultrasonic receiver sampling a focused wavefield.
• An incoherent array receiver sampling a focused wavefield.
• A profiled, incoherent, receiver array sampling a focused wavefield.