CA1314970C - Measurement of corrosion with curved ultrasonic transducer; rule-based processing of full echo waveforms; preprocessing and transmitting echo waveform information - Google Patents

Abstract

ABSTRACT OF THE DISCLOSURE
An acoustic transducer having a curved surface to match the inner or outer surfaces of a target surface such as a cylindrical borehole tubular is disclosed. A sequence of excitation pulses causes the transducer to launch a series of high-frequency, short duration acoustic energy pulses toward the tubular. The characteristics of the tubular surfaces, including the location, area and depth of any corrosion pits on the tubular surfaces, are determined from a full echo waveform.
A rule-based Expert System analyzes the full waveform of the echo pulses returned from the tubular to determine the most likely characteris-tics of the surfaces which produced the full echo waveform. More specifically, the Expert System uses expertise about the acoustic properties of the target medium, as well as constraints chosen by an expert, to determine which signal structures are informational and which signal structures are confounding, The present invention also involves preprocessing and transmitting information. The present invention reduces the absolute amount of information that is transmitted without reducing the amount of meaningful information that is conveyed. The present invention transmits more meaningful information than conventional transmission systems. It also reduces the amount of information processing at the destination.

A41.1.WP

Description

1 3 ~

TITI~ OF THE INVENTION

MEAS~REMENT OF CORROSION WITH CURVED ULTRASONIC
TRANSDUCER; RULE-BASED P~O OE SSING OF FULL
EC~O WAVEFORMS; PREPROCESSING aND T~ANSMITTING
20~HO ~AVEF~RM INFORMATION

BACKGROVND OF THE INVENTION

1. Field of the Invention This invention relates to devices and methods for dekecting and analyzing ~surface imperfections 30using ultrasonic techniques. More specifically, the invention relates to mea~urement of corrosion pits on the surfaces of metal tubulars within the boreholes of oil wells and subsequent analysis of ultrasonic echos using a rule-based artificial 35intelligence technique.

2. Related Art a. The Corrosion Problem Corrosion in oil well tubulars has long been a source o~ difficulties for the oil production industry. In a t~pical oil well, the metallic tubular is disposed within a casing. The interior 45o~ this metal tubular passes crude oil from a 1 3 ~

ths naturally available pres~ure of the "gas cap."
The production o~ oil ~ay be artif icially enhanced during "artificial lift" by pumping gas down the exterior o~ the tubular ~ithin the casing to S replenish the ga~ cap.
Corrosion pits can ~row to eat thrsugh the side o~ the tubular 80 that the seal between the oil in the interior o~ th~ tubular and the gas exterior to the tubular is broken. Gas penstrates into the oil passaqeway instead of going to the "gas cap.l- The escaping gas (or ~luids within the casing~ frustrate the replenishment of the gas cap.
Depending on the ~everity of the corrosion-caused hole in the tubular, the ability of the pumping lS ~echanism to pump oil ~rom beneath the earth is dîminished or even prevented.
When a corrosion pit has eaten completely through a tubular, it was once necessary to xeplace the tubular. This was obviously a very expensive process, considerin~ both the~ direct cost of the replacement as well as the ]Lost revenues due to "downtime" of the oil well. Chemical tr~atment methods were then developed to inhibit the ~urther progres~ o~ existing corros:Lon pits. Although chemical treatments proved less expensive than the actual replacement of a downhole tubular, significant costs were still incurred from the application of the chemical proce~s itself, in addition to the lost revenues due to downtime o~
the oil well.
There~ore, even given the ability to prevent and repair corrosion pi~s be~ore they become ~atal to the operation of the oil well, it is crucial from a cost standpoint that the chemical treatment ~3~7~

be applied only in circumstances when such chemical application i8 necessary, so a~ to prevent needless downtime of the oil well.

b. ~

There are variou~ known device which attempt to monitor the condition o~ the interior o~ o~l well tubular~.

i. T~a~ Kinley ~ali~E

A first known device, commonly known as the Kinley Caliper, is a device which is drawn up the interior of the tubular. The Kinley Caliper comprises a plurality ~usually less than 48) of wire probes which extend radially from the caliper device to contact the inner wall~ of the tubular.
Springs provide outward force on these wire probes to maintain their contact with the tubular face as the devicz is drawn up the tubular. As a corrosion pit is encountered by one or more of the wire probes, the spring which urges the probe(s) outward causes displacement of the probe(s~. A
quantitative measurement of the displacement of these probe(~) ia made in the device, and this di~placement measurement is internally recorded on metal drum.
Th~ Kinley Caliper had the disadvantages that, a~ the probes were dxawn along the borehole's interior, the probes' outward pressure caused scratches on ~he tubular surface. Al~o, depending on the draw rate and spring strength, the wire pro~es had a tendency ~o "skip over" smaller ~3~ ~7~

corrosion pits or give misleadingly shallow indications of deepPr corrosion pits. Finally, the ef~ective coverage of the probe~ was limited, in the sense that the surface ~rea of the tubular which happened to fall b~tween the probes was not tested for corrosion pits.
ii. Ul~rasoni$~

Advances in the fields of ultrasonics and microelectronics allowed more thorough, less damaging ~ea~urement of corrosion pits, as well as the ability to transmit greater amounts of information uphole in a limited amount o~ time so as to speed the logging process.
A. ~The BHTV

~n acoustic logging de~ice is disclosed in ~0 U.S. Patent No. 3,503,038 to Baldwin. One embodi-ment of this device has ~eaome known as the Borehole Televiewer (BHTV). The BHTV is a device which i~ designed to be dra~n up the interior o~ an uncased boreh~le. The BHTV comprises a rotating transducer which defines a hellcal pathway as the entire device is drawn up the borehole. Although the BHTV was originally intended for use in detecting ~ractures in uncased boreholes, the device has been tested for measurement of casing corrosion. [C. Carson and T. Bauman, "~se of an Acoustic Borehole Televiewer to Inve~igate Casing Corrosion in Geothermal Wells,~ Paper #408, Corrosion 86, Houston, Texas, March 17-21, 1986~.
The following description present~ an explanation ~ 3 ~
--5~

of the ultrasonic operation of the device, as it could be applied to detection o~ corrosion in tubulars~

Referring to Figure 1, a known transducer 102 having a flat surface 104 is enclosed wi~hin a sonds body 106. The ~onde 106 is located withln a circular tubular 100, a portion of which is indicated in Figure 1. Transducer 102 is excited by an ex~itation pulse. The excitation pulse causes an acoustic wave to emanate from transducer surface 104 toward the inner surface 108 of tubular 100 radially along lines 110, 112, and 114. After striking the inner surface 108 of tubular 100, a portion of the acou~tic ener~y is reflected from the tubular (as an "echo"~ back toward transducer surface 104. This echo is detected by transducer 102. Electronic circuitry w:Lthin the sonde body 106 converts the echo~s acoustic energy into electrical signals indicative of the instantaneous magnitude o~ the returned ;acoustic energy for transmission uphole.
It can be seen from Figure 1 that the path traversed by the portion of the acoustic wave along pa~hs 110 and 1~2 in i~,8 round-trip from transducer surface ~04 to the surface 108 o~ tubular 100 is shorter than the round-trip path traveled by the portion of the acoust~c wave trav~ling along path 114. ~he difference in path lengths is due to the curva~ure of the inner sur~ace 108 of tubular 100 which is not matched by the flat surface 104 o~
txansducer 102.

~3~P~

The difference in distance traversed by di~ferent portions o~ the acoustic pulse results in a returned echo which is dispersed in time. The dispersed echo pulse i~ smaller in magnitude and longer in duration than th~ original pulse emitted by the transducer. This dispersion results in increased uncertainty ~ to the distance traversed ~y the acoustic wave. This increased uncertainty results, first, ~rom the r~duced amplitude of the received echo, since an echo of smaller amplitude is inherently more difficult to detect. Second, the increased uncertainty as to the exact time o~
arrival of the pulse is also caused by the widening o~ the returned echo.
To those skilled in the art of acoustics, it is well known that the degree of a~t~nuation o~ an acoustic pulse is proportional to its ~requency.
(As is known in the art, "pulse" actually denotes a pulse-modulated sinusoid; the "pulse" is actually the envelope determined by the peaks o~ the modulated sinusoid. It is t:he frequency ~ the sinusoid, and not o~ any pulse train, which is referred to when speaking of l'~r~guency'l in this discussion. Figure 8, described in greater detail below, illustrates pulse-modulated sinusoids 804 of returned echo pulses with their accompanying envelopes 802.) It is also known that, in acoustic eahography, a higher degree of resolution can be attained usinq pulse3 composed o~ higher ~re~uency acoustia waves. Therefore, in selecting acoustic frequancies, tber~ has traditionally been a tradeof~ between the competing desires for high resolution and for low attenuation.

~ 3 ~*~

It is therefore desirable to produce acoustic pulse echos which are large in magnitude and short in duration, and which are capable of improved resolution.
C. ~=S~

The rotating transducex emits periodic ultrasonic excitation pulsas which impact the tubular surface and cau~e a series of echos to be returned to the transducer. A ~full echo waveform"
comprises the set o~ all echo~ which are returned from a targe~ surface in response to a single incident excitation pulse. ~The process by which a series o~ echos, or "full echo waveform", is formed from a single excitation pulse is described in detail in the dis~ussion of Figure ~, below.) The present discussion serves as an exposition of the limitations of known devices which do not utilize the full echo waveform. In the known device described above, only ~he first of the series o~ eahos was used to determine the distance of the transducer grom the tubular ~urface. The information in the ~ull ~cho waveform beyond the 2$ first echo is not ~tiliæe~. The distance determination was ~ade by ~eans of a simple :multiplication of (1) the i~own speed o~ the acoustic pulse ~and echo) in the li~uid medium within the borehole by (2) the mea~ured one-way ~ravel time of the firs~-received acoustic pulse.
I~ the transducer were situated adjacent a corrosion pit larger than the transducer sur~ace, the round-trîp travel time, and therefore the measured distance between the transducer and the bottom of the pit, would be larger than would be the case if the transducer were ~ituated above a non-corroded area of tubular.
For this known device, a small pit is datected only if the entire ultrasonic beam enters the pit.
In this case, the first echo comes from the bottom of th~ pit and the increase in arrival time can be used to measure the depth. If, however, an~
portion o~ the beam ~ntercept~ the tubular beyond ~he edge of the pit, an echo will be generated by the uncorroded inner wall of the pipe. If this echo is large enough to trigg~r the device's timing circuits, it will be recorded as the ~irst echo, and the pit will go unde~ected. Since a smooth area of uncorroded inner wall will generate a strong reflected echo, very little of the beam has to intercept the inner wall ~or a significant echo to be generated. Thus, for this known davice, the ultrasonic wave must impact totally within the pit area for pit detection to occur.
Similarly, if the transducer were positioned adjacent a ccrrosion pit which wa3 smaller in area than the area of the txansducer, then (assuming for simplicity that the bottom of the pit is flat) two echo signals would be received by the transducer.
The two echo ~ignals would correspond to the two different round-trip travel times of the acoustic : pulse in re~lecting off the uncorroded tubular surface and th~ corrosion pit surface. The analysis of only the front-surface echo, even in the presen~e o~ an "ideal" (flat) pit, wastes a great deal of information regarding the progress of corrosion. Thic waste of ~nformation is a ~ 3 ~ 7 ~

limitation characteristic of the BHTV, described above.
As will be seen in the discussion of F~gure 10, below, the complexity of the ~'full e ho waveform" occurring after the front ~urface echo also shows the disper6ion effects of the rough surace of real-world corrosion pits. The acoustic energy which falls on a surface which iæ not flat and perpendlcular to the incldent acoustic wavefront follows a path which is less predictable.
Referring to Figure 3, paths of various portions o~ an acoustic pulse emanating from a transducer are illustr~ted. Element 102 (also in Figure 1) designates a transducer according to a known device, whereas element 202 (also in Figure 2) designates a transducer according to the preferred embodiment of the present invention.
Since the following discussion relates to the reflective properties of a corroded ~ubular 100, the particular transducer chosen has little bearing on this conceptual discussion.
It is co~mon to have ~he complex pit surfaces reflect the acoustic energy in a way which simply disperses the returning echos over a greater time period so that a single, detec~able peak is not rormed at all. Path 308 illustra~es this occurrenca. This complex surface roughness is indicat~d in a full echo wave~orm only as a reduation in the amplitude of existing peaks, and a corresponding rise in the "noise floor." This dispersion, of course, makes it more difficult to discriminate returned echo pulsa peaks from the noi~ which is superimposed on the full echo wave-~orm.

13~7~

In instances of actual corrosion, then, the surface o~ the pit is very ~eldo~ flat. Therefore~
a second "clean" echo is unlikPly to be produced because of the sideway~ deflection (and eventual absorptlon) of much of the acoustic energy which is incident upon the rough ~urface of a real-world corrosion pit.
Furthermore, som~ acoustic energy is lost altogether, such as t~at which travel~ path 310.
Thi~ energy might otherwise have positively contributed to an acho peak. The lo~s of this energy makes detection and discrimination o~ peaks more difficult.
The echos received by the transducer will thus be much more complicated than that r~ceived from an "ideal" (flat) corrosion pit. Information i~
therefore lost unless a more sophisticated means of interpreting the full echo waveform, including the plurality of echos which follow the first, "front sur~ace" echoO
Human error may result in lack of proper diagnosi of corrosion, especially when consider~ng the massive number of full echo waveform~ which are produced in a single ~Irun~ up a 20,000-~oot oil well.
It is thersfore de irable to per~orm expert, reliable, automated analysis of an extremely large number o~ waveforms compri~ing echos which arrive a~ter the front-surface echo.

~L 3 ~ 7 a D. ~e~uency Domain verSus Ti~ne Domain Ana~sis Analysis techniques ~ay roughly be divlded into two ~lasses.
The first class analyzes echos in the time domain~ A measurement of th~ amplitude o~ the ~ront-surface echo (described abovs3 to determina lo the presence of fractures (and, perhaps by implication, pits) is probably the simplest example o~ a time-domain tech~ique. A smaller front-surface echo presumably indicates a ~arger ~racture (or pit), since any acoustic energy which entered a lS fracture (or pit) would not be returned to the transducer surface coincident in time with tha front sur~ace echo. See Carson et al., Corrosion 86, cited above.
The second class of analysis is that which 2Q converts the incoming echos illtO functions o~ ~re-quency. Frequency domain analysis techniques involve the analysis of the sjpectral components of a returned echo signal to dedluce information about the re~lection surfaces wh~ch must have caused th~t particular array of spectral ~omponents.
Fr~guency domain techniques have been used in order to avoid certain proble~s ~n the analysis o~
the raw, returned echos in the time domain. Raw, time-domain erhos were often indistinct due to dif~usion o~ the wavePront caused by complex and unpredictable intexaction o~ the wave~ront with the irregular sur~ace of corrosion pits~ The ef~ect of this di~fusi4n of pulse echos was to maXe it more di~ficult to distinguish when ona echo end~ and another begins, or even whether a given echographic ~ormation constituted one echo or a sum o~ echos ~ 3 ~
-12~

which traversed different paths but happened to arrive at the transducer at the same time.
Frequency-domain techniques have met with some success. Frequency domain technigues have been most commonly employed for measuring chan~es in the thickn~ss of large areas of a wall, rather than for detection of localized pits and meas~ring their depth. But because a time-to-frequency transformation is nQcessary, the intuitive grasp of individual echo components was removed from the analysis process. What wa~ actually analyzed was a frequency spectrum, not an untransformed representation of what was actually occurring in the physical world.
It is therefore desirable to reliably analyze pulse echos "directly," in the time domain.
In summary, it is desirable to efficiently use a full echographic waveform whose individual echo components have improved resolution so as to facilitate time-domain analysis.

~ 3 ~

SUMMARY OF THE LNvENTIoN

In the field of time-domain echography, the present in~ention contemplate~ that imprcved resolution of pulses may be achieved by the production of high frequency, short-duration pulses by a transducer whose outer ~ace i8 ~haped to match the contour o~ the sur~ace on which i.rregularities ~such a~ corrosion pit8) are to bes detected and guantita~ively analyzed. ~he improve~ pulse echo resolution achieved by the present invention allows substantial elimination o~ "false echos~' which are associated with the mismatch of known flat transducers with the curved innex surfaces o~
tubulars, resulting in greater accuracy in the measurement of the location, depth, and area of surface irregularities.
The ability to intuit and visualize the echographic structure during the course of the ~0 echographic analysis i.q joined with the reliability and predictability o~ computer-based analysis through use of a rule based computer ~ystem. The analysis is not limited to the first received "front sur~acel' echo, but also embraces the "îull wave~orm" o~ all echos which follow the first echo.
The expertise of a human expert is progxammed into rules. The rules are opportunistically "fired" to automat~cally yield a set of conclusions about the set of echo peaks in a full waveform. The system's conclusions possess a high accuracy not possible with manual analysis by thos~ individuals with less expertise. Tha system' conclus~ons also possess a consistency not possible with manual analysis by any individual, including experts.

~ 3 ~

The rules act opportunistically ~o as to find and eliminate confounding signal structures which mask the echos which would otherwise cleaxly indicate the location, depth, and area of surface irregularities. The rules t~en act on the remaining infor~ational echos to actually calculate their depth and area, and associate those calculations with a particular location. BPcause information ~s to ~oth the inner and outer surface of a ~ubular ~ay be inherent in a ~ull echo wave~orm, the Expert System may draw conclusions as to the condition of either or both surfaces.
This rule-based automated system, an Expert System, has an "explanation facility" which outputs the sequence of steps ~path of rule firings) through which it reasoned to arrive at its conclusions regarding the nature of each peak in the echographic waveform. This output may be used in conjunction with the conclusions about the target surface themselves to allow the user to accept or reject the conclu~ions formed by the Expert System.
In the field o~ the echography, the present invention i~ a system and method of preproaessing and transmitting inform2tion (~uch as echo waveform in~ormati~n) ~rQm a sourc~ ~such as a transducer) to a destination (such as a rule-ba~ed computer system). The present invention reduces the absolute amount of information that is transmitted without reduc~ng the amount of meaningful information that i5 conveyed. In thi~ way, the present inv~ntion is effectively able ~o ~ransmit more meaningful information than would be expected ~or conventional transmission systems. It also ~3~Y~

acts to reduce the amount o~ information processing that mus~ be done at ~he des~ination.
The present invention has particular applicability to the transmission o~ digitized echo S wave~orm information ~rom a transducer used to per~orm pulse echo resolution of the surface irregularities of metal tubulars within the borehole o~ an oil well. This digitized information i 8 transmittad up the borehole (called I'uphole transmi~sionl') to a des~ination at the top of th~ borehol~ (such as a rule-based computer system) where it i5 analyzed. The preprocessing results in le~s data being transmitted for the information conveyed than would be the case if conventional transmission technigues were used. It should be appraciated that the present invention has applicability beyond the preferred field of use. It can be used in any envlronment where there would be a need or bene~it for a transmission system or method where the amount o~ information that must be transmitted i~; more than can be transmitted using conve~tional transmission t~chnique~ and any given transmission media.
In the preferred mode, the preqent invention processes an echo wave~orm derived from the echo of an acoustic pulse launched toward a target surface.
The ~cho waveform i8 converted into a monopolar wave~orm using an envelope detector and then emoothed using a low pass ~ilter. The smoothed echo waveform is then passed through a logarithmîc ampli~ier where it is compressed. The compressed and smoothed ~onopolar wavefor~ i~ th~n digitized, and the digitized data i~ stored ~n ~ recirculating memory. ~he recirculating memory ~ill5 Up and begins to overwrite itsel~. ~hile the digitiæation is taking place, the echo wav~form is simultaneously put through a threshold detector.
Approximately 20 microsectinn~ after the amplitude of the echo wave~oxm rises above a predetermined thr~shold the recirculating memory is frozen and the data within is prep~red ~or transmission.
A binary counter measures the tim~ period from the launching of the acoustic pulse to the time when the echo wave~orm reaches the predeter~ined threshold. This measured time period is multiplexed with the digitiæed echo waveform data frozen in memory and gain information from an automatic gain control within in the system.
These three data ars subsequently transmitted in a multiplexed fashion uphole to the rule-base computer syste~.

7 ~

BRIEF DESCRIPTION O~ THE DRAWINGS

The invention is be~t understood ~y reading the following detailed description in conjunction with tha accompanying drawings, in which li~e reference numerals refer to likQ elements throughout, and in whicho Figur~ 1 represents a top plan view depicting the acou~tic properties of a known transducer within a borehole tubular.
Figure 2 represPnts a top plan view depic~ing the acoustic properties of a transducer according to the pre~ent invention.
Fig~re 3 repr~sents a top plan view of the acoustic properties o~ an aoustic transducer when acoustic pulses are applied to a corroded tubular surface.
Figure 4 presents a comparison of actual waveform envelopes of signals reflected from uncorroded pipe, illustrating the improvement in quality of signals to be expected when using a curved tranducer accordinq to the preferred embodiment o~ the present invention.
Figure 5 represents an actual measured ultrasonic beam pattern for a curved rectangular transducer accord~ng to the present invention.
Figures 6A and 6B represent examples of a coveragQ d~agram demonstrating the nonoverlapping and overlapping, respectively, of acoustic "spots"
schematically superimposed on a ~olded-out interior surface o~ an oil weIl tubular.
Figure 7 is a timing diagram represen~ing an excitation pulse and suksequent returned echos according to the present invention.

~ 3 ~

Figure 8 details the high-frequency components and envelopes of typical acoustic signala.
Figure 9 represents a timing diagram indicating the full ~cho wave~orm produced from a singlQ ~xcitation pul~ on a sur~ace having an "ideal" (flat-bottomed) corro~ion pit.
Figure 10 represents an actual full echo wave~orm to ~e expected when a transducer according to the present invention operates on a corroded tubular surface.
Figure 11 is a block diagram of a preferred uphole transmission system used to transmit full echo waveform data (or extracted echo ~eature data) from a sonde to the rule-based analysis system according to the pre~erred embodiment of the present invention.
Figure 12 i~ a flow diagram illustrating in very general form the relation of Xnowledge sources in the rule-based system according to the pre~erred embodiment of the present invention.
Figure 13A is a ~low chart illustrating the echo feature extrac~ion routine according to the pre~erred embodiment of the 'IExtract Features..."
block in Figure 12.
Figure 13B illustrates a preferred data file structure produced by the echo feature extrac~ion routine diagrammed in Figure 13A.
Figure 14 shows in very general form the structure of the rule-based system according to the preferred embodiment of the present invention, ~howing the relation of the knowledge sources to the working memory ~'lblackboard").
Figure 15 i~ an activation diagram for the Control Function knowledg~ ~ource o~ Figure 12.

13~7~

Figure 16 is an activation diagram for the MULTIPLE knowledge source of Figur~ 12~
Figure 17 is an activation diagram for the "El" portion of the ECH08 knowledge source o~
5Figure 12 which identifies the dominant echoO
Figure 18 is an activation diagram for the "E2" portion of the ECHOS knowledge source in Figure 12 which identifi~s echos other than multiple structure echos, assuming a multiple 10structure has been identi~ied, Figure 19 is an activation diagram for the "E3" portion of the ECHOS knowledge source of Figure 12 which identiies echos in the absence of any identified multiple structures.
15Figure 20 is an activation diagram for the ULTRASONICS knowl~dge source of Figure 12.
Figure 21 illustrates a user-interface display for the processing example illustrated in Figure 10 .
ETAILED DESCRIPTION O~T~E P~FER~E~ EMBODIMENTS

1. Generation o~ FuLl Echo W,~veform 25a. Curved Transducer Re~erring now to Figure 2, a transducer 202 acc~rding to the preferred embodiment of the present invention is disposed in a sonde 206. The 30tran~ducer 202 of the present invention is characterized in that its outer sur~ace 204 is cux~ed so as to match the curvature of its target surface. In the specific case of an oil well tubular 100, the tran~du~er is curved only in a ~ 3 ~

direction perpendicular to the axis o~ the tubular.
Along lines parallel to the tubular axis, the preferred embodiment o~ the transduc0r i8 straight.
Thus, the transducer outer surfaae 204 may be de~ined mathematically a~ a ection o~ a right circular cylinder. Curved transducers suitable ~or embodying the present inv0ntion are co~mercially available from, for example, Ultran L~boratories in State College, Penn~ylvania.
In roughly ths manner described with respect to the known transducer 102 (Figure 1), an excitation pulse causes an acoustic wave to emanate from transducer sur~ace 204 toward the inner surface 108 of tubular 100 to cause the wave to hit the inner sur~ace 108 "in phase." An acoustic echo is thus returned to transducer sur~ace 204 ~or detection and eventual ~nalysis. In a preferred embodiment of the pre~ent invention, a rectangular excitation pulse of 300 volt magnitude, 50 nanosecond duration, and l-millisecond repetition period i~ employed for ease of detection and greater resolukion of the returned echos. The transducer emits an ultrasoni~, wave with a center ~requency o~ 4.5 ~egahertæ ~MHz).
Of course, these values are typical values.
The values may be varied according to the demands of particu~ar experiments or applications.
Generally, the invention may be practiced with excitation pulses anywhere in the 10~1000 volt range. Pulse durations may vary from 10-100 nano-seconds for a 4.5 NHz tran~ducer~ Repetition peri~ds may be as long as convenience requires, and may be as short as permits non~interference of excitation pulses with r~turned echos from previous citation pulses. Standoffs (transducer-tubular separation) ~ay be chosen as required, 50 long a~ a recognizable signal i8 returned from the target surface. The ultrasonic wave ~requency ~ay be any frequency which the environment allows, ~ut ~s advantageously in exce~s o~ the 1 ~Hz signals used by others in ultrasonic evaluations.
A dif~erence in performance of the ~ransducPrs 104 and 204 (Figures 1 and 2, respectively) is that the curved transducer face 204 ensures that the round-trip travel time of all portions o~ an acoustic pulse is substantially ~dentical, regardless of whether path~ 210, 212, or 214 are traversed. To achieve this equality of travel time, the center of curvature of the transducer sur~ace must be the same point as the center of curvature of the in~er wall o~ the tubular.
~he choice o~ identical cen~ers o~ curYature results in a noticeable improvement in the "crisp-ness" of the returned echo pulse. (A "crisp" pulse is characterized as having large amplitude and small time duration~) Increased ampli~ude allows easier detection of each individual echo pulse, and decreased pul6e width facilitates the discrim~nation o~ pulses ~rom adjacent pulses.
(Although it i~ not evident from Figure 2, the signal received by transducer 202 i8 comprised not only of a s~ngle QChO, ~Ut rather multiple echos~
The production o~ these other echos w~ e de~cribed in more detail below~ with regard to Figure 9.) Figure 4 show~ waveforms reflected fro~
uncoxroded pipe with a 5.5-inch inside diameter.
The top panel of Figure 4 ~hows a waveform of echo~

~ 3 ~

in an experiment employing a curved, .75 x .63-inch rectangular transducer with a 5 ~Hz center ~r~quency. The. bottom panel of Figure 4 shows a wave~or~ of echos produced in an experiment employing a flat-faced one-inch diameter transducer with a 5 MHz center frequency.
The di~ference in the two waveforms is substanti~l, especially considering the fact that the target ~urface wa~ smosth and uncorroded. Both waveforms demonstrate the "front sur~ac~ echo~" 402 and 414 which, in most circumstances, are the echos o~ greate~t magn~tude. The remalnder of each full echo wave~orm also contain~ a "multiple structure".
Multiple structures are formed as a result of internal reflectio~s within the tubulax (see Fi~ure 9). Multiple structures manifest as substantially equally-spaced sets of peaks, ~ince the ac~ustic wav~ is presumably reflecting internally between the same two surfaces be~ore a portion o~ the energy is returned to the transducer for detection.
The multiple structure ~or the full echo waveform produced by the curved tran~ducer is shown as peak elements 404, 406, 408, 410, and 412. ~h2y are ea~ily di6cernable above the "noise floor" of the ~ull echo waYeform.
The multiple structure o~ the full echo waveform produced ~y a ~lat tr~nsducer is much more difficult to discern. The multiple stxucture (as could later have been discerned) comprises peaks 416, 418, 420, and 422. ~he presence of this multiple structure is ma~ked by the presence o~
other, "falsel' echos 424, 426, 4~8 and 430.
Although, in ~he present experiment, it was known a priori t~at the sur~ace o~ tha pipe was ~3~Y~ ~

uncorroded, the presence of false echos such as 424, 426, 428, and 430 could have led an observer to belie~e that the tubular surface was in fact corroded. ~ny o~ the falce echo peaks could reasonably have been interpxeted as echos cau$ed by re~lection of acou5tic energy against the botto~
surface of one or more corrosion pits. As described abov~, the round-trip travel time of an acoustic wave which enter~ a pit is increased because of the greater distance ovex which the acoustic wave mus~ travel in order to reach the bottom of the pit, as compared with the distance from the transducer to th~ uncorroded front surface of ~he tubular.
To interpret these false echos as pit echos would be reasonable. Such a formation of peaks could be caused by the conversion o~ longitudinal acoustic waves to shear acouætic waves, in a process called mode conversion which is well known in the art. Also, it wou:Ld be reasonable to ascribe this formation of i~alse echo peaks to internal reflections within the converted wave~, as indicated above in the di.scussion related to Figure 3.
The difference between the front surface echos 4 02 and 41~ produred by curved and flat transducers, respectively, i~ also apparent ~ro~
Figure 4. A ~ront-surface echo 402 produced by a cuxved transducer is shown as being a "crisp" echo with ~ide lobes o~ much reduced amplitude. In contrast, the front-sur~ace echo 414 produced by a ~lat transducer exhibit~ th~ broadening effe~t characteristic of pulse dispers~on, as descx~bed above in the Background of the I~vention s~ction.

1~ 3 ~ Y~ ~

Also, the side ~obe~ 432 and 434 are oP much larger magnitude than those o~ curved transducer front~
surface ~cho ~102. Conceivably, in an actual downhole scenario, it would be indeterminate whether, for example, echo 432 was the front-surface echo and echo 4~4 was a pit echo. Although a sophi~ticatad technigue of analyzing fllll echo waveforms could possibly eliminate the interpretation of echo 4~2 as a front-surf`aca eoho, it can be seen ~rom Figure 4 that the 51uality of full echo waveform i~ much enhanced throuyh the use of a s::urved transducer. The full echo waveform of enhanced quality allows greater reliability, regardless of the degree of sophistication oiE the analysis method subsequently employed.

b. Coveraqe o~ he Tarq~et Sur~ace Figure 5 portrays an ultrasonic beam pattern for a rectangular transducer having a transducer surface with a curvature matching a tubular front surface. (In actuality, Figure 5 shows peal~
amplitude as a function of target refl~ctor position relative to the c:enter of this transducer.
For production of data contributing to this contour plot, a .125 inch diameter steel rod with a rounded end was used as a target reflec:tor.~ Figure 5 is thus essentially a ~ap o~ the acoustic energy which would strike the surface of a tublllar. It is to bP
emphasized that this diagram og acoustic energy, called the "spot size," relates to the tubular surface, and not the size of transducer itself.
The particular data shown in Figure 5 was obtained from a .75 x .63 inch transducer having a ~ 3 ~

"standof~ separation ~rom the target surface) of 1.5 inches, a convex radius of 2 inches, and a 4.2 MHz peak ~requency (~5 MHz at 6 dB down).
The numerical decibal ~igures indicate the reduction ~n amplitude ~rom th~ point o~ maximum amplitude which i~ located in the center of the "spotO" The degree of attenuation at which to define the edge of the spot is somewhat arbitrary, but the -12 dB contour ha~ been found to be a lo reaqonable choice.
An accurate determination of the size and shape of the spot is necessary in order to determine the coverage of successive applied acoustic pulses as the sonde moves up the tubular.
It is also important for the calculation of pit ~rea, as will be discussed in detail below.
Figures 6A and 6B depict the ~olded-out inner surface of ~ tubular, in a way which is common in the ar~. The tubular is assumed to have a vertical orientation. The left and riyht extremes of the diagram are understood to be ~oined together so as to express the transverse circular continuity of the cylindrical tubular.
Spots of successively applied acoustic pulses are represented on the inner surface o~ the t~bular. ~he transduc~r tand therefor~ the spot~
according to a pref~rred em~odiment of the present invention may traverse what (in three dîmensions) is a helical path.
~h~ helical path in three dimensions is represent~d in both Figures 6A a~d 6B as upwardly sloping path~, exa~ples o~ which are indicated at 612 and 624. In Figure 6~, an intervening strip 625 is present. Examples of a ~eries of successive 11 3 ~ f~

spots are shown at 602, 604 t and 606. It is to be understood that the spots continue to be applied across the entire pathway 612 and then continue on pathway 624, where further exemplary consecutive spots 614, 616, and 618 are depicted. Intervening spots, and additional pathway6 beneath 612 and above 624, have been omitted Por pu~poses of simplicity. Also, it i~ to be recognized that the igure is not drawn to scale, but the spots are drawn disproportionately large 1n comparison to the tubular for diagrammatic clarity.
~he calculation of the coverage of the sequence o~ spots can be accomplished readily, as follows. We first define the appropriate variables:

Rotation rate (rev/~ec) = R
Transducer height (inches) = H
Transducer wldth (inches) = W
Excitation Pulse Frequency (cyc/sec) = F
Tubular Inner Circumference (:Lnches) = C
Logging Rate (inches/seo) - L

Assume f~r now that the spot areas do not overlap.
2~ This case is show~ in Figure 6A.
First, the area of the uncovered patch 625 formed by the non~overlapping tops and bottoms of the ~pot-areas t~ounded by lines c & d) may be calculated. We first note that the upward travel of the sonde per revolution = L~R. Triangular portions of the spot-area on either side of the horizontal center, labeled a & b in Fi~ure 6A, can be reflected through the cen~er o~ the ~op o~ ~he spot to give an uncovered strip bounded by the scan ~ 3 ~

lines c ~ d. The Vextical dimension of the strip 625 ic equal to:
Upward travel - spot height - (L/R - ~) For each r~volution we can ma~ the ~trip a per~ect rectangle by drawing a line at the bottom which is at the lower end of, and perpendicular to, the upper helical path ~d). The resulting triangle 627 i~ moved ~o ~he other end to 62g to form a rectangle between lines c, d, and triangles 627 and 629.
The area o~ this strip, per revolution~ is just the vertical dimen~ion times the circumference, (L/R - H) * C. :
Now for the area uncovered between (horizontally) 6uccessive spot areas:
Time to move to next spot-area o l/F
Rotat~on velocity = Rev/sec * inch/rev = R * C
Distance between ~pot centers = (R * C) / F
G = Width of area uncoverèd per spot = ~(R * C) F) ~ W
Nu~ber of spot a~eas per rotation =
~#spots;sec)~(Rev/sec) - ]F/R
Area uncovered per spot = G * ~H
Area of patches not covered between successive spots per rev. = G * H * F / R
The total area availa~le to be covered per revolution i~ the product of the upward distance : and the circumference:
Area availa~ls for coverage per r~v. = C * ~ / R
Thu~ ~he total uncov~red area a~ a fraction is:
(L~ ~ Hl*~ *C/~_- Wl*H*F/R
C*L/R
Dividing through by the denominator gi~es:
Fraction uncovered =
(1 - ~*R/L~ F*W/~R*C)~*H*R/L (Eg- 1 1 3 ~

Also, the fraction covered - 1 - fraction uncovered = 1 - (1 - H*RJL) + (1 - F*W/~R*C))*~*R/L (Eq- 2) Now let A ~ H~R/L and B ~ F*WJR*C. The covered area can be expressed as a percentage of the total area~
Covered Area (~) =
lo 100. * ~ - t(l - A~ + A * ~1 ~ B)] ] (Eq. 3) The first factor, A, is a dimensionless number determining the amount of vertical ga~ ~or overlap). The second îactor, B, ~s a dimensionless number deter~ining the amount of horizontal gap (or overlap). The interaction of these two number~
determines the amount of coverage.
Note that if A = H*R~L = 1 and B = F*W/(R*C) =
1 then there are no uncovered areas. This corresponds to the case o~ adjacent spot-areas which touch at top and bottom. In this case, the values for ~ and W are the ~in.Lmum values ~or which there is 100% coverage, given R, L, C and F.
Note also that if A~l, B~l then there i more than 100% coverage. The terms in parenthesis in Eq. 3 bec~me negative and add to the coverage.
Eq~ 3 can be further simplified by canceling the unity and A terms. This leaves the following equation for the coverage.
Covered Area (%) - 100. * A * B
= 100. * H*R * F*W
L * R*C
= 100. * H*F*.W
L*C (Eq. 4) Thi~ equation implies there is no dependence of coverage on the rotation rate. This lack of - ;2 9 -dependence on R is due to the way coverage has been de~inad~ For these equations9 any time an additional area is probed, it is included in the coverage calculation, regardl~s~ ~ thi~ area had already been probe~ by another spot~area. ~hus, Eq. 4 show~ ~hat 100% area coverage can be o~tained by rotating the transducer at any speed so l~ng as H*F*W/L*C - 1. It does not imply that all areas o~
the pipe ara covered, however. To insure 100%
10 co~erage with all areas of the pipe being scanned, term A must equal unity and term B must equal unity.
In practice, coverage in exces~ of 100% is desirable, to gain added information about given 15 areas of the tubular ~urface. However, in this context of overlapping coverage, the very concept of "coverage'l beco~es a~blguous. (If two spots overlap, but there is still an area completely un-touched by any spots, i~ that greater or less than 20 100% coverage?) When conlsidering overlapping ~pots, then, re~ort is best made to diagrams, such as Figure 6B to determine the nominal transducer rotation rate R. In any event, it is possible to adjust the rotation rate, R, and excitation 25 requency F, in conjunction with the pull or logging rate, L, ~o ensure total coverage of a tubular surface during a survsy.
As can be seen from the above ~or~ula, the tubular circumference (C) is the only variable 30 which is not independently controllable by the sonde designer. Of the controllable variahles, it i~ the transducer size which is most directly relevant to the quality of resolu~on which is to be expected ~rom the device. ~deally, a small ~ 3 ~

transducsr height H and width W would be employed, so as to provide measurement of smaller pits on the tubular surface. In a practical system, however, the data rate o~ t~¢ in~ormation which must be transmitted uphole i~ a limiting ~actor on the minimu~ size o~ the transducer, even assu~ing that the other parameters in the equation are optimiæed for maxi~um resolution of detail. A smaller transducer surface size H*W require3 a greater bandwidth of information s~nce a greater number of full echo waveform~ must be detected and transmitted in a given time period.
Re~erring to Figure 6B, the concept of coverage of greater than 100% is indicated by areas surh as 626, which is covered by two spots. Area 628 has been covered by four spots. Area 630, like area 626, is covered by two spots, but the overlap comprises overlap of consecutive spots t rather than by spots in the next path. 0~ course, the amount of coverage as shown by the overlap of paths 624 and 612 (which overlap is defi.ned by lines 610 and 622) is easily controllable. By manipulation of any of ~he parameters o~ the above formula, the size of the overlapping zones 626, 628 and 630 can easily be ch~sen.
Although the transducer rotation rate R does not appear in Eq. 4, R could be adjusted to obtain deslred coverage J in addition to ad; usting the pulse repetition rate F and the logging rate L.
As the percentage of coverage increases, the condition of the tubular sur~ace Gan be more accurately determined. Thi~ more accurate deter~ination is made possible by the fact that ~ 3 ~

in~ormation about any given area may be gained Prom more than one Bpot I S acoustic echos.
Referring now to Figure 7, a timing diagram indicating the occurrence o~ a typical se~uence o excitat~on pulses and the envelopes o~ the resultant acou~tic echo~ are prese~ted. In a preferred embodiment, an excitation pulse of typical magnitude 300 volts and duratlon 50 nanosQcond~ i~ applied every 1000 microseconds.
This 1 kHz excitation pulse ~requency has been found to be a practical ex~itation pulse ~requency, given the fact that the xeturned acoustic echo pulses resulting ~rom a *irst excitation pulse 702 must die out before the appllcation of the subse-quent excitation pulse 704. I~ ths echos had not subsided, the transducer, which, in the preferred embodiment, both transmits the~ excitation pulse and receives the echos, would not be able to accurately receive the echos bscause of the overwhelming effect o~ the much stronger ~xcitation pulse.
Also, echos o~ a fir~t excitation pulse which had not been substantially attemlated would interfere with accurate reception o~ t~e echos of a subsequPnt excitation pulse.
It is to be understood that, regarding echo wavefo~ms such as thos~ illustrated in Figure 7, the echo waveform actually comprises a set of pulse-modulated sine waves 804 (Figure 8). What is typically represented in echo wa~forms (as in Figure 7l i~ merely the envelope 802 of these pulse modulated sine waves ~pulse modulation as used herein need not imply a pulse of a single magnitude, but may be an analog shaped pulse).
Also, it i~ not meant to be impl~ed that the ~ine waves in the various pulses 802 are meant to be in phase, or even well-defined sine waves per se. The high~r-frequency signal~ within th~ envelope may be the sum of interfering sine waves which are returned from a rough or corroded tubular surfac~
so as to be com~ined in al c:omplex way.
The present inventlon contemplates a particular embodiment in which 5 ~Hz sine waves are employed to achieve greater resolutlon than transducers of known syætems, which used frequencies of approximately 1 MEz or below.

c. Bandwidth RQduction The pre~erred embodiment o~ the present invention uses narrow, high-frequency pulses in conjunction with a pulse wave~orm analysis method which can determine the area and depth o~ corrosion pits much smaller than the spot size. For detecting pits of a required minimum size, larger transducers may be used 1:han was previously possible. The use of large:r transducers causes embodiments o~ the present inven~ion to posæess an inherent bandwidth compression, as compared with known systems.
In the pre~erred method according to the present invention, the full echo waveform is ~easured. The li~itation noted above in the Background of tha Invention section i8 avoided by recording both the front~surface and pit-sur~ace echos in the full ec~o waveform. In the case where the acoustic wave intercepts both the fron~ surface and the pit surface, the individual echos are analyzed to detect the pit and measure its depth.

Thus, a very small pit can be detected even thou~h the ultrasonic wa~e may impact an area beyond the per~me~er o~ the pit itsel~.
One advantage o~ the method according to the present invention is the reduction in data rate necessary to transmit data describing a given area.
Although the present ~ethod requires transmission of echo data for all significant echos in the wave~orm, this is still les~ data than that associated with known methods. In the following example, the data rates for the two methods are compared. Here, it is assumed for simplicity that the spots are square, and that the smallest pit area to be detected is a square 1/8-inch on a side.
It is also assumed that successive spots are immediately adjacent (with no overlap and no gaps, either horizontally or vertically). Finally, it is assumed that any surface c:oincident with the "effectiv2" edge o~ the spot area will be detected, whereas any just ou~side the edge will not be d~tected. In general, de~ection of sur~aces near the spot edge will depend on the slope o~ the sensor response contour (Fig. 5) and the detection threshold. The 'Jeffective~' size of the spot (considering the edgest possible echo charac-teristics) may thus be larger than t~e spot dimension as de~ined by the -12 dB contour (Figure 5). This enl~rged effective spot size may affect the following bandwidth compression calculations accordingly.
In known methods, the transducer spot area must be much smaller t~an the l/8 inch x l/8 inch square pit. For the pit to be detected, the spot area must co~pletely ~it inside the pit ~or at least one reading. Sinoe, in practical downhole scenarios, there i5 no control over the actual pla~ement oP the spot center in relation to the pit center, the "worst-case" placement must be a~sumed (i.e., the beam must be assumed to be center~d in the pit). Over ~his square pi~ all 8 spots around the center ~pot intersect non-pit areas at the edge~ (pit not detected~. Thus, to insure that at least one spot covering the pit doeR not intersect the uncorroded tubular sur~ace surround~ng the pit, the area o~ the spot mu~t be at most l/9th that of the pit area. For the square pit this gives a spot area of O.0017 sq. inches. Two readings, (1~ the first echo time, and (2) the first echo amplitude, must be recorded for each pulS2. ~hus, the number of readings per sq. inch is ~1/.0017) x 2 = 1,152.
This data rate will provide detection and depth me,asurement down to square pits 1/8 inch on a side.
For the method according ~o the presant invention, a much larger area is sensed for each pulse. In practice, a s~uare pit 1/8 inch on a side can be detected using a beam which is 40 times larger. Thus, the spot size for each pulse can be 40 x 1/8 x 1/8 = 0.625 ~q. inshes. Tha number of readings per pulse depends on the number of ~chos exceeding the threshold. On average there will be about 10 echos to be record~d~ Three features per echo must be measured: ~1) echo time; (2) echo amplitude; and (3) echo width. Thus, there are 30 values per pulse. The number of readings per sguare inch is 10 x 3~.625 = ~8---a factor of 24 smaller than that of th~ prior method.
The reduction in da~a rate for the new method is achieved at the cost o~ spatial resolution on ~3~P~ ~

the pit. The prior method will give a ~uch better localization of the pit on the pipe surface.
However, for most practical applications, the maximum pit depth for a ~iv~n region is of primary concern, rather than excruciatingly localized spatial resolution9 ~he maximum pit depth has been judged to be a good indicator of the condition of the tubular and the corrosion progress in the tubular.
lG
d. Full Echo Wav~forms i. A Simple_~xample Figure 9 i8 a timing diagram which illustrates with specificity the formation of a full waveform of pul~e echos. The horizontal position indicates time, and the vertical posi.tion indicates the physical position of the various acoustic wavefront~ travelling betwlsen the transducer .~ surface 204 and ~he front ,surface 108 of the tubular 100~ Figure 9 also de,picts the wave~ronts re~lected ~rom the surfac~ O:e an "ideal" (flat) corrosion pit which ha~ eaten into the tubular to the depth indicated at 924. (It should be no~ed that the tubular i8 presented in Figure 9 as a straight line, sinc~ this is a timing diagram.
Figure 9 is not a physical diagram as is, for example, Figure 2.) The ~ull waveform of acoustic echos is formed as follows. An excitation pulse excite~ transducer 202 so a~ to cause an acoustic pulse o~ high ampli~ude and short duration to emanate from the curved surface 204 o~ transducer 202. Assuming a ~3~ ~$P~ ~

sinusoid of 5 MHz ~re~uency, only two cycles o~ the sine wave ara generated. ~hQ wavefront travels at its known speed (1300 m/~econd in oil) through the li~uid medium occupying tha fipace between the transducer 202 and the tubular Pront surface 108, as indicated along path 902. Much of the acoustîc energy i8 reflected from the tubular's front surPace 108, and is returned a~ a "~ront ~urface echo" toward ~he transducer 202 along path 904.
The front ~urface echo ls indicated as peak 928.
Not all of the aGOUStic energy incident on path 902 is re~lected back ~oward transducer 202.
Some of the acoustic energy is transferred to the tubular material itself. That energy which is transferred traverses a path 906 at a speed in metal (typically 6100 m/second) which is different than th~ speed of the aaoustic energy in the liquid (1~00 m/second in oil) intexnal to the tubular.
Therefore, the slope of path 906, indicative of the speed o~ the acoustic wavefront, is di~ferent from that o~ path 902.
A portion of the energy which travels alon~
path 906 is re~lected at the outer surface 938 of the tubular 100 to be returned toward the tubular front sur~ace 108 along path 908. Some o~ this energy along path gO8 i8 ~n turn transmitted alsng path 910 toward the tran ducer 202, whereas some of it is re~lected from tubular front surface 108 back into the tubular material along path 912.
A process of repeated and alternating transmi~sion and re~lection (indicated along paths 914, 918 and 920) is contlnued until virtually all the acoustic energy has been lost to the environment and to conversion to thermal ener~y.

Various acou~tic echos produced ~y energy escaping thQ repeated internal reflections within the tubular 100 are transferred back toward the transducer 202 along paths glO, 916, and 922.
Since ener~y is being continually diminished in the internal reflectio~ process at the tubular surfaces interfaces 108 and 938, the magnitude of acoustic pulse wavefronts are continually decreasing in power as time pa~ses. This diminis~ment in energy is indicated by the roughly exponentially decreasing magnitudes of echo peaks 930, 932, and 934 which are shown as repre~ent;ng the acoustic energy returned along paths 910, 916, and 922, respectively. Collectively, the set of peaks ~30, 932, and 934, etc. are referred to as a 7'multiple structure" (M5). By a convention used in this di~closure, the front surface echo is not considered a part of the multiple structure.
It is being assumed in this explanation that an ideal corrosion pit, having a fla~ surface located at 924, occupies an alrea on tubular front surface 108 which is smaller than the spot size.
The acoustic energy which enters the tubular's corrosion pit to a depth indicated by 924 will take a longer time to traverse the round-trip pathway 902 and 926 than the front surface echo alon~ paths 902 and 904~ The returned 'pit sur~ace echo"
produces a peak of acoustic ~nergy at 936.
This pit 5urface echo 936 occurs at a point in time which is determined by tha distance between tubular front surface 108 and pit surface g24.
This distance is independent of the thickness of the tubular (unless, of course, the corrosion pit ha~ eaten entirely through the tubular to actually 7 ~

form a hole in it). The pit surface echo can therefor~ occur at any point between, or even be superimpos~d on, the multiple structurs'~ peaks 930, 932, and 934. ~h~ depiction in Figure 9 of pit surface echo 936 a~ oacurring substantially between ~ultiple echo peak~ 930 and 932 is thus arbitrary.
The acoustir energy deteated by the transducer is first envelope-detected to reduce the signal bandwidth (to approximately 2 MH~ in the preferred embodiment). Th~ envelope is then sampled at a high frequency (typically 5.0 MHz) in the preferred embodiment. Thes~ discrete-time samples are digitized so as to allow formation of a full echo waveform such as that indicat~ in Figure 9.
ii. Act~al Example Figure 10 presents an actual full echo waveform which is characteristic of those which are likely to be encountered in practice. The dif~iculty in deciding whether a particular con~iguration constitutes a peak (and, if so, how to distinguish whether a particular formation comprises only one or a plurality of peaks), is that confounding signal 6tructures (such as multiple structure~ ~rom both the front surface echo and ~rom any pit sur~ace ~chos) are mixed in with the directly reflected surfac~ echos. It is the collection of all these peaks, whether sharply defined or disper~ed, which aonstitu~e the full echo waveform. It ~5 ths ~ull echo waveform that i~ analyzed to det~rmine which assumed peak is the front surface echo, which echo(s) are pit s~rface 13~`6 ~

echo(s), and which echo(s) are multiple ~tructures as ociated with either the front ~urface or plt surface~s). (I~ it is the ~ultiple ~tructures which are khe sign~ls of interest, for example, in determining remaining wall thickness, then the front ~ur~ace echo and any pit surface echos would more properly be called the ~confounding signal structures.") In th~ preferred e~bodiment of the invention, the acoustic energy which i~ detected by transducer 202 is continuously measured and digital representations of the magnitude of the r~turned acoustic energy are made at regular ti~e intervals.
on the basis of this discrete-time digital information, the rule-based Expert System according to a pre~erred embodiment o~ the present in~ention operates to determine the most likely tubular sur~ace condition which could be responsible for the digital full echo waveform under sonsideration.
Based upon analysis of the xeturned, full echo wave~orm shown in Figure 10, echo 1002 was dete~mine to be a ~ront-surf,aae echo. Regularly timed echos 1004, 1006, 1008, and 1010 form a multiple structure. Various echo peaks never achieve a threshold value 1012 which is s~gnificantly greater than the noise floor indicated by the slight perturbations above 0.0 on the vertical axi~. Multiple ~tructure echo 1010 is, in this example, one of thosR echos with diminished amplitude.
According to a preferred em~odiment of the present invention, such a threshold 1012 may bP
establi~hed so as to eliminate ~rom consideration 'Ispurious" echos which are of un~nown origin. I~

'p~

the threshold 1012 had been ~et slightly higher, echo 1020 would hav~ been eliminated from con~ideration. Although it i5 known in thiR
example that echo 1020 wa~ in ~act a spuriouæ echo, echo 1020 could rea~onably be interpr~ted as a pit echo. Smaller~amplitude echos are advantageously eliminated from consideration because the amount of ener~y which contributed to their formation is correspondingly small, and are more likely to have been caused by the dispersive ef~ects o~, for example, mode conversion.
Echos other than tha front-~urface echo 1002 and the known detected multiple structure ec~os 1004, 1006, and 1008 are first as~u~ed not to be false echos. Echos such as 1014, 1016, and 1018 wexe determined to be pit echos.
A comparison of a ~ull echo wave~orm o~ Figure 10 with the simple full echo waveform of Figure 9 demonstrates the utility of a more sophisticated analysis method for determining the cause of each echo in the full scho waveform.
In the discussion relating to Figure 21, the ~ull echQ waveform portrayed in Fi~ure 10 will be used as an example to demonstrate the process by which the ruIe-based sys~em, according to the present ~nvention, isolates and identifies each echo exce~ding a known or calculated threshold.

2.
Information about the condition of the tubular is derived from the full echo wave~orm. Tha depth and area of corrosion pits is valuable information.

7 ~

a. ~it De~t~

First, the depth o~ a plt may be derived from the time separation between the front suxface echo 928 and the pit sur~ac~ echo 936 (Figure 9~. The dep~h of the pit Dpi~ i8 equal to one~half th~
product o~ the speed of sound in the liquid Vl and the time diffarence Tp ~ between the front surface echo 928 and the pit surface echo 936.
Symbolically, Dpit = . 5 * Vl * Tp_f The tubular thickness can be datermined from multiple structures which are present in the full echo wavefor~. The ~nly multiple structure illustrated in Figure 9 is that which comprises peaks 930, 932, and 934. This multiple structure is based on reflections between the inner surface of the tubular 108 and the t~ular's outer surface 93~. Based on this multiple structure, the tubular thickness ~s equal to one-half the product of the speed of sound in the tubular material ~presumptively m2tal) and the time separation of the multiple echo peak~.
Th~ eparation of adjacent peaks in a given multiple s~ructure should be constant, regardl~ss of which paix o~ ad~acent peaks are chosen; that is, the ti~a separation between peaks 930 and 932 should be identical to the time separation between 932 and 934. A more accurate estimation of the tubular thickness can be aained by using a weighted average of the various peak separa~ions. That is, an average of the 928-930, 930-93~, and 932-934 ~ 3 ~

separations could be used, with greater weighting being given to the earlier (and pres~mably stronger) peaks.
Determination of this averag~ peak separation lends itself readily to solution by a rule-based system. The advantages of a rule-based solution become ~Yident in considering the case when a pit echo 936 occurs nearly coincident in time with a multiple structurs peak. In this case, the location of the multiple ~tru~ture peak is indeter-minate because the pit echo distorts the multiple structure peak. A rule-based system cou~d easily analyze the multiple structure's characteristics to determine if one or ~ore of the multiple ~tructure peaks was anomalous. In this case, the anomalous peak could be ignored in the calculation of the multiple ætructure peak separation~
Althou~h not specifically illustrated in Figure 9, multiple structures may be associated with a pit acho 936 as well as a front surface echo 928. In order for a multiple structure to be formed by a pit, there must be present a large flat pit whose bottom is perpendic:ular to the incident acoustic energy. Due t9 the nature of "real-world"
corrosion, such large, flat pits (and their accompanying multiple -~tructures~ are sta~istically uncom~on. (Any multipl~ stxuctures associated with pit echo 93~ have been omitted ~rom Figure 9 for purposes of simplicity.) A measurement of the remaining wall thickness (RWT) (the distance between surac~s 924 and 938) behind a corrosion pit can be determined in the same ~anner as described above, for the case of a multiple structure defined by the front sur~ace.

13~9 ~3 In the case of a pit multiple ~tructure, the minimum remaining wall thickness can be determined from the multiple ~tructure having the minimu~
separation of all multipl~ 8trUcture8 encountered.
Thi~ minimum remaining wall thicknes~ gives a direct indication of how close a given sorro~ion pit has come to the outer ~urface o~ the borehole tubular. This direct indicatlon of RWT can be used in conjunction with thQ direct measurement o~ pit d~pth (descri~ed above) to determine the progress of a gi~en corrosion pit with a higher degree o~
confidence. Increased (or decr2ased) coniden~e derives from an ability to verify (or contradict) one measurement by comparing it with another. ~his confidence consideration is ideally suited for solution by a rule-based sy~tem.
Ideally, the depth of a given corro~ion pit and the remaininy wall thickness associated with that corrosion pit should total the nominal thickness of the borehole tubular. Any deviatlons or inconsistencies between thQ pit depth indication and the remaining wall thickness indication may be analyzed using a rule-based system so as to quantify the confidence level in a given conclusion a~ to the corrosion's progress~ The interaction of the pit depth ~easurement and the remaining wall thickness measurement in the rule-based analysis is deæcribed in greater detail below in the section on Rule-Based Analysis of Full Echo Wave~o~m.
Although the pre~erred embodiment o~ the present invention con~emplate~ ~etection of i~perfections on the inner surface o~ a borehole tubular, information concerning the outer surface i B also easily determined.

13 .3. '~

In practice, corrosion pits almost never take on the "ideal" character indicated in Figure 9.
That is, the pit surfacQ of an actual corrosion plt is usually much more complsx, as was shown in Figure 3. A real-world corrosion pit does not have a single depth, as was ind~cated by the di~tance between surfaces 108 and 924 in Figure 9. An ac~ual corrosion pit is likely to have a range of depths, so that a single p~t echo (936 in Figure 9 is not to be expected.
Referring again to Figure 3, paths of various portions of an acoustic pulse emanating from the curved surface 204 o~ transducer 202 are illustrated.
Paths 302 and 306 depict the paths which are followed by the acoustic energy which is incident upon a corrosion-free area of the tubular surface 108. The fuli waveform acoustic signals Prom these portions of the wave ~ront are the same as those produced in front surface echo 928, and ~ultiple echo structure 930, 932, and 934 tFigure 9).
It is po~sible that some portion of a corrosion pit may be flat and perpendicular to the wavefront, so that a portion of the acoustic energy would be directly refleeted bacX toward transducer face ~04. This portion of the acoustic energy would travel along path 304 (Figure 3, described above, in the Background o~ the Invention section~, and would produce an echo signal in the same way as peak 936 (Figure 9) wa~ produced~
To di~criminate what is a p~ak Prom what i~ a random nolse spike, a "detectlon threshold" may be establishQd. Any potential "peaks" whiah are smaller in. magnitude than this threshold are r3J ~

-~5-ignored, in accordance with a preferred embodiment of the inven~ion~ In practice, it has been ~ound that the "detection threshold" may advantageously ba determined as midway between: ~13 the noise level when no echo ignals are present, and (2) the magnitude o~ the fifth echo in a mult~ple structure for uncorroded pipe~
This preferred threshold, as i~ evident from its def~nition, i8 ~ voltage which i~ der~ved empirically. The threshold's empirical derivation implies its caloulation need not explicitly depend on such factors as excitation pulse magnitude, transducer characteristics, and attenuation of wavefronts in the liquid medium.
It is possible that the pit surface will cause acoustic energy to be inally reradiated in a direction which does not inte:rsect the transducer face 204 at ~11. This eventuality is indicated as path 310 ~Figure 3, above)O This loss of ac~ustic energy would be evidenced in a full waveform only by an apparent loss o~ to~al energy when integrating over the entire full echo waveform~
In practice, a very high correlation between pit depth and true dept~ has been obsexved using the above techniques. Pit depths can be determined to within a tolerance of 20 mils (.02 inches), even for very convoluted surfaces.

b. Pit ~re The area of a given corrosion pit can also be determined from the echo waveform. Briefly, the pit area as a per~entage o~ the txansducer spot size area could be determined fro~ the amount of ~ 3 ~
-46~

acoustic energy which enters the pit as a fraction of the total acoustic energy incident upon the inner sur~aoe o~ the borehole tubular. ~owever, not all the ener~y which enters a corrosion pit is refle~ted back to the transducer for detection, a~
was indlcated ~n the discussion of Figure 3. If ths pit area measurement depended on the amvunt of energy reflected fro~ the pit, the calculation of pit area would yield an incorrectly low value. A
significant portion of the acoustic energy which enters a real-world corro~ion pit is absorbed or deflected and never returns to the transducer for a reflected energy calculation.
It has been found that a much more reliable determination of pit area can be derived from the amount of acoustic energy which does not anter the pi~. The amount of acoustic energy which does not enter the pit is, by implic:ation, the acoustic energy whi~h. falls upon the uncorroded tubular surface. This energy which falls upon the uncorroded tubular sur~ace i~ largely reflected back towards the transducer to be detected. That portion of the tubular surfac6! which is uncorroded reflects incident acoustic wave~ with much more predictability than the pit surface. Therefore, the amount o~ reflected acoustic energy which the transducer receives ~rom a tubular sur~ace ~nown to be corrosion-~ree provides an excellent gauge as to the ~aximum amount of energy which can be expected to b refleoted from actual tast surfaces. ~he amount o~ acoustic energy which is detected by the transducer at a time when a fron~ surface echo is expected is therefore inversely related to the area of corrosion falling under the transducer's "spot."

~3~ ~P~

~he calculation of pit ar a Apit a~ a fraction o~ tran~ducer spot size area Aspo~ can be summarized as a calculation o~ 1.0 minu~ a guotient whose numerator is the ~easured magnitude o~ the front surface ~cho HFSE, and whose denominator is the predetermined or precalculated ~agnitude of the front surPa¢e echo for a corrosion-free area of ~ubular HFs~,O-Symbolically, A i~ 0 ~ HFSE/HFSE,O) Aspot The quotient represents the ~raction o~ acoustic energy which is reflected from (and ideally wa~
incident upon) the uncorroded surface. Subtracting this quotient from 1.0 therefore yields a ~raction indicative o~ the remainder of the acoustic energy which, by implication, entered the pit and was not reflec~ed in the front surface acho.
This pit area measurement technique directly measures only the area oP corrosion within the transducer spot~ The precise shape of the corrosion pit, (and, in some instances, whether ~5 there is more than one corro~ion pit in he spot) is not determined. (The presence of plural pit ~urfaces ~.s easily detected in most circumstances ~rom analy~is o~ the echos in th~ full waveform, provided the pit sur~aces are at different depths.) But such determinations are not necessary in ~ost practical applications, given the spot size ~.75 x .75 inches) and spot overlap ¢apabilitie~ o~ the preferred e~bodiment.

~ 3 ~

The denominator in the quotient (uncorroded front surface echo magni~ude) may ~e determined by a variety o~ techni~ues. ~ Gin~le value may be determined either heoretically or empirically prior to a logging run. However, this method does not allow ~or any long-term differences in such parameters a tubular diameter and inadvertent de-centering of the sonde within the tubular as it ascends. A variation in either of these parameters would cause a misl~adingly large or small ~ubse-~uent calculation of pit area.
Another ~ethod of calculating the maximum expected front surface magnitude is to continuously monitor what this maximum magnitude should be for uncorroded pipe at any given depth in the oil well.
This maximum expected magnitude may be continuously redeter~ined based on, ~or example, a moving average ~or a weighted moving average) of ~he mos~
recently encountered front sur~ace echos so as to yive a continuously updated indication. Factors such as a deviation from true circularity of the tubular can bQ accounted for by considerin~ only those previous front surface lecho magnitudes which lie directly below the transduc~r spot presently under eonsideration. Allowance may also be made for the transition ~rom one section of tubular to another through advance preparation for the transition, which transition may entail a mea urable change in tubular diameter because of differences in diameters in different sections of tubular a~ ~hey are ~anufac~ured, or misalignment as they are installed. Açcounting for variations in such paramsters l~nds itself readily to analysis by a rule-based system.

-4~-In practice, front sllrface echos are not always ideal in shape. Front sur~ace echos are often distorted due to ~llght surface irregulari-ties, called "surface roughne~sN for this discussion. Surface roughness may actually comprise ths beginning~ of a corrosion pit, but have caused so little deterioration in the surface that it cannot yet be called a corrosion pit.
Surface roughness may cause a disp~rsion (decrease in ma~nitude and increase in width) of the front surface ~cho because of separate re~lections ~rom the imperfect surface, as was indicated in exaggerated form in the Figure 3 corrosion pit.
In the case of surface roughness, two overlapping peaks may constitute the front sur~ac echo. In this case, since al:L the acoustic energy arrives back at the tran~ducer at a time when a front surface echo is expected, the magnitudes o~
any two peaks which occur :in this time window should be added to obtain the numerator in the above quotient.
The ~uestion arises a-~ to how to determine when an echo which is the second echo to arrive ceases to be a co~tribution to the front surface echo and begins to be a pit surface echo. Stated another way, the issue i8 how to determine when ~wo ad;acent peaks are overlapping, and when they are not overlapping.
Two peaks axe de~ined to overlap when the time separat~on o~ the ~irst and the second echo is less than the average width o~ the two echos~ The width of echos in turn may be de~ined in terms of the width o~ the pulse envelope at the "detection ~3~ d~

threshold", described above , in the section entitled l'Pit Depth'~.
~en two eohoE~ overlap 60 as to conceal o~e lobe of t~e pulse erlvelope, the ~'edge't of the echo may be defined a~ the point closest to the peak at which there is a change in envelope slope polarity.
A preferred method o~ implementing the dete.rmination of the edge of the echo pulse is described in greater det il below~ in the section on Rule-BasF~d Analysis of Full Echo Waveform~.
Less ri~orous correlation has been observed with regard to the pit area measurement than with pit depth measurement, above. An uncertainty of approximately 30~ o~ the spot size must be allowed in pit area determinations as described above. But given the small spot size (.75 x .75 inches~ in a preferred embodiment of the invention, as well as the fact that pit sizes smaller than the spot are indeed reliably detected, the improvement of the preferred embodiment of this invention over known systems is substantial.

3. Transmission of Full Echo Waveform formation The acoustic in~onmation received by the transducer, in the form o~ a full echo waveform, must be transmitted to the means for analyzing the full echo waveform. ~his transmission o~ full echo waveform information may be accomplished as de~cribed in this section.
Re~erring to Figure 11, an uphole trans~ission system according to a praferred embodiment is presented in block diagram fo~m. The illustrated uphole tr~nsmission system may ba integrally O
~5~

connected with the means which produce the excit~tion pulse ~eguence which cause~ the full echo waveforms which are later analy2ed.
Timlng and control block 1102 emit~ a pulse generat$on ~iignal on line 1110. ~he pulse gensration ~ignal on lina 1110 travel~ to pulse generator 1104 for the gen~ration o~ the short-duration, high-magnitude excitation pulse which is sent through rotary transformer 1106 to curved transducer ?02. Rotary transformer 1106, known in the art, makes the bidirectional electrical connection between the transducer and the remainder of the downhole circuitry. The pulse generation si~nal on line 1110 is also input to the start input of counter 1128, whose purpos~ and ~unction will be described below.
A motor causes both rotary transformer 1106 and transducer 202 to rotate as the entire sonde is drawn up ~he tubular. A magne~ometer 1107 with associated electronics of a type known in the ar~
emits a signal once per revolution. Thi~ signal is sent uphole on the wireline 1134, advantageously b~ing sent in coordination with ~h~ full echo waveform information, through a serial data transmission syst~m 1132. The coordination of the signal from the magnetometer 1107 with the full echo waveform tnformation allows verification of the vertical alignment of 6pots on successive passes. It can also provide absolute directional orientation information. For an example of a known rotary transformer and magneto~eter arrangement, ~ee U.S. Patent 3,503,038 to Baldwin.
AB described above, in the section entitled "~eneration of ~ull ~cho Waveform,~ transducer 202 i ~ 3 ~

emits an acoustic pulse which strike~, and i8 re~lected from, a target surface such as an oil well borehole tubular. The transducer senses the set of echos which comprise the full echo waveform and co~unicates this information in ~he Porm of an analog voltage through ro~ary trans~ormer 1106 to variable-gain amplifier 1108.
The ~ull echo wave~orm i5 passed to the timing and control block 1102 Yia lines 1122 and 1136.
Within the timing and control block 1102, an automatic gain control circuit, ~AGC) detects the peak amplitude of the wave and generates a ~ain control sign~l. The AGC circuit insures that the output ~rom variable gain amplifier 1108 does not fall outside o the dynamic range o envelope detector 1114. The AGC circuit also insures that a waveform of relatively constant peak amplitude is applied to the input o~ envelop detector 1114. AGC
circuitry is well known to those skilled in the art.
Timing and control bloGk 1102 controls the gain o~ varia~le-gain amplifier 1108 by sending a gain control signal along line 1112.
Adva~tageously, timing and control block 1102 produces a gain signal on ~ins 1112 so that the amplifier makes full use o~ its dynamic range without clipping the signal input to it from rot~ry transformer 1106.
Line 1138 is used to send a digitized value for the gain control signal, as applied to variable-gain amplifier 1108, to th~ serial data transmis~ion syste~ 1132, where it is multiplexed with other data to be sent to the surface. ~his value is later used to convert the transmitted ~3~ 7~

wave~orm voltages back to the or~ginal signal values that appeared at the input of variable gain a~plifier 1108. Where the optional Echo Feature Extraction and screening block 1103 ~ used, the digitized gain control signal may instead be applied to block 1130 ~or down-hole preprocessing.
The full echo wave~orm thus gain-adjusted iq input to envelope detector 1114. In a preferred embodiment this envelope detectox 111~ has a dynamic range of 60-~0 dB, 80 that even very small peaks in the waveforms are preserved. Envelope detector 1114 ef~ectively reduces the bandwidth of the su~stantially sinusoidal ~ull echo waveform ~rom ampli~ier 1108 to a monopolar waveform of substantially lower bandwidth. In a preferred embodiment, the received ultrasonic echo~ comprise short pulses of approximately 5 MHz center frequency. By rectifying the oscillatory pulses and sffloothing the envelope-detected signal with the low-pas~ filter 1116, the bandwidth is reduced to approximately 2 MHz.
The smoothed monopolar full echo wave~orm is next input to logarithmic amp:Lifier 1118 to reduce the eff~ctive dyna~ic range of values of the full echo waveform magnitude so as to optimize ~he use of the wireline transmission data and to ensure maximum resolution capabilities o~ high-speed digi-tizer 1120, which receives its output. Typically, the 80 dB range o~ envelope detector 1114 is reduced to the approximately 40 d3 range of an 8-bit digitizer advantageously employed as element 1~20. High-speed digitizer 1120 convert~ the logari~hmically compressed, smoothed monopolar ~ull echo waveform from its analog representation to a ~ 3 ~ 7 ~
~-54-digital representa~ion. Preferably, digitizer 1120 operates at a 5 MHz sampling rate, so that an average echo pulæe encompasses about f iVQ samples~
The digitized, logarithmically compressed, smoothed monopolar full ~cho waveform is stored in a memory 1121 which receiYe~ tha output of digitizer 1120~ Advantageously, ~emory 1121 is a recirculating memory which continuously store-~ the digi ized information. ~The control of the memory 1121 is accomplished by the arrangement o~ the timing and control blsck 1102 and threshold detector 1124, described in greater detail below.) The memory 1121 has enough storage locations to store slightly more than the amount of data comprisin~ one full echo wav~form. The continuous storage of data from digiti~er 1120 is halted at a time determined by the arbitrarily chosen "end" of the full echo wave~orm (an interval usually on the order o~ 20 microsecond~, as indicated in element 706 of Figure 7). What remains in memory 1121 when storage is halted i8 thus a digitized full echo waveform, including data from a small amount of "lead time" be~ore the full ~cho waveform, due to the slight oversizing of the memory. It is the information in thi~ digital r~presentation o~ the full echo waveform, "frozen" in circulating memory 1~21, which is transmitted uphole.
Thls digital representation of the full e~ho waveform may be transmitted uphole in at least two forms, depending on the degree o~ sophistication of the downhole electronics. The echo ~eature extraction bloc~ 1130 is indicated in dashed lines to indicate the choice which the designer has in 1 3 ~

including or omitting it hefore transmitting the information uphole.
In a irst, preferred embodlment, echo feature extraction and ~creening block 1130 is present.
Under control of the timing and control block 1140 block 1130 analyzes the waveorm digitally according to principles presented below, in the discussion related to Figures 12 and 13. In this embodiment, iniormation as to the quantity of echos in a given full echo wave~orm, as well as the height, width, and '~location" (time of occurrence) of each echo in the wave~orm, is transmitted uphole. Also, wavaforms which are clearly characteristic of an uncorroded tubular may be screened out downhole, so as to optimize the long-term utilization of available transmission bandwidth. After being transmitted uphole, the extracted echo features may ba analyzed by the rule-based system according to a preferred embodiment of the present invention.
The ~lrst embodiment ~ust described presumes a ~air degree of computational ~ophistication to be inherent in the sonde body. ThiS embodiment could be implemented using microprocessor or special purpose electronics technology well within the ability o~ one o~ ordinary skill, upon a rea~ing of the text related to Figures ~2 and 13.
In a eeaond, alternative embodiment, the raw ~amples of the full echo waveror~ temporarily frozen in memory 1121 may be immediately transmltted uphole 80 that the echo ~eature extraction method and screening ~unction may be performed there, perhaps by the same computer which execute~ the rules of the rule-based system. In ~L 3 ~ r~1 ~

the second embodiment, echo feature extraction and screening block 1130 i~ not present, and th~
digitized full echo wave~orm ~rozen in memory 1121 is ~ent vla serial data tran~mission ~ystem 1132 up the wire line 1134 without the echo height, width, and location first being extract~d.
A~ter the digitized full echo waveform has been read ~rom memory 1121, timing and control block 1102 ~ends a restart signal to the memory 1121. The restart signal "unfreazes" the circulating memory so that it begins to conti-nuously stora new digitized waveform information from digitizer 1120 in preparation for the next full echo wave~orm.
In the first embodiment, the restart s~gnal is generated after echo feature extraction and sareening block 1130 has ~inished reading the current digitized ~ull echo waveform from the memory for analysis. In the second e~bodiment, the res~art signal is generated alfter the serial da~a transmission system has read the raw digitized full echo waveform for transmission uphole.
Ti~ing and control block 1102 is advantageously constructed of a micropr~cessor, but may also be implemented u~ing large scale integrated circuit (LSI) hardware. A hybrid of microprocessor, L~I and discrete components also lies within the scope of the present in~ention.
Advantageously, the same processor h~rdware which implements the timing and control circuit 1102 may also execute the ame functions a~ echo feature extraction and screening block 1130.
Generally, the ~unction~ performed within the echo feature extraction and screening block 1130 7 ~

are more complex and ti~ewise demanding than those functions performed with$n timing and control block 1102. Timing and control block ~102 serves the main purposes of the comparatiYely long-duration timing functions, such a~ generating the excitation pulses on lin 1110 or the choice o~ the ~ain factor on line 1112~ Block 1102 al50 per~orms the timing delay ~easurement commenced upon receipt of an input signal ~rom thr~shold de~ector 1124 SQ as to produce signals which "~reeze" and '1unfreeze"
the contents o~ memory 1121. Th ability ~o implement these ~unctions using existing technology lies well within the scope o~ one of ordinary skill, and will not be discussed further.
Referring again to Figure 11, the following discussion describes the interaction of timing and control block 110~, threshold det~ctor 1124, and counter 1128 wlth the blocks already described.
The main purpose of threshold detector 1124 and counter 1128 are to ensure that only the full echo waveform itsel~ is sent uphole, and not the generally ~lat signal before it. Referring briefly to Figure 7, there is a delay on the order o~ (in one particular example) 60 mic:roseconds between the excitation pulse 702 and the earliest possible arrival time of an echo from the target surface.
During this time, no meaningful information could be received from the transducer so that its output (and the output o~ digitizer 1120) should be ignored by ser~al data transmission system 1132.
It is only during the time o~ the ~ull echo wave~orm 706 that meaningful in~ormation can be received ~rom the transducer.

.~ 3 ~
-58~

In operation, ti~ing and control block 1102 (Figure 11) ~ends a pu152 generation ~ignal along line 1110 to start counter 1128 counking. Counter 1128 counts ~o as to measure the time interval between the pulse generation ~ignal ~which creates the excitation pulse) and thQ ti~e whan the ~irst echo rises above a thre~hold. The time when this first echo rise~ above a threshold i8 determined by threshold det~ctor 1124 upon n21ysis o~ the smoothed monopolar ~ull echo waveform input to it along line 1122 from low pass filter 1116. The threshold detector output stops counter 112X at a tima when the use~ul in~ormatio~ has started to b~
received. The value of counter 1128 at this tim~
thus marks the time o~ the first-received echo in relation to the original excitation pulse.
This information fro~ counter 1128 is u~eful in determining, in real ~er~s, how far away from the tranfiducer ~he nearest portion of the target surface was. The threshold dectector's si~nal indicates that the period o~ useful information on the full echo waveform has be~un. Time period 706 (Figure 7~ indicates this period of useful infor~ation to be approximately 20 microseconds~
Regardless o~ which embodi~ent is chosen ~reflecting the inclusion, or exclusion, o~ the echo ~eature extraction and screening block 1130 in the downhole electronics), the information reflecting the dalay batween the excitation pulse and the rising o~ the ~irst returned echo above the thre~hold must be transmi~ted uphole. If the first embodiment is employed (echo ~eatura extraction a~d screening block 1130 being present downhole~, then the value of counter 1128 is ef~ectively ~ 3 ~ 9~3 incorporated into the location and data entries output from echo feature extraction block 1130, I~
~he second embodiment i~ employed (echo feature extraction and creening block 1130 be~ng uphole), then the value of counter 1128 is multiplexed with the raw digitized full echo waveform information passed directly from digitizer 1120 to the serial data transmiss~on system 1132~
Serial data transmission system 1122 i~volves communications apparatus, and modulation and multiplexing schemes which are within the ability of one of ordinary ~kill in the art. See~ for example, U.S. Patent No. 4,415,895 to Flagy, entitled "Well Logging Transmission System." The ability to perfor~ the uphole transmission of full echo wave~orm data, in whatever form, is within the ability o~ one of ordinary skill, reference being made to the present di~cussion and the related di~cussions re~erenced within it.
As described above in the section on Bandwidth Reduction, the present invention utilizes an amount of bandwidth which is much ~maller than would be usad by known systems, ~or a given requirement as to minimum-sized detectable corrosion pit. Thus, for a given bandwidth restriction imposed on the downhole-uphole co~munication~ channel and a ~iven minimum detectable corrosion pit 6ize, the use of the present invention 18 larger transducer size and optional downhole echo feature extraction and screening functions allows more rapid coverage of the tubular 6urface ~han would have been possible by using the smaller transducer necessitated by use of the known system described in the 8ackground of the Invention section.

1 3 ~

Standard transmission methods ~ay be employed.
If the transmission bandwidth imposed by the physical configuration o~ the downhole-uphole transmi~sion channel prove~ ~ burdensome limitation, various parameters may be changed so that the desired amount of information may be transmitted uphole. For example, the logging rate L may be reduced, p rhaps in conjunction with a reduction in the excitation pulse frequency F.
Also, the frequency at which the analog ~ull echo waveform is sampled could be reduced so that, in the embodiment in which the raw digitized samples are transmitted uphole, an e~fective reduction in the amount of data may be achieved.
Altern~tively, if the extracted echo features are the information w~ich is actually transmitted uphole, the threshold below which echos are to be ignored may be raised, so that, statistically, less information will have to be transmitted uphole. O~
coursa, any o~ these methods o~ reducing the amount of bandwidth will compromise some other performance feature, such as downtime o~ 1:he oil well (because of a slow~r logging rate), or reduced sensitivity to smaller corrosion pits (due to raising o~ the threshold). Arbitration among these various competing factors is well within the ability o~ one of ordinary skill in the art and will not be detailed further here.

4~ Rule~ ed An~lysis of ~ cho Waveforms What follows is a description of the preferred embodiment o~ the rule-based system which reliably and auto~atically analyzes and interpret~ the ~3~

sampled full echo waveform~ returned from the transducer.

a. Texminology S
The following table $~ o~er~d to present to the reader the m~aning of various terms which will be used throughout the di~cus~ion of the rule-based system according to a pre~erred embodiment o~ the present învention.

List of Terms Used to Describe Echos and Echo Structures Signal - Response of ~ensor (transducer) for a single location with respect to the pipe tubular wa:Ll. Composed of a series of echos at various times.
Signal contains in~ormation on corrosion only over a localized area on the pipe (typically 3/41- by 3/4") equal to the "spot area".
~5 Echo - A distinat porl:ion o~ the ~ignal characterized by a rapid increase in values with time, a peak value, and a following rapid decay in value.
Echo - Index (a.g. "sample #26"3 o~ thP peak Location value within the echo. The term implies a "location" in time, not physical location on the target sur~ace.
Echo - Maximum value within echo width.
Height Echo - The nu~ber of sample points between Width the start and end sampl~s of the echo as defined by the echo extraction program.

Pit - A surface of the inner pipe wall which is deeper than the uncorroded surface.
There can be ~everal pit~ iden~ified from one ~ignal.

Front - An echo which is generated ~y Surface rsflection ~rom the first surface lo Echo en ountered by the ultrasonic s~gnal.
This echo will be generated by the uncorroded innar surface except in the case where an area lar~er than the ~pot size i~ aorroded away.
Leading - An echo which i the ~irst echo in a Echo multiple structure. It i~ gener~ted by the front side of the sur~ace causing the multipIe structureO
Dominant - A "dominant echo" must be some factor Echo (typically 2) larger in value than any other echo.
Multiple - A multiple structure is characterized Structure by a "strong'7 leading echo followed by a series o~ echos with successively lower values at locations which are separated by a fixed time (sample) interval.
Neighbor - Echo #1 i5 a neighbor o~ echo #2 if Echo there is no other esho between these two.
Spuriou~ - A spuriou~ echo :15 onc that has been Echo generated by some process which is unknown and for whlch ther~ is no prior:information. Most of these are caused by interferenee of the signals at a surface (peak splitting). They : are not pertinent ~o t~ e interpret~tion, are normally small and should not be identified as pit echos.
S~ructure - A set of echos having physical and~or acoustical sign~icance (e.g., a multiple structure, front surface echo, pit echo, etc.).

_~3_ b. General In~I~ductio~ ~L~ 5 Intelli~ence ~AI~

The general concept of a rule-based sy te~ i8 known in the art. ~uch system~ fall under the general description of "artificial intelligence"
systems. Tha ter~ "artificial intelligence"
reflects the fact that AI systems emulate the conscious (and perhap~ subcons ious) reasoning process actually occurring in the mind of a human expert.
Generally, expert systems are conceived as comprising three major components: the ~nowledge base (including knowledge sources and a "blackboard"); an inference engine; and a user inter~ace.
The knowledge base comprises the raw declarative and procedural knowledge of a human expert. The procedural knowledge takes the form of "rules'~. Each rule compri~es a set o~ "IF"
conditions and a set of corresponding "THEN" action statements. When all of the "IF" aondition~ are satisfied, based on analysis of the declarative knowledge lor "blackboard"), then the inference engine may cause the rule to "fire" ~be ~xecuted~
When a rule is ~ired, the l'THEN'' actions are performed by the computer, usually resulting in a change o~ state of some ~n~ormation on the blackboard. Thus, during the operat~on ~f the expert system, the amount and nature of knowledge on the blackboard will change.
The second majox component of an expert system is the in~erence engine. ~s the ~erm "engine ~plies, the in~erence engine is the motive forc~
which actively uses the faats ~declarative ~3~P~

knowledge) and the rules (procedural ~nowledge) to arrive at a solution to the proble~ at hand~ Very ba~ically, the in~erence engine ls that part of the expert sy~tem that determine~ wh$ch, i~ any, of the ruleR have their "IF" condition-~ met. Based on some predetermined criterion, such a~ the number of "IF" conditions present in the rule, the inference engine decides which one of rules, all of whose "IF" conditions are met, i~ ~ired next. That one rule is said to have "priorityl'. .
The third ma~or component of an expert system is the user interface. Briefly, the user inter~ace is the means by which humans co~municate with the expert æystem. ~ore exotio embodiments of user inter~aces may be artificial intelligence systems in and of themselves. User interfaces may comprise such "input" (sensory) functions as speech recognition and computer vision ~patt~rn recognition). User interfacles may also oomprise such "output'l (expressive) functions as speech synthesis.
According to the preferred embodiment of the present inv~ntion, it is th~ knowledge ~ase (the knowledge sources plus the blackboard on which they operate) WhiCh is the focu~ of discussion. A
standard inference engine, -~uch a~ that present in the rule-based programming language OPS5 (available from Digital Equipment Corporation, ~aynard, ~A), may be employed. Also, les~ exotic embodiments of user interfaces are employ~d in the pre~erred embodiment, as will be seen below. However, it is to be understood that other variations of the components of expert system~, whether they be now ~3~7~

known or hereafter developed, lie within the contemplation of the present invention.
According to a preferred e~bodiment of the rule-based approach of the present invention, "forward chaining" of rules i~ uæed. That is, conclusions about echo identity and pit characteristic6 are made on the basis of signal features of ~he sampled ~ull echo waveform.
Initial numeric cons~raints ~n the blackboard, as well as various predetermined control functions, determine which rules are activated at any given time.
The rule based approach to problem solving has two prima~y advantages over sequential programming technigues. First, the execution o~ the program is oppor~unistic, rather than sec~uential. Processing speed is thereby increased, since no time is con~umed a the processor st:eps through logical tests which do not apply to the present situation.
Second, the rules of the expert sy6tem according to the present invention are ~lodular, and may be easily modi~ied without changing large amounts of preexisting code. ~he program ~an be built up incrementally, simply by adding more rules.
The rule-based system according to a preferred e~bodiment of the present invention i5 implemented in OPS5. OPS5 was chosen because it is well suited for ~orward-chaininy. ~e~ Brownstone ~ , Addison-Wesley, ~eading, NA, 1985. The knowlsdge base according to the pre~erred embodi~ent of the present invention may be more easily understood with reference to any of various publications on the general topic. See generally, Rich, Artificial~ntelliqçnce, McGraw-~ 3 ~

~66-Hill, Inc., 1983). For a description of the more specific concept o~ "blac~board ~ystem~", see Nii, "Blackboard Systems: The ~lackboard ~odel of Proble~ Solving and the Evolution of ~lackboard Architectures," Part I t The~ ~qazi~e, S~er, 1986; "Blackboard 8ystem~, ~lackboard Application System~, Blackboard Systems ~rom a Rnowledge Engineerlng Perspect~vQ,N Part II, The ~I Ma~azine, Augustl 198S. For a dQscription of an expert system directed toward machine analysis of acoustical signal~, see Maksym et al., "Machine Analy~is of Acoustical Signals," ~attern Recoqni-tion, Vol. 16, No. 6, 1983.

c. Overall 50ftware Flow The flow diagram for the system is shown in Figure 12. Each of the rule-based modules is marked with a symbol (C1, El, E2, E3, ~1, Ul) that refers to a connection point in later figures (Figs. 15-20).
Processing begins with the input o~ sampled waveform ~alues ~rom a file. 'rhe ~irst module 1204 extracts specific features of the waveform which are used hy the knowledge sources. The height, ~idth and location ~in samples3 o~ each echo peak are determined. Echo peaks of height (amplitude) below a set threshold are ignoredO
Screening function 1205 eliminates ~rom further consid~ration the vaRt majority o~
waveforms ~n wh.~ch there is no traGe of structures which could hav~ been caused by corrosion. Such a screening function could eliminate from consideration those waveforms in which a large echo ~7-is followed by a single sequence oP properly spaced echos, and in which no other ~chos are present.
What echo spacing wa6 deemed "proper" would be determinad by the nominal thicknass o~ ~he tubular, since the Bpacing o~ ech~c in a multiple structure is detexmined by the two-way tra~el time of an acouctic pulse between the tubular's inner and outer sur~aces. Advantageously, a certain tolerance ~e.g.l 5 percent of the nominal interpulse spacing) i~ allowed for the slight irregularities of measured echo spacing to ~e expected even in a per~ectly uncorroded area o~
tubular.
Advantageously, this screening function may be performed by a streamlined, sequential (non-rule-based) program module, since it is so specialized in nature. However, it is to be under~tood that thi~ ~creening module 1205, as well as ~he echo feature extraction modul~ ~.204, could also be implemented using non;seguential, rule-base~;
software.
In Figure 12, box 1230 surrounds those modules which, ~n a preferred embodiment, are impl~mented in rule-based software. Box 1234 surrounds those modules which could be implemented in rule-based softw2re, even though they are preferably implemented with sequential programming techni~uesO
Box 1232 indicates the extension of rule-based software to the User Interface in the interactive output of ¢orrosion infoxma~ion, described below.
In a preferred embodiment, modules ~204 and 1205 are executed downhol~, so as to minimize the amount of data that must be transmitted uphole on a wireline commu~ication~ channel o~ necessarily ~ 3 ~

limited bandwidth. 0~ course, one or both of these modules could be executed uphol~, perhaps in the same computer which executes the rule-based so~tware in boxe~ 1230 and 1232.
Control function~ are appl~ed at 1206 which direct the application of the knowledge sources to working memory elements.
The ECHOS Rnowledge Source (KS) determine~ at 1212 whether therQ is present a Ndominant echo"
(de~ined in Ta~le 1). Depending on the conclusion reached by the ECHOS KS at ~212, rule activation can ~ollow two paths.
A left path 1203 is taken when the echo ~eatures indicate the presence of a dominant echo (and therefore, probably a multiple structure) in the signal. All multiples, the front-surface echo, and any pit echos are identified using the "MULTIPLES" and 'I~CHOS~' knowle.dge sources at 1214 and lZ16. When all echos have been identified, the UL~RASONICS KS determi~e~ at 1226 the corrosion characteristics.
If no dominant echos are found, then control passes along the right path 1210 (Figure 12) so that th~ XCHOS XS (E3) at 1218 identifies echos when no multiple structure has been found. Again, when all echos are iden~ified the ULTRASONICS XS
ta~es over at 1226.
Execution terminates a~ter the corrosion information is output at 1228. The system is initialized for ~he next signal.
Although Figure 12 would appear to indicate sequential proce~sing, the system is highly opportunlætic. For example, should the processing along left branch f~il to identify a multiple 1 è~ 7 struc:ture, the rules on the right branch would automatically ta3ce over, a3 lndicated by dotted line 1220. Li}cewise, if the initial conditians for a possible ~nultiple skructure are not presen~ in the input features, non~ of t~ie rules involvQd with multiples would be activated.

i. ~ch~ Featur~ ~xtractiQn The extracltion of echo features (height, width, and location) may be performed for each scho in each signal uslng a se~auential (e . g ., Fortran) program. A flow chart for the program is shown in Figure 13A.
The sampled data is read from an input file at 1302 and the echo features are determined and eventually written to an output ~ile at 1308. ~rhe output f ile i8 then read by the rule-based program .
The ~ymbols used in Figure 13A are deîined as follows:
INPUTS:
NPT Number of points in the waveîorm.
Y(l to NPTS~ P~rray of waveform magnitudes at eaoh sample point) THRESH Threshold below which peaks are ignored . ~ O < THRESH <
largest Y) XMTN Criteria ussd to find WIDTH.
Value is a given number of points. Default=3.
NPK~ Maximum number of peaks to be found (< or = to NPq~) 3 5 OUTPUTS;
NPK Nlamber oî peak found in the waveform. Will be ~ or = to N3PKMAX .
PX ( 1 to NPK) ~CATION ~ Value is a sample index (from 2 to NPTS-1).

PY(l to NPK) HEIGHT = Y at location PX(i):
i - l to NPK
PW(1 to NPK) WIDTH in sample points = P - N.
N & P ar~ ~ound a~ ~ollows.
N (Neg. pointar) ~tarts at peak then mdves to lower sampl~ indexes until l) cr 2~ ~g true:
1) Y(N) < ~HRESH or N i~ at ~ir~t po~nt (N = l~
or 2) N is at least %MIN pts ~rom peak XMIN ~ PX- N AND
valu e ha~ ~tarted inereas$ng Y(N~ > Y(N l) P (Pos. pointer) ~tarts at peak then moves to higher sample indexes until l) or 2) is true:
l) Y~P) < T ~ S~ or P is at last point (P = NPT) or 2) P is at least X~IN pts fro~ peak XMIN < ~ - PX
AND value has started increasing Y(P) > Y(~

The deteotion of a peak is based upon a change from positive to negatlve i~ the local slope of the values~ For a valid peak, the amplitude must exceed a threshold Yalue. This value is typically set to a value midway between the height of the fifth multiple echo and the balakground noise level when no corroæion is present.
Once a peaX ~s detected, the amplitude at the peak (height) ~nd the loca~ion in ex ~sample#~ are stored i~ an arr~y. In a preferred embodiment, the pea~ width i~ obtained by ma~hing down the sides of the peak and looking for either of two conditions to be met. The edge of ~he peak is ~ound when l3 a sample Yalue is below the threshold, o~ 2) a slope change occursO ~he width of a peak i5 taken to be the difference in sample number between its two edges. ~he routine ends when the array values have been output to a file.

~ 3 ~

The following ~s a detailed de~cription of Figure 13A for Echo Feature ~xtraction.
The sampla value~ are input from a file as shown by block 1302. Once these values have been input ~o ~he progra~, certain flags and pointers are initialized in block 1304. The number of peaks ~ound in the waveform, NPK, is first set to zeroO
In addition, flag 'I~EN~ UP" is initialized to ~alse and the positive pointer "P," the index on the currant sample value~ is Bet to l.
The routine then enter~ the loop indicated by LOOP 1. The test in block 1306 determines if the number of peak~ ~ound, NPK, is less than the maximum number of peaks to be found, and if the positive pointer is less than the total number of point~ i.n the waveform, NPT. The maximum number of peaks to be ~ound is initially set to the total number o~ points in the waveform. If this logical test is false, the program immediately outputs all values for the echo attribut,e~ to a file in the ~ormat described below. If this logical test is true, then the posltive pointer on the sample index is incremented by one.
The ~et of logical tests 1312, 1314 and 1316 are used in the detection o~ an echo peak. The ~est indicated by ~lcck 131~ determines if the next value Y~P) in the ~ample data is grea~er ~han the given threshold. I f this is false, ~hen the program i~mediately goes to LOOP 1 so that the next sample value can b~ considered. If the test is true, the program branche~ to 1314, where a ~est is made to determine if the new ~ample value Y(P) i8 greatex than the prev~ OU5 sample valua~ If this is true, then the flag I'~ENT-UP'~ is set to true and 3 7 ~
-72~

bra~ch is made back to LOOP 1. If the test is false, then the flag i~ tested in 1316. If the Xlag "WENT-UP" was ~al5e at thQ ~ime of the test indicated in 1314, then a branch is ma~e, again, to LOOP 1.
I~ the flag "WENT-UP" was equal to true in 1316, several other program variables in block 1320 are initialized and the lower portion of the figure i5 enkeredO This means that a new peak has been detected in the ~ignal because the slope of the sampled data changed from positive to negative.
Block 1320 increments the countar on the number of peaks found, NPK. In addition, the flag "WENT-UP'!
is set to false. The location ~or the new peak in the sign~l iB set to the previous pointer "P~' minus one sample index, and the height of this peak PY is set to the value at index PX.
The next step i8 to set "N,l' the pointer for the negative direction, egual to the index for the peak. In this lower sec-ion of the ~igure, the pointers begin at the index o~ the peak and march down both sides o~ the peak in order to determine ~he width oP the echo ~ust identified.
First, the negative pointer l'N" i8 decremented in 1324. The logical test in 1326 determines if N=l or i~ the value at the new negative pointer is les~ than the threshold. I~ N=l, then we have progressed to the left to the very beginning o~ the echo waveform. If the #econd part of thQ test is t~ue, then a left edga o~ this identi~ied echo has been identi~ied~ Thus, i~ N=l or the second condition is true, we branch to the lower right side of the figure where a right-hand edge is identifiedO If condition 1326 is false, the ~ 3 ~

following test in 1328 determines i~ the left-hand, or low-time, edge o~ the ec~o can be identified by a change in slope in the signal. This test handles the case where the left edge o~ th~ ~cho does not cross the threshold but rather encounters another peak to the le~t of the current peak.
To separately identi~y thece peaks, test 1328 tests if ths slope going down the left side of the echo changes to a positive slope. I~ no change in slope is indicated, the logical test ~ails and we return to block 1324. Here the negative index N is dearemented again to consider points further to the left of the current index. Should either logical test 1326 or 1328 be true, then the left-hand edge of the echo has been discovered, and a search for the right-hand point, or high-time point, is then initiated.
For the identificatio~ of the right-hand point, the first step i~ to set th~ positive pointer index (PX~ a~ual to the locat~on of the peak value (block 1330). A loop is then entered which is identical to that previously de~crib~d for the negative pointer. Blo~k 1332 succeæsively increments the pointer to look at values at increasina time on the right ~ide o~ the peak. As in the cas~ for the left-hand edge o~ the echo, blocXs 1334 and 1336 are used to identify the right-hand side o~ the peak. In thi~ case, the right edge o~ the echo is identified by a thr~shold crossing, or a change in slope for successive ample values from negativ~ to positive. I~ either of these conditions ara true, then the xight-most edge of the echo has been determined. If both of ~he conditions are falsey the positive pointer is ~ 3 ~

incremented again to consider the next sample value.
Once the right edge of the echo has been determined, the width for the echo is calculated in 1338 as ~he differ nce between the positiv~ pointer P and the negative pointer N. In addition, the pointer ~or the current location ~n the sample data i~ set to an index two (2) lower than the right-mos~ edge Or the echo. LOOP 1 i8 en~ered again and execution goes back to logical t~st 1306.
This process continues until all peaks in the signal have been identified ~nd their left- and ri~ht-most edges determined. Logical test 1306 tests ~or the case where the indQx P is equal to the total numb~r of points in the waveform. At this point, all the echo attributes for peak value, width and location are output to a ~ile and th2 process terminates.
Thi~ ends the detailed discussion of Figure 13A.
The preferred output ~ile has a file header line ~shown in F~gure 13B) which contains: 1) a descrip~ion o~ the data se~ he date recorded, 3) ths number of signals, 43 the number of samples/second. 5) the total number of samples, and 6) the threshold value.
Each sampled signal in the output file has a record header (shcwn in Figure 13B~ containing data on th~ location of the spot area on the pipe and the ~umbar of echos in the signal which exceed the ~hreshold. Pollowing the header are data entries, each comprising the sample location ~or the peak echo value, the echo width ~n samples, and the height (peak digitiged voltage value)~

~3~9 d ~

The data file thus arranged is ready for analysis by the rule-based analysis system, generally indi~ated at 1~30/1232 (Figure 12), possibly after record~ are selectiv~ly eliminated by screening ~unction 1205 (Figure 12~.

ii. Ini~ializa~ion an~ Çonstraints When the rule-based system i~ run, the slement classes o~ workiny memory (~'blackboard") are defined. Once defined, no additional element classe~ can be defined, but any number o~ working memory elements of a given cla~s can be created or modi~ied.
Table 2 shows the element classes and associated attri~utes used for the system according to the preferred embodiment~ For example, the class "echo" (the third entry in Table 2) has 10 attributes (signum echo_num width loc...) which are used to describe individual echos. There is one working memory element o~ class "echo" for each echo. Each such working memory element has the same pattern as that indicated in Table 2.

~ 3 ~

Workin~ Memory ~lement-Classes and ~ttributes ELEMENT
CLASS ELEMENT ATT~I.BUTES
signal set name date nu~sig usec/samp numsam threshold x offset y offset siynal signum num_o~ echo~ xang ydis~
max_loc ~ax he~ght echo ~ignwm echo num w~dth loc h~ight type atatus tag multiple_of m~rk pit signumber type pitnum depth area osig match depth match xcen ycen xlow xhigh ylow yhigh control name status que~tion name asked answer constraint name value tag computed name value dom echo signum echo_num loc struc~ure type genera~or._num number f_mult status mult list -task name datal data2 counter name count plot ter}ll_type status cur echo signum echo_num :1 oc diff echo signum ech 1 ech ~ di~f offset h_ratio limit~ lower upper exact ~verlap ech_l ech_2 lowlim uplim corrosion type echo_num depth size ~3~7~
~77--D~sc:~i~tio~Qf~the ~t~ributas SI~L_SET One WM~3 with in~ormation on measurement parametQrs frc:m input file name Identif ier ~or data set containing feature~ of sampled signal date Date on which ~ampled data recorded numsig Number of individual signals in the input file usec/samp The time intervP.l between sampled values in microseconds numsam The f ixed number of amplad values ~or each signal thresho:Ld Height below which sampled data values are ignored x_o~fset Scan parameter for the circumferential distance between spots y_offset Scan parameter for the vertical 3 O distance between spot areas WME ' s of this class hold informat~on on each signal in se~

signum The number of a partiS-ular signal ~1 to numsig) num of echos The number of distinct echos identi~ied in the signal xang Angular of ~s~t of spo1: center from north on pipe 4 5 ydi~t Depth location of spot center on pipe max loc Maximum index value ~or signal data (=numsam) max_hei~ht Value of height for largest echo ~ 3 ~

~Q ~MEIs o~ this cla5s holds information on echo ~eatures and type signum Signal number to which the echo WME
belong~
echo num Tbe numb~r o~ the echo (1 to num_of~echo~3 width Width feature for the echo (number of ~amples -see ~eatures) loc Location of the echo peak (sample lS index - 1 to numsam) height Value o~ the sampled data at location of echo peak type Identity of the echo (e~g. pit, front-s~rface, spurious, etc~) status Fla~ for echo plotting by the user-interface tag If the echo is a multiple, the tag is set to the fir~t echo number mark Flag used to indicate that the width of the echo has been counted ~I~ WME to hold in~ormation on identi~ied pits ~ignum T~e number of a par~icular signal for whi¢h a pit has been found type Type of pit (e.g., pit within multiple) pitnum Echo number which ~ndieates presence o~ pit depth Depth o~ pit as ~ound using rulas in Ultrasonics KS
area ~rea of pit as foun~ using rules in Ultrasonics XS

-7g-xcen Circumerential location o~ ~pot (pit) on pipe ycen Vertical loration of spot (pit) on pipe CONTRO~ These WME8 are used to control the activation of some rule~
name The pecific type of control WME
~e.g., signal_done ?~
status Flag to indicate ~tatus for the control function (Typ. yes/no) OUESTION These WMEs hold process information from the user interface name Particular gue~tion asked asked Flag to indicate if the question has been asked yet answer The answer provided by the user cg~E~ r These WMEs hold information on physical and measurement variables name C o n ~ t r a i n t t y p e ( e . g . , "p~pe_inner cir" for pipe inner circum~erence) value Curr2nt value of the constraint tag Flag used to indicate status of updates ~or constraint values ~0 COMPUT~D This cla~s holds temporary calculated values for each ~ignal nam~ The nam~ for the value computed by the rU~Q
value Th~ stored value ~-~o -WM EC~o A single WME of each sign~l which identi~ies the dominant echo signum The signal number for which the echo i5 dominant echo~nu~ The number o~ the echo which is the dominant ~cho for the signal lo loc The index for the peak value of the echo ~same as "loc" above) ~la~5~E~ WME class to hold information on identi~ied structures in signal type Type of ~tructure identified ~e.g.
"Multiple") generator The number o~ the echo g1ving rise to num the structure number Number of echos in~olved in a "multiple" structure (only) f mult The echo number for the first multiple in a "~lultiple" structure status Flag to indica~e, that the labeling o~
multiple echos i.s complete mult list Array of values holding the echo numbers ~or each multiple TASK Thes~ WME ' s ~ocus attention to specif ic tasks (removed when dQne) name ~he task name ( "get new signal", "determ~ne_ corros1on") COUNTE~ These WME I 5 hold results of various counting operations name Name o~ specific counter (e.g.
'tcounter_front_ ~urface_average") count The actual count at any time ~. 3 ~

PLOT ~ single WME of thi~ class has information ~or the user interface term_t~pe Tha terminal typ8 i5 speci~ied here status Flat to lndiaate that echo Peature plotting ha~ been in~tialized ~UR_ECHO ~ single WME ~or each signal identi~ie~ the current echo signum ~he number of the ~ignal being investiyated echo num Number for the current echo (change~
when looking for multiples~
loc The location ~or the current echo WME's to store time separation~
between echos for multiple search signum The number of the signal being investigated ech 1 The echo number for the echo with lower location value ech_2 The echo nu~ber for the echo with higher location value di~f The di~ference in location values (higher-lower) offset Ths di~ference between the exact multiple separation and "diff"
h_ratio The guotient of the height o~ ech_2 and ech_l ~IMITS One W~E of this class for each possible first ~ultlple lower Location diff. ~first mult.-dominant - constraint_loc multiple) upper Location dif~. (first mul~-dominant + constraint loc multiple) 13~Y'~
~2-exact Location di~erence lfirst mult.-dominant~

OVERLAp Th~se WME~ R hold information used to identi~y overlapping echos ech 1 Number for one echo which may overlap lo another ech 2 Number ~or another ~cho which may overlap ech_1 lowlim Location of ech 1 minus the aver~ge of the two echo widths uplim Location o~ ech 1 pluc the average of the two echo widths CORROSION WME~s to hold information on the nature o~ the corroded sur~aces type Descriptor ~or corrosion naturs ~e.g.
"smooth front sur~ace'i) echo num Echo number givin~ rise to the conclusion on surface nature depth ~aximum depth of any pitting in the ~pot area size Maximum eize of any pit opening within the spot area For example, in OPS5, a particular working : me.mo~y element of class 'iecho" might contain the following ins~ances o~ a~ributes:

signum 5 (fi~th signal wavefo~m on a run) echo num 5 ~fi~th echo in that Wav~form) width 20 (20 samples "wide"--time duration) loc 144 (144th sample in the wave~orm~

8~ 3 ~

height 102 ~units of voltage~
type pit (identi~ication--per-haps ~entative) ~tatus plotted (fl ag for user interface~
tag n~l (as yet undefined~
multiple_of mark ~il (as yet undefined) Constraints are used to guide the activation of rules, controlling the sensit~v~ty of the syste~
during identification o~ echos and structures.
Table 3 shows actual constraints. The numeric values held in element-class '~constraints" are lS applied in the condition part o~ a rule to determine if it ~hould be activated. These constraint values are part of the heuristic nature of the rules. Both rule-oonstraints and ultrasonic parameter values are stored as "constraint"
elements.
Particular constraint valules, such as those in Table 3, are chosen by a sy~;tem operator. The constraint values may be modiÆiled during the actual processing of a signal.
T~BLE 3 List of Preprogrammed Constraints for the Rule-Based System CONSTRAI~T NAM~ ~ALUE
Value ~heiaht) constraints:
constraint on value dominant 2 For an echo to be a dominant echo, the height must be this ~actor greater than any other.
constraint on l~ading value% 80 This constraint sets the height value used in the identi~ica~ion o~ a dom~nant echo. The constraint is only used when an echo overlaps the maximum height echo. IP the height o~
thi~ other echo exceeds the above percentage ~ 3 ~
-8~-of the maximum echo hei~ht, and comes before the echo with maxi~um height, then the echo i~
identified a~ the dominant echo (rules '~ax_overlap, wide do~inant, and wide dominant fir~t") height ratio Pit/spuriou~ 2 This constraint ets the height value used in the identification of a spurious echo. ~he constraint is only u~ed when a pit echo has been identified within a ~ultiple ~tructure.
For identification as a ~purious echo, the height value of the unidentified echo must be less than the guotient of the pit echo he~ght and the constraint value above.
ratio max/large_pit 4 This constraint sets the height value used in the identification of a pit. The constraint is only used when no multiples echos have been identified. For identification as a pit echo, the height vaIue of an unidentified echo must be greater than the quotient of the largest echo height and the cons*raint value ~bove.
ratio max/spurious 5 This constraint sets the ~eight value used in the ide~tification of a spurious echo. The constraint is o.nly used when no multiples echos have been identified. For ~dentification as spurious echo, the height value of an unidentified echo must be greater than the quotient of the largest echo height and the conqtraint value above.
constraint on:height ratio for~multiple 2.2~
~he valuè of this constraint is used to place an upper limlt on the hei~ht of the next possible multiple. If the ratio of the heights of 1) the next possible multiple echo and 2) the current :mu:Ltiple exceedæ this con~traint, then the ~ext echo cannot be identiied as a multiple echo.
default front sur~ace height 175 The value of this constraint is used as default value for the height of the front surface echo. It is only used for the calculation of the area of the pit for the v~ry first signal,:.hould this signal indicate a pit. Later calculations use the running avarage of front surface echo heights (for no corros~on).
constraint_on overlap multiple 1.6 The value o this constraint is used to test ~or ~ pit echo overlapping a multiple echo.

~ 3 ~

Other Gonstra ints:
nu~ber_of echos for multiple6 2 q~hi~ cons~raint ~ets the minimum required number o~ multiple echos for a valid multiple ~tructure. If a lower number of ~equal-time-spaced" echos are identified, then no structure i~ ~tored.
number of echo~ ~or large_multiple~et 4 I~ the 6ear~h for anumber o~ potential multiple structure~ reache~ a point where one structure has a number of mul~iple echos equalling this constraint, the search for . other multiple ~tructures is cancelledO
constraint on location ~or_multiple 17 This constraint is used to ~et the tolerance on the spacing between echos in a possible multiple structure. For two echos to be multiples, they must be separated by a fixed number of ~amples, plus or minus the number of samples given by the constraint, Physical const~aints transducer ~pot ~ize(~q.inches) 0.5625 Actual area of sensor spot for given sensor to pipe wall ~eparation.
pipe inner cir(inches) 22 Pipe inner circumference. :
spot y_offset(i~ches) .75 Sensor spot dimension in direction o$ pipe long axis.
spot_~_o~fset(inches) .75 3S Sensor spot dimension in along circumference . o~ pipe.
de~ault area small_pit(sg. inches) 0.0123 Smallest area to b~ reported ~or an identi~ied pit.
min nominal wall_thickn~ss(inches) 0.300 Minimum wall:thickness expected ~or the pipe.
microseconds/sample 0.02 The time interval between sampled values.
speed in 6teel~inches~microseco~d~ 0.223 Sound speed in the pipe material.
speed in liquid~inches/microsecond) 0.0583 Sound ~peed in the liqu~d between sensor and pipe wall.

--f~6--As an example of a constra~nt, th2 rule ~'LEADING PI~ FRONT~ ~1812, Fi~ure 18) uses the constraint "min nominal wall khickness~' to determine if the leading echo should be idantified as a pit echo. The natur~ ~nd ~eans of execution oP this rule i8 described below, in the ~ection on Firing of Rules in the Knowledge Base.

iii. Firin~ vf_lRulçs in th~ ~nowled~
Rase This section ~irst presents an in-depth description of the process through which an 1~ individual rule may be fired. Special attention will be paid to the use of classes and attrlbutes of working memory elements (Table 2) and o~
preprogrammed constraints (Table 3). Then, this section will present a synopsis of how the rules interact with one another, in preparation for the detailed analysis in Structure of the Knowledge Base, and in the Appendix.
The ECHOS LEADI~G PIT FRONT rule is exemplary.
This rule looks for the case ~a~ a large ~lat pit.
In this case, mul~iples can }~e generated between the top ~urface of the plt and the outer wall o~
the pipe. If a large pit is present, then the r~.maining wall thickness, as determined ~rom the multiple separation, must be less than the "min no~inal wall thickness." An English condensation of the ma;or ~eatures of th~ rule is as follows:

~L 3 ~
-~7-There i8 complete ~ultiple ~tructure Sall multiples found) and ~n echo has been identi~ied as th~ leading echo of this ~truoture a n d t h e c o n ~ t r a i n t N m i n -nominal wall thickne~s~ has value MNWT
and ~he r~maining wall thickness calculated from the multiples i~ ~ ~NWT
and there are no echo~ except those in the multiple structure THEN
Identify the leading echo as an echo from the pit ~urface and create: a working memory element of class "structure" for the pit echo.

In the preferred embodiment, the rule is implemented in OPS5 as follows:
echos leading~pit front (structure ^type multiple ^generator num <de#>
^~_mult <fmult> ^status labeled) (<ech> ~echo ^signum ~s#~ ^echo num cde#~
^type leading ^tag c~mult~) ) ~constraint ^name min nom-Lnal wall thickness ^value ~mwt>~
. (computed ^name remaining_wall thickness ^value (< ~mwt>3) -(eaho ^~ignum ~#~ ^tag ~> ~fmult~) -~echo ^si~nu~ ~s~> ^type nil) __>
(modify ~ech> ^type pit ^s~atus nil) S b i n d < t >
Found_structure "pit dominant as front_-surface" for Pit) ~write ~crl~) ct> cde#>) ( ~ a k 8 8 ~ r u c t u r e ^ t y p e p it dominant as_ front surfaae ^generator num ~de#> ^number <de#>) This rule, ECHOS LEADING PIT_FRONT, will nex~
be expl~in~d in detail ~o~that the use of Table 2 and Ta~le 3 in ac~ual program~ing by one of ordinary skill will be facilitated. Of course, the particular progra~ming language, rule organization, ~ 3 ~

rule content, rule interaction, working ~emory element classes and attribute~, and ¢onstraint identity and value are pre6ented purely by way of illustrative example, and do not limit the scope of the present invention.
As in many rule-baeed programming languages, OPS5 may be descri~ed generally as Pollows. The rule comprises a left side and a right ~ide. ~he left ~ide compri~es (in our cas~) four afirmative condition6 and two negative conditions. The right side ~omprises four actions which are performed when all six conditions have bePn met.
On the left (conditional) side of th2 rule are located a list of six conditions. ~ch condition i5 a pattern which the inference engine attempts to match against the elements of working memory. Each condition is of the form:

Class Aattribute atom/<var.Lable> ~attribute atom/<varia~le>

Reference may be ~ade to the classes and attributes in Table 2 and its accompanying des~ription for a deeper undexstanding of each of the six condition~ in this rule. For example, the four di~ferent clas~es which are invoked by the six conditions o~ thi~ rule are "structure", "echo"
(three occurrences)~ i'constralnt", and "computed".
~asically, this first condition ensures that a multiple has already been dePined at that polnt in the execution of the rule-based system.
The ~irst condition i8 that there must be, somewhere in working memory, a ~orking memory element of class "structure" having the four attributes specified in the rule above, namely:
attribute~ "typP", i'generator_num", ~If_mult~ and ~8t~1~U51~ -The ~pecificat~n of cla~s "~tructure"
eliminates from consideration all working memory elements which are not of that class.
"Type multiple" en~ure~ that, o~ all worki~g memoxy elements o~ class ~tructure" ~ee Table 2), only those working memory elements having ~lass "~truct~re" and of type "multiple'~ are further considered. (Other "types" of ~tructures include pits and front surface.) Also, for the working memory element of class "structure" matching the pattern demanded by this condition of this rule, the generator number is read into a local variable, cde#>.
For ~he ~atching working ~emory elements, the value of the attribute "f mult" is placed in the local variable ~fmult>.
Finally, the "status" of the working memory element must be ~he ato~ "labelledl'.
A second ~onditîon causes the inference engine to loo~ through the working m~emory for a working memory element of class llecho", as shown immediately to the right of the left parenthesis "(". (The ~ymbol ~ech~ will be explained below.) The first attribute o~ this working memory element can be any value, and is assigned the value <s#> and takes on a value which is used later in the rule~ The requirement "echos_num <de#>l' perform~ a test as to whether the particular at~r~bute in the working ~emory element u~der consideration has the 6ame value as was read in ~ 3 ~ P~ ~
~90-during processing of the previous condition, in the attribute ~generator_numH.
The "type leading" requirement ensures that the working memory element under consideration i8 a leading echo.
Finally, the fourth require~ent ~f the second condition of the rule is that the attribute ~'tag"
b~ the same a~ the variable "fmultll read in during the execution of th~ firæ~ condition above.
Basically, thiæ ~econd condition of the rule checks for the consistency ~hat there is a leading echs qenerated within a multiple structure. This condition effectively labels the particular echo which is identified as the leading echo. This labeling is accomplished by the <ech> which is at the extreme left of the second condition of the rule. This ~ech> is used in the first action (on the right hand side of the rule~ so that the particular working memory elemen~ detected in the second condition is t;~le one whose type is mndified from "leading" to "pit".
The third condition of the rule is that there be a working memory element of the class "constraint" (~ee Table 3) with a name "min nominal wall thickness'l having a value >Il .
The fourth condition of the rule is that there be a working memory element of class "computed'g having a name "remaining wall thickness" and whose value i~ strictly less than the value "mwt" read in during the third condition. "mwt" is the minimum wall thickness which is to be expected (including a ~mall degree of tolerance, of course).

-'31-Ba~ically, conditions 3 and 4 test ~or the condition that the computed remaining wall thickness (based on a calculation based on multiple separation, carried out in another rule3 is les~
than ~hat which would be expec~ed for an uncorroded pipe. Essentially, this i~ a cxucial test for datermining whether a mult~ple has been areated by internal reflection~ from a pit surfa~e, rather than ~rom the front surface o~ the tubular.
The fifth and sixth conditions of the rule are negative conditions. That is, none of the actions on the right side of the rule will be executed i~
the fifth and sixth condi~ions (to the right o~ the minus ~i~ns) are met.
The ~ifth condition prevents firing of the rule i~ the "signum" (echo index number) equals the local variable "s#" read in during the testing of the second condition of this rule. Also, the tag for this "echo" working ~emory element must not be equal to th~ value of the variable "fmult" read in during the processing of the firs~ condi~ion o~
this rule. Basically, the fi~`th condition of this rule ensures that there are no other echos in this signal except thosa in the multiple structure. The 2~ absence o~ other echos is ensured ~ecause, in another rule, tAis tag attribute is ~et equal to the echo number of the ~irst multiple, so that if there are no other echo~ with tag.~ other than "fmult", there must be no other echos in the signal besides those in the multiple structure.
The sixth condition of the rule, also a negative condition, ensures that there are no unidentified echos in the signal. Unidentified echos are characterized in that their working ~ 3 ~

memory elements have a type ~Inil~. Presence of "nil" indiaates that no information ha~ been added to that attribute.
If all ~ix conditione have been met, then ~11 four of the actions (on the right side ~f the rule) will be taken.
ThP first action which is taken in the event that all ~ix condition~ are ~et is that the "type"
of the working memory element discovered in the matching test performed during the processing of the second condition is modified from "leading" to llpit~l. Th~ ~StdtUSII (used for plotting purposes in the user interface) i8 changed to "nil" so that a bookkeeping rule detects the ~act that this echo's status has been changed and plots it as a pit echo rather than a leading echo.
The second and third actions on the right side of the rule are e~fectively a print command which tells the user that a pit echo ~rather than a front sur~ace echo) indicates the presence of the observed multiple structure.
The ~ourth and final action to be taken when this rule is fired is that a working memory element of class "structure" is created. A structure type is "pit_dominant as front_surace" whose "generator numt' and "number" are both the value "de#" assigned during the processing of the first condition of the rule, above.
This concludes the detailed description of the rule ECHOS LEADING PIT FRONT. Tables 2 and 3, the detailed descript~on of the rules given in the A~pendix, and the sequence o~ rule firinys in Table 4 facilitates the unders~anding and praotice of the invention. ~hanges in any of the rules or their 7 ~

interactions may be e~fected without departing from ths ~cope of the present lnvention.
~aving completed a detailed description o~ a particular rule, the following axposition is presented as a more global de~cription o~ the preferred embodiment.
In order to describe the struoture and interaction of the rule~ in the various knowledge bases, "aotivation-diagrams" are shown in Figures 15-20. These dia~rams are ~i~ilar to "and/or"
diagrams used to describe search trees See, e.g., Winston, P.H., rtifiçlal Intelliaence. Addison-Wesley, Reading, Mass., 1977. The activation-diagram is used to ~pecify which rules must have "fired~' (been activated) before the conditions for any other rule can be ~atisfied. In Figures 15-20t circles, lines and arcs are used to connect rules.
~ule-connection points are indicated by the circles. Lines between two rules indicate a dependence of the lower rule upon the action part of the upper rule. ThUS, when a rule is dependent on two or more rules there are two or more lines intersecting at the top of the dependent rule.
When these lines are connected by an arc (for example, 1524 in Figure 15), all of the higher : level rules mu~t have been activated ~an "AND~
function). As an example, consider the conditions for rule CO~PUTE AVERAGE FRONT SURFACE-_HEIGH~ ~1522 ~n Figura 153. The higher level rule, SET FRONT SURFACE HEIGHT 1510, must fire be~ore this rule 1522 oan be activated. In addition, a rule MULTIPLE NO CORROSION from 1520 (2024 in Figure 20) in the ULTRASONICS knowledge source must also have iEired. Otherwise, in the ~ 3 ~

absenoe of an arc any one o~ the higher level rules could hav~ been activat~d ~or the conditions o~ the lower rule to be met (an ~OR'I function).
~ny one of activation-diagrams in Figures 15-20 may contain rules ~rom ~any knowledge sources.
They have not been constructed to show exclusively the activation~ o~ single ~nowledge ~ources.
Rather, they re~lect the interactions o~ the knowledge ~ources and the opportunistic nature of the rules.
A complete description o~ the rules in the knowledge sources is given in the Appendix. The basic solution methods used for each XS are described in the following sections.
iv. 0 u t p u t o f C o r r os i on Information--~he User Interface ~0 ~he presentation of corrosion in~ormation is a function of the user interface of the expert system. In the pre~erred embodiment; the output of in~ormation may be on any standard output device, such as a catho~e ray ~ube (CRT) or on a pri~ter.
The particular in~ormat:ion which is output varies with the degree of confidence which the user has in the reasoning ability o~ the expert system on a given task. I~ the user had high confidence in the ability of the Qxpert system to arrive at correct conclusions, th~ expert system may display or print only its ultimate conclusions, such as whether there is any corrosion in a given section o~ pipeO
If the user had less confidence in the ability of the expert system to arrive at correct . .

, conclusions all of the time, the expert system may be commanded to present lts conclusions, as well a the reasonin~ by which it arrived at those conclusions. The "reasoning o~ the expert sy~tem"
is, of cour~e, the path taken through the rules of the knowledge ~ource~ to arr~ve at a given conclusion. Commonly used e~pert ~ystem ~hells, such as OPS5, have their own ~eans of "expressing their reasoning". ~hese ~eans are commonly called the "explanation facility", any embodiment o~ which is contemplated by the present invention. A
preferred embodiment of the explanation facility is that illustrated in Table 4, described in the section entitled "Processing Example." tSee also the discussion o~ Figure 21 in that section.) For example, in areas o~ the target surface where there is known to be corrosion, the user would likely wish to verify the accuracy of the expert system's conclusion. ~he actual presentation o~ the analysis by the expert system of an actual full echo waveform (that of Figure 10) will be Aescribed in detail below, in the section entitled "Proce~æing Exa~ple". Of course, the presentation given below is offered to demonstrate only one possible embodiment of the way in which ~nformation could be presented to a user.
Variation o~ the ~tructure, sequence, and organization o~ the method of output of information lies within the contemplation of the present invention.
In addition to presenting its reasoning, the expert sy6tem ~ay create a graph of the inspected ~urface which it has concluded has caused the ~ull echo wave~orm under consideration. The graph may ~3~P~

be either a "facial" (tran~ducer-eye) view, or a cross sectional view, of the target surface. A
~acial view advantageously color-codes di~ferent areas of the target surface according to the different depths of corrosion pits which have been detected. ~he user can then judge the credibility o~ the 6ystem 1 8 conclusions~ perhaps based on his personal experience.
The human user's interactive input of information into the expert syætem according to the preferred embodi~nt of the present invention may be accomplished by any of a variety of input devices known in the art, surh as a keyboard, "mouse", or any of the more exotic devices described generally above in the section entitled ~General Introduction to Artificial Intelligence."
The user interface may allow the user to change certain intermediate conclusions in the reasoning process by which the expert system arrived at its own conclusion, so that the user can investigate hypothetical interpretations at will.
The section entitled "Processing Example"
illustrates but one exampla of the means by which a human user can view the process by which the expert system arrived at a ~et of conclusions regarding a particular full erho wave~orm. At any point in the seguence of rule firings listed in Table 4, above, the user could interrupt the sys~em and substitute a diff~rent constraint or attribute value so that he could view the effect whid~ that change had on the final conclusion arrived at by the expert system. Thus, the ability to input constraints or attribute values during the actual execution facilitates the judicious choice o~ ~onstraints for ~uture runs. In this way, the expert system acts as its own ~o~tware development system.

d. Structure of the Knowledqe Base Figure 14 i~ a diagra~ indicating the structure o~ the rule-based sy~tem ~ccording to the preferred embodiment of the present invention.
Generallyt evarything that i~ illustrated in Figure 14 ~sxcept perhaps Echo Feature Extraction routine 1~08) may be considered part of the knowledge base of the expert system. The knowledge base ~omprises two ma~or components. The first component is the "blackboard'l ~or, more commonly, "working memory"). The second component is the set of knowledge sources generally indicated at 1414.
~he ~et of dotted lines with arrows on their ends indicate the means by which information in the working memory is (or may be) modified by rules in the Xnowledge sources firing.
Working memo~y 1424 comprises declarative information in the form of constraints 1412, as well as oth~r declarativ~ in~ormation 1416, 1418, 1420 and 1422 regarding the signal and the physical structure a~out which conclusions are to be made.
In the preferred embodiment, the physical stru~ture about which conclusion~ are to be made is the ~urfaae o~ a borehole tubular in an oil well.
The sa~pled signal data 1422 is the full echo wave~orm received by the acoustic transducer. The echo ~eatur~ ~420 include the heiyht, width, and location (time o~ occurrence) of each particular echo pul~e. The echo ~tructures 1418 comprise such information a~ whether individual echo~ are pit echos, front ~ur~ace ~chos, sr member~ of a multiple ~tructure. Corrosion pit~ 1416 contain the ultimate conclusion ae to the nature of any dePormities in the target ~ur~ace, ~uch as the location, depth and area of corrosion pits in the tubular.
Working ~emory elements are hasically di6crete items of declarative information which are created, and po~sibly modified, and even deleted, by the rules as they fire. The progression of the expert sytem from waveform in~ormation to target surface conclusions is accomplished by the chain of intelligent firing of the rules. Working memory elements are of the "classes" and patterns of "attributes" listed in Table 2.
A preferred embodiment o~ the constraints in constraints memory 1412 of the working memory 1424 are listed in Table 3. Generally, constraints take on a single value (although some may be changed during operation). Constraints are usually parameters derived experien~ially, and lead the expert system to ef f iciently arrive at its conclusions. Also, the constraints memory 1412 allows flexibility in applying the system to a variety of physical æituations. For example, the "pipe_inner cir" (pipe inner circum~erence) can be altered simply by changing one value in a constraints memory rather than changing the value throughout all the rules in which the tandoff is used.
The knowledge sources act upon the declarativc knowledge in ~he working memory. All of the ~nowledge ~ources 1402, 1404, 1406, 1408 and 1410 are described immediately below. The e knowledge 13 ~ J~ ~
_99_ ~ources rontain the rules, ~uch as those described above in the eection Firing of ~ules in the Knowledge Base, which effectuate changes in the working memory element~ ~o that conclusions as to the nature of the target sur~ace are ultimately reached.
The arrows with dashe~ lines, indica~ed as 1426, 1428, 1430, and 1432, indicate the actions taken (on the "right hand ~ide" of the rule~). The rules are selectively executed, based on the present contents of the working ~emory, in a manner well understood in ~he art.
It should be understood that ~igure 14 is very schematic in nature, and does not represent an l~ exhaustive description of the present invention.
It also does not show how even this particular application of the invention need be implemented.
It is presented mainly as an aid in understanding the particular descriptions o~ the knowledge sources presented imm~diately below, as well as the individual rule descriptions in the Appendix.
Figure 14 is to be contrasted with Figure 12, which represents a flow diagram. Figure 12 indicates generally the timewise order of activation of the various functions of full echo : wavefor~ analysis. With this understanding o~ the ~tructure and flow of the analysis in Figures 14 and 12, the ~oll~wing more detailed descriptions may be placed in context.

i. Con~rol Euncti~

Once the echo datz and constraints are stored as working memory element~, proces~ing begins with the execution of "control" rules (F~gure 15).
These rules act to guid~ further rule activation down the two branches 1208 and 12~0 (Figure 12) discussed pr~viously ~or the "multiples" and ~'no multipl~s" cases.
In addition, rules in the control function are used to determine when all echos have been identified, and to compute the average height for the "~ront-surface'i echo. This average hPight is advantageously a xunning average o~ front 6urface echo heights ~or physi~ally successive signals from uncorxoded areas.
Also, rules in the Control Function allow predispositions to be built into the rules for the analysiR of a given full echo waveform signal, if part o~ the spot corresponding to the present signal was covered by an overlapping 8pot from a previously-analyzed ig~al (note areas o~ spot overlap in Figure 6B3. For example, suppose that analysis of a first signal indicated the presence 25 ~ of a pit of a given dep~h, and analysis of a second signal ~whose ~pot overlapped that of the first signal) indicated an ambiguity as to whether a certain eoho ~was a pit e~ho or a purious echo.
Rules in the control function corroborate the presence of a pit of that depth in the econd spot.
This corroboration ~ay,~for example, taXe the form of a pr~sumptive initial ~ssignmen~ of an echo : "type" attribute of t'pit'i.

~ 3 ~

ii. ~TIP~ES XnowLedge SouEce The MULTIPLES XS (Figure 16) is an important element of the rule-based E;ystem (Figure 14)o Once a multiple structure ~ 8 ~ound and eliminated from consideration, further processing is simplified due to the reduced number of echos to be identified.
Any echo which is not part of the multiple structure i~ a potential pit echoO
~ "mul~iple structure~, as defined above with respect t~ Figure 9, is a series of substantially equally timed echos resulting from acoustic energy reverberating within a target obj ect such as an oil well tubular. If the only information one is seeking is al~out pit echos, then the echos in a multiple structure may be considered "confounding signal structures" on the :Eull echo waveform, inasmuch as they may masquerade as pit echos (the true, "in~ormational signal st~uctures" on the full 2 O echo waveform) .
In more general terms, once the confounding signal structure has been detected, identified, and (at least temporarily, or for ~ome purposes, permanently) eliminated from con~ideration, then the informational signal structures, if any, are laid bare for further analysis, free of the corrupting influence of the confounding signal structure. The substantially egually-spaced echos of a multiple structure are thus but one example of a confounding æignal structure.
Other examples o~ confounding signal structures might include the delayed versions of a received rad~o or radar ~ignal which are superimposed on the received xadio or radar signal ~ 3 ~
~10~

it~elf, in the pre~ence o~ multipath interference.
D~tection and ~ub6equent elimination from consideration of these delayed multipath ~ignals is readily accomplished through variation of the particular ~mbodiment described in th~s disclosure.
Instead of ~ocu~ing on ~limination from consideration of e~ually-timed echos, the rules would focus on eliminatlon ~rom consideration (or perhaps time-correcting and constructive superimposing) of delayed ~ignals on a stronger, primary ~ignal. The prima~y ~ignal without the delayed signals (or strengthened by the construotive superimposing of the time-corrected delayed signals) would then constitute the informational signal structure.
Similarly, the expert removal of confounding æignal ~tructures finds application in ~eismic exploration. Wave energy returned ~rom geological formations i8 embodied in complex waveforms which contain signal structures w~ich are confounding signal structures (for example, echos reverberating within certain geological fo~ations at di~ferent depths). Isolation, and! elimination ~rom consideration, of these mutually con~ounding signal structur~s facilitates analysis of the individual, informational signal structures.
A cingle signal structure ~ay be both a confounding ~ignal structure (~or one purpos~) and an informational signal ~tructure (for another purpose~. Specifioally, in the oil well tubular application, multiple ctructures are confounding ~ignal 6tructures (and pit echos are information signal structures) when the purpose o~ analysi~ is, ~or example, to determin~ the presence of a small ~03- ~ $ ~ 7 ~

pit. On the other hand, pit echss are the confounding ~ignal ~tructure~ ~nd multiple 6tructures are the informational signal structures) when the purpose of analysis i8, ~or ¢xample, to determine the remaining wall th~cknes~ directly from the separation o~ multiple ~t~ucture echos.
With the under6tanding that the application of the rule-based analysl~ technique to ~ull echo wave~orms of acoustic echos returned from the surface of oil well tubulars is merely the preferred embodiment of the invention, the following discussion describes the preferred means o~ identi~ying multiple structures for purposes of exemplification, ~nd not limitation.
The identification of a multiple structure is based on the ~act that there must be a fixed separation between multiple echos. Once a dominant echo for the possible multiple structure has been identified, the system looks for a series of later echos which occur at a ~ixed (but as yet unknown) ~eparation. An effective rlecursive loop 1609 between MULTI FIND 1608 and MULTI DIFF 1606 do~s most of the work. (~ desaription of the rule interactions for the identification of multiples is given below. In addition, the rule firings ~or the example ~ignal o~ Figure 10 are presented in ~able 4, and are explained in the section "Processing Example," balow.) The recursion starts with the calculation of the separation between the dominant echo and the next echo. . Rule MULTI DI~F 1606 performs the actual ~ubtraction necessary to calculate the separation. This next echo i~ marked the ~irst-"current echo, '~ Rule P~ULTI FIND 1608 then looks for another, later echo which is separated ~rom the "current" by the 6ame separ~tion~ If such ~n echo i5 found, ~t beco~es the ~current" and another, later echo is investigated7 If enough of these echos ar~ found, a multiple ~tructure is identified.
As a final step ln the recursion~ the ~irst~
"current" echo after th~ dominant is marked the ~first ~ultiple." If there are no echos after the fir~t-~current~ echo which are separated by the proper interval, the rule MULTI START restarts the process. The next echo after the previous ~irst-"current" echo becomes tha ~Icurrent~ echo.
Possible multiple structures at this new, larger separation are investigated.
Since every echc~ after the dominant is considered as a possible first multiple, ~everal multiple structures can be id~entified in a complex signal. Rules within the `~ULI'IPLES KS decide which potential multiple structure is the most probable.
Once a single multiple: tructure has been identified, the individual echos are l~beled. The system, according to the preferred embodiment, does not spend any more time trying to identi~y the labelled multiple structure echos.
iii. ~HOS_Knowl edge_~ource The E~HOS Xnowledge source is a grouping of rules which are used ~or echo iden~ifi~ation. The three parts of this knowledge source have been shown previously as 1212, 1214, and 1218 in Figure 12. The parts are connected to other KS^s at ~3~49~0 connection poin~s ~, E2 and ~3 as indicated in Figures 12 and 17-19.

A.
~lL

Irhe ac~tivation diagram rOr the first part OI
the ECHOS KS i8 given in Figure 17. This group of rules i~ primar~ly involved with the identi~ication o~ a "dominant echo" (defined in Table I), The dominant echo is important ~ecause its presence in the signal indicates that a multiple tructure may be present.
In ~ore general ter~s, a dominant echo can be conceivecl as an indicator that a confounding signal structure (in the casa of oil well tubulars, a multiple structure) may be present. Different confounding signal structures (for different applications, such as multipath communications or seismic interpretation) ~ay have respectively different confounding signal structure indicators.
A single confounding signal structure ~ay have more than one confounding signal struature indicator. Conversely, a confounding signal structure may be of such a nature that the only confounding signal structure indicator is the confounding ~;ignal structure itself. Thus, a preferred embodiment'~ identi~Eication of a ~Idominant echo" as a con~ounding ignal structure (multiple ~;tructure) indicator ~s presented by way of example, and not limitation.
In a pre~erred eDlbodiment, the rule FIND DONINAN~ determines $~ a dominant echo exists.
This rule is activated if an echo has a height which is some predetermined ~actor (e~g., 2) ~3~9~0 ~106-greater than any other. Khowledge about the ultra~onic ~ethod would ~ugqest that such a large echo could have been generated by a large flat surface. Similarly, this flat ~urface may form a ~ultiple reflection path with the outer wall of the pipe.
The other rule in this part test for the presence of 6everal dominant echos. This case can occur when thexe is minor surface roughness on the inner wall of the pipe. The roughness can result in the generation of large ~eighboring echos at the beginning of the signal, a phenomenon called peak splitting.

B. Identification of Echos Other than an Identified_ Multi~Lq Struct~re tE2~

Once a multiple structure has been found, the second part of the ECHOS KS deals with the identification o~ all other echos tconnection point E2). The activation diagram ~or this second group o~ rules is 6hown in Figure 18. Two rules 1804 and 1806 from the ULTRASONICS KS must fire befsre any rules in this portion of the ECHOS knowledge source ar~ activated. These two rules are used to determine the remaining wall thickness, based on the ~eparation between ~ul~iples. They are considered part o~ the ULTRASONICS KS because they contain knowl~dge concerning the ultrasonic process. However, since the rule-based program is non-sequential, they can fire at any time. The activation of these two rules sets up ~or the identification of echos within the multiple structure.

~314970 In addition to ~he identification o~ echos from corrosion pits, both ~front-surfa4e" and "spurious" echos are ~dent$~ied. Each rule is used to identify these three echo types in different situations. The lettar~ FS, ~, S below a rule indicate the identi~ication of a "front-~urface,"
"pit" or "spurious" echo respectively. The "front-surface~ echo is generated by a reflect~on from the first material .urface encountered by the ultrasonic signal. "Spurious" echos are any echos which do not play an important role in the inter-pretation of the signal. For example, rule OVERL~PS DOMINANT (1836 in ~i~ure 18) ~inds spurious echo~ which are very near the dominant echo and are normally generated by minor pipe surface-roughness.

C. Iden~i~ication_of Echos in tk~
Absence of. Identified Mult~le ~truct.ures (E3~

The third part of the ECHOS KS handles the case when no multiple ~;tructure has been identified. The activation diagram is shown in Figure 19. In addition to 1:he ide~tification of front-surface and pit echos, cchos which overlap the front-sur~ace echo are investigated. When all echos are identified, control functions re~urn processing to the ULTRASONICS KS.

iv. U~TRASONICS Rnowledqe Source The U~TR~SONICS XS is a grouping of rules which contain knowledge about the ultrasonic reflection process. In addition, the determination 13~497~

of the characteristi~ of the coxrosion pits is performed by this KS. The activation-diagram is chown in Figure 20 with a connection point Ul to the control functions.
The top-level rules in Figure 20 are use~ to determine the pit characterist~cs for various types of pitted ~urfaceec ~he ~'ANDed" connections to the rules in the ECHOS XS indicate the dependence on the type o~ pit echo identified.
First, the depth ~f the pit is calculated from the time separation between the front~surface and pit echo using a formula:

Dpit = 0-5 * Ve * Tp_f where Ve is the acoustic velocity in the liquid, and Tp f is the time di~ference between the pit echo and the front surface echo. See the Principles of Analysis, Pit Depth section above~
Next, the location of the ]pit on the pipe wall is taken to be the location of the spot area. This location is read in with the ec]ho features.
Finally, the pit area is calculated using a Pormul~:

A it = (1-0 ~ ~FSE ~ ~FSE,o3 spot where Asp~t is the spot area, ~FSE is the present measured fr~nt surface echo height, and HFSE 0 is the mean front ~urface echo hei~ht as calculated or determined in the Principles of analysis, Pit Area section above. The front-surface echo height is thus used as an ~ndirect ~easure of the area of the pit.

~3~ ~7~

At this point, ~mory ~lements for detected pits have been designated, and the location, depth, axea and t~pe attributes assi~ned. This data is output to ~ ~il8, and the ~ystem i~ re-$nitialiæed for the next signal.

Thi6 6ection describes th~ opera~ion of the rule-based system for the actual signal shown in Figure 10. The rule ~ctivations for this typical real-world signal are list d in Table 4.
In addition, Figure 21 shows a copy of the user-interface di~play ~or :thi~ example. This display is a "snapshot" of the display after identification of the corrosion is complete. Each echo of the full waveform has been repres2nted as a rectangle in Figure 21. The height, width and location (time) o~ these rectangles correspond to the actual echos shown in Fiqul-e ~0. ~he echos are numbered sequentially as indicated below each rectangle. The identification of the echos changes as the signal interpretation proceeds.

~3 8 ~ e;
,, P 5~ ~ ~ ~U

o ~D~OOOOOO~OOO~O~ OO~O O a C o~,,,II,,~ ,,,,,~ ,,~, , 0~ O O C~ O O O~r~ I; O O O Q O~ O O O O ~r O I --~ c ~ 0 ~ t ~ ~ 0 ~ u a~q) o ~
r~ I~ ~.U ~-rl ~ rJ U a~ a) c _3 o ~ ~
t3~ ~ N ~ U ~
.~ 1~ l o ~ o ~ o ~1 ~ l ~1 (~ ~ o o~ o t~ ~ a)h ~1~1 o Iq ~ o ~ C
o o~ o a o o ~ ~ o o o o o ~ ~ o o o o ~ o m ~ e1~ ~00UO~u~1~U~VOo~ U~U~O 0 o ~ ~o ~oooooo~ ~o~o~o~ ~ooo~oo~

~ ~ ~ 0 ~ 0 ~ 0n ~
~U~ OO~OOO~ OOOOOO~ UOOOOO~OOL~
U~ ) C.l C) U C~ O V-- IY C) O C) ~ P~ V ~ h ~:1 1~

a) ;;
~1 I I I~n a ~ a ~ ~ z 2H~ 14 F': ~ ~ h h ~ h P4 ISi ~ 4 ~ Z 54 æ P
~; O H 1~1 E I H H H H H H E~l H 1~1 H H H E ~ H H 1~1 1-1 H 1~1 H O E-l I~ tq U~ H H H H l-t H H H H H t~l 1-1 H H H H 1-1 H H H H H H

ul o ~l o u~ o ~

3~97~
, ~q ~ I` f~ ~ ~ OD ~ O
O O O ~n ~ O o ~D ~ o t~
o ~ ~ ~ o 9 ~ o o ~ o o o o ' ~ 0 ~ ~ ~
u a~ 0~ c P 3 ~ 0 0 0 0 0 0 O O O ~ ~ O O rl ~ O ~ ~q~
~ ~ æ ~ h 1~3 ~ ~ ~

~u ~ ~ u O ~ u t) O
:> ~ h ~ h ~ ~ a! 0 ~ O
~i ~ h ~ ~ 2 ~ ~
41 q-JO ~ ~ ~ O ~ S O ~ h .r~ -rl ~ J-rl-r~ ~ ~-~1 ~ .C JJ ~ J~ ~ ~ ~ ~ ~
1~ ~ u~ ~U uuu~uou~ ~ ~J~
o ~ . P ~ In o 5: C ~: C.~ h t) U~ ~O o~?
O . ~ "~ h B`
I O O O ~ ~ O O ~ ~ O ~ ~ 0 ~O ~D 0 ~
. I ~ ~ ~ ~ OU U
.~ I ~P.~ P~O~ ~o~ ~oooooooa~
~11:~ 13 E3 0 ~ 13 13 0 .C ~ 13 0 .C ~ .C ~ R ~S Q
,~ ~ v u ~ _ ~ u ~ 8 ~ ~ ~ ~

~ ~ ~ , h ~ E~

zl ~4 h 1~ E~IE1 E~ E~ El E~ E~ ~ æ æ E I m El El m E~ E~
HHH E~HH E~H h1i3$~IYW~ 'CH~ HC~ H:~
a~ Q ~ n ~ ua ~ P~ ~ ~ P~ P~
~ H H H H i~ 1 H H H H H H H 1-1 H H H 5-1 H ~; ~ H p~ ~ H p; ~
u ~ Z

u) o u~ o In O ~

U~
~ _ o o ,~
~a o .~ ~ o ~.
~ ~ ~ o ~
O ~ æ ~

~ ~ ~ O ~ O O ~ O
h Q3 Q~ C~ ~1 .3: rC U ,t~
~a o v a~
rl O O H
o ,C ~q ~ ~ .c c~
o :~ so ~ O 0 ~ 0 rc ~ ~ ~ u ,l O u u 5~
rl ~ a\ o ~ RS h h I ~ ,s s~ 10 4~ ~ O O E-l . ~ ~ O O ~ ~ ~ ~ ~ h ~ ~ H

Qi ~ ~ O O ~ h I ~ U O h E3 t~ U t) ~ O ~1 0 O O :~
I ~ O t~ rl ~ O
D 0 1 11~ rl rl ~1 0 ~ O O O O ~ ~rl . ¦ ,~ ,~ rl .~ rl ~rl ~rl r~ r l r~l ~1 C~
t ~ v H 1 ~3 13 13 h 1~ ~ E3 13 ~ 13 ~ r~l Ul H rl H Z
~ I c ~ ~ ~ 8 ~ ~, ~3 w ~3 H H 1-1 H ~ U .t~ H

~ ~ P~ ~ PC,~ .
H
o ~ ~ E4 a ~
1-1 H H i~l ~) O P~ Z ~1 1~1 H 1~1 t l H ~ ~ ~; H

H1~ 1 H H E~1 ~ Z Z Z Z;

0 ~ Z Z ~ ~ ~¢ p3 Q ~ ~ ~
~ x ~ ~ æ ~ a p~
r p~ H fi ~j O ~ H H H P1lH ~ H H ~ H
~ U ~ U ~ u ~ u ~ 8 ~
v ~ u t~ ~ P P P~ ~ P~ U ~ P P ::~ P

o In o -113- ~ 3~ ~97~

~he lnterpretation steps of the systems are best understood by looking at Table 4 and Figure 21. Once the echo features are extracted and tored as working memory elements, rule activation S starts with CONT~ DOM CO~PUT~. 5ince the example signal was xecorded from a pipe area that was only about 50% corroded, multiples from the good pipe surfaces are present. The recursive looping of MnLTI DIFF and MULTI FIND is uæed to identi~y a multiple structure consisting of echos 1-4-6-8.
Echos 2, 3 and 5 are identi~ied as echos from pit surfaces, and echo 7 aæ a spurlous echo. Woxking memory elements for the pits are created, and the deepest of these (5~ i~ sized and located. The deepest pit in a sensed area is of primary concern in a practical embodiment o~ the invention.
Echo 7 is identified as a spurious echo due to a low height compared to the pit echo heights.
This echo was a reflection from a deep val~ey close to the edge of the ~ensed are~a for this signal.
Since only a very small portion of this valley was intercepted by the ultrasonic w,ave, the echo height is small. Thus the interpretiation as a spurious echo is appropriate. The s~nsed area fox a later scan would be expected to detect this corrosion valley.
The actual example signal discussed here is representative of the majority of the signals which are produced by real corrosion pitting. The system according to a pxeferred embodiment of the invention has been tested on a wide variety o~
signals from both simulated and real pits, and the system's per~ormance appear~ to be as good as manual interpretation by a human expert.

-114 ~3~497~

AP~ENPI~
The important operational rules of the preferred embodiment o~ the Expert System are described in this Appendix. Other rules, whose presence a~d implementation would be well known to t~ose skilled in the art, are not described here.
(Such rules include the housekeepins rules which clear or reset attributes before or after activation of the important operational rules.
Also, means of inputting and outputting information, either textual or graphic, lies well within the ability of one o~ ordinary skill and will not be detailed here. Finally, the ability to gain information about the outer surface of the tubular may be gained using ext~ensions o~ the descriptions contained in this Appendix and in the above Detailed Descr$ption section.) -115- ~3~ ~9~ ~

~ULES IN "MULTIPLE" KNOWLEDGE SOURC~E
(Figure 16: the "MUL~I n prefix has been omitted for brevityO ) 1604. 5T~RT:
Begins ~earch for a multiple structure within echos after a dominant echo. Sets the current possible ~ultiple to the first un-marked echo after the-do~ina~t. Computas a location ~eparation limit egual to the dominant-current echo separation which is used to test later echos ~or inclusion in multiple structure (DIFF and FIND). Marks the current echo.
Creates a working memory element of class "structure", type l'multiple" to denote a potential multiple tructure defined by this separ~tion limit.
1606. DIFF:
Computes the location difference between the current echo and each successive unidentified echo after the current echo. Location differences are computed sequentially, in order of recency in ~he working memory, until MULTI FIND fires.
1608. FIND:
Teæts an echo a~ter the current ~or inclusion in multiple structure by computing the upper location di~erence with upper and lower location limits. ~f the location difference aomputed by DI~F falls wi1:hin these limits (within a~pecified constraint defined by "constraint on location ~or ~ultiple" (Table 3)~, the acho-i~ included in the struct~reO
If several possible echos are within the cons~raint, ~he one having a lscation difference closest to the exact location (~ltimately d~termined ~y START, above) i5 chos~n. ~ets ~he current echo to the new multiple echo and marks it. This rule ~orms a recursive structure 1609 with DIFF to search through all echos *or a multiple structure associated with the dominant echo.

1610. FIND_LAST:
Completes the id~n~i~ication of multiple echos when there are ~o echo~ a~ter the current echo to be consldered. ~his rul~ looks at the last echo, whereas FIN~ looked at all bu~ the last.
Thi~ rule relaxes ~he condition, present in FIND, of having still another multiple structure eaho after the current echo. This rule only look~ "backward" to corroborate the last echo' e inclusion in a multiple ~tructure.
1612 FOUND:
Completes the search for a multiple structure.
There are no unidentified echos between the dominant and first muItiple which have not been considered ~o~ inclusion in some multiple ~tructure. The number of mu~tiple echos for the curr2nt ~tructure must ex~eed an initial constraint "number of echos_for~multiples~
(Table 3), which contraint i8 an integral number of echos, typically ~.
1614. SET REMOVE_NIL:
Removes from working memor~ multiple structures which had been g~neratad by START
which have no identi~ied multiple echos.
1616. SET_REMOVE NUMBER:
I~ there are two or more ~ultiple structures associated with a single leading echo (an impossible circumstance for an uncorroded outer ~u~ular wall), this rule removes the one wit~ fewer identified echos. : :
1618. SET REMOVE LAST:
I~ (1) there are $wo or more multiple structures associated with a ~ingle leading echo, and (2) these tructure~ have the same numbe~ ~ multiple~, remove from working memory the multip}e ~tructure with least locat~on 6eparation between the multiple echos.

~117- ~ 3 ~

1620 SET COMPLE~E~
.
Label~ (i.e., changes the "type" atkribute) the dominant echo ~s the leading echo when only sne multiple s~ructure for this dominant echo remains.
1622. SET FOUND ~ANY: -Completes the search for multiples and labels the leading echo a~ ~oon as the number of identified multiple~ echo6 exc~eds a given Gon~tra~nt '~number o~ -echos_for~large_multiple_set" which is an integer number of echos, typically 4.
1624. CLEAN DIFF.
~emoves from working memory the location difference values computed by DIFF after the multiple structure is complete. 1.
1626. LABEL START~
Init~ates the labeling of the identified multiple echos ~or subsequent rules.
1628 LABEL:
La~els (i.e., defines ~he 1'type" attribu~e of) the identi~ied multiple echos in a completed multiple ~tructure.
1630. L~BEL DONE:
Completes the labeling process ~i.e~, sets the "statu~" attribute to "labelled" in the appropriate working memory element o~ class "structure").

~118- ~ 3 `~ ~ 9 ~ ;~

(Figure~ 15 and 18; the "C." pref~x has been omitted for brevity) 1504. DOM COMPUTE:
Initiate~ the ~earch for a dominant echo by computing a co~straint on tha height of this echo. The constrain~ is ~he guotient of the height of the echo with the greate~t height and an initial div$sor con~traint_on --value_domlnant (~ypically 2.0~.
1506. HEIGHT COMPUTE NO MULTIPLESo This rule calculates height values used to :test for pit and spurious echos when no multiples are present. The v~lues are the quotient of the maximum echo height and ~nitial divi~ors. Typical multipliers for pit and spurious echos are 4.0 and 5.0, respectively (constr~ints "ratio_~ax/large ~it" and "ratio_max~spurious"
in Table 3).
1508. DONE ALL ID~NTIFIED:
_ _ This rule fires as soon as all echos have been identified ~i.e., there are no unidentified echos). The ULTRASONICS RS's task of determining the corrosion from the identi~ied e~hos is i~itiated.
: 1510~ SET F~ONT_SURFACE HEIG~T;
If the current signal is the first to be considered, ~he running average value for the heiqh~ ~f the ~ront-~ur~ace echo is taken ~o b;e the value in constraint "de~ault_front_ ~urface height". This value is obtained from readings on signals previously determined to be from uncorroded pipe. The average front-surface height is used in the ULTRA50NICS KS
to estimate the pit opening area for pits in subsequent signals. This rule fires only once~ to define the initial default value for the average ~ront surface height.

-1 19~- ~L 3 ~

1522. COMPUTE AVERAGE FRONT SURFACE HEI¢HT:
The avera~e front-surfacQ echo height i6 updated when the current 6ignal indicates that no corro~ion is pre6ent. The dif~erence~
between the current ~ront-sur~ace hei~ht and the average ~ computed ahd divided by:the acaumulated count on the number of updates.
Thi~ quotient i~ then added to the past a~era~e to ~i~e ~he new aYera~e, ~na the update count i~ incremented by one.
1834. PIT OVE~L~PS MULTIPLE_Co~PUTE ~Figure 1~):
I~ ~1) no pit has been identified and (2) a multiple structure has been identi~ied, then compute the sum o~ the widths o~ the multiples, and mark each multiple as the width is summed. Thi~ rule is preparatory ~or the ~ollowiny rule, 18420 1842. PIT O~ERLAPS_~ULTIPLE COMPU~E DONE (Figure 18):
If the sum of the multiple widths has been computed, then compute a value to be used to ~est for a pit overlapping a multiple. The value is calr.ulated by dividing the sum by the number o~ multiple echo~, and multiplying by the constraint "con~traint on_overlap_o multiple" (typ. 1.~. The computed value i~
used by ECHOS: PIT OVERLA.Ps MU~TIPLE (1848 in Figure 18).
1844. PIT OVERLAPS ~UL~IPLE COMPUTE W NE R~MARX
(Figure 18):
Un-marks multiple echos (i.e., sets the ~Imark~
working attribute of ~he echo element ~o 'inil") once the sum o~ ~he multiple widths has been computed.
~838. PIT IN MULTIPLE COMPUTE (Figure 18):
If a multiple structure and pit ~cho have been ~ound, compute a constraint ti'~purious height in multiple") on the height of a spurious echo within the structure. ~he constraint is oomputed a~ the quotient o~ the pit ~cho height and the constraint "height ratio Pit/spurious" ~typically 2~.

-120- ~ 3 ~

~ .s ~ cho~ E~O ~ DG~ ~QUR OE ~
(Figures 17, 18, 19; the "E:'l prefix has been omitted for brevity) 17~8~ FIRS$
If an ~cho has the ~mallest location attrlbute value for any achos in the ~ignal, then change the value of the "type" attribute to "first'~.
1706. FIND DOMINANT:
I~ the height of an echo exceeds the current constraint on dominance tas computed by DOM_COMPUTE 1504), the echo is identified as both the dominant echo ~nd ~he current echo.
Thi~ rule initiates t~e search for ~ultip~es associated with ~he dominant ~cho.
1704. MAX OVERLAP:
If both a maximum hei~ht echo and a neighboring echo exist with heights exceeding the curren~ constraint cn the dominant, prepare for testing o~ an overlap condition.
B~th upper and lower values are computed as follows. The upper value is the location of the ~aximum height echo plus the average o~
thè echo widths. ~he low~!r value is the location o~ ~he maximum h~ight echo minus the averaqe of the echo widths. I~ the location of the lower height echo falls within these bounds, the echo is determined to overlap the ~irst ~maximum heig~t) echo.
1712. WIDE DOMINANT:
I~ a neighboring echo overlaps the maximum height echo, the neighboring echo is identified as a spurious echo and the maximum height echo is labeled the dominant. The neighboring echo must be less than some pèrcentage of the ~aximum height echo ("constraint_on_leading value" (Table 3), and iB typically BO%) but greater than the constraint on dominant value (i'constraint_on_value dominant" from DOM COMP~TE 1504). This rule initiates the -121- ~ 3 ~

~earch for multiples a~soclated with the dominant echo~
1714. WIDE_DOMINANT~FX~ST:
I~ (1) a neighborlng echo overlap~ the maximum heiqht echo, and (23 the neighboring e~ho is greater than ~ome ~ultipie of the maximum hei~ht typ. O.8, ~rom ~Iconstraint_on_value dominant" from ~OM CO~PUTE 1504), then identi~y the neigh~oring echo as th~ dominant echo and label the ~aximum height echo ~purious. This rul~ lnitiates the ~earch for multipl~ associated with the dominant echo.
1812. LEADIN~ PIT FRONT:
I~ (13 th~ remaining wall thickness calculated by COMPUT2 REMA~NING WA~L THICKNESS 1806 for a multiple structure is less than a constraint : determined by 'imin nominal wall thickness", typically 0.3 inches and (2) ther~ are no echos except those associated with the multiple structure, then identify the leading echo as a pit surface. Thi~ is the rule analyzed in deta-il in the section entitled "Firing of Rules in the K~lowledge Base".
1848. PIT OVERLAPS ~ULTIPLE:
3~
This rule identifies a pit echo which occurs at a location so alose to that of a multiple that no distinct pit echo i~ detected. If (~) there are no echos in the signal other than those alr~ady ascribed to a multiple structure~ and (2) a multiple echo has a width whieh exceeds the value computed by CONTROL:
PIT_OVERL~PS MU~TIPLE COMPUTE DONE 1842, then identi~y the multiple echo as a pit echo.
1816. DOMINANT OVERL~P COMPUTE:
When tl~ a ~ultiple structure and dominant echo exist ~nd (~) an unidentified echo neighbor~ the domlnant, then compute the overlap Gondition~ for these echos ~ee MAX OVERLAP 1704).
-122- 1314~7~

1836. OVERL~PS DO~INANT:
I~ (1) an unidentified echo neighborE the dominant echo in a multiple ~tructure, and (2) the n~ighboring echo overlaps the dominant ~ov~rlap conditions atisfied), then identi~y the neighboring echo as ~purious.
1818. PIT_IN_MULTIPLæ:
1) an echo ~xists within a multiple structure, and ~2~ it i~ the largest unidenti~ied echo~ then Identify this echo as a pit echo (i.e., change its working memory element's "type" attribute to "pit'l~.
1846. PIT IN MULTIPLE_SPURIOUS:
I~ (1) a pit echo and an unidentified echo exist within the ~ame signal as a multiple structure, and (2) the height o~ the unidentified echo is less than the constraint "spurious h0ight_in multiple", then identify the echo as spurious. The constraint is typically se~ to 0.2 ~imes the height of the largest echo ~or the ~ignal (see CONTRO~:
PIT IN_ MULTIPLE_COMPUTE 1838~.
1822. FRONT PRECEDES D0MINANT:
When (1~ the ~ir~t echo of a signaI is located before the leaqing echo of a multiple structure and ~23 this ~ir~st echo is of lower height than the leadlng echo, then identify . the echo as the front-sur~ace echo. However, the rule FIRST IS SPURIOUS COMPUTE (1826, described below) has the "higher priority"
then this xule (fir~s first if the conditions -of b~th rules are satisfied).
1826. FIRS~ IS SPURIOUS_COMPUTE:
When t~le fir~t echo i~ located be~ore th~
leading echo o~ a multiple structure, compute a location constraint for the echo to be identified as spurious. The ~onstraint value is calculated as:
2 ~ (nominal ~ell thicknoss - remainin~ ~ell thiekness) ~ ~ampl~s/second) ~
~spoed of $ound in liquid) 123~ 7~

This constraint value i~ the location difference between (1) a po~ ible front-sur~ac~ echo ~n~ ~2) the leading echo calculated on the ba~is of the mea~ured re~aining wall thickne~s. The nominal wall thicknes~, ~amples/second, and ~peed o~ sound in liguid are pre-program~ed, as described in the di6cu~sion related to Table 3. The remaining wall thickness is calculated by U CO~PUTE REMAINING WALI, ~HICKN~SS 1806.
1840. FIRST_IS SPURIOUS:
If the separation ~etween the ~irst echo and the leading echo exceeds the constraint calculated ~n FIRS~ IS SPURIOUS COMPUTE (1826, above), then the first echo cannot be . ~ront-surface echc. (I~ it were from the front sur~ace, then the le~ding echo would have to have come from a surface beyond the pipe outer wall, which is impossible.) The ~irst echo is identified as spurious and the leading echo is marked as the front surface echo.
1828. FIND FRONT DOMINANT-If there are (1) no unidentified echos and (2) no echos before the leading echo o~ a multiple structure, then identify the leading echo as the front-surface echo.
1830. FIND_FRONT SPURIOUS:
If (1) there are no unid~ntified echos and ~23 thQre is.a single spurious echo before the leading, then:iden~ify thi~ echo as the ~ront-surface echo and generate a structure 'l~puriou~ as front surface." The rule F~RST_IS_SPURIOUS COMPUTE 1826 has higher priority.
lgO4. F~ONT_NOT MAX NO ~ULTIPLES: (No ~ultiple Structure) Xf ~1) there is an echo before the echo with maximum height with a height which exceeds the constraint ~n pit height computed in HEIGH~ COMPUTE NO ~ULTIPLES (1506, Figure 15), and ~21 there are no echos ~efore this echo with heights greater than the constraint ~124- ~3~7~

~'6purious height no ~ultiples" (from HEIGHT CO~PUTE NO MULTIPLES 1506), then identify the echo as the front-surface echo.
1906. FRONT_MAX FIRST NO ~ULTIPLES: (No ~ultiple Structure~
If there ~re no echos before the echo with ~axi~um heiqht which have height~ greater than the constraint ~'~purious height no multiples"
(from HEIGRT COMPU~E NO MULTIPLES 1506), then identify the maximum heiqht echo as the front-sur~ace echo.
1908. FRONT NOT M~X FIRST NO ~ULTIPLES: (No ~ultiple Struc-ture) If there is an echo before the echo with ~0 ~aximum height which has a height greater than the constraint "s~urious height no ~ultiples"
(from HEIGHT ~OMPUTE NO ~ TIPLES 1506), then identify this echo as the front-surface echo.
1914. FRONT SURFACE OVERLAP COMPUTE:
I~ ~1) a front-surface echo (or spurious ~cho ~fter the front-sur~ace) have been found, and (2) there is an echo which precedes or is after this echo, then compute the location for e~ho overlap with the ~roJIt-sur~ace or spurious echo (see MAX_OV~RL~P 1704).
1916. FRONT_SURFACE OVERLAP:
I~ ~1) a front-surface echo has been identifi~d and (2) th~re is an echo which precedes the front-sur~ace or is after the ~ront-surface ~nd 13) its location: i5 within the limits for overlap with the ~ront-surface, then (1) mark the echo as a spurious echo and (2) add a working memory element of class I'structura" with type "wide front surface."
: The limit~ for overlap ~re computed by FRONT SURFACE OVERLAP COMPUTE 1914.
1918. FRON~ SURFACE OVERLAP SPURIOUS:
I~ ~1) there is an overlap between a front-surface echo and following spurious echo and (2) there is an echo which follows the -125- ~3~7~

~purious and overlaps it, then ~ark the echo as a ~purious echo and (add a working memory element o~ class ~tru~ture"3 and type "wide front_sur~ace." Note thi~ rule forms a recursive structure 1920 with FRONT SURFACE OVERL~P COMPUTE.
1922. FRONT MhX_PIT NO MULTIPLES: (No Multiple Structure) If ~1) a ~ront-sur~ce echo ha been identified~ (~2) no structure has been found other than "wide front-6urface," (3~ an echo of height ~reater than the constraint "pit_ height_no multiplesl' occur~ after the front surface echo (CONTROL: HEIG~T CO~PUTE NO MUL-TIPLES 15~6) and ~4) there are no unidentified echos with greater height than this echo, then identify the echo as a pit echo and generate a structure (i.e., add a working memoxy element of class "structure") "large pit after_-front_~urface_no multiplesl'.
1924. FRONT NEW PIT NO MnLTIPLES: (No Multiple Structure) If ~1) a structure, "large pit_after front surace no multiples"
(from 1922) has been identified and (2) there is an echo after the front ~urface with a height ~ than the constraint "spurious height no multiples" (from HEIGHT CO~PUTE NO MULTIPLES 1506~, then identify this as a "pit echo".
192B. FRONT PI~ SPURIOUS_NO_MULTIPLES: (No ~ult~ple Structure) If (1) a ~tructure, "large Pit-a~ter ~ront_sur~ace no multiples"
has been identified ~in 192~j and (2) there is an eaho after the front surface with a height less tha~ the constraint "spuxious height no multiples," then identify these thi as a ~purious echo.

~ LTRASONICS RNO ~ DG~S-~R OE ~
(Figures 18 and 20; the ~U: n prefix is omitted for brevity) s For the following rule6, the depth o~ the pitted surface gi~ing rise to an identified pit echo is determined as:
t (pit cho location ~n Js~ples) - ~tront-eurf~loe echo locotion) ) _ _ t 2.0 e 5~pl~ cond) / tspeod of ~ourld in liquid) ) $5 ~he factor of two account~ ~or the "round-trip"
path taken by the signal as ~t travels between the front-sur~ace and the pit ~ottom through the liquid~
The pit opening area is calculated as follows:
~1.0 - (tfront sur~s~e height)it~verHI~e f s height))) ~ po~-~rea) 1804. COMPUTE MULTI~LE SEP~RA~ION:
This rule aomputes the average location separatio~ between echos in a multiple structure once a multiple ~;tructure has been ~ound. The separation is aomputed as the difference between the last: multiple echo and tbe leading echo of the multiple structure, divided by the tot`al number o~ multiples.
1806. COMPUTE REMAINING WA~L THICKNESS~
_ _ -Onae the multiple stxucture's echo separation has been computed ~1804)~ the remaining wall thickness can be calculated based on the multiple structure. If tha average multiple separation is giv2n in terms of the number of equal-spaced time ~amples, the remaining wall thickness can ~e calculated as:
t~verase seporotion in s~mples) ~ ~speed of sound in pipe ~oterisl) ~ 2.0 ~ ~number of tiQC s~mple~ per ~econd) ~

The factor of 2~0 i~ the denominator correa~s ~or the fact that the ~ignal passes through the pipe wal~ twice for each multiple reflection.

2006. S~ALL_PIT_OVERL~PS_MULTIPLE:
If both a multiple ætructure and a structure tipit overlaps mul~iple" haYe been ~ound, then use formula Al, above, to determine the ~epth of the pit ~rom the loca~ion di~ference between the front ~urface and the overlapped multiple. Create a working ~emory element of class np~ t" with thi~ calculated depth inserted ~nto the "de~th" attribute (i.e., store the depth as ~ new pit attribute).
2010. PIT D~PTH IN MULTIPLE:
If a pit echo has been identi~ied within a multiple structure which does not overlap an e¢ho in the multipl~ structure, then determine the depth of pit ~rom location difference between t~e front-sur~ace and pit echo.
Create a working memory eI~ment o~ cla~s "pit"
with this calculated depth inserted into the "depth" attribute (i.e., store the depth as a new pit attribute).
2016. PIT DEPTH LARGE PIT:
I~ a structure ~large Pit after front_surface' or "large_pit a~ter front surface no multiples"
has been found, then determine the depth o~
pit from thè location difference between the front surface echo and pit echo. Create a . working mèmory element of claæs "pit" with this calculated depth inserted into the "depth" attribute~ e., store the depth as a new pit attribute).
2020. PIT IS FRONT SURFACE-~0 I~ a ~tructure "pi~ dominant as front_surface"
has been ~ound, then compute the depth of the large pit ~s the difPerence between ~he nominal pipe wall thickness and the measured remaining wall thickness. In additio~ mark the pit ~s type "front surface." Create a working memory ale~ent of class "pit" with this calculated depth inserted into the "depth~ attribute (i.e., store the depth as a new pit attribute).

-128- ~3~

2024. MULTIPLE NO ORROSI~N:
If (1) a multiple ~tructure has been ~ound and (2) there are no identified pits and (3) the rema~ning wall thlc~ness ~s greater than the nom~nal wall thickness, ~hen conclude that there ~s no corro6ion, and create a new working memory element o~ class "corrosion"
and attribute "no_corrosion1'.
2028. PIT LOCATION:
If ~1) a pit has been found and (2~ it is the deepest ~or the signal, then store the location o~ the center of the 6~nsor spot-area a a pit attribute~ (~he ~ensor`spot center is obtained when the signal data is read into the program at 1202 (Figure 12).
2038. PIT AREA:
If ~1) a pit has been ~ound and (2) it is ~he deepest for the ~ignal then compute the openlng area o~ the pit as a fraction of the sensor spot-area according to the ~ormula given above. Store the opening area as a pit attribute.
2034. PIT AREA SPLIT LEADING:
If (1) a pit has been found and (2) i~ i~ the deepest ~or the signal and (3) the front-surface echo overlaps the leading echo (a working mèmory ele~ent of structure "spurious_as~front surface" has been ~
previously identified~, then add the heights of the spurious echo to the front-surface height before computing the opening area of the pit. Store the opening area as a pit attribute.
2042. PIT AREA_NO_FRONT:
I~ a pit of type "front-sur~ace" has been found, then set the opening area to 100% o~
the ~pot area. Store the opening area as a pit attribute.

L3~970 ~ON~ILUSIOM

While ~rariou~ ~Eea~ure~i; of particular embodiment~ according to the preE;ent invention have been pre~ented ~bove, it i~; to be understood that they have been presenlted by way o~E example, and not 1 imitation . Thus, the breadth and ~cope of the invention are to be defined not by the above exemplary embodiment~;, but only in accordance with the following claim~; and their equivalents.

Claims (30)

1. An acoustic logging device, comprising:
excitation pulse generation means for generating an excitation pulse;
transducer means, having a transducer surface and connected to said excitation pulse generation means, for receiving said excitation pulse and for launching in response to said pulse, an acoustic signal from said transducer surface toward a target surface, said target surface reflecting a portion of said acoustic signal back toward said transducer means, full echo waveform detection means for receiving acoustic energy returned from said target surface and for generating a full echo waveform including echo heights, widths and relative positions in time, analysis means for analyzing said full echo waveform to determine the condition of said target surface, said analysis means comprises a rule-based Expert System for analyzing said echo heights, widths, and positions so as to determine the condition of said target surface, and said rule-based expert system having a knowledge source comprising one or more opportunistically executed rules.

2. An expert system for time-domain analysis of a waveform, comprising:
an electronic blackboard comprising:
(i) time-domain waveform feature data comprising representations of the waveform as a function of time, and (ii) constraints, said constraints relating to said waveform feature data and to the ways in which said waveform feature data may have been formed;
a first knowledge source containing knowledge relating to confounding signal structures in the waveform;
a second knowledge source containing knowledge relating to informational signal structures in the waveform, whereby items of said waveform feature data may be at least tentatively identified as confounding signal structures or as informational signal structures;
a third knowledge source containing knowledge relating to the behaviour of wave energy in a physical environment, whereby said at least tentative identification of said items of said waveform feature data as confounding signal structures or as informational signal structures contribute to an at least tentative determination of the nature of said physical environment; and a fourth knowledge source containing knowledge relating to the application of the knowledge in said first and second knowledge sources to identification or tentative identification of items of said waveform feature data, whereby the application of the knowledge in said first and second knowledge sources is controlled;
wherein each of said knowledge sources comprises one or more opportunistically executed rules, each of said rules comprising:
(i) one or more conditions related to information stored on said blackboard, with (ii) one or more action statements which are executed when the conditions are satisfied.

3. An expert system for analysis of a time-domain waveform, comprising:
an electronic blackboard comprising time-domain waveform feature data comprising representations of the waveform as a function of time; and a procedural knowledge base comprising opportunistically executed rules for selectively responding to said time-domain waveform feature data and for selectively modifying said time-domain waveform feature data, each of said rules comprising:

(i) one or more conditions related to information stored on said blackboard, with (ii) one or more action statements which are executed when the conditions are satisfied.

4. The system of claim 3, wherein said procedural knowledge base comprises plural knowledge sources.

5. The system of claim 4, wherein said knowledge sources comprise:
a first knowledge source containing knowledge relating to confounding signal structures in the waveform; and a second knowledge source containing knowledge relating to informational signal structures in the waveform, whereby items of said waveform feature data are at least tentatively identified as confounding signal structures or as informational signal structures.

6. The system of claim 5, wherein:
said first knowledge source contains knowledge relating to confounding signal structures which comprise multiple structures formed by wave energy reverberating within a physical structure; and said second knowledge source contains knowledge relating to informational signal structures which comprise wave energy echos formed by the smooth and/or rough surfaces of said physical structure.

7. The system of claim 5, wherein:
said first knowledge source contains knowledge relating to confounding signal structures which comprise wave energy echos formed by the smooth and/or rough surfaces of a physical structure; and said second knowledge source contains knowledge relating to informational signal structures which comprise multiple structures formed by wave energy reverberating within said physical structure.

8. The system of claim 5, wherein said knowledge base further comprises;
a third knowledge source containing knowledge relating to the behaviour of wave energy in a physical environment, whereby said at least tentative identification of said items of said waveform feature data as confounding signal structures or as informational signal structures contribute to an at least tentative determination of the nature of said physical environment.

9. The system of claim 5, wherein said knowledge base further comprises:
a fourth knowledge source containing knowledge relating to the application of the knowledge in said first and second knowledge sources to at least tentative identification of items of said waveform feature data, whereby application of knowledge in said first and second knowledge sources is controlled.

10. The system of claim 5, wherein said blackboard further stores constraints relating to said waveform feature data and to ways in which said wave-form feature data may have been formed, whereby said first and second knowledge sources are guided in the at least tentative identification of said items of said waveform feature data.

11. A method of determining the condition of a target surface, comprising the steps, substantially opportunistically executed, of:
launching an acoustic pulse from a transducer surface toward a target surface;
receiving echos of said acoustic pulse from said target, wherein the waves of said acoustic pulse travel paths of substantially identical lengths between the target surface and the transducer surface when the target surface is free if physical defects;
transmitting full echo waveform data describing said acoustic pulse echos in the time domain;
storing said transmitted full echo waveform data on an electronic blackboard;
ascertaining the presence or absence of a confounding signal structure indicator in the time-domain full echo waveform comprising representations of said waveform as a function of time;
distinguishing, in the ascertained presence of said confounding signal structure indicator, confounding signal structures from informational signal structures, if any;
identifying said informational signal structures, if any;
applying knowledge relating to acoustic wave properties in a physical environment in which the target surface is located to said signal structures;
forming conclusions, from said applying of knowledge, regarding the condition of the target surface; and displaying said conclusion-; for human perception;
wherein said distinguishing, identifying, applying, and forming steps each comprise, opportunistically firing rules, each said rule comprising:
(i) one or more conditions related to information stored on said blackboard, with (ii) one or more action statements which are executed when the conditions are satisfied.

12. A method of expert analysis of a time-domain waveform, comprising the steps, opportunistically executed, of:
ascertaining the presence of a confounding signal structure indicator in the time-domain waveform comprising representation of said waveform as a function of time;

distinguishing, in the ascertained presence of said confounding signal structure indicator, confounding signal structures from informational signal structures, if any; and identifying said informational signal structures, if any;
wherein said distinguishing and identifying steps each comprise opportunistically firing rules, each said rule comprising (i) one or more conditions, with (ii) one or more action statements which are executed when the conditions are satisfied.

13. The method of claim 12, wherein said ascertaining step comprises ascertaining the presence or absence of an echo whose magnitude exceeds a predetermined multiple of the magnitude of all other echos in the waveform.

14. The method of claim 12, further comprising the steps, opportunistically executed, of:
applying knowledge relating to acoustic wave properties in a physical environment in which a target medium is located; and forming conclusions, from said applying of knowledge, regarding the condition of the target medium.

15. The method of claim 14, further comprising the step of displaying said conclusions for human perception.

16. A method of expert analysis of a time-domain waveform, comprising the steps, opportunistically executed, of:
ascertaining the presence or absence of a confounding signal structure indicator in the time-domain waveform comprising representations of the waveform as a function of time;
distinguishing, in the ascertained presence of said confounding signal structure indicator, confounding signal structures from informational signal structures;
if any;
identifying said informational signal structures, if any;
applying knowledge relating to wave properties in a physical environment in which a target medium is located to said signal structures;
forming conclusions, from said applying of knowledge, regarding the condition of said target medium; and displaying said conclusions for human perception;
wherein said distinguishing, identifying, applying, and forming steps each comprise opportunistically firing rules, each said rule comprising:
(i) one or more conditions, with (ii) one or more action statements which are executed when the conditions are satisfied.

17. A method for preprocessing and transmitting echo waveform information derived from an echo of an acoustic wave reflected from a target surface, comprising the steps of:
digitizing an echo waveform produced in response to a launching of an acoustic wave;
detecting amplitude of said echo waveform;
continuously storing said digitized echo waveform in a recirculating memory;
freezing information stored in said recirculating memory during a predetermined time period beginning after the detected amplitude of said echo waveform rises above a predetermined threshold;
measuring a time interval between said launching of the acoustic wave and the rise of said detected amplitude of said echo waveform above said predetermined threshold; and multiplexing the time interval measured in said measured step and the waveform information frozen in said recalculating memory in said freezing step for transmission to a location remote from said acoustic wave.

18. A method for preprocessing and transmitting echo waveform information derived from the echo of an acoustic wave reflected from a target surface, comprising the steps of:
converting an echo waveform produced in response to a launching of an acoustic wave to a monopolar waveform;
smoothing said monopolar waveform, whereby a smoothed monopolar waveform is formed;
compressing the dynamic range of the magnitude of said smoothed monopolar waveform, whereby a compressed, smoothed monopolar waveform is formed;
digitizing said compressed, smoothed monopolar waveform, whereby a digitized waveform is formed;
storing said digitized waveform in a recirculating memory for transmission to a location remote from said acoustic wave; and freezing after a predetermined time period, information stored in said recirculating memory, said predetermined time period beginning when the amplitude of said monopolar waveform, after said launching of said acoustic wave, exceeds a predetermined threshold.

19. The method of claim 17 including, before digitizing said echo waveform, sensing the magnitude of said waveform, determining the gain required to adjust said magnitude to a desired range, amplifying said waveform by the determined gain and digitizing the determined gain value.

20. The method of claim 19 including multiplexing said digitized, determined gain value with said time interval and said stored waveform.

21. The method of claim 19 including, after amplifying said waveform by said determined gain, but before digitizing said waveform, logarithmically amplifying said waveform, whereby the dynamic amplitude range of different waveforms digitized is compressed.

22. The method of claim 19 including, before digitizing said waveform, reducing the bandwidth of said amplified waveform by passing it through an envelope detector.

23. The method of claim 22 including, before digitizing said waveform, further reducing the bandwidth of said amplified and envelope-detected waveform by rectifying said envelope-detected waveform and filtering it through a low-pass frequency filter.

24. The method of claim 17, including before digitizing said waveform, logarithmically amplifying said waveform, whereby the dynamic amplitude range of different waveforms is compressed.

25. The method of claim 17 including resetting said recirculating memory, after said multiplexing step, for continuously storing incoming waveform information.

26. A method for preprocessing and transmitting echo waveform information derived from the echo of an acoustic wave reflected from a target surface, comprising the steps of:
digitizing a full echo waveform produced in response to a launching of the acoustic wave, whereby digital waveform data is formed;
detecting amplitude of said full echo waveform;

continuously storing said digital waveform data in a recirculating memory for transmission to a location remote from said acoustic wave; and halting said continuously storing step after a predetermined time period beginning when the amplitude of said monopolar waveform, after said launching of said acoustic wave, exceeds a predetermined threshold, whereby a digital representation of said full echo waveform is frozen in said recirculating memory.

27. The method of claim 18 including, before said converting step, measuring the magnitude of said wave form, determining the gain required to adjust said magnitude to a desired range and amplifying said waveform by the determined gain.

28. The method of claim 18 including converting said echo waveform to a monopolar waveform by passing it through an envelope detector.

29. The method of claim 18 including smoothing said monopolar waveform by rectifying it and filtering said rectified waveform through a low-pass frequency filter.

30. The method of claim 1,3 including compressing the dynamic range of said smoothed, monopolar waveform by logarithmically amplifying it.