On-line analysis of hydro-abrasive erosion of pre-crackedmaterials by acoustic emission

A.W. Momber a,*,1, R.S. Mohan b, R. Kovacevic c

a WOMA Apparatebau GmbH, P.O. Box 141820, D-47208 Duisburg, Germanyb The University of Tulsa, Department of Mechanical Engineering, Tulsa, OK, USA

c Southern Methodist University, Department of Mechanical Engineering, Dallas, TX, USA

Abstract

Acoustic Emission (AE) sensing technique is used as a tool for on-line monitoring of hydro-abrasive erosion (HAE)

of pre-cracked multiphase materials. As reference materials, ®ve types of concrete materials were used for the exper-

imental study. Compression tests were performed to determine the mechanical properties and the failure behavior of

these materials. Erosion parameters, such as abrasive particle velocity, local exposure time, and abrasive mass ¯ow rate

were varied during the experiments and AE-signals were acquired. The trends exhibited by the time domain and fre-

quency domain AE-signals with change in process parameters and material properties were analyzed. The results

indicate that acoustic emission signal is capable of revealing the di�erent material removal mechanisms occurring in

pre-cracked multiphase materials when subjected to hydro-abrasive erosion. Visualization studies performed on the

erosion site provide more insight into the physics of the process and verify the observations made from the AE-signals.

Finally, it is concluded that due to its capability to quantify the amount of material removed, AERMS could be con-

sidered as a parameter for monitoring the material removal process. Ó 1999 Elsevier Science Ltd. All rights reserved.

1. Introduction

As a new manufacturing process, hydro-abra-sive erosion (HAE) has been very e�ective forcutting di�cult-to-machine materials. The toolthat is used for the material removal process is ahigh-speed abrasive-water jet. From the point ofview of jet generation, the technique could becategorized as injection jets or suspension jets.Injection jets are the most commonly used type for

practical applications. An injection jet is formedby accelerating small abrasive particles throughcontact with a high velocity plain water jet. Thevelocity of the plain water jet can be estimated byapplying the Bernoulli's law of pressure constancy,

vw � l

������2pqw

s: �1�

The mixing of abrasives, water and air takesplace in a mixing chamber, whereas the accelera-tion process occurs in a focus tube. The abrasiveparticles leave this focus at velocities of severalhundred meters per second. The abrasive particlevelocity can be approximated by assuming a sim-ple momentum balance in the mixing chamber.

Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

* Corresponding author. Tel.: +49-2065-304 380; fax: +49-

2065-304 200.1 Visiting Scholar at The Southern Methodist University,

Department of Mechanical Engineering, Dallas, TX, USA.

0167-8442/99/$ ± see front matter Ó 1999 Elsevier Science Ltd. All rights reserved.

PII: S 0 1 6 7 - 8 4 4 2 ( 9 8 ) 0 0 0 6 2 - 7

Neglecting the mass ¯ow rate of the air which issucked with the abrasive material, the abrasiveparticle velocity is,

vA � uvw

1� � _mA= _mw� : �2�

In Eqs. (1) and (2), the parameters l and u aremomentum transfer coe�cients of water andabrasive acceleration, respectively.

Hydro-abrasive erosion (HAE) is one possiblesolution for the manufacture of high qualityworkpieces of brittle multiphase materials, suchas ceramics, high-strength concrete, and com-posite materials, of special shapes and forms. Dueto their tendency for microcracking, these mate-rials are considered to be hard to machine byconventional tools. Despite this need, the numberof serious investigations on the machinability ofthis type of materials by HAE is comparativelylow.

Kim et al. [1] investigated ceramic stock re-moval using HAE. They conducted piercing andcutting experiments on alumina ceramics anddiscussed the in¯uence of several process param-eters on the piercing process. They demonstratedthat the HAE could be e�ectively used to ma-chine even high-strength ceramics. Neusen et al.[2] carried out cutting experiments on aluminum±silicon carbide and observed smearing of alumi-num on the cut surface as well as embedding offractured abrasive particles. They identi®ed wearby individual abrasive particles as one of themicro mechanisms of material removal. Later,Freist et al. [3] reported about the application ofHAE for three-dimensional machining (milling)of alumina ceramics. This work included the de-velopment of a kerf geometry model as well as alimited parameter study. Hamatani and Ramulu[4] made use of HAE for piercing of ceramics.They found random damage on the top surface ofthe pierced holes as well as non-linear hole tapers.They also detected a notable increase in the targettemperature and concluded that HAE-machiningof ceramics may not be totally free of thermale�ects.

Zeng and Kim [5] conducted SEM-observationsto evaluate the behavior of sintered aluminumoxide ceramics subjected to HAE under di�erent

jet impingement angles. They observed inter-granular cracking and plastic ¯ow as the twomajor material removal mechanisms. The inter-granular fracture dominates the material removalmode in the normal impingement, whereas forsmall angle impact, plastic ¯ow also has equalsigni®cance. In an advanced investigation [6], theydeveloped a material removal model as well as akerf cutting model for brittle materials processedby HAE. To verify their model, they carried outcutting experiments under di�erent process con-ditions, such as di�erent pump pressures, traverserates, abrasive mass ¯ow rates, and focus diame-ters. Nevertheless, it was found [7] that inter-granular fracture dominates the removal processonly in materials that are characterized by a low-strength matrix and ®ne inclusion grains. Formaterials with comparatively high matrix strengthand coarse inclusions, cracking occurs through thematrix as well as through the inclusion grains andthe failure is dominated by transgranular fracture.The quantitative model for cutting brittle materi-als by HAE developed by Zeng and Kim [6] is usedin this study for discussing the erosion behavior ofthe investigated materials. Ramulu et al. [8] ma-chined composite ceramics by HAE at shallowimpacting angles and observed erosion by micro-cutting in the matrix whereas the inclusions wereremoved by the shoveling action of the incomingjet.

In contrast to the kinetic process discussions,Kahlman et al. [9], who observed that the forma-tion of high localized temperatures (about 1280°C)on the surfaces of ceramics during the HAE-pro-cess, introduced thermal spalling failure of ceramicmaterials. They assumed that cooling by the water¯ow creates local stress ®elds. Based on SEM-photographs and wear volume measurements,these references de®ned the thermal shock resis-tance, RS, as the main resistance parameter againstHAE and de®ned a machining limit for ceramics atRS� 15 W/mm. Above this value the material re-moval rapidly decreases.

Hochheng and Cheng [10] reported about vol-ume removal studies on aluminum oxide and sili-con nitride ceramics. In their study, theyinvestigated the in¯uence of several process pa-rameters, such as pump pressure and abrasive

2 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

mass ¯ow rate, on material removal rate and sur-face quality. Nevertheless, no attempt was made todiscuss the presented experimental results in termsof basic material removal mechanisms.

Momber et al. [11] investigated the behavior ofrefractory ceramics machined by HAE. ThroughSEM-studies they found that the material removalmode changes from predominantly transgranularfracture in the upper wall region to a combinationof inter-granular fracture and matrix removal, inthe lower region of the wall. This observation wasexplained by the loss of the abrasive particle ki-netic energy during the erosion process as theo-retically treated in Refs. [12,13]. They found thatthe machinability of ceramic materials by HAE isin¯uenced by their capability to plastically deform.In a unique contribution, Momber and Kovacevic[7] could show that similarities exist in the behav-ior of pre-cracked materials during the conven-tional compression test and during the HAE-machining. For this reason, Momber and Kov-acevic [14] found surprisingly good agreementbetween the depth of cut in concrete materials andtheir mechanical properties, such as compressivestrength and the Young's modulus.

Despite these studies, the knowledge about thebehavior of pre-cracked multiphase materialssubjected to HAE is just preliminary. One majorreason could be that `o�-line' methods, such asSEM-studies, geometry measurements on the ero-sion site, and surface topography measurementsare used for the analysis and observation of theprocess. These methods require an interruption ofthe material removal process to obtain suitableinformation. These `post-process' results are thenlinked to the real process. Therefore, the applica-tion of `on-line' methods, which allow a directreal-time observation of the microscopic andmacroscopic features of the material removalmechanism, may provide additional informationabout the HAE-process. Experimental on-linemethods which could be applied to HAE arephotoelasticity [15], jet and workpiece reactionforce measurements [16], and AE sensing tech-nique.

AE-technique was used spontaneously in the®eld of water jet cutting [17], solid particle erosion[18], and abrasive water jet cutting [19]. The ®rst

systematic investigation of AE-signals generatedduring the HAE-machining process was performedin Refs. [20,21]. Using advanced signal processingtechniques, these authors detected functional re-lationships between the AE-signals acquired dur-ing the cutting of metals by HAE on one hand,and the erosion depth and the energy dissipated inthe specimen on the other. They also found thatAE-technique may be a useful tool for HAE ma-terial removal process monitoring and control formetals.

Acoustic emission, in the sense of this paper,refers to elastic waves produced by microscopicdeformations occurring in materials, as they arestressed. It contains part of the elastic energy re-leased during deformation. Sometimes called stresswave emissions, acoustic emissions are associatedwith dislocation movements, crack growth, defor-mations and other mechanisms. These are sourceswhich are related to the eroded materials. Othertypes of sources, such as pressure pulses, turbu-lence, and cavitation, depend on the ¯uid dynamicsof the process.

2. Experimental setup and procedure

2.1. Experimental setup

The experimental setup consists of an AWJcutting system, AE-sensor, pre-ampli®er, A/Dconverter, AE monitoring system, PC/AT withsuitable software and workpiece. The AWJ cuttingsystem used for conducting the experiment consistsof a high-pressure intensi®er pump, AWJ cuttinghead, abrasive metering and delivery system,abrasive hopper with garnet as abrasive, catchertank, and X±Y±Z positioning system controlled bya CNC controller. A schematic of the experimentalsetup is shown in Fig. 1.

The generated AE-signals were detected andprocessed by Model AET 5500 Acoustic EmissionMonitoring System which consists of AET 5500mainframe (signal processing unit), graphics ter-minal (interface, data storage and display) and theaccessories (sensors, pre-ampli®ers etc.). When anacoustic emission caused by an induced stressoccurs in a test specimen, the sensors (resonant

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 3

frequency 2 MHz), ®xed on top of the specimen,convert this acoustic emission wave into a voltagesignal which is ampli®ed by the pre-ampli®er andsent to the mainframe (16-bit microprocessor) forpost-processing. The distance between cut/kerfand sensor was between 5 and 90 mm. The sensorwas ®xed on the samples with a water resistantepoxy-adhesive. The raw signal from the AE-monitoring system was acquired at 1 MHz sam-pling frequency (with a gain of 5.0) using a suitablePC-based data acquisition system for spectrumanalysis. Each data set consists of 1024 discretetime domain points. Five data sets of the FastFourier Transform (FFT) of the time domainsignals acquired under the same experimentalconditions were averaged and the representativesignal was obtained for future analysis.

The HAE-parameters, namely abrasive particlevelocity, local exposure time, and abrasive mass¯ow rate were varied and linear cuts were made onthe workpieces. The abrasive particle velocity wascalculated by Eqs. (1) and (2). The momentumtransfer parameters in these equations were takenfrom Ref. [22]. The local exposure time was esti-mated by

tE � dF

vT

: �3�

The abrasive mass ¯ow rate was varied by amicro-screw in a special metering device. Aftercutting, the dimensions of the generated cuts, such

as depth of cut, cut width, and cut length, weremeasured as average values of ®ve measurements.The process parameters of these experiments areshown in Table 1. The 14 experimental conditionsin Table 1 were repeated for each workpiece ma-terial.

For all 70 parameter combinations, a theoreti-cal depth of cut has been calculated based on themodel in Ref. [6]. These values are then comparedwith the experimentally estimated depth valuesand discussed in terms of deviation between themodel and experiment.

2.2. Test materials

For the investigation, ®ve di�erent concretemixtures (#1±#5) were created and six specimensof each mixture were manufactured. The me-chanical properties of these materials were in-¯uenced by changing the water±cement ratio andthe aggregate ®neness (Table 2). The strength ofthe cement matrix as well as the average inclu-sion diameter were increased in the order from#1±#5. After mixing and placing, these mixtureswere cured and hardened for 28 days underwater. After hardening, three cylindrical speci-mens of each mixture were tested to estimate themechanical properties, such as compressivestrength, Young's modulus, absorbed fractureenergy, and crack velocity. The dimensions ofthe cylindrical specimen used for these tests were30.8 cm in length and 15.24 cm in diameter. Thecompressive strength was measured according tothe ASME Standard C 39, and the Young'smodulus was estimated as the static chord mod-ulus according to the ASTM Standard C 469.

The energy absorbed during the compressionfracture test was approximated as follows. Thestress±strain curves of the materials were deter-mined by loading and unloading the strain-gaugeequipped specimen 11-times and then expressed bya parabolic regression. The area enveloped by thestress±strain curve and the ultimate strain was thencalculated by integrating,

EAB �Zeult

0

A � �eÿ eult�2 � rult

h ide: �4�

Fig. 1. Experimental setup.

4 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

Table 1

Experimental process parameters

Part (a) Constant parameters

Abrasive waterjet cutting

Abrasive material Garnet

Abrasive mesh size #36 (dP� 425 lm)

Abrasive particle shape angular (random)

Ori®ce material Sapphire

Ori®ce diameter 0.457 mm

Mixing nozzle length 88.9 mm

Mixing nozzle diameter 1.27 mm

Stand-o� distance 6.00 mm

Method of feed suction

Abrasive condition dry

Jet impact angle 90 degrees

Workpiece details

Material Mixtures # 1±# 5

Material thickness 310 mm

Length of cut 50±80 mm

Part (b) Variable parameters

Erosion conditions

Cut No. Pressure [MPa] Traverse rate [mm/s] Abrasive mass ¯ow rate [g/s]

1 100 4.0 7.41

2 150 4.0 7.41

3 200 4.0 7.41

4 250 4.0 7.41

5 300 4.0 7.41

6 350 4.0 7.41

7 200 2.0 7.41

8 200 6.0 7.41

9 200 8.0 7.41

10 200 12.0 7.41

11 200 4.0 6.12

12 200 4.0 10.58

13 200 4.0 14.32

14 200 4.0 19.00

Table 2

Composition and mechanical properties of the specimens

Concrete

mixture

Aggregate a/

cement b

Sand c [%] Water/cement YM [GPa] rM [MPa] EAB [MJ/m3] qM [kg/m3]

#1 2.24 80 0.85 10.7 4.0 1.5 1962

#2 2.51 70 0.71 24.1 12.5 7.9 2107

#3 2.13 60 0.45 33.0 27.1 27.0 2107

#4 2.25 50 0.38 34.3 34.2 40.9 2135

#5 2.26 40 0.32 42.3 47.6 72.8 2343

a coarse limestone aggregates (maximum grain size 19 mm) and river sand (maximum grain size 4.7 mm).b Portland Cement, Type I.c river sand (maximum grain size 4.7 mm).

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 5

In order to obtain information about the me-chanical behavior of the di�erent type of materialsused in this study, the standard compression testwas combined with an analysis of the materialfragments which remained after the specimenfailure. These fragments were collected and clas-si®ed into two groups as suggested by Momber[23]. The ®rst group contained the main debrisfrom the top and the bottom of the specimenwhich are usually cone-shaped (see Fig. 2),whereas the second group contained the ®ne-grained material from the periphery of the speci-men. These grain collections were investigated bysieve analysis.

3. Experimental results

3.1. Compression tests

The mechanical properties obtained by thestandard compression tests are listed in Table 2. Itcan be seen that the mechanical properties increasewith the number of the material mixture. Theconcrete #1 exhibits comparatively low values andconcrete #5 high values.

The results obtained from the analysis of thefragments of the cylinder compression tests arediscussed in detail in Ref. [23]. It was foundthat the concrete mixtures with the low me-chanical properties (concrete #1±#3) are char-acterized by a cone-type failure with a highdegree of intergranular fracture on the conesurfaces. This failure mode is illustrated inFig. 2(a) showing two cones generated duringthe testing of concrete #1. In contrast, thehigh-strength concrete mixtures (concrete #4and #5) showed a splitting mode failure,mainly by transgranular fracture. This is illus-trated in Fig. 2(b), which shows the fragmentsof the cylinder specimen made from concrete#5. It can be seen that the number of frag-ments is higher. Also, the broken aggregategrains can be clearly distinguished. As the lat-ter failure mode requires more energy due tothe high speci®c surface energies of the frac-tured aggregate grains, these observations arein good agreement with the measured absorbedfracture energies (Table 2). It was also foundthat the ®ne grained material generated duringthe compression test has characteristic features.Most signi®cantly, the average particle diameterof the ®ne grained material increases with anincrease in the mechanical properties of thematerial. Also, with increasing materialstrength, the shape of the grains changed fromalmost spherical (concrete #1) to sharp edged(concrete #5), indicating di�erent failure be-haviors. The results obtained from the com-pression test suggest that with increasingstrength properties, the failure mode changesfrom a `crumble' intergranular fracture for low-strength materials to a `spalling' transgranularfracture for high-strength materials.

Fig. 2. Fragments generated during the compression tests of the

concrete specimens: (a) Fragments (conical) generated during

the compression test of concrete #1; (b) fragments (splitting)

generated during the compression test of concrete #5 (several

broken inclusion grains are shown).

6 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

3.2. Cutting experiments

The in¯uence of the investigated process pa-rameters on the erosion depth in the materials isillustrated in Fig. 3. Fig. 3(a) shows the in¯uenceof the abrasive particle velocity. Here, two inter-esting features could be noticed. First, the rela-tionship is highly non-linear with a decreasing

trend at higher particle velocities, indicating thatthe material removal process is less e�cient athigher velocity levels. Second, a critical abrasiveparticle velocity, vTHR, could be noticed at the in-tersection of the velocity axis and the depth func-tion which describes the minimum impact velocityrequired to erode the material. A similar parame-ter is also known from dry solid particle erosion

Fig. 3. Process parameter in¯uence on the depth of cut in the investigated materials: (a) Abrasive particle velocity; (b) local exposure

time; (c) abrasive mass ¯ow rate.

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 7

[24] and is considered as the material `thresholdvelocity'. In the present case, the values are be-tween 200 and 240 m/s. It can be seen that thisparameter increases with an increase in the mate-rial strength properties. But interestingly, even iftwo materials are having almost identical thresh-old velocities (concrete #1 and #2), they will be-have quite di�erently once the erosion process isintroduced and as they are attacked by highervelocities.

The in¯uence of the local exposure time onthe erosion depth is shown in Fig. 3(b). Similarto Fig. 3(a), a threshold value could be noticed.This `incubation time' illuminates the fact that acertain number of impacting particles may berequired to introduce the erosion process. Thematerial removal process starts once this value isexceeded. The incubation periods are located ina narrow range between 0.025 and 0.05 s. Therelationships in Fig. 3(b) are non-linear and theprogress drops at long exposure times. As theresults show, the erosion process is very sensitivein the range of short exposure times. In thisrange (tE < 0.1 s) just small variations in theexposure time yield signi®cant changes in thedepth of cut. Interestingly, the initial slopes ofthe functions depend on the material type. It isvery high for the low-strength concrete #1 anddecreases as the concrete strength increases.Obviously, low-strength materials are less sensi-tive to damping e�ects in this stage. Later, theslopes of the di�erent materials are almost equal.For very long exposure times, it can be assumedthat the cutting process has some similaritieswith the piercing process. In particular, dampingprocesses may play a signi®cant role in deter-mining the material removal process. Therefore,it is assumed by several authors that the depthof cut in a material has a constant value for longexposure times. It can be assumed that thefunction may approximate this maximum possi-ble depth. Again, one feature could be observed:although the incubation times for all mixturesare very close together, the materials show avery di�erent erosion resistance. Obviously, thedi�erent conditions for damping, friction andother energy consuming processes are not es-tablished during the incubation period.

Fig. 3(c) illustrates the in¯uence of the abrasivemass ¯ow rate on the cutting process. As the ma-terials could be eroded by plain water jets thefunctions do not start at zero. For low abrasivemass ¯ow rates (mA < 6 g/s) the erosion depthincreases rapidly with an increase in the abrasivemass ¯ow rate. For the investigated ¯ow raterange, the in¯uence of the abrasive mass ¯ow rateon the erosion depth is weak. The reason may bethat for the given process condition, includingparticle mixing and material removal, a criticalabrasive mass ¯ow rate exists. Beyond this critical¯ow rate, the number of abrasives is too large fore�cient acceleration and cutting. This leads to lossof e�ciency in the mixing chamber as well as onthe erosion site due to friction, damping, particlecollision, and particle fragmentation. Therefore, afurther increase in the abrasive ¯ow rate will notsigni®cantly increase the depth of cut. The fre-quency of impacting abrasive particles has beencalculated by an approximation method; it is be-tween 180 and 460 kHz. Interestingly, concrete #1does not follow this trend consequently. As theabrasive mass ¯ow rate exceeds the value of 14 g/s(350 kHz), an additional increase in the erosiondepth can be noticed. This could be due to theabsorption of particle impact energy by the ma-terial through several mechanisms, such as mic-rocracking, crack branching, crack arresting, etc.Thus, high-frequency loading (i.e. a high numberof impacting particles) will be required to over-come these dissipation mechanisms. A very similarphenomenon was observed in Ref. [11] for HAE ofrefractory ceramics.

Fig. 4 shows the experimental results plottedagainst the depth values calculated with the brittleerosion model [6]. A reasonable agreement can benoticed for depths up to 70 mm. As this value isexceeded, the correlation drops drastically. A fur-ther striking feature is that concrete #1 does not ®tthe model at all over the entire parameter range.Concrete #2 behaves similarly but is showing goodagreement for shallow cuts.

3.3. Acoustic emission measurements

Typical time domain AE-signals obtained forthe mixtures #1 and #5 for identical experimental

8 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

conditions are shown in Fig. 5. It can be notedthat the AE-signal is predominantly of a continu-ous type for concrete #1 which is typical for asteady material removal mechanism; whereas, inthe case of concrete #5, the AE-signal is primarilyburst emission type which indicates the presence ofunsteadily appearing energy sources. It can also benoticed that the signal peak-to-peak-amplitude isalmost twice in the case of concrete #5 comparedto concrete #1. The FFT of the time domain sig-nals obtained for various process parameters areplotted in Figs. 6±8. The energy of the signal in-dicated by AERMS and the FFT-peak frequency aswell as amplitude for various process parametersand di�erent workpiece materials are given inTable 3.

The frequency domain AE-signals with changein the abrasive particle velocity for three concretematerials are plotted in Fig. 6. It can be seen thatthe amplitude of the FFT-signal increases withincreasing particle velocity showing a trend similarto that of the erosion depth (Fig. 3(a)). There is agradual reduction in the amplitude of the signal(acquired under the same conditions) from con-crete #1 to #4, and an increase with concrete #5.In fact, the amplitude of the AE-signal is higher inconcrete #5 than concrete #1. Concrete #5 is thus

more sensitive to impact velocity variations com-pared to concrete #1, which could be due to dif-ferent material removal modes. Given the samematerial removal modes (of concrete #1±#4), in-crease in the magnitude of the mechanical prop-erties (Young's modulus, compressive strength,and absorbed fracture energy) causes decrease inthe erosion depth and hence quantity of materialremoved. This leads to a drop in AE-signal am-plitude.

The frequency domain AE-signals with changein local exposure time for three concrete materialsare plotted in Fig. 7. It can be noted that the AE-signal generated in concrete #1 is much moresensitive to time variations than concrete #5. Thismatches the observations in Fig. 3(b) and could beattributed again to di�erence in material removalmodes. There is an exponential increase in theamplitude of the AE-signal with increase in expo-sure time for concrete #1 which could be expectedfrom the trend shown by the depth (Fig. 3(b)).

Fig. 4. Accuracy of the brittle erosion model for the investi-

gated materials.

Fig. 5. Time domain AE-signals for two di�erent concretes and

photographs of wall structures showing di�erent material re-

moval modes. Experimental conditions: vA� 445 m/s, tE� 0.32

s, _mA � 7:41 g=s�: (a) Concrete #1, intergranular fracture; (b)

concrete #5, transgranular spalling.

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 9

Tab

le3

AE

RM

Sa

nd

FF

T-P

eak

of

the

AE

-sig

na

lsfo

rv

ari

ou

sp

roce

ssp

ara

met

ers

an

dm

ate

rials

Cu

tN

o.

Co

ncr

ete

#1

Co

ncr

ete

#2

Co

ncr

ete

#3

Co

ncr

ete

#4

Co

ncr

ete

#5

AE

RM

S(V

)F

FT

Pea

kA

ER

MS

(V)

FF

TP

eak

AE

RM

S(V

)F

FT

Pea

kA

ER

MS

(V)

FF

TP

eak

AE

RM

S(V

)F

FT

Pea

k

Fre

q.

(kH

z)

Am

pl.

(V)

Fre

q.

(kH

z)

Am

pl.

(V)

Fre

q.

(kH

z)

Am

pl.

(V)

Fre

q.

(kH

z)

Am

pl.

(V)

Fre

q.

(kH

z)

Am

pl.

(V)

10

.43

51

14

2.6

0.0

959

0.4

275

109.4

0.0

864

0.1

870

129.9

0.0

490

0.3

012

274.4

0.0

703

0.6

773

127.0

0.1

495

20

.49

31

15

0.4

0.1

115

0.7

007

131.8

0.1

585

0.3

416

105.5

0.0

855

0.3

541

176.8

0.1

094

0.8

153

149.4

0.1

682

30

.66

59

13

8.7

0.1

531

0.7

178

139.7

0.1

441

0.5

177

142.6

0.1

523

0.3

238

159.2

0.0

656

1.0

360

142.6

0.1

969

40

.89

01

14

7.5

0.2

240

0.7

415

134.8

0.1

521

0.5

347

144.5

0.1

328

0.3

618

167.0

0.0

734

1.2

953

147.5

0.3

771

51

.03

52

13

8.7

0.2

250

0.8

747

105.5

0.2

043

0.5

566

140.6

0.1

367

0.4

182

165.0

0.0

641

1.3

540

138.7

0.4

609

61

.19

30

11

7.2

0.2

951

0.9

527

119.1

0.2

755

0.5

894

127.0

0.2

330

0.4

318

152.3

0.1

031

1.4

900

161.1

0.5

031

71

.93

83

15

6.3

0.3

805

1.0

750

120.1

0.1

192

1.1

500

158.2

0.2

734

0.1

828

194.3

0.0

347

0.6

027

169.0

0.1

414

80

.44

82

15

9.2

0.1

004

0.4

913

160.2

0.1

356

0.6

668

149.4

0.1

400

0.2

724

208.0

0.0

470

0.4

708

131.8

0.0

891

9N

AN

AN

AN

AN

AN

A0.5

447

129.9

0.0

864

0.1

666

142.6

0.0

370

0.3

484

166.0

0.0

566

10

0.1

514

11

5.2

0.0

371

0.4

205

126.0

0.0

951

0.4

954

162.1

0.1

242

NA

NA

NA

0.2

300

146.5

0.0

406

11

0.6

719

13

7.7

0.1

734

0.2

714

127.9

0.0

631

0.5

606

137.7

0.1

095

0.1

848

180.7

0.0

312

0.6

105

123.1

0.1

367

12

0.5

997

16

4.6

0.1

354

0.6

132

102.5

0.1

372

0.7

152

130.9

0.1

578

0.1

364

193.4

0.0

250

0.9

135

143.6

0.1

982

13

0.5

748

11

3.3

0.1

203

1.9

03

125.0

0.2

724

0.6

913

100.6

0.1

543

0.1

885

132.8

0.0

352

0.8

565

109.4

0.1

941

14

0.5

879

10

8.4

0.1

354

0.4

438

128.9

0.0

775

0.7

812

117.2

0.1

697

0.4

660

213.9

0.0

722

1.1

292

148.4

0.2

520

10 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

Interestingly though, for the mixtures #2±#5 theincrease in exposure time was leading to an in-crease in the AE-amplitude till an exposure time of

tE� 0.32 s and it reduced for the exposure time oftE� 0.635 s. Visualization studies performed onthe kerf geometries o�er an explanation to thisdi�erence in trend. The results of these studies arediscussed in Section 4.1. It may also be noted thatthere is a gradual reduction in the overall ampli-tude of the AE-signal from the mixtures #1 to #4,and then an increase for concrete #5. This trend is

Fig. 6. Frequency domain AE-signals with change in abrasive

particle velocity: (a) Concrete #1; (b) concrete #4; (c) concrete

#5.

Fig. 7. Frequency domain AE-signals with change in local ex-

posure time: (a) Concrete #1; (b) concrete #4; (c) concrete #5.

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 11

similar to that observed in the case of abrasiveparticle velocity variation.

The frequency domain AE-signals with changein abrasive mass ¯ow rate for three concrete ma-

terials are plotted in Fig. 8. The overall amplitudeof the frequency domain AE-signal drops gradu-ally from concrete #1 to #4, and then increaseswith concrete #5. There is no appreciable trendobserved in the amplitude of the AE-signal (simi-lar to that of erosion depth) with increase inabrasive mass ¯ow rate for any of the concretematerials. The causes for this behavior are dis-cussed in Section 4.1.

3.4. Optical microscopy investigations

Results of the optical microscopy images ta-ken from the upper parts of the erosion sites areshown in Fig. 9. Fig. 9(a) illustrates the structureof concrete #1 after it is eroded by HAE. It isclearly seen that the material is removed by in-ter-granular failure. Any cracks that have beenactivated, run through the weak interface be-tween cement-matrix and aggregates. Severalcompletely exposed grains are still visible. Thelocations where the aggregate particles are re-moved from, can also be seen. The same featurescan be observed in Fig. 9(b) illuminating thevery similar failure behavior of concrete #2. Thesituation looks di�erent in Fig. 9(c) that showspart of the erosion site of concrete #4. Here, amixed mode between inter-granular failure andtransgranular fracture can be detected. Whereasthe left part of the photograph shows locationswhere inclusions have been exposed and removedcompletely, respectively, the right part exhibitsinclusions that are partially broken. The trans-granular fracture runs through the aggregategrains. The same observation can be made inFig. 9(d) showing on its lower left part a verytypical example of a large aggregate grain that isfractured. Additionally, microcracks are visiblein the upper center and lower right area of thephotograph, which could be a sign of crack ar-resting. As the crack went through an inclusionwith high speci®c surface energy it might havegot exhausted and could not run further. But incontrast to the images discussed previously, noundamaged, embedded or completely removedinclusion grains are visible in Fig. 9(d), indicat-ing that the material failure is predominantlydue to spalling fracture.

Fig. 8. Frequency domain AE-signals with change in abrasive

mass ¯ow rate: (a) Concrete #1; (b) concrete #4; (c) concrete

#5.

12 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

4. Discussion

4.1. In¯uence of process parameters on the AEsignals

AE-signal acquired from the workpiece can beconsidered as a quanti®ed measure of the kineticenergy dissipated on the erosion site with somedamping e�ect (caused by turbulence of the highspeed slurry), as suggested in Ref. [21]. The areaenclosed by the FFT-curve or AERMS could be usedfor this purpose. The part of the kinetic energydissipated in a cavity with the depth h is given [13] by

EDIS � v�h� � _mA � _mw�2

�vA ÿ vTHR�2tE: �5�With increase in the abrasive particle velocity morekinetic energy is exposed to the workpiece mate-

rial. Thus the higher amplitude of the AE-signal athigher velocities is an indication of the presence ofa more powerful material removal process andlarger amount of energy utilized for the erosion.

An increase in the local exposure time causesmore energy to be dissipated at the cut. This leadsto deeper cuts, more material removal, and hencehigher amplitude for the AE-signal. Interestingly,the AE-signals for the mixtures #2±#5 show aslight decrease in amplitude at large exposuretimes. This may be an indicator of the onset ofdamping processes at the bottom of the cavity asdiscussed earlier and shown in Fig. 3(b). Althoughmore energy is dissipated during the material re-moval process because of the longer exposure, theenergy of the acquired signals seems to be reduced.As shown in Ref. [20] for AWJ cutting and in Ref.

Fig. 9. Optical microscopy images from erosion sites: (a) Concrete #1; (b) concrete #2; (c) concrete #4; (d) concrete #5.

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 13

[18] for solid particle erosion, a ¯uid ®lm on thesurface of eroded material can dampen AE-sig-nals. This drop in the AE-signal was not observedon concrete #1 (see Fig. 7(a)) indicating that thismaterial mixture was not sensitive to damping ef-fects. In fact, large cavities were observed in thekerfs of concrete #1 caused by intense washing outof the hardened cement matrix. This e�ect isshown in Fig. 10(a). These cavities created addi-tional space for the abrasive±water mixture to es-cape and probably prevented signi®cant dampinge�ects. In contrast, as shown in Fig. 10(b), the cutsin the high-strength materials do not show theselarge cavities. Therefore, the kerf walls of thesematerials may restrict the high-speed slurry in thekerf which creates the damping e�ects. These re-lations are in agreement with the results from theerosion depth measurements (Fig. 3(b)) as it wasfound that the initial progress of the time-functionis extraordinarily high for concrete #1 indicating areduced damping e�ect.

The in¯uence of the abrasive mass ¯ow rate onthe AE-signal characteristics is not very signi®-cant. This observation is in agreement withFig. 3(c) which shows that the abrasive mass ¯owrate has just a minor in¯uence on the erosiondepth in the investigated range. As the abrasivemass ¯ow rate is much lower than the water mass¯ow rate (in this study: _mA= _mw � 0:1), Eq. (5) in-dicates that variations in the abrasive mass ¯owrate will not change the dissipated energy signi®-

cantly. The increase in the integral abrasive parti-cle mass and thus of the higher impact frequency,due to the higher abrasive mass ¯ow will be com-pensated by the decrease in the mean particle ve-locity (Eq. (2)). Therefore, the AE-signals in Fig. 8are an indicative of real conditions.

4.2. In¯uence of material properties on the AE-signals

Figs. 5±10 and Table 3 contain information onthe in¯uence of the material properties on the de-tected AE-signals. Based on the assumption that theenergy of the AE-signals are related to the energydissipated during the material removal process [21],it may be ingenious to discuss the in¯uence of theenergy absorbed during the compression test on thestructure of the measured AE-signals.

The relation between the AERMS and the energyabsorbed during the compression test is shown inFig. 11. The AERMS corresponding to di�erentprocess parameters are indicated by the points.Two trends could be noticed from these ®gure.First, the AERMS generally reduce with increase inthe mechanical properties of the materials. Second,the magnitude of AERMS for concrete #5 is com-paratively high.

Fig. 10. Visualization of kerf geometries indicating di�erent

material removal mechanisms: (a) Concrete #1, large cavities in

the lower range; (b) Concrete #5, comparatively straight kerfs.

Fig. 11. Relation between absorbed fracture energy and

AERMS.

14 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

The ®rst trend could be explained by the obser-vations made in the previous chapter. There, it wasfound that the intensity of the AE-signals generallydecreases with a drop in the energy dissipated dur-ing the material removal process. With increasingabsorbed fracture energy, less material is removedfrom the workpiece. The material removal mecha-nisms at the erosion site can be investigated throughAE-visualization studies of the specimens. TheHAE signals from the wall structure for two con-crete mixtures are shown in Fig. 5. The structuresfor concrete #2±#4 are similar to that of concrete#1. Fig. 5(a) clearly shows that the material inconcrete #1 is removed by HAE due to intergran-ular erosion mechanism which is characterized byhigh degree of undamaged aggregate grains. Thefailure path is along the interface between aggregategrain and hardened cement matrix. The matrix isremoved separately which could be due to the actionof the high velocity water ¯ow. These assumptionsare also supported by observations made on thecompressive test specimens (see Fig. 2(a)) and themacro-geometry of the erosion site (see Fig. 10(a))as described earlier as well as by microscopic in-vestigations on the specimens (Fig. 9(a)). The ma-terial failure is almost 100% inter-granular forconcrete #1. Although the amount of intergranularerosion decreases with the absorbed fracture energyof the eroded materials (see Fig. 9), it dominates thematerial removal mechanism at least for the mix-tures #1±#4. Assuming a fairly identical uniformmaterial removal mode for these concrete mixtures,less energy is dissipated during the erosion processin the samples with higher strength properties whichleads to a drop in the AERMS and also in the FFT-peak. In contrast, the AE-structure shown inFig. 5(b) clearly illustrates that material failure inmixture #5 is primarily due to trans-granular frac-ture; there is an inclusion fracture of almost 100%.This is most probably due to the homogenization ofthe structure. Since the cement matrix is compara-tively hard and the inclusion grains are large, thestructure in general is more homogeneous. Thishomogenization e�ect becomes active as the mate-rial is subjected to high stress rates. During thematerial removal process, cracks run preferablythrough the aggregate grains. Although this mech-anism may shorten the path of the cracks as there is

no crack de¯ection or crack branching, the aggre-gate grains absorb a high amount of energy forfracture. Typically, the speci®c surface energy of theaggregate material is one order of magnitude higherthan the speci®c surface energy of the interface be-tween aggregate and hardened cement paste [25]. Asthe di�erences in the erosion depths between thespecimens #4 and #5 are not very large, the e�ects ofaccelerated energy dissipation due to the aggregatefracture may explain the increase in the AE-signalintensity in the concrete material #5. The type of thetime domain AE-signal also provides a clue to thepresence of di�erent material removal mechanisms.Whereas the AE-signal for the concrete #1(Fig. 5(a)) shows a fairly continuous behavior, theAE-signal of the concrete #5 (Fig. 5(b)) is charac-terized by burst emissions, illustrating the suddenenergy release, most probably due to the spallingfracture of the structural elements.

Thus, the AE-signals seem to be sensitive tochanges in the material removal regime duringHAE. Probably, the regions of minimum AE-sig-nal intensity in Fig. 11 indicate a transition rangebetween a granular erosive material removal modeand a spalling fracture mode.

The di�erent failure behavior of concrete #1 asdiscovered by the AE-studies is further illustratedin Fig. 4. The large deviations of the measureddepths from the model line suggest that themechanism of crack-network generation, that issuggested in Ref. [6], as a result of stress wavesinduced by impacting abrasive particles, is notvalid for concrete #1. There is strong evidence tosuggest that the erosion of this material type issupported by action of high-speed water ¯ow inthe eroded kerf. This ¯ow will wash the very weakmatrix away and the undamaged inclusion grainswill then be separated. As this additional erosivemechanism is not considered in the model, thecalculation procedure for concrete #1 fails.

5. Summary and conclusions

Acoustic emission measurements on pre-cracked multiphase materials indicate that AE-sensing technique is a promising tool to identifyon-line the di�erent material removal mechanisms

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 15

occurring during HAE. Material removal domi-nated by intergranular fracture is typically char-acterized by a continuous type of AE-signal;whereas material removal dominated by trans-granular fracture is characterized by burst emis-sions illustrating the sudden energy release duringinclusion fracture.

The results obtained from the compression testsindicate that with increase in mechanical strengthproperties of pre-cracked multiphase materials, thefailure mode changes from `crumble' intergranularfracture for low strength materials to a `spalling'transgranular fracture for high-strength materials.This is also true for HAE. The AE-signal is ca-pable of identifying this trend very e�ectively.

Any change in process parameter which in-creases the erosion depth and quantity of removedmaterial, causes the energy dissipated in workpiecematerial to increase. AE-signal is capable of de-tecting this change in the energy dissipated in thematerials and hence AERMS can be used as aquantitative measure of this dissipated energy.However, at large exposure times the dampinge�ect of the erosive slurry on the erosion reducesthe AE-signal amplitude in brittle materials whichare characterized by spalling trans-granular frac-ture.

Acknowledgements

The authors are thankful to the Alexander vonHumboldt Foundation, Bonn, Germany, for the®nancial support.

References

[1] T.J. Kim, G. Sylvia, L. Posner, Piercing and cutting of

ceramics by abrasive waterjet, PED-vol. 17, ASME, New

York, 1985, pp. 19±24.

[2] K.F. Neusen, P.K. Rohatgi, D. Alberts, Abrasive waterjet

cutting of metal matrix composites, in: Proceedings of the

Fourth US Water Jet Conference, ASME, New York,

1987, pp. 175±182.

[3] B. Freist, H. Haferkamp, A. Laurinat, H. Louis, Abrasive

jet machining of ceramic products, in: Proceedings of the

Fifth American Water Jet Conference, Water Jet Technol-

ogy Association, St. Louis, 1989, pp. 191±204.

[4] G. Hamatani, M. Ramulu, Machinability of high temper-

ature composites by abrasive waterjet, Trans. ASME J.

Engng. Mat. Technol. 112 (1990) 381±386.

[5] J. Zeng, T.J. Kim, Material removal of polycrystalline

ceramics by a high-pressure abrasive water jet ± a SEM

study, Int. J. Water Jet Technol. 1 (1991) 65±71.

[6] J. Zeng, T.J. Kim, Development of an abrasive waterjet

kerf cutting model for brittle materials, Jet Cutting

Technology, Kluwer Academic Publication, Dordrecht,

1992, pp. 483±501.

[7] A.W. Momber, R. Kovacevic, Principles of Abrasive

Water Jet Machining, Springer, London, 1998.

[8] M. Ramulu, S.P. Raju, H. Inoue, J. Zeng, Hydro-abrasive

erosion characteristics of 30vol.% SiCP 6061-T6 Al com-

posite at shallow impact angles, Wear 166 (1993) 55±63.

[9] L. Kahlman, S. Karlsson, R. Carlsson, C.G. Nilsson, Wear

and machining of engineering ceramics by abrasive water-

jets, Amer. Ceram. Soc. Bull. 72 (1993) 93±98.

[10] H. Hochheng, K.R. Cheng, Material removal analysis in

abrasive waterjet cutting of ceramic plates, J. Mat. Proc.

Technol. 40 (1994) 287±304.

Nomenclature

A regression parameterAERMS RMS of the acoustic emission signaldF focus diameterEAB energy absorbed during compressive test-

ingEDIS energy dissipated during erosion processh cutting (erosion) depthmA abrasive mass ¯ow ratemw water mass ¯ow ratep pump pressureRs thermal shock resistancetE local exposure timetTHR incubation timevA abrasive particle velocityvT traverse ratevTHR threshold abrasive particle velocityvW water jet velocityYM Young's modulus

e straineult ultimate strainv(h) energy dissipation function/ momentum transfer coe�cientl momentum transfer coe�cientqM target material densityqw water densityrM compressive strengthrult ultimate stress

16 A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17

[11] A.W. Momber, I. Eusch, R. Kovacevic, Machining

refractory ceramics with abrasive water jets, J. Mat. Sci.

31 (1996) 6485±6493.

[12] S.P. Raju, M. Ramulu, Predicting hydro-abrasive erosion

wear during abrasive water jet cutting, in: Manufact.

Science and Engineering, vol. 1, ASME, New York, 1994,

pp. 339±351.

[13] A.W. Momber, R. Kovacevic, Quanti®cation of energy

absorption capability in abrasive water jet machining, J.

Engng. Manufact. 209 (1995) 491±498.

[14] A. Momber, R. Kovacevic, Test parameters analysis in

abrasive water jet cutting of rocklike materials, Int. J. Rock

Mech. Min. Sci. 34 (1997) 17±25.

[15] M. Ramulu, Dynamic photoelastic investigation in the

mechanics of waterjet and abrasive waterjet machining,

Optics and Lasers in Eng. 19 (1993) 43±65.

[16] R. Kovacevic, R. Mohan, Y.M. Zhang, Cutting force

dynamics as a tool for surface pro®le monitoring in AWJ,

Trans. ASME J. Eng. Ind. 117 (1995) 340±350.

[17] T. Okamura, R. Oshima, K. Satoh, J. Sato, Erosion

intensity of submerged cavitating jets, in: Proceedings of

the 9th International Symposium on Jet Cutting Technol-

ogy, BHRA Fluid Engng., Cran®eld, 1988, pp. 33±48.

[18] W. H�ubner, E. Leitel, A. Rolle, J. Lehmann, G. B�ulow,

Schallemissionsanalyse erfasst Feststo�teilchen-Einschl�age,

Materialpr�ufung 34 (1992) 13±16.

[19] R.A. Tikhomirov, V.F. Babanin, E.N. Pethukov, I.D.

Starikov, V.A. Kovalev, Gidrorezanje Sudostroitel'nikh

Materialov, Sudostroenie, Leningrad, 1987.

[20] R. Mohan, A.W. Momber, R. Kovacevic, On-line moni-

toring of depth of AWJ penetration using acoustic emis-

sion technique, Jet Cutting Technol., Mech. Engng. Publ.

Ltd., London, 1994, pp. 649±664.

[21] R. Mohan, A. Momber, R. Kovacevic, Detection of energy

absorption during abrasive water jet machining using

acoustic emission, MED-vol. 2-1, ASME, New York,

1995, pp. 69±85.

[22] A.W. Momber, R. Kovacevic, Energy dissipative processes

in high-speed water solid-particle erosion. HDT-vol. 231/

FED-Vol. 233, ASME, New York, 1995, pp. 243±256.

[23] A.W. Momber, The fragmentation of standard cylinders

during the compressive testing, Lett. Fracture Micromech.

74 (1998), in press.

[24] A.G. Evans, M.E. Gulden, M. Rosenblatt, Impact damage

in brittle materials in the elastic-plastic response regime,

Proc. Roy. Soc. Lond. A 361 (1978) 343±365.

[25] S. Ziegeldorf, Fracture mechanics parameters of hardened

cement paste, aggregates, and interfaces, in: Fracture

Mechanics of Concrete, Elsevier, Amsterdam, 1983, pp.

371±409.

A.W. Momber et al. / Theoretical and Applied Fracture Mechanics 31 (1999) 1±17 17