accoustic eye details

8/10/2019 Accoustic Eye Details

http://slidepdf.com/reader/full/accoustic-eye-details 1/34

TECHNICAL WHITE PAPER

PULSE REFLECTOMETRYTUBE INSPECTION SYSTEM

V.2d



TECHNICAL WHITE PAPERPULSE REFLECTOMETRY TUBE INSPECTION SYSTEM

CONTENTS

1.0 SCOPE 3

2.0 TECHNICAL CHALLENGES IN HEAT EXCHANGER TUBE INSPECTION 3

3.0 TECHNOLOGY OVERVIEW – APR + UPR 4

4.0 ACOUSTIC PULSE REFLECTOMETRY (APR) 5

A. Physical principles of APR

B. Optimizing the APR excitation for maximum SNR

C. APR signal interpretation

5.0 ULTRASONIC PULSE REFLECTOMETRY (UPR) 16

A. Physical principles of UPR

B. UPR signal excitation and measurement

C. UPR signal analysis

6.0 COMBINING APR AND UPR 26

A. Implementation of a combined probe

B. Combined signal interpretation

C. Examples

7.0 SUMMARY 33

________________________



Technical White Paper 3

1.0 SCOPE

The purpose of this white paper is to provide a technical understanding of the underlyingphysics which enable AcousticEye's (AE’s) Acoustic Pulse Reflectometry + UltrasonicPulse Reflectometry (APR+UPR) technology, how it is applied and how it addresses thekey challenges in heat exchanger tube inspection.

2.0 TECHNICAL CHALLENGES IN HEAT EXCHANGER TUBE INSPECTION

Heat exchangers operate in harsh environments of many different types. Exposed to avariety of corrosive fluids, which may also hold contaminants, their performance is prone

to two main forms of deterioration: 1) fouling, through sedimentation, scale, sludge andeven shellfish growth; 2) tube wall degradation, as the result of erosion, corrosion,thermal shock, cracks and more. The resultant damages ranges from loss of efficiency tocatastrophic failure. Nevertheless, due to their high cost, heat exchangers normally stayin service for decades, and therefore require regular maintenance to address theseissues.

Ideally, a tube inspection technique should provide several key properties: highsensitivity and accuracy are extremely important, at the same time providing a high levelof consistency regardless of the operator. System design should facilitate shortinspection time while providing objective and highly consistent data interpretation criteria,and be applicable to a wide variety of tube materials and dimensions. Finally, it should

require minimal technical knowledge and experience to implement properly andconsistently, and should require minimal pre-inspection preparation of the tubes.

Inspecting and reporting using current techniques is a lengthy process. Several suchtechniques exist today, though they all suffer from various drawbacks. The fastest, eddycurrent (ET), can inspect about 60 tubes per hour, but it is limited to non-ferromagnetictube materials. It is also heavily reliant on technician expertise – a study by MTI andEPRI has demonstrated fault detection scores varying between 87% and 50%,depending on the technician using the equipment. Ferromagnetic tubes require the useof alternative electromagnetic methods, mainly remote field testing (RFT), which isslower, less accurate and also depends heavily on the technician expertise. Finally, theInternal Rotating Inspection System (IRIS), based on ultrasound, is also accurate, butextremely slow and requires a very high degree of tube cleanliness – down to the baremetal. In addition it cannot be applied to tubes having walls thinner than 0.9 mm.

Overall, current methods are too slow to enable full inspection, and are limited withrespect to tube materials or tube dimensions. All these methods are also very limited intheir ability to inspect challenging configurations such as bent, spiral or finned tubes.This situation leaves much room for improvement in the field.

In this light, several core principles governed all the technological decisions indeveloping the combined APR+UPR system:




1. Non traversing inspection2. Short inspection time per tube3. Detection of all ID/OD faults relevant to the industry.4. Minimal sensitivity to tube material, dimensions and configurations5. Minimal dependence on operator judgment.

The following describes how the APR+UPR solution operates and how it has thepotential to meet these requirements.

3.0 APR + UPR TECHNOLOGY OVERVIEW

The APR+UPR system is a natural evolution of AcousticEye's already well-established Acoustic Pulse Reflectometry (APR) tube inspection technology. The principle behind APR is to inject an acoustic pulse into the air enclosed by the tube, which propagatesdown the tube axis. Any changes in the internal cross-section (caused either by wall lossor obstructions) create reflected waves which propagate back up the tube. Thesereflections can then be measured and analyzed, as discussed in detail in the followingsection. This non-traversing technology is extremely fast and has been successfullydeployed for several years. However it is inherently limited to identifying InternalDiameter (ID) defects and is blind to defects on the Outer Diameter (OD). The combined

APR+UPR system was therefore conceived to overcome this drawback and also widenthe range of detectable defects. The system employs two complementary technologiessimultaneously: the existing APR technology along with the newly developed UltrasonicPulse Reflectometry (UPR) technology.

Ultrasonic Pulse Reflectometry (UPR) is a variant of the well-known Guided Waves(GW) technique. In contrast to APR, it involves injecting ultrasonic waves into the tubewall via an array of dry coupled transducers placed just inside the tube end. It istherefore non-traversing, similarly to APR, and very fast. UPR technology has the abilityto detect OD fretting, ID/OD axial and circumferentially oriented cracks, ID/ODcorrosion/erosion, or ID/OD pitting along the entire length of a tube regardless of itsconfiguration, i.e. twisted, u-bends, spiral, etc.

By combining APR and UPR technologies a system was conceived which could detectboth OD and ID defects simultaneously at the same high speed and efficiency as the

APR alone. A general overview describing the capabilities and characteristics of APRand UPR separately is presented in Table I.




Table I : A general overview of APR and UPR capabilities

ADVANTAGES SHORTCOMINGS APR Fast inspection time: 5 to 10 seconds

per tubeDetects only ID defects

Independent of tube material Requires a relatively high level of cleaningHigh sensitivity to through holes Less sensitive to small pits than currently

conventional methodsHigh sensitivity to blockages Cannot detect cracks

Provides no circumferential information

UPR Detects both OD and ID defects Cannot distinguish between OD and ID defectsFast inspection time: 5 to 10 secondsper tube

More complex to implement: requires moreadvanced hardware and signal processing

Can detect cracks, regardless oforientation

Has difficulties in distinguishing deep pits fromholes

High sensitivity to pits Cannot detect blockagesProvides circumferential data

The next sections present each component technology in detail, followed by adescription of how they are combined, both from the viewpoint of implementation andsignal analysis.

4.0 ACOUSTIC PULSE REFLECTOMETRY (APR)

A. Physical Principles of APR

Acoustic Waves in Tubes

Acoustic waves in air are longitudinal waves: particle velocity is parallel to thedirection of wave propagation. In free space, acoustic waves can propagate in alldirections. However, in a confined space, such as a tube whose transversedimensions are small with respect to the minimal wavelength, such waves willpropagate solely along the tube axis.

Up to a certain "cut-on frequency", a wave propagating in a tube can beconsidered a plane wave, i.e. wave fronts are flat and the pressure fluctuations areuniform over the cross section of the tube. This kind of wave is the mostconvenient wave to measure, since it suffices to measure it at one point in thecross section. Most commonly this is performed by a microphone embedded in thetube wall so as not to create a disturbance in the tube. Above the cut-onfrequency, higher order modes of propagation are excited. These modes havedifferent wave velocities, and in addition when they occur the pressure is no longeruniform over the cross section. It is difficult both to excite these modes in acontrollable manner and also to measure them, therefore they are usually avoided




in APR systems. The plane wave mode of propagation is also referred to as thelowest order mode.

The cut-on frequency of the first higher order mode is given by the equation:

(1) d c /84.1

where c is the speed of sound in air, and d is the inner diameter of the tube. Thisfrequency is determined by tube diameter and the speed of sound, becoming loweras the tube becomes wider. To avoid the complications created by higher ordermodes, APR systems are usually designed to create an excitation signal that islimited to a maximal frequency that is below this cut-on.

Attenuation of Acoustic Waves in Tubes

Acoustic waves propagating in a tube will experience attenuation due to friction atthe tube wall. The equations governing attenuation as discovered by Kirchoff andlater formulated in more mathematically tractable approximations by Keefe (1984),show attenuation to be dependent mainly on the ratio of wavelength to tubediameter. Attenuation increases with frequency, therefore a wideband pulseexcited at one end of the tube will gradually lose its high frequency content,becoming gradually more smeared in the time domain, as shown in Figure 1.

Figure 1 : A wideband acoustic pulse as excited, and after propagating down successively longerlengths of tubing

Today most measurements on acoustic waves are digitized for storage andprocessing in computers, therefore it is most convenient to describe theattenuation through a digital filter rather than in the more classical form of

2500 3000 3500 4000 4500

-0.2

-0.1

0

0.1

0.2

0.3




differential equations. Amir et al (1996) were the first to provide a formulation ofattenuation as a digital filter, which is now commonly used in APR applications.

Defects as Discontinuities in Tubes

Acoustic waves within a uniform tube will propagate down the tube, experiencingonly the gradual attenuation described above. However, any internal change incross section of the tube will split the wave into two components: a reflected and atransmitted component. Several types of cross-sectional change can occur: anincrease in cross section due to wall loss, a through hole, and a reduction in crosssection due to full or partial blockage. When dealing with the lowest order mode,only the change in overall cross section has an influence on the reflected andtransmitted wave, regardless of the particular shape. In the case of a hole, forinstance, the area of the hole determines the reflection, whether it is round orelongated. The same holds for reduction in cross section – whether it is localizedor uniformly distributed over the circumference of the tube has no importance.

The reflection and transmission caused by an abrupt change in cross section canbe modeled easily through the reflection and transmission coefficients. Given awave propagating down a tube with cross section S1, which then encounters atube with cross section S2, the reflection coefficient R is given by:

(2)21

21

S S

S S R

And the transmission coefficient T by:

(3)21

12

S S

S T

From (2) it can be seen that an increase in cross section (S2>S1) causes anegative reflection, whereas a decrease in cross section (S2<S1) causes apositive reflection. In heat exchanger tubes, typical defects such as blockages and

wall loss cause local changes in cross section, as shown in Figure 2. A typicalblockage, for example in Figure 2(a), will be composed of two successivediscontinuities: a reduction of cross section at the beginning of the blockage, andan increase back to the nominal cross section where the blockage ends. A wallloss defect, as shown in Figure 2(b) is the opposite: an increase in cross sectionfollowed by a decrease. Furthermore, the amplitude of a reflected pulse isdetermined by the value of the reflection coefficient R, thus it can be used todetermine S2 if S1 is known.




Figure 2 – Schematic depictions of local cross section changes caused by (a) blockage and (b) wallloss defects

From equation (2) and Figure 2 it can be inferred, that the reflections fromblockage and wall loss defects will have typical signatures. Assuming a positivepulse is sent down the tube, when it encounters a blockage it will cause first apositive reflection followed by a negative one, whereas a wall loss defect will causethe opposite: a negative pulse followed by a positive one. These are shownschematically in Figure 3.

Figure 3: Schematic depictions of typical signatures corresponding to the defects in Figure 2

Finally, a hole in the tube wall can be regarded as a small branched tube openinginto the atmosphere. The equations describing the effect of such a hole have beendiscussed in detail in several academic papers (Sharp and Campbell, 1997). The

Impinging pulse

Reflection from alocal blockage

Reflection from a local wall loss

Reflection from ahole




result is illustrated in Figure 3, showing also the typical signature of a hole: anegative pulse that starts as a sharp reduction in pressure followed by a long tail.

To summarize, it becomes evident that different defects commonly found on theinternal surfaces of heat exchanger tubes create characteristic reflections. Thebasic idea of APR is therefore to excite a pulse at one end of the tube and recordthe ensuing reflections, then interpreting them in order to determine the types andsizes of the defects that caused them.

Pulse Bandwidth and Its Effect on Defect Resolution

The pulses in the above figures represent a schematic picture of the excitationpulse. To enhance axial resolution of an APR system, an ideal pulse would be asnarrow as possible. This is because it is necessary to distinguish reflections fromclosely spaced defect. If for example the pulse is too wide, the positive reflectionfrom the beginning of a blockage might merge with the negative reflection from theend of a blockage, in effect cancelling each other to a large degree. Basic Fouriertheory tells us that to obtain a narrow pulse in the time domain, its spectrum mustbe as wide as possible. The effective limit on the bandwidth is determined both bythe loudspeaker creating the pulse and on the first cut-on frequency in equation(1). In the AcousticEye system, the spectrum of the pulse extends to about 8kHz,

corresponding to the inability to distinguish between defects separated by lessthan 2cm. This limitation on the axial resolution is difficult to overcome in any APRsystem. In addition, resolution gradually decreases down the tube due toattenuation, since higher frequencies decay more rapidly than lowerfrequencies.Thus the reflections from further defects become smeared in the timedomain.

Ensuring a large bandwidth may still not be sufficient. Whereas a pulse with a trulyflat spectrum will have a tall narrow structure in the time domain, an excitation withlarge bandwidth but irregular phase and amplitude characteristics may suffer fromexcessive ringing and smearing. Such characteristics maybe be imposed by the

frequency response of the driving loudspeakers, and are hard to avoid in "openloop" implementations. The AcousticEye system, however, remediates this to avery large degree using proprietary pulse shaping techniques. This is performedduring system setup: before initial use, a test measurement is performed on aknown, defect free tube. The characteristics of the raw pulse are recorded andcarefully analyzed. Subsequent pulses are shaped to compensate for loudspeakerirregularities, resulting in a large improvement in their properties.




B. Optimizing the APR Excitation Signal for Maximum SNR

In a practical APR system, as in any physical system, background noise willalways be present. The common measure for quantifying the disturbance causedby noise is SNR – Signal to Noise Ratio, which is simply the RMS of the signaldivided by the RMS of the noise. SNR is usually quantified in decibels, or dB:

(4) )log(20][noise

sig

A

AdBSNR

To improve SNR, the average signal amplitude must be increased as far aspossible, though there are practical constraints on the attainable value. Increasing

pulse width is one option, though as seen above, this has a detrimental effect onresolution. Another option is to increase pulse height, though the amplifier andloudspeaker capabilities limit this option too.

One method that has been used mainly in research setups is to carry out ameasurement on a single tube repeatedly, and average all the resultingmeasurements. Averaging does not affect the informational part of the signal, but ifthe noise in successive measurements is incoherent (random) it will result in alower noise level. SNR increases in proportion to the logarithm of the square rootof the number of averages, thus performing 100 measurements and averagingthem increases SNR by a factor of 10 (20dB). Since the computation is

logarithmic, increasing SNR by 40dB requires 10,000 repetitions, which is clearlyimpractical: typically, sending a pulse and waiting for the reflections to die outtakes 1/10 of a second, thus 10,000 repetitions would result in a totalmeasurement time of 1000 seconds, almost 20 minutes, per tube.

AcousticEye's APR implementation employs another method that combines theadvantages of repeating a measurement multiple times, yet nevertheless keepsmeasurement time down to a few seconds. This method is based on the use of asignal called a "Maximal Length Sequence" (MLS), a form of pseudo-noisecomposed exclusively of the values +1 or -1. The theory behind MLS sequences iswell known and used also in other applications. An MLS sequence is always oflength 2 N-1, where N is an integer. For example, if N=10, the sequence will be1023 samples long, taking up only 23 thousands of a second to transmit, at atypical sampling rate of 48 kHz. For N=14 the sequence will by 16,384 sampleslong, taking about 1/3 of a second to transmit. The value of N can be selected insoftware in the AcousticEye system. Typically it is set to 13, and the measurementrepeated several times, giving a total measurement time of approximately 10seconds.

Extracting the pulse response from the measured MLS signal requires acorrelation computation. Mathematically, it is a linear operation, and thus anynonlinear distortions in the system will create spurious peaks in the resultant




signal, which could be misinterpreted as defects. Evidently it is very important tokeep nonlinear distortions to a minimum. The component most susceptible to suchdistortions is the loudspeaker, which becomes nonlinear when driven at highamplitudes. Therefore there are two conflicting demands on the excitation signal:on the one hand, it is beneficial to increase signal amplitude in order to increaseSNR, yet on the other hand, increasing it too much leads to nonlinear distortions.To strike the correct balance, upon startup the AcousticEye system performs ascan on a range of signal amplitudes. The optimal amplitude is the one at whichthe nonlinear noise and background noise balance to achieve the highest overallSNR.

C. APR Signal Interpretation

After acquiring the measurements it is necessary to analyze them carefully in orderto extract all the available information regarding defects and tube condition.Interpretation currently remains one of the major challenges for most existing tubeinspection methods, with a heavy emphasis on visual analysis of each and everysignal by a skilled technician. This is both subjective and time consuming, leavingmuch room for improvement through automatic analysis methods and otherprovisions for streamlining the process.

In contrast to measurement acquisition, which is constrained mainly by the

physical principles of APR technology, the quality of signal interpretation isdetermined largely by two factors: 1) the capabilities of the algorithms andheuristics developed to automate the process, and 2) by the degree to which thesoftware interface facilitates this process. Both of these are described below.

Automatic Interpretation

Automated signal analysis is relatively new to the NDT field, and since there aremany potential benefits to performing this task automatically AcousticEye hasinvested heavily in this procedure. Automatic signal analysis can speed up theanalysis considerably, in addition to rendering it more objective. In contrast,traditional tube inspection methods are based on visual analysis, which issubjective and very taxing when carried out on a large number of tubes. However,it is critical to develop a highly accurate automatic analysis in order for it to beaccepted in the field. AcousticEye has developed a two pronged approach toaddress this issue: First, a large amount of R&D resources have been allocated tooptimizing the algorithms and heuristics behind this process. To maintain acompetitive edge, a large part of this methodology remains necessarily proprietary,however it has been verified over a large number of measurement sessions and isbeing constantly improved upon and updated. A general outline of the process ispresented below. Second, the software interface has been designed to present theresults as clearly as possible so that a technician can screen the results rapidlyand accurately, employing a multitude of on-screen visual aids. This is also

described below.




Signal interpretation is composed of four stages:

1. Preprocessing and registration

2. Detection

3. Classification

4. Sizing

Preprocessing and registration : The raw measurements recorded by themicrophone contain a mixture of reflections caused by structural properties of thetubes and by any defects that might be present. It is imperative to distinguishbetween the two types of reflections in order to report only true defects. Forbundles of straight tubes, all tubes are identical, therefore any reflections fromstructural features will be identical in all the tubes. These reflections originatemainly from the discontinuity between the probe and the tube, from expansionsdue to rolling into tube sheets and baffles, and from the far ends of the tubes. In u-tube bundles there is an added layer of complexity: first, there are reflections fromthe weld between the straight and U segments, and in addition tubes from differentrows will different overall lengths, since normally each row will have U segments ofdifferent lengths. However, if each row in such a bundle is treated as a separate"mini bundle", all the tubes in such a mini-bundle will once again be identical.

The most straightforward method to remove features that are common to all the

tubes is to subtract from all measurements a prototypical measurement taken on atheoretical clean tube having no defects or noise. In this sense APR is applied asdifferential method, evaluating the deviations in each measurement from themeasurement on such a pristine tube. However, obtaining such a measurement inthe field is clearly impossible, since no single tube in a given bundle can beassured to be free of defects or fouling. The approach taken in the AcousticEyesystem is statistical: averaging over measurements from a large number of tubes(in practice, more than 30 tubes is considered sufficient) will cause the deviationsto average out, resulting in a signal that is very closed to what would be obtainedon the pristine tube mentioned above. The averaging process is somewhat morecomplex than a pure algebraic average, since large defects in even one tube mightstill bias the result. A proprietary process to remove outlying values in thisaveraging process ensures that such effects do not occur. Figure 4 demonstratethis procedure. Part (a) of the figure shows the raw measurement including theoriginal pulse, reflections from a U section in the middle, and the reflection from theend of the tube. Part (b) shows the same measurement after the reference valuehas been subtracted, on the same scale. The reflections caused by structuralfeatures are reduced considerably.




Figure 4 : A raw measurement (a) and the same measurement after subtracting the reference (b)

As shown in Figure 5, zooming in vertically on the same measurements leads totwo conclusions: 1) some remnants of the structural features are visible, resultingin a small decrease of sensitivity in their vicinity; 2) two defect signatures atapproximately 3.6m and 4.6m become clearly visible on this scale, whereas theywere previously almost indistinguishable.

Figure 5 : The signal from figure 5(b), after zooming vertically

Defects




An additional element in the preprocessing stage is "registration." Smalldifferences in tube length or apparent tube length (due to fluctuations intemperature, affecting the speed of sound), can cause tube ends across thebundle to be misaligned. Subtracting the reference in such a case might causesome tubes to exhibit large spurious peaks in the vicinity of the tube end, whichcould appear as faults. To avoid this, all measurements are stretched or contractedslightly to precisely the same length.

Detection : The main challenge in the detection phase is to decide which featuresof a given signal represent actual defects, as opposed to random fluctuations dueto ever-present background noise. Several factors can contribute to this noise:ambient noise, internal noise and fluctuations caused by reflections off residualfouling and tube surface roughness. Regardless of the source, it is necessary todetermine the actual background noise level and use it to determine a threshold ofdefect detectability, which we term here the "noise threshold." Calculating thisthreshold is performed by carrying out a statistical analysis over the entireensemble of measurements. Any reflections crossing this threshold are consideredto represent defects. This is demonstrated in Figure 6, where 20 superimposedmeasurements are shown. Clearly, most of the signals fluctuate close to thehorizontal axis, while some of them exhibit large peaks which appear to representdefects. The noise threshold, drawn in both positive and negative sign on thisfigure (the thick red lines) makes it possible to distinguish random fluctuations fromdefects. It is noteworthy that the noise threshold varies with distance along thetube, mainly due to reflections from residual fouling which is not necessarily

uniform. Finally, the statistical analysis used to determine the noise threshold canbe carried out in several ways. The simplest is to calculate the standard deviationacross the ensemble of measurements at each point along the tube, howevermore complex methods can be used.

Figure 6 : A group of 20 superimposed measurements of, along with the noise thresholds(heavy red lines)

It is important to stress that the single most important goal of the detection phaseis to ensure that any possible defect is flagged. This is crucial in order to enablethe technician to skip the lengthy visual analysis of all the signals over their entire




length (as performed in eddy current, for example), focusing instead only on thesignatures flagged by the software.

Classification : The second stage of the analysis is defect classification. Peaksthat extend beyond the noise threshold are classified by comparing them tosignature templates derived from the schematic examples shown in Figure 3. Thisprocedure is complicated by the fact that there still remains a large degree ofvariability in reflection shape due to variations in axial length of the defects, forexample, or irregularity in defect morphology.

The difficulties encountered in this phase are usually related to the degree ofcleanliness of the tubes being detected. Excessive debris and fouling can create amultitude of spurious reflections that can interfere with the reflections off defects,especially small ones. As in classification, the heuristics behind the algorithmicclassification routines are the result of a large R&D effort combining informationfrom many different measurements in the lab and in the field. The details of thismethodology remain proprietary.

In applying APR, as in any NDT technology, tough calls can occur. As long as thenumber of such cases can be kept marginal, the best policy for dealing with themis to flag them and bring them to the operator's attention, rather than forcing themto fit into one of the existing categories.

Sizing : Acoustic theory enables accurate simulation of all defect types detectableby an APR system. Wall loss and blockage signatures can be calculated based on

equation (2), while holes can be simulated based on the works of Sharp et al.(1997). The idea behind sizing is therefore straightforward: after defect signaturesare detected and classified, they are matched to signatures derived from thetheoretical simulations. This matching process can achieve a high degree ofaccuracy, as demonstrated here in several figures. Figure 7 shows a measurementof an ID pit created artificially (blue line) together with the simulation of this pit (redline) – both curves are nearly identical. Figure 8 shows the same comparison of anactual and simulated through-hole having a diameter of only 0.5mm. Such smallholes are nearly undetectable by any other tube inspection technology.

Figure 7 : Actual and simulated signature of an ID pit




Figure 8 : Actual and simulated signature of a through-hole

The major challenge in applying this methodology is speeding up the searchprocess. Typical defects can have one or more degrees of freedom, thus carryingout an exhaustive search to determine the sizing parameters can be timeconsuming. Various proprietary methods have been developed to speed up thisprocess considerably, though the basic methodology of sizing remains the same.

A final stage of visual verification is often desirable, and is fully supported by thesoftware. This issue is discussed in detail at the end of section 6, where thecombined APR+UPR signal interpretation is presented.

5.0 ULTRASONIC PULSE REFLECTOMETRY (UPR)

The first part of this paper has shown that although APR has many advantages ascompared to other tube inspection techniques, it inherently suffers from two mainlimitations: insensitivity to OD defects and limited accuracy in sizing wall loss defects.To provide a comprehensive solution that retains the speed and ease of use of APR, it ismore practical to employ a complementary technology, rather than start from a cleanslate looking for a totally alternative technology. The most natural choice is to retain thebasic approach of pulse reflectometry, by sending ultrasonic pulses in the tube walls inaddition to the acoustic pulses within the tubes. This is referred to in this document asUPR – Ultrasonic Pulse Reflectometry. This method is also referred to in the literature as"guided waves." It has been applied previously mainly to pipeline inspection rather thanheat exchanger tube inspection. Though tube inspection applications of guided wavesdo appear in the literature, this method on its own is also limited and therefore does notappear to have met with much success. However, with properly designed hardware, thetwo types of pulses (acoustic and ultrasonic) can be generated and recordedindependently, with no sacrifice in inspection time. The capabilities of these two methodsthen provide a superset that is comparable or superior to what is found in most othertechniques.




A. Physical Principles of UPR

Ultrasonic Waves in Tube Walls

UPR employs elastic waves in the solid tube wall. These behave differently fromwaves in air used in APR. The fundamental difference is that waves in air and anyother fluid are necessarily longitudinal, meaning that particle velocity is alwaysparallel to the direction in which the wave is propagating. In solids, the direction ofparticle velocity is not constrained. This adds a large degree of complexity to boththe excitation and analysis of waves in solids. However, when this complexity isharnessed properly for NDT applications, such waves can provide information that

is not accessible to APR.In general, any waves propagating in a medium having one dominant dimensionare termed guided waves. In this sense APR also employs guided waves,propagating in a long and narrow cylinder of air enclosed by the tube wall. UPR, onthe other hand, propagates waves in a cylindrical shell , having both an internal andexternal boundary.

As mentioned in the previous section, above a certain frequency acoustic waves inair can propagate as higher order modes. These are difficult to control andmeasure, and are therefore avoided in most APR applications. In very generalterms, a similar distinction can also be made in ultrasonic waves. Ultrasonic wavesbelow a given frequency will have uniform properties across the cross section ofthe tube wall. When implementing UPR using only such frequencies, both ID andOD defects will be detected, but no circumferential information will be available. Asin APR, this compromises the ability to accurately assess the depth and extent ofdefects. However, in contrast to APR, higher order modes can be generated andmeasured more reliably in solids, at the cost of adding considerable complexity tothe system.

Ultrasonic Wave Mode Families

Since elastic waves propagating in a cylindrical shell are not necessarilylongitudinal, it has been shown [Ref – Rose] that such waves can propagate inthree distinct families of vibrational modes: torsional, longitudinal and flexural.These are often denoted as T, L and F families, respectively. More recent studiessuch as the study by Sun et al. (2005) employ a more convenient notation in whichthe F family is subdivided into F L and F T families, since each group of flexuralmode is related to either a single torsional or longitudinal mode. Generally, each ofthe above families is comprised of an infinite number of modes (the F modes aredoubly infinite), which can be summarized as follows:




T(0,m) – where m=1,2,3….

F T (n,m) – where n=1,2,3… and m=1,2,3….

L(0,m) – where m=1,2,3….

F L(n,m) – where n=1,2,3… and m=1,2,3….

"Dispersion curves" are very useful in describing the properties of the differentmodes. Specifically, these show the phase velocity of each mode at eachfrequency (Note: an alternative representation is group velocities, which are simplythe derivatives of the phase velocities). A typical example of phase velocities in apipe (Schedule 40 steel, OD: 88.9 mm, WT: 5.49 mm) is shown in Fig. 9, taken

from the paper by Sun et al (2005). This figure shows how the flexural modes aregrouped around each torsional or longitudinal mode, converging asymptotically tothe same phase velocity.

Figure 9 : Example of dispersion curves for a Schedule 40 steel pipe- OD: 88.9 mm, WT: 5.49 mm

One of the main challenges in applying UPR is determining which combination ofmodes and frequencies should be excited in order to obtain the most informativeand unambiguous results. This is performed by examining the constraints imposedby the physical layout of the transducers in conjunction with the above dispersioncurves. Putting aside the F modes for the moment, the first constraint is to choosemodes which can detect the entire range of expected defects. In this respect, Tmodes have the advantage over L modes, in that they can detects axial cracks,which L modes cannot. The second constraint is to choose only modes in whichthe particle displacement is uniform over the thickness of the tube wall. This is




because the transducers that record these waves are located on the tube surface,and are unable to detect displacement throughout the depth of the wall. Thereforeit would be impossible to distinguish between modes having the samedisplacement on the tube wall surface but different displacements further down.For T modes, this translates to exciting only the T(0,1) mode. Putting aside the Fmodes for the moment, Figure 10 shows the dispersion curves for the T and Lfamilies as a function of m in a tube having a diameter of 20mm, wall thickness of2mm, made of stainless steel. T(0,1) can be excited from DC up, and theconstraint above limits the available bandwidth to the spectrum between 0Hz andthe cut-on frequency of T(0,2). From the figure it becomes obvious that using the Tmode makes the spectrum from 0Hz to nearly 0.8 MHz available. Generally, thecut-on frequency of T(0,2) is determined mainly by wall thickness: as wall

thickness increases, the cut-on frequency becomes lower.

Figure 10 : Dispersion curves for symmetric modes only: stainless steel, OD 20mm and WT 2mm. Arrow: available bandwidth for torsional mode T(0,1).

In the above example, utilizing the full available bandwidth of the basic T(0,1) modecorresponds to an axial resolution on the order of 6.5 mm. However, this is uniformover the circumference of the tube, therefore of limited usefulness on its own. This iswhere the flexural modes can be applied. Figure 11 shows the T(0,1) mode with itsassociated flexural modes. These are not axisymmetric, thus providing successivelymore detail over the circumference. Specifically, since m is constrained here to be 1,the flexural modes shown are F T(n,1) – where n=1,2,3…A complicating factor whenusing these modes in a UPR system is that the group velocity for each mode ateach frequency is different. This issue is discussed further below.

0 1 2 3 4 5 6 7 8 9 10

x 105

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

Frequency [Hz]

P h a s e v e

l o c

i t y [ m / s ]

L(0,2)

L(0,1)

T(0,2)

T(0,1)




There are two constraints on the number of flexural modes that can be utilized. First,since the cut-on frequency of T(0,2) imposes a global maximum on the bandwidth,only flexural modes having cut-on frequencies below this maximum can be used.This is determined mainly by tube radius: the larger the radius, the lower the cut-onfrequencies of the flexural modes. The second constraint on the available number ofmodes is the number of sensors: as a rule of thumb these two are equal. In a UPRimplementation having 12 circumferential sensors, there are 11 flexural modesavailable; the 12 th flexural mode is indistinguishable from the T(0,1) mode. Thus, inthe example shown in Figure 11, the overall usable bandwidth is limited by theflexural modes to approximately 700 kHz.

Figure 11 : Dispersion curves for T(0,1) and the associated F T(m,1) modes only: stainless steel,OD 20mm and WT 2mm

To summarize, in a practical application of UPR for tube inspection, the modes to beemployed are T(0,1) and F T(n,1), n=1,2,3…with the range of feasible values for n determined by tube diameter and the number of sensors. Furthermore, to avoid anyambiguity in signal interpretation, it is important to generate and measure only thesemodes, to the extent that this is possible. Any other spurious modes will beinterpreted as noise. This type of noise is typically referred to as "coherent noise",since for a measurement under given conditions it is caused by the excitation itself,and therefore reproducible, rather than random.

0 1 2 3 4 5 6 7 8 9 10

x 105

0

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000

Frequency [Hz]

h a s e v e

l o c

i t y m s

T(0,1

T(0,2

FT(0,1-12)




Attenuation of Ultrasonic Waves in Tube Walls

Ultrasonic waves experience attenuation while propagating in the tube walls,similarly to acoustic waves propagating inside the tube. However, in the frequencyrange typically used by UPR, attenuation per unit distance is approximately an orderof magnitude less than the attenuation of the APR signals. Therefore, over the rangeof lengths of heat exchanger tubes attenuation has a minor influence on defectdetectability when using UPR.

Defects – Discontinuities in Tube Walls Revisited

The basic principles of defect detection using UPR are similar to those used in APR. As long as the tube wall is uniform along its length, the ultrasonic waves will simplypropagate down the tube. As in APR, any discontinuities will cause reflected wavesto propagate back up the tube, however a certain defect may appear very differentlyto APR and to UPR. This merits a more detailed discussion of each separate type ofdefect. Below we discuss the reflections created by various defects whenconsidering only the T(0,1) mode, which is fully axisymmetric. Incorporating higherorder modes into the analysis is discussed in a later section.

Wall loss: to APR, wall loss appears as a local increase in cross section, whereasfor UPR it appears as a decrease in cross section of the wall itself. Thus, an ID wallloss defect will show up in APR as a negative pulse follow by a positive one (asshown in Figure 3 above), whereas in UPR the polarities will be reversed. Moreimportantly, the amplitude of the UPR reflection will be relatively larger. This isbecause the cross section of the wall itself (the medium for UPR waves) is muchsmaller than the cross section of the air enclosed in the tube (the medium for APRwaves). Considering a typical wall loss defect such as a 50% deep pit with a 5mmdiameter in a 1" tube with 2mm wall thickness (Figure 12): At the pit's widest point, itdecreases the wall cross section (which is what UPR detects) by 6.5%, whereas itincreases the internal cross section (which is what APR detects) by only 1.2%.Thus, though the basic detection methodology is similar in both techniques, thepolarity of the peaks is reversed, and we can expect UPR to be more sensitive towall loss defects than APR. In addition, UPR is also able to detect OD defects. Onits own, however, it will not be able to distinguish OD from ID defects, but this canbe established by cross referencing APR and UPR detection results.




Figure 12 : A 5mm ID pit, 50% deep, in a 1" tube (not drawn to exact scale)

Blockage: As shown above, blockages cause a reduction in cross section and showup in APR signals as a positive pulse followed by a negative one, nearly regardlessof the material causing the blockage – scale, debris, sludge or even pooled water.This is because the characteristic impedance of all these materials is much largerthan the characteristic impedance of air. In UPR, however, the extent to which ablockage will show up depends on how firmly it is attached to the tube surface andon its characteristic impedance relative to that of the tube wall. Generally speaking,a hard blockage rigidly attached to the tube will cause a local increase in crosssection, showing up therefore as a negative pulse followed by a positive one. Thissignal cannot be used to estimate the size of the blockage reliably, however APR onits own is very accurate in this respect.

Through holes: to UPR, a through hole is simply a more extreme case of wall loss,to be distinguished from a pit only by its depth. Even a slight error in estimating thisdepth can result in wrongly flagging a through hole as a deep pit. In APR, however,as discussed above, there is a large qualitative difference in the signature of athrough hole vs. the signature of wall loss. Therefore in a combined system, APR isused to disambiguate holes from pits.

Cracks: Hairline cracks present a discontinuity only in the wall itself, therefore theyare detectable by UPR but not by APR. The reflections they create depend on theirorientation. Considering the extreme cases, a circumferential crack causes a largedisruption in cross section over a short axial distance, whereas an axial crackcauses a small disruption in cross section over a large axial distance. Both of thesecan be detected by UPR, and details of the reflections such as peak heights andinter-peak distances must be used to correctly diagnose this type of defect.




Upper Bounds on Defect Resolution

Axial resolution is determined by the shortest wavelength, i.e. highest frequency, inthe excitation signal. As shown above, the optimal implementation of UPRconstrains us to excite only the T(0,1) mode and the associated F T(n,1), n=0,1,2,…modes (The practical issues of how to ensure that only these modes are excited arediscussed in a later section). The T(0,1) mode is non-dispersive and can thereforebe excited at any frequency. However, to avoid exciting T(0,2) we must stay belowits cut-on frequency. This frequency depends mainly on the bulk speed of sound inthe tube and on the tube wall thickness. It is therefore different for different tubematerials and dimensions, however a range of typical values for heat exchangertubes is shown in Table II.

Table II: Axial resolution limitations in several typical heat exchanger tubes

OD WALLTHICKNESS

MATERIAL APPROXIMATE T(0,2)CUTON FREQUENCY

SHORTESTWAVELENGTH

3/4" 0.065" AISI 304 steel 1 MHz 5.14 mm3/4" 0.083" AISI 304 steel 0.8 MHz 6.43 mm3/4" 0.109" AISI 304 steel 0.6 MHz 8.57 mm

The axisymmetric T(0,1) mode in itself provides no circumferential resolution. Thisresolution is provided by the F T(n,0) modes, which are non-axisymmetric, havingprogressively more nodes on the circumference.. Observing Figure 8, it can be seenthat the cut-on frequencies of these modes are successively higher as n increases.

The cut-on frequencies of these modes depends mainly on the bulk speed of sound,but in contrast to T(0,2) they depend on the diameter rather than the wall thickness.Theoretically we would wish to excite as many such modes as possible, howeverthe constraints on this number are partly theoretical and partly practical:

1. The cut-on frequency of the highest F T(n,1) used must remain below thecut-on frequency of T(0,2).

2. To excite the n'th mode, n separate transducers are required over thecircumference of the tube. Practical limitations dictate how manytransducers can be fitted onto the circumference, especially when they areapplied on the ID. AcousticEye's current implementation employs 12 suchtransducers.

Assuming all 12 possible modes can be excited, the circumferential resolution(measured by the azimuthal angle) will be 360/12, i.e. 30 degrees. In some rarecases, particularly narrow tubes with thick walls, not all these modes will beavailable.

Overall, the axial resolution of UPR is generally better than that of APR, the issue ofattenuation is less significant and it provides circumferential resolution lacking in

APR.




B. UPR Signal Excitation and Measurement

Looking back once more to APR for comparison, exciting and measuring theacoustic waves in the air inside the tube is relatively straightforward: it can beperformed using loudspeakers and microphones found in widespread production.However the equivalent task in UPR is much more complex, for several reasons.First, the higher order modes necessary for obtaining circumferential resolutioncannot be excited or measured by a single transducer, therefore an array oftransducers is necessary. Second, the transducers must be coupled to the tube wall,an issue which is non-existent in creating acoustic waves in air. Finally, generatingthe correct signals in each of the transducers in order to excite the required modesin a controlled manner is also challenging. Each of these issues is discussed below,though some proprietary aspects are not fully disclosed.

Transducer Selection and Coupling

Two basic types of transducers can be employed for exciting ultrasonic waves in thetube wall: magnetostrictive and piezoelectric. The magnetostrictive effect has beenknown for a long time, however the ability to use magnetostrictive transducers fortube inspection appears to be limited by a wide range of patents owned by theSouthwest Research Institute in San Antonio, Texas. This type of transducerappears to suffer from other limitations which are not discussed here, since

AcousticEye chose not to pursue this transducer type at all.

The transducers employed in AcousticEye's UPR system are piezoelectric. Thesetransducers are employed in many different applications of ultrasound, from NDT tomedical. The piezoelectric effect is well known and has been exploited for manydecades, having two reciprocal aspects: 1) applying an electric field to certainmaterials results in a mechanical strain; 2) applying a force to the same materialswill create a voltage differential. Therefore piezoelectric transducers can be usedboth for exciting and for measuring the acoustic waves. Specifically, in order togenerate the torsional modes outlined above, "shear mode" piezoelectrictransducers are used, as opposed to "compression mode" transducers used inmany ultrasound applications.

The first problem in transferring the mechanical energy from the transducer to thetube wall is the issue of coupling. In medical ultrasound applications, for instance, acoupling gel acts to match the impedance between the transducer and the tissuebeing examined. Clearly, in tube inspection this is not at all practical, therefore amethod for "dry coupling" must be implemented. This requires shaping thetransducers to obtain a small contact surface which must be pressed against thetube wall with sufficient pressure to create appropriate acoustic coupling, withoutcausing deformation of the transducer or damaging the tube. In the related world ofGuided Wave pipe inspection, transducers are either glued to the pipe underinspection or pressed against it from the outside by an inflatable collar. A solution forheat exchanger tube inspection, on the other hand must couple the transducers to

the inside surface of the tube in a manner that will enable each measurement to




take place in the space of several seconds. In the AcousticEye system, thetransducers are placed on a flexible printed circuit, encircling an inflatable bladderand held within a slotted metal cylinder. When the bladder is emptied it contractssufficiently for the transducers to recede through the slots into the metal cylinder.When the bladder is inflated to a known pressure, the transducers are pushedthrough the slots against the tube wall, with a force that can be easily calculated. Acompressor and reservoir in the Data Acquisition unit enable both inflation anddeflation of the bladder in less than a second each, thus lengthening the entiremeasurement process per tube by no more than 10%.

A further aspect of transducer design is the use of proper piezoelectric crystals,backing materials and contact layers between the tube and the transducer to attainhigh durability and the desired band-width. Some ultrasound systems employpiezoelectric crystals having a sharp resonance, which creates a signal having anarrow bandwidth around this resonant frequency. The AcousticEye systememploys a wideband signal spanning a large portion of the spectrum up to cut-on ofthe T(0,2) mode. Therefore the structure of the transducers is designed to avoidpronounced resonances. The details of this design are proprietary.

Transducer Layout

As stated in a previous section, the azimuthal resolution is determined by thenumber of transducers spread over the tube circumference. The currentimplementation provides a resolution of 30 degrees, by using 12 transducers. Theirdimensions are too large to allow a ring of 12 such transducers to fit into tubes as

small as 3/4" in diameter. Therefore they are arranged in two separate rings in astaggered orientation, thus sampling the circumference at 12 points, as shown inFigure 13. The signal processing software must compensate for the difference inaxial locations of these two rings. To separate right and left propagating waves, twosensors are placed in each slot, separated by an axial distance that is smaller thanthe smallest wavelength propagating in the tube. A weighted and delayed sum of thetwo signals in such a pair enables isolation of these two waves, thus eliminatingmultiple reflections off the tube sheet.

Figure 13 : The slotted cylinder holding the piezoelectric transducers,with two staggered rings of six slots each




Transducer Excitation

The signal used to excite the transducers is an important issue. Conventional guidedwave systems employ narrowband signals centered on relatively low frequenciessuch as 100 kHz or less. This reduces the axial resolution considerably. In contrast,the AcousticEye UPR system is intended to gain the maximal possible informationfrom the bandwidth and modes available, therefore the excitation signal is awideband signal stretching nearly from DC to a frequency as close as possible tothe cut-on of T(0,2). In practice, resonances in the piezoelectric materials and theircoupling points to the tube tend to limit this bandwidth. Currently the AcousticEyeUPR system achieves bandwidths of approximately 400MHz.

C. UPR Signal Analysis

UPR is essentially a pulse reflectometry method, therefore the initial signal analysisis very similar to the analysis performed in APR. UPR signals must first go throughthe preprocessing and registration stages: determination of a reference signal andestimation of background noise. Most of the following steps – alignment, detection,classification and sizing, are carried out on the APR and UPR signals jointly,therefore they are discussed further in the next section.

6.0 COMBINING APR & UPR

A. Implementation of a Combined Probe

The previous sections show that the capabilities of APR and UPR arecomplementary in many ways. To take advantage of this, both measurements mustbe conducted on every tube to be inspected, preferably within a single instrument.Theoretically speaking, there is no obstacle to carrying out the two measurementssimultaneously: APR and UPR operate in different frequency bands, propagatingwaves in different media. When inspecting heat exchanger tubes, however, accessis possible only to the ID, where real estate is extremely limited. Therefore there is aconsiderable engineering challenge in squeezing transducers for both APR andUPR into the confined space and communicating with them at high data rates.Though a full discussion of the issues involved is beyond the scope of this paper,

several highlights of the AcousticEye implementation are worth noting:

1. The UPR transducers must penetrate the tube beyond the thickness of thetube sheet, otherwise the waves they generate will propagate into the tubesheet instead of the tube itself. The transducers are also expected toexperience a certain degree of wear and tear as they are repeatedlypressed against the inspected tubes. Therefore they are mounted on anassembly termed the "probe" unit that can be detached from the body ofthe handheld "gun" unit. Probes are manufactured with different axialspacers, to accommodate various tube sheet thicknesses. These rangefrom less than 1" up to 12".




2. A further reason for using a detachable probe, is that the UPR transducershave a limited radial travel distance as they emerge from the slots in theprobe and press against the tube wall. Thus a probe with a given diametercan be used to examine IDs that span a limited range of approximately4mm. Probes are manufactured with different diameters to accommodatedifferent ranges of tube IDs.

3. All stages in the process of measuring a single tube: bladder inflation,acquiring the APR and UPR signals, followed by bladder deflation, arecarefully designed to achieve an extremely short inspection time. Currentlythis is approximately 10 seconds per tube.

4. Ergonomics of the probe and gun have been considered carefully, both toshorten measurement time, to ensure that measurements are acquiredcorrectly, to flag any faulty measurements if they occur, and to enable theoperation of the system by a single technician.

B. Combined Signal Interpretation

Information found in the APR and UPR signals is partially complementary andpartially overlapping. Exploiting this information fully requires careful design of the

signal processing, algorithmics and user interface from the ground up. This involvesall the stages described in the above discussion of APR: registration, detection,classification and sizing.

Registration : In addition to aligning tube ends across the measurements for thebundle, APR and UPR measurements must be aligned with each other. Thisprocess takes into account that fact that the point of origin for the two types ofmeasurements is not identical. This is because the microphone which measures

APR signals is not located at the same axial distance along the probe as the UPRtransducers; in addition, signals from the tube endings do not arrive from the samelocation. APR end-of-tube signals occur at the very end of the tube, whereas UPRend-of-tube signals occur where the tube meets the tube sheet. The latter dependson the thickness of the tube sheet, which varies from bundle to bundle. Theparameters influencing registration are either known to the system from the structureof the probe, or entered by the user (e.g. tube sheet thickness). Registration is thenperformed automatically.

Detection : Initial detection is performed on both signals separately, then crossvalidated. The methodology used for detection based on the APR signal has beendescribed above. Detection based on the UPR signal is very similar, employingmainly the T(0,1) mode. Cross validation at this stage consists mainly of notingwhich defects were detected in common by both APR and UPR, and which defectswere detected only by one of them.




Classification : Initial classification is also performed on each signal separately.Classification as carried out on the APR signals has been discussed above in detail;classification based on the UPR signal again follows the same general outlines, withadjustment for the polarities in the UPR signals, as described above also. Crossvalidation is more complex at this stage, to ensure that each detected defect isclassified as correctly as possible and to resolve any possible conflicts. The fullalgorithm is not disclosed here, but several general principles are as follows:

1. At locations where APR indicates a hole and UPR indicates wall loss, thedefect is classified as a hole.

2. At locations where both signals indicate wall loss, classification is ID wallloss.

3. At locations where only UPR indicates wall loss, classification is OD wallloss.

4. Blockages will normally be detected by APR only, therefore no indication isexpected from UPR.

5. Close lipped cracks and OD cracks will be detected by UPR only, involvingmore complicated analysis then observing just signals peaks and polarity.

Sizing : As in the previous stages, sizing is also performed separately on each signal

type. Sizing based on APR signals follows the discussion in the previous section. Itis used mainly for holes and blockages, where no azimuthal resolution is required,and the accuracy of UPR is lower or nonexistent. Sizing based on the UPR signalsis used mainly for wall loss and crack defects, and it is mainly here that theinformation found in the F T(m,1) modes is used. A final stage of cross validation iscarried out at the end of this stage also. The defect table is then analyzed forconsistency, merging defects if necessary.

The entire process of signal interpretation is summarized in a flowchart in Figure 14.




Figure 14 : A flowchart summarizing the combined signal interpretation process

Visual Verification

The automatic interpretation described in the above sections is highly successful inproviding a complete picture of the defects found in a tube bundle. Nevertheless, atube inspection system cannot be treated as a "black box" system to be relied uponblindly. The AcousticEye APR+UPR system therefore has many design featuresincorporated into the user interface for scanning through and verifying the resultsvisually, as well as creating a detailed, comprehensive final report. The guidingprinciple behind the software design is achievement of a streamlined visualverification process, in order to attain the highest possible throughput.While it is beyond the scope of this paper to provide a complete user manualdetailing all the options that can be found in the software GUI, several importantfeatures are highlighted here:

Initialization: Data input, Registration

Detection - APR Detection - UPR

Detection cross validation

Classification- APR Classification- UPR

Classification cross validation

Size estimation APR Size estimation UPR

Size estimation cross validation

Defect table post processing:merging, reassignment if necessary




1. Summarizing defect tables, tabbed by defect type

2. Rapid scanning through the flagged defects showing actual signaturesalong with calculated templates, and providing keyboard shortcuts forapproving or declining

3. Fault specific thresholds

4. Report generation A sample screen shot of the GUI is shown in figure 15. The top graphic

panel shows the APR signal. The signature of a through hole is highlightedin light blue, with a calculated template of a through signature super-imposed in red. The bottom graphic panel shows the UPR T(0,1) signal,with two OD defects also highlighted in light blue: an OD pit and an ODcircumferential groove.

Figure 15 : A GUI screen shot showing APR and UPR measurements on the same tubes, with threedefects highlighted in light blue.




C. Examples

In this section we present several sample measurements taken on a heat exchangermockup with tubes containing defects typically found in heat exchangers. The tubeswere 3/4" Carbon steel, with wall thickness of 0.083" and a length of 1.2 meters.Some defects show up in both APR and UPR modalities, though often more stronglyin one than in the other. Other defects show up in one of the two only. Table IIIshows the layout of the defects in the tubes presented here.

Table III: Layout of the defect in 3 mockup tubes

TUBE # DISTANCE FLAW SIZE FLAW TYPE NOTES

A (2) 70 cm 60% 7/64" diameter OD pit UPR100 cm 80% 5/64" diameter OD pit UPR

110 cm 0.052" Through wall hole BothB(4) 65 cm 40% 3/16" diameter OD pit under baffle plate UPR70 cm 10% Internal Blockage APR

100 cm 40% 0.01" wide, 1/2" long OD circ notch UPR

C(8) 70 cm 60% 7/64" OD pit UPR

105 cm 40% 3/16" ID pit Both

The APR and UPR measurements from each tube in Table III are presented belowalong with a short discussion.

Tube A : APR and UPR signals for this tube appear in figure 16.

Figure 16 : APR (top) and UPR (bottom) signals from tube A, with defects highlighted in light blue.




The APR signal reveals only the through hole located at 110 cm., having a strongsignature that lasts till the reflection from the end of the tube. The UPR signal showsall three defects, at 70, 100 and 110 cm.

Tube B : APR and UPR signals for this tube appear in figure 17.

Figure 17 : APR (top) and UPR (bottom) signals from tube B, with defects highlighted in light blue

The APR signal reveals the blockage located at 70 cm, again having a strongsignature. The UPR signal shows the remaining two defects at 65 and 100 cm.

Tube C : APR and UPR signals for this tube appear in figure 18.

Figure 18 : APR (top) and UPR (bottom) signals from tube C, with defects highlighted in light blue.




The APR signal reveals the ID pit located at 105 cm. The UPR signal shows boththis ID pit and also the OD pit at 70 cm.

7.0 SUMMARY

Pulse reflectometry has two properties which hold a special appeal for tube inspection:high speed and the fact that it eliminates the need for traversing tubes with a probe.These properties are important advantages over currently more widespread methods.

In this paper we have reviewed the physical principles behind two modalities of pulsereflectometry, APR and UPR. We have shown that neither of these methods on its ownoffers a sufficiently comprehensive solution to heat exchanger tube inspection, however,their capabilities are largely complementary. Therefore a system combining bothtechnologies can detect all the defects found in heat exchanger tubing, while retaining theabove advantages and suffering from fewer limitations than electromagnetic orconventional ultrasonic systems.

Beyond the underlying physics, we have described the challenges involved inimplementing such a dual system, and shown how these challenges have been overcomein the AcousticEye APR+UPR system. This system exploits both the complementaryaspects of the two technologies and their overlapping capabilities in order to provide thebest possible detection, classification and sizing of defects.

REFERENCES

Amir, N., Shimony, U., Rosenhouse, G. (1996), "Losses in tubular acoustic systems – theoryand experiment in the sampled time and frequency domains," Acustica – Acta Acustica, Vol.82, 1-8

Sharp, D. B., Campbell, D.M., (1997), "Leak detection in pipes using Acoustic PulseReflectometry," Acustica, Vol. 83(3), 560-566.

Keefe, D. H. (1984), "Acoustical wave propagation in cylindrical ducts: transmission lineparameter approximations for isothermal and nonisothermal boundary conditions", J. Acoust.Soc. Amer., Vol. 75, 58-62

Rose, J.L. (1999), Ultrasonic Waves in Solid Media , Cambridge University Press.

Sun, Z., Zhang, L., Rose, J.L. (2005), "Flexural torsional guided wave mechanics andfocusing in pipe", Transactions of the ASME, Vol. 127, 471-478



AcousticEye Ltd.www.acousticeye.com

accoustic eye details

Documents