Eb

Ma

b

a

ARRAA

KEDSTLS

1

tssiwei

htsisto

0d

Wear 267 (2009) 1605–1618

Contents lists available at ScienceDirect

Wear

journa l homepage: www.e lsev ier .com/ locate /wear

xperimental assessment of droplet impact erosion resistance of steam turbinelade materials

. Ahmada,∗, M. Caseya, N. Sürkenb

University of Stuttgart, Institute of Thermal Turbomachinery and Machinery Laboratory, Pfaffenwaldring 6, 70569, Stuttgart, GermanySiemens AG, Energy Sector, Fossil Power Generation, Steam Turbine Technology, Rheinstr. 100, 45478 Mülheim an der Ruhr, Germany

r t i c l e i n f o

rticle history:eceived 18 June 2008eceived in revised form 16 March 2009ccepted 5 June 2009vailable online 16 June 2009

eywords:rosionroplet impactteelitanium

a b s t r a c t

The droplet impact erosion resistance of five different but highly relevant steam turbine blade materialsis investigated with the help of an erosion test rig. The rig adapts wetness and droplet impact speedconditions in the last stages of condensing steam turbines in such a way that the material degradationis greatly accelerated in order to establish monotonic saturating material loss gradients—ideally withina testing time interval of 50 h. Repeatability and reproducibility of the evaluation method is ensured tofacilitate the representative ranking of materials based on droplet impact erosion resistance being a keymaterial property for durable steam turbine blade designs.

A selection of three blade steels (X20Cr13, a steel similar to X5CrNiMoCuNb 14-5, X5CrNiCuNb 16-4) andone titanium alloy (Ti6Al4V) is tested and analysed. Additionally, X5CrNiCuNb 16-4 in a laser-hardenedcondition is investigated. Besides the influence of droplet impact speed and droplet impact angle on

aser hardeningteam turbine

erosion, the generated surface jaggedness, the level of material degradation as well as the material lossgradients are discussed and utilised for further deductions. Among the high yield strength blade steels,the laser-hardened X5CrNiCuNb 16-4 exhibits the best erosion resistance while Ti6Al4V exhibits a highererosion resistance than all the steel alloys tested.

Finally, a simplified but functional model is inferred from the test data to estimate the droplet impacterosion resistance of alternative steel and titanium blade materials relative to the materials discussed in

this text.

. Introduction and review

In order to enable a classification of the subsequent wear inves-igation, it is deemed sensible first to supply some applicationpecific background information for the reader not involved inteam turbine erosion problems. The physical phenomenon lead-ng to a material specific degradation process is discussed together

ith the motivation that triggers the scientific endeavour to rem-dy or at least mitigate the erosion process. A brief historical reviews also given.

Water droplet impact erosion of last stage steam turbine bladesas been a well-known and at times aggravating phenomenon inhe steam turbine and power utility community for a century. Theteam is expanded to a low pressure and temperature in order to

mprove the thermal efficiency of the plant and this causes theteam to expand below the saturation line leading to the forma-ion of droplets in the flow. It is commonly agreed that this kindf droplet erosion is unavoidable when a steam turbine is oper-

∗ Corresponding author. Fax: +49 71168569440.E-mail address: ahmad@itsm.uni-stuttgart.de (M. Ahmad).

043-1648/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.wear.2009.06.012

© 2009 Elsevier B.V. All rights reserved.

ated under wet steam conditions. Only the extent of material lossover time may be positively influenced by various means being dis-cussed later in the text. The droplet impact erosion leads to theloss of blade material and, especially in the blade tip region withhigh impact speed (450–600 m/s), this changes the aerodynami-cally optimised blade geometry and noticeably disturbs the flowaround the blade profile. This in turn adversely influences the per-formance of the machine eventually leading to a need for turbineblade replacement. In addition, the last few decades furthered thedevelopment of highly efficient low-pressure steam turbine designsfeaturing significantly increased exhaust areas. These lead to highaspect ratio blades with enormous tip speeds approaching 750 m/swhich may possibly result in an increased droplet impact erosionpotential. Among a few other restrictions, droplet impact erosionmight therefore be considered as service-life relevant for steam tur-bine blades. Moreover, as the extent of blade leading edge erosionis, for thermodynamic reasons, chiefly related to the actual oper-

ation of the power plant, it is deemed almost impossible to setforth a comprehensive and fool-proof blade erosion protection con-cept from a manufacturer’s point of view. However, significant andsensible erosion mitigation measures such as properly chosen andtreated blade materials have been developed.

1 ar 267

tsidfvsuosTicuwem[

strbbbfipetsiaflwe

t2ttpcHrIwotcmf

p

witsom“ss

606 M. Ahmad et al. / We

Having explained the motivation for the present investigation,he source of detrimental moisture leading to droplet impact ero-ion is now outlined. In the last stages of steam turbines, steams expanded below the saturation line and a part of it is con-ensed into primary droplets with typical sizes of 0.2–2.0 �m. Araction of these primary droplets deposits on the stationary guideanes where it may eventually form rivulets or water films. Thesetructures grow in size, move towards the trailing edges, becomenstable due to aerodynamic forces and finally convert to a sprayf coarse secondary droplets of up to 1500 �m in diameter. Thispray travels in the wake downstream of the vanes’ trailing edges.he large droplets eventually enter a region of higher steam veloc-ty where they are broken further into smaller droplets, known asoarse water droplets, of the order of 100 �m. They accelerate grad-ally with the steam and finally hit the downstream rotating bladesith an impact speed of less than, but at times close to the periph-

ral speed of the rotating blades. The result of this droplet impactay then be erosion, i.e., structural damage of the blade material

1–5].Since the recognition of the phenomenon leading to blade ero-

ion, several mitigation measures have been adopted to minimisehe erosion of steam turbine blades. These include simple geomet-ic design considerations such as an increase in the axial spacingetween stator and rotor to allow the droplets to be accelerated androken up. Thinner trailing edges of the stator vanes are thought toe advantageous as they produce smaller initial secondary dropletsrom the water film. Moisture extraction between the blade rowss considered to be a more sophisticated and efficient method byroviding suction slots on the guide vane surface as well as byvaporating the water film and rivulets by internally heating uphe stationary guide vanes. The latter is the most efficient ero-ion mitigation measure known to date [6]. Additionally and mostmportantly, attention is paid to ensure that the blade leading edgesre more resistant against erosion. Laser treatments, induction orame hardening of blade materials as well as shielding of bladesith Stellite or tool steel, have been used to improve the leading

dge erosion resistance [1,7,8].Historically, the erosion of steam turbine blades became the

opic of scientific interest and research in the beginning of the0th century when the tip velocities of the rotating blades of steamurbines became sufficient to cause erosion. The material degrada-ion of steam turbine blades had been explained by every possiblehenomenon including chemical attack, oxidation, solid particlesarried by the steam except liquid droplet impact (Coles, 1904).owever in the 1920s, experiments had been carried out to cor-

elate erosion of steam turbine blades with droplet impact [9,10].n 1928, Cook presented his famous water-hammer equation in

hich he estimated the pressure generated when a liquid columnf water impacted on a solid surface. In his theory, he showedhat the pressure generated at liquid solid impact is sufficient toause erosion of steam turbine blades [10,11]. According to Hey-ann [12], Cook’s water–hammer relation may be extended as

ollows:

impact = �lclvimpact ·(

1 + k · vimpact

cl

)(1)

here the droplet impact pressure pimpact depends on the dropletmpact speed vimpact, the liquid density �l and the liquid acous-ic speed cl which, with some limitations, represents the shockpeed in the compressed liquid. Here k is a constant depending

n impacting liquid properties and according to Heymann’s experi-ents, its value approaches to 2 for water. The first term denotes the

classical” water hammer pressure derived from momentum con-iderations while the second term reflects the variant nature of thehock speed. It is important to note that the magnitude of impact

(2009) 1605–1618

pressure is independent of droplet size, its duration, however, isdependent on droplet size and geometry [13].

In the period from the 1960s to the 1990s, a lot of scientificresearch was undertaken in the field of droplet impact erosion, seee.g. [4,10] and [14–20]. The basic finding was that, as the dropletimpacts on a solid surface, a pressure wave is generated within thedroplet at the point of contact which travels back inside the dropletwith the speed of sound. This shock wave remains in contact withthe solid surface as long as the contact velocity is higher than theshock velocity and the liquid remains compressed within this shockenvelope. Later on, shock speed overtakes the contact edge velocityand the shock wave detaches from the contact surface. At this point,a lateral jetting is observed with velocities many times higher thanthe impact velocity. The impact pressure then reaches its maximumvalue (about three times the water hammer pressure). Shock speed,jetting time and impact pressure remain the topic of interest duringall these investigations.

A variety of dedicated erosion test rigs have been constructedin the past where erosion caused by repeated droplet or jet impacthas been studied. Generally, these experiments show that dropletimpact erosion depends on the impact count and hence is a timedependent process. It starts with a so-called incubation period withno or very minor material damage, followed by an accelerationperiod where the rate of erosion increases rapidly to a maximumvalue, followed by a deceleration period where the erosion ratedecreases to some fraction of maximum erosion rate (1/2 to 1/4)and finally a steady terminal erosion condition where the erosionrate remains almost constant. The erosion rate is found to be sensi-tive to impact velocity preferably described by a power law equationRe ∼ Vn where the value of n is reported to be 4–5 for ductile mate-rials and 6–9 for brittle materials. In accordance with theory, biggerdroplets produce more erosion while the impact angle is found tobe most significant in terms of erosion damage at perpendicularimpact to the target surface. Moreover, dependencies of the liquidproperties on erosion are observed as erosion rate varies at 2ndto 2.5th power with liquid density and 1/2 to 3/4 power with theinverse of the liquid viscosity. An increase in temperature of theimpacting liquid generally increases the erosion slightly. This effectis attributed to the increased shear damage of the surface causedby the evolving lateral jet flow [1,17,18,21].

Extending the definition of erosion test rigs to real steam tur-bines, the impact count is chiefly related to the local wetness value.The droplet size may be related to the trailing edge diameter ofthe stationary vane and the local density of the steam, which isproportional to the local steam pressure. Besides the axial spacingbetween stator and rotor, the relative droplet impact speed maybe related to the steam density as well. However, more significantis the rotor blade tip speed [6,22]. As in the past, the rotor bladematerials have not been varied to a great extent, the describedmulti-variable system is often condensed to a set of semi-empiricalcharacteristic numbers that determine the amount and strength ofnecessary countermeasures. This pure phenomenological approachessentially requires a large fleet experience as a key factor to suc-cess.

From the stressed blade material’s perspective, it seems desir-able to correlate the erosion resistance of a specific material to awell-defined set of macroscopic mechanical properties. It is foundthat hardness, resilience, toughness, tensile strength, ductility andstrain energy can significantly affect the ability of a material towithstand droplet impact erosion. However, none of them provesto be a single material parameter to whom erosion resistance can

be related uniquely [1,13,17]. Hardness proves to be the most reli-able material property to assess the erosion resistance. It is foundthat erosion generally varies with the 2nd to 2.5th power of Vickershardness number. However, for materials of different categories ormetallurgic structures, this simple relation may not hold [9].

ar 267 (2009) 1605–1618 1607

deuayseMesthetfSoesba

rwmadics

2

2

tdaTww

larger disc in diametrically opposite positions. Each specimen ismounted in a tool steel guard ring to prevent fragmentation and

M. Ahmad et al. / We

Bearing in mind that erosion is a complex process with a lot ofifferent parameters and physical processes involved, a normalisedrosion resistance has been proposed which is defined as “The vol-me loss rate of a test material, divided by the volume loss rate ofspecified reference material similarly tested and similarly anal-

sed”. Austenitic stainless steel with a hardness of 170 HV, stainlessteel type 308 and some other materials have been used as ref-rence materials [1]. None of them, however, is widely accepted.any authors have proposed theoretical parameters to define the

rosion resistance of materials [23]. But these parameters do notolve the problem as they are either too difficult to evaluate or failo predict the empirically observed dependencies [1]. Some authorsave tried to correlate erosion resistance with fatigue strength, see.g. [17] and [13]. Although repeated stress pulses may be thoughto be common in both erosion and fatigue, the idea to correlateatigue strength with erosion resistance did not gain popularity.everal authors tried to find the dependence of erosion resistancen surface microstructure. A smooth surface is found to be morerosion resistant than a roughened surface whereas a surface withmall grain size has more inherent erosion resistance. Inter-atomicond strength as well as the size and distribution of surface flawsre found to play an important role in erosion resistance [17].

Although many scientists have tried both empirically and theo-etically to predict the erosion in steam turbine blades, none of themas able to tackle the problem in a comprehensive way. Keeping inind that a lot of parameters affect erosion in steam turbine blading

nd that these parameters may also depend on each other, the pre-iction of steam turbines blade erosion is a serious challenge which

s yet to be solved. The prediction of erosion becomes even moreomplicated when varying operating conditions of the individualteam turbines is taken into account [24–32].

. Description of experiments

.1. Test rig set-up

In 1959, an erosion test rig (Fig. 1) was constructed in Newcas-le upon Tyne by C.A. Parsons to assess the erosion resistance of

ifferent materials. As a result of the merger between C.A. Parsonsnd Siemens, this test rig was relocated to the Institute of Thermalurbomachinery at the University of Stuttgart, Germany, where itas completely overhauled and upgraded with modern drives asell as state-of-the-art automation and control logics.

Fig. 1. Erosion test rig at ITSM, University of Stuttgart, Germany.

Fig. 2. Specimen holder and sprayer; arrows indicate the direction of motion ofspecimen and spray disc.

The test rig has two separately driven continuously adjustablecontra-rotating shafts, each carrying a mild steel overhung disc.Four specimens are mounted on the perimeter of one disc andtwo sprayers on the other. With both shafts running at maximumspeed, a droplet impact speed of up to 660 m/s can be obtained(Figs. 2 and 3). Both, jets and specimens, are enclosed in a vacuumchamber where a pressure of 10 kPa (absolute) can be obtained bysteam packed shaft glands. Each shaft is driven from a two-pole,water-cooled induction motor rated at 67 kW at 200 Hz. The fourspecimens are mounted equispaced around the perimeter of the

unrepresentative weight loss at the edges. To ensure well-defineddroplet impact angles the specimen holder can be tilted according

Fig. 3. Velocity triangles of spray droplets. us: velocity of spray nozzle, usp: velocityof specimen, wj: relative velocity of spray jet, cj: absolute velocity of spray jet, wsp:droplet impact velocity.

1608 M. Ahmad et al. / Wear 267

Ft

tfl

attmfsotrw(awvu

2

tipcmtts

iseEwidaismtmw

hmsata

therefore, emphasises the larger droplets and reads:

ig. 4. From left to right: new specimen, specimen guard ring and specimen afteresting of 50 h.

o the combination of the two shaft speeds and the actual waterow rate (Fig. 3).

Water is fed along the hollow centre of the sprayer shaftnd passes through radial holes into two nozzles mounted onhe smaller disc, the water flow being equally divided betweenhem. The pressure-atomising Watson fan type sprayer nozzles are

ounted diametrically opposite to one another. The droplet sizesrom these nozzles have been measured in a static test rig at atmo-pheric conditions using a laser diffraction technique. The resultsf this investigation will be discussed in a later section of thisext. The water mass flow to the sprayer nozzles is measured andecorded. From this, the radial velocity component of the ejectedater droplets is calculated. Bearing temperature, casing pressure

vacuum) and temperature, lubrication oil pressure, shaft speedsnd flow to the nozzles are monitored throughout the testing. Theater mass flow rate can be continuously adjusted including to a

alue of 0.057 kg/s, which is the standard flow rate for one nozzlesed during the testing in the erosion test rig.

.2. Erosion resistance assessment procedure

The specimen materials are taken from real turbine bladeshat are manufactured according to strict technical supplier spec-fications including fulfilment of all major macroscopic materialroperty requirements. Any variation of an individual property thatomplies with the requirements set forth in an original equipmentanufacturer’s (OEM) documentation can be judged as a contribu-

ion to natural scatter and is consequently not evaluated for theested materials. Only the actual Vickers hardness values of thepecimens are sampled separately for each batch of material.

In each experiment, four materials can be tested simultaneouslyn the test rig. Specimens of each material are manufactured in atandard shape and placed in a guard ring to avoid chipping at thedges. The specimen test face is 13 mm in diameter and 3 mm thick.ach probe is placed in a specimen holder, which is tilted in such aay that the droplet impact angle to the pristine material surface

s 90◦ (perpendicular impact). The specimens are tested for a totaluration of 50 h in 5 h intervals. After each 5-h test, the specimensre taken out, cleaned and weighed. A test specimen before test-ng, a guard ring and a test specimen after 50 h exposure time arehown in Fig. 4 for illustration. The 50 h standard test duration isotivated by the rig’s test intent specification to greatly accelerate

he erosive material loss found in real machines in such a way thatonotonic saturating material loss gradients can be establishedithin a feasible time frame.

The weight loss after each time interval is interpreted with theelp of the materials density to reflect the volumetric erosion ofaterial during that time interval. As a second key figure, the area

pecific first time derivative, i.e. the erosion rate Re,50 in [m/s] being(material loss) velocity by nature, is derived and evaluated. As

he erosion rate Re,50 reflects the tendency of a material to erodet a given erosive environment, its reciprocal value RE,50 in [s/m]

(2009) 1605–1618

will reflect the resistivity of the material to erode in terms of timetaken by the erosive environment for a given degree of materialdegradation. For the materials tested at the same conditions for aconstant time period, RE,50 can be utilised as a criterion to assessthe material inherent erosion resistance and is given as:

RE,50 = 1

Re,50= Aspecimen

1/50 h ·∫ t=50 h

t=0V̇(t)dt

= 50 h · Aspecimen

�V50(2)

where Aspecimen is the pristine specimen surface area, V is the spec-imen’s volume and �V50 is the volume loss after 50 h of testing.

Of course, this erosion resistance assessment criterion may seemsomewhat arbitrary and rig specific at first view. However, asthe blade materials approach a saturating erosion rate, this crite-rion will only slightly overestimate the material loss when beingutilised for real turbine blading designs. Hence, conservatism of theapproach is minimized. An observed erosion rate shall be definedto be “saturating” if the area specific second time derivative of thevolume loss of the tested material referenced to a reference mate-rial (here: X20Cr13 at standard testing conditions) is smaller than10−5 after completion of a minimum testing time interval of 25 h.This definition explicitly includes greater negative values of the sec-ond time derivative. In the rare but still possible case, that thesesaturating erosion rates cannot be established within the specified50 h standard testing interval, the test needs to be prolonged untilthe fulfilment of this condition. Otherwise, conservatism for designpurposes cannot be granted.

Given the possibility of precise repeated testing of a specificmaterial, preferably at well determined and, if desired, differ-ent testing conditions, a meaningful relative ranking of differentmaterials becomes feasible. This fundamental and most importantfinding may be supplemented and accompanied by metallographicanalyses and macroscopic surface structure investigations addingto the comprehension of the determined erosion resistance.

2.3. Droplet size measurements

Cook’s water–hammer equation and a variety of its descendants(see e.g. Eq. (1)) suggest an independency of droplet impact ero-sion on droplet size. However, Cook presumes that the deterioratingeffect may also be related to a droplet’s size by “the area of attackand the duration” [10] of the water–hammer effect. Together witha distinct influence on its acceleration and break-up characteristicin a steam turbine flow environment, a droplet’s size is thereforeconsidered to be of significant importance for the complete descrip-tion of the droplet impact erosion phenomenon. Hence, the dropletsproduced by the sprayer nozzle inside the rig deserve appropriateattention.

Since a droplet size measurement inside the erosion rig at oper-ating speed proves to be difficult, similarity theory is employed todesign a static test set-up at atmospheric conditions where the noz-zle characteristic at different flow rates is assessed using a MalvernParticle Sizer. It shall be assumed that the droplets Sauter meandiameter, at a spray length corresponding to the distance betweensprayer tip and specimen impact surface, is indicative for the sprayitself as well as for its detrimental effect. Moreover, in doing so, alink to the real steam turbine wetness values may be established[33]. Contrary to the arithmetic mean diameter, the Sauter meandiameter characterises the dispersion of the liquid phase by relat-ing the integral droplets volume to the integral droplets surface. It,

DSauter =∑M

i=1(Ni · D3droplet,i)∑M

i=1(Ni · D2droplet,i)

(3)

M. Ahmad et al. / Wear 267 (2009) 1605–1618 1609

Table 1Comparison of the Weber number inside and outside the erosion test rig.

Influencing parameter Dimension Static test rig atatmosphericconditions

Rotatingerosion rigat vacuum

Mass flow rate per nozzle [kg/s] 0.057 0.057Surface tension of water � [N/m] 0.073 0.073Environment pressure [kPa] 96 10Rotational speed [1/s] 0 145Initial velocity of droplets

relative to environmentvrel,0

[m/s] 72 224

Primary droplet diameterD0

[mm] 1 1

DW

wiDttrmtt

wifbfdato

ifsbDjzirsba

dalcseTtbec

Fig. 5. A picture of the spray at the flow rate of 0.057 kg/s showing the coordinatesystem of the spray measurement. A fan type spray is evident in the picture with theangle of divergence of about 55◦ . The spray is analysed at different positions alongz-axis at y = 0.05 m which is the true spray length in the erosion test rig.

eter of about 90 �m at standard testing conditions. With the spraybeing fan shaped and hence thin, the variation of droplet size inthe third direction (spray breadth x) is already incorporated in theexisting laser diffraction result.

ensity of environment �air [kg/m3] 1.141 0.111eber number:We0 = (�air · v2

rel,0· D0)/�

[–] 81 76

here M denotes the number of droplet size classes and Ni thendividual droplet counts within such a droplet class i of a diameterdroplet,i. From this, and taking into account the major contributors

o spray characterisation to be the density of the ambient (air) �air,he surface tension of water �, the spray nozzles aperture D0, theelative velocity of the droplets at the aperture and at the speci-en’s surface, vrel,0 and vrel,0.05, respectively, and finally the spray

ravel distance y, the set of similarity parameters reads as followso be:

DSauter,0.05

D0= f

(�air · v2

rel,0 · D0

�;

vrel,0

vrel,0.05;

y

D0

)

= f

(We0;

vrel,0

vrel,0.05; y∗

)(4)

here ‘0’ and ‘0.05’ denote quantities at the nozzle tip and the spec-men surface, respectively. The Weber number We0 relates inertiaorces to surface tension and hence is a measure for a droplet’s sta-ility, i.e., its ability to maintain a spherical geometry while shearorces try to tear it apart. Whereas the Weber number We0 and theimensionless spray travel distance y* are comparatively easy tossess and sustain, the deceleration ratio of the droplets is not. Sohis latter influence may be tackled by engineering judgement inur case.

In order to assess to what extent the droplet sizes measuredn the static test rig at atmospheric conditions are representativeor droplet sizes in the rotating test rig at 10 kPa ambient pres-ure, the Weber numbers for the two set-ups are calculated toe 81 and 76, as shown in Table 1. Adapting the “Blob Primaryroplet Break-up Method” [34,35], the diameter of a drop ejected

ust at the nozzle outlet is considered to be equal to the noz-le aperture in both cases. As the corresponding Weber numbers nearly the same, the droplet sizes measured in the static testig are judged to represent the droplet sizes in the rotating rigufficiently. Any supplementary shear force exerted by the tur-ulent flow inside the rotating rig is, however, neglected in thispproach.

Figs. 6–8 illustrate the spray characteristics. Fig. 6 shows theroplet size and number distributions measured in the static rigt the standard mass flow rate of 0.057 kg/s at the spray’s centre-ine (z = 0 m) at spray length coordinate of about y = 0.05 m, whichorresponds to the distance between the spray nozzle and thepecimen surface in the rotating test rig. The generated dropletsxhibit a Sauter mean diameter according to Eq. (3) of 88 �m.

he volumetric (or mass averaged) mean diameter is calculatedo be 150 �m, the maximum of the liquid volume distributioneing found at about 190 �m. The relation between the differ-nt diameters remains approximately similar throughout the sprayharacteristic and should be born in mind while discussing the

Fig. 6. Droplet size and number distribution in the spray at y = 0.05 m and z = 0 mmat standard testing conditions.

influence of droplet diameters. Note, that the volume specificdroplet count is on a logarithmic scale.

Droplet sizes have also been assessed at different positions alongthe spray width (z-direction, see Fig. 5) at a spray length coordinateof y = 0.05 m (Fig. 7). In the core of the spray (±0.01 m from thespray axis) the generated droplets exhibit a Sauter mean diameterof about 90 �m. Having a specimen surface diameter of 13 mm, itis inferred that droplets impacting on the specimen surface are ofuniform and reproducible dispersion having a Sauter mean diam-

Fig. 7. Spray Sauter mean diameter measured for the nozzle along different positionsalong z-axis at y = 0.05 m. Droplets in the spray core are reasonably uniform in sizeand are in the order of 80–90 �m. Flow rate is 0.057 kg/s.

1610 M. Ahmad et al. / Wear 267 (2009) 1605–1618

Fz

flrtaasecdsdWdmnrbsrb

2

cd(hrbctjv

2

pmftdtpm

ig. 8. Spray Sauter mean diameter at different initial Weber numbers at y* = 50 and* = 0 (spray characteristic; squares: static test rig results).

Note, that the droplet mean sizes decrease with increasing sprayow rate (Fig. 8) because of higher shear forces present at higherelative velocities. Here, the spray flow rate is reflected by the ini-ial Weber number We0 on the abscissa whereas the diameter ratioccording to Eq. (4), depicted on the ordinate, reflects the spraytomisation capability (droplet break-up). In order to appraise thehear forces at the location of the specimen surface, the local decel-ration rated Weber numbers We∗

0.05 are also plotted as an array ofurves. From these curves, it may be deduced that the measuredispersions are reasonably stable as the stability threshold, whichhall be associated with the Weber number of the maximum dropletiameter measured at the standard flow rate, is estimated to bee∗

0.05 = 25. Moreover, as stated before, assuming that the dropleteceleration ratio is of the same order of magnitude and bearing inind the low gradient of the droplet size with the initial Weber

umber, the Sauter mean diameters presumably present in theotating rig may also be estimated to be 90 �m. In addition, wheneing evaluated and compared to reported coarse water dropletize distributions (e.g. [36]), the rig’s droplet sizes are judged to beepresentative for the coarse water in real steam turbine last rotorlades.

.4. Gauge R&R study for the measuring balance

An important factor to evaluate the results is to assess the pre-ision of the balance used to measure the weight loss of the probesuring the testing. For this purpose a classical gauge R&R studyrepeatability and reproducibility study) of the measuring gaugeas been pursued. Ten different specimens covering the completeange of investigated materials have been weighed three times eachy two different appraisers using the same gauge. The result indi-ates that the measuring gauge (balance) adds a variance of 0.01%o the observed weight measurements. The balance is thereforeudged to be perfectly suited for the given task in terms of absolutealues and discrimination.

.5. Reproducibility of erosion rig results (procedure validation)

It is an important quality requirement to verify that the resultsroduced with the test rig are reproducible. For this reason, speci-ens made from the same reference material but partly stemming

rom different casts are tested and evaluated at different times at

he same testing conditions. As the reference material’s propertieso not vary to a large extent, this procedure is repeated from timeo time in order to detect long term changes (aging) of the com-lete experimental set-up and to assess the influence of significantaintenance measures like the exchange of sprayers.

Fig. 9. Scattering pattern of the test rig results obtained by testing four X20Cr13specimen for 50 h at different times at constant testing conditions (standard devia-tion: 2.6%).

Fig. 9 shows the results for four different X20Cr13 specimens.The figure shows that the results obtained from the test rig arereasonably reproducible. The standard deviation of the volume lossduring a 50 h test run is calculated to be 2.6%. This remaining scatterlevel is attributed to variations in material properties as well as tovariations in the testing procedure itself and is considered to beacceptable.

2.6. Test settings and investigated materials

For the current investigation, five different but highly relevantsteam turbine blade materials have been selected for testing in theerosion rig according to the procedure described in Sections 2.1 and2.2. Relevant material properties are shown in Table 2. The testedmaterials are:

• X20Cr13 (key: 1.4021).• X5CrNiCuNb 16-4 (key: 1.4542); short term “X5 base”.• X5CrNiCuNb 16-4, laser treated condition (key: 1.4542); short

term “X5-Laser”.• Blade steel similar to X5CrNiMoCuNb 14-5 (key: cf. 1.4594); short

term “X5-equiv”.• Ti6Al4V (key: 3.7165).

The following test conditions are maintained during the mea-surements:

• spray water flow rate: 0.114 kg/s (supplies two sprayer nozzles);• testing environment (inside rig): 10 kPa (absolute) at 313 K;• droplet Sauter mean diameter: approx. 90 �m;• test duration: in total 50 h per series (minimum: 25 h).

In order to assess their influence on the erosion progress, thefollowing test conditions are varied:

• droplet impact speed (impact pressure);• droplet impact angle relative to pristine specimen surface.

3. Results

3.1. Volume loss and erosion rate comparison at standard testingconditions

From a generic viewpoint, Fig. 9 illustrates the classical erosionbehaviour of a steam turbine blade material. To illustrate this state-ment, the first time derivative of the volume loss, being referred toas the erosion rate, is inferred from the test results and depicted in

M. Ahmad et al. / Wear 267 (2009) 1605–1618 1611

Table 2Mechanical properties of specimen (at 313 K; standard testing conditions).

Material Density[kg/m3]

Surface hardness[HV-10]

Young’s modulus E[GPa]

Modulus of resilienceU∗

r /U∗r-X20Cr13

Erosion resistanceRE,50/RE,50-X20Cr13

X20Cr13 7.710 271 216 1.00 1.00X5CrNiCuNb16-4 7.760 328 200 2.61 1.17X5CrNiCuNb 16-4, laser treated 7.760 420 200 4.28* 2.58X 200 3.38* 1.78T 110 3.95 3.23

*

tiadcismad

iifrar

taXSeoX1h

lXimdi

i[t

Fv

5CrNiMoCuNb 14-5-equiv 7.750 373i6Al4V 4.420 334

Values inferred from hardness values acc. to Eq. (5).

he graph as well. A short incubation period with very little damages observed, followed by a rapid acceleration period which lasts forpproximately 2 h. After that the erosion rate decreases again. Thiseceleration period may eventually lead to a saturation which isommonly regarded to be the terminal erosion condition. Assum-ng that the 50 h test results represent the gradient observationatisfactorily, the dimensionless terminal erosion rate may be esti-ated to be 0.003. Although there is significant uncertainty in this

ssumption, it may well serve as a qualitative mean to compareifferent materials.

Fig. 10 shows the dimensionless cumulative volume loss and thenferred erosion rates for the five investigated materials. The dropletmpact speed is 488 m/s, the water mass flow rate is 0.114 kg/sor two sprayer nozzles. All results are normalised by the cor-esponding values of X20Cr13 after 50 h of testing. For claritynd comparison, the X20Cr13 result discussed already in Fig. 9 isepeated.

Being exposed to standard testing conditions, Ti6Al4V is foundo exhibit the most favourable erosion behaviour of the five materi-ls tested. Among the high yield strength steels, the laser-hardened5CrNiCuNb 16-4 is found to have the highest erosion resistance.ince the depth of the laser hardening is much greater than therosion depth, it is inferred that the complete testing intervalf 50 h has been pursued on the laser-hardened metal surface.5CrNiMoCuNb 14-5-equiv is third in ranking while X5CrNiCuNb6-4 base material is in fourth place after 50 h of exposure. X20Cr13as the lowest erosion resistance.

All tested materials show the erosion behaviour described ear-ier in this section. However, it is observed that the laser treated5CrNiCuNb 16-4 and the Titanium show a considerable reduction

n erosion acceleration during the initial hours whereas the otheraterials do not. It may be inferred that this preferable behaviour

uring the initial period (incubation and erosion acceleration) is

ndicative for a beneficial higher erosion resistance.

From another perspective, the volume loss characteristics aren good agreement with the work of many other authors (e.g.1,13,21]) and may form another modest contribution to the his-orical database.

ig. 10. Dimensionless cumulative volume loss and erosion rates for the five investigatedalues of X20 after 50 h).

Fig. 11. Erosion resistance as a function of material’s surface hardness.

3.2. Erosion resistance and macroscopic mechanical properties

The mechanical property “hardness” shall be understood as theability of a material to resist an incursion or penetration of anothersolid or liquid body and is closely related to the property “strength”that describes the resistance of a material against deformation byexternal forces. As a consequence, and reflected by the data sum-marised in Table 2, the surface hardness is again confirmed to be amost significant mechanical property that can be utilised to assessthe erosion resistance of a (steel) metal. Generally, a material with agreater surface hardness has a higher erosion resistance. The results(see Fig. 11) suggest that the erosion resistance of steel materialsvaries according to a power law equation of the form RE,50∼Hn

V ,

where the value of n is found to be 2.3 which is in good agreementwith the previous investigations [1].

However, this observation does not explain the high erosionresistance of Ti6Al4V which is about three times higher than thatof the X5 base material while both exhibit almost the same surface

materials (impact speed: 488 m/s; water flow rate 0.114 kg/s; results normalised by

1 ar 267

htfhhci

(itaaoUtrnttcRa

U

wsorrtwtrmtpstterm

aia

Fs

the non-dimensional volume loss is depicted as a function of the

612 M. Ahmad et al. / We

ardness (see Fig. 11). This suggests at least that a direct connec-ion between the erosion resistance and the hardness exists onlyor materials that feature similar metallurgical structures, whichas already been contemplated by Honegger in 1927 [9]. Hence,ardness alone is not enough to assess the erosion resistance whenomparing metals of different categories. At the minimum, a secondnfluencing material property may be needed.

As Young’s modulus of elasticity (E) of Ti6Al4V is about 110 GPathat for steel is about 210 GPa) and the yield strength of Ti6Al4Vs almost of the same order as that of steel alloys, it seems indica-ive that the higher erosion resistance of titanium, even if featuringlower surface hardness level, may be explained by its ability to

bsorb more energy in an elastic deformation process. This sec-ndary macroscopic property is commonly referred to as resiliencer and is represented by the area beneath the elastic fraction of

he stress versus strain curve of a material. The idea of treating theesilience in connection with the erosion resistance of a material isot new and many other authors have also tried to use this propertyo evaluate the erosion resistance of materials (e.g. [37]). However,his idea is not amongst the most popular ones despite of its signifi-ant explicative potential, especially if the material’s yield strengthp0.2 is rated by its surface hardness HV as proposed by the currentuthors:

∗r = (Rp0.2 · (HV /HV0))2

2 · E(5)

here the index (*) shall indicate that the resilience is rated by apecific hardness ratio with HV and HV0 being the surface hardnessf the, what ever way, treated material and the datum mate-ial respectively. Note, that the actual hardness changes the ratedesilience quadratically. In addition, it may be directly inferredhat changes in hardness or yield strength are gaining importancehen high yield strength materials with comparatively lower elas-

ic moduli, like titanium, are considered. Of course, the (elastic)esilience may not supply perfect explanations if being applied toaterials of significantly different toughness or metallurgic struc-

ure. While “toughness”, being defined as the amount of energyer volume that a material can absorb before rupturing, may be ofignificant importance once the droplet impact pressure exceedshe elastic resilience value and material degradation enters in plas-ic range, the metallurgic constitution may influence the overallrosion characteristic of a material. However, important erosionesistance characteristics of relevant steam turbine blade materials

ay already be reasonably reflected by the resilience.Fig. 12 shows the erosion resistance of the tested materials

ccording to Eq. (2) as a function of the rated resilience accord-ng to Eq. (5) at standard test conditions. For comparison, all valuesre referenced to the corresponding values of X20Cr13. Again, the

ig. 12. Erosion resistance as a function of rated resilience according to Eq. (5) attandard test conditions.

(2009) 1605–1618

results show that Ti6Al4V featuring a lower hardness but a higherrated resilience than most of the steel materials has a higher erosionresistance than steels of the same or even greater surface hardness(Table 2). Among the high yield strength blade steels, the laser-hardened X5 has the highest erosion resistance. For orientationand to foster further discourse, three copper precipitation hard-ened steels are connected by a line suggesting a linear relationshipbetween erosion resistance and rated resilience. Indirectly inferreddashed lines are inserted for X20Cr13 (same gradient) and Ti6Al4V(twice the gradient). Despite being educational to some degree, nostatement about the accuracy of such a deliberation can be madeat present.

A direct link between the macroscopic material property“droplet impact erosion resistance” and the previously discussedconventional macroscopic material properties seems from a mate-rial designer’s point of view desirable and, at first glance, feasible.However, the current investigation does not sufficiently supportthis idea. On the contrary, the results suggest that “erosion resis-tance” against liquid droplet impact should be treated as anindependent material property. Once this additional degree of free-dom in material design is accepted, a dedicated search for erosionresistant materials offers new chances of success.

3.3. Dependence of erosion on droplet impact pressure

Having looked at the general erosion characteristic from a mate-rial property perspective (“defence”), the viewpoint shall now beslightly shifted to the damaging aspect connected to the imping-ing water droplets (“offence”). If the detrimental impact pressureis varied by varying the droplets’ impact velocity according to Eq.(1), it is interesting to investigate how the erosion characteristicof copper precipitation hardening alloys and titanium changes. Forthis purpose, supplementary tests series are performed at dropletimpact speeds of 366 m/s and 580 m/s corresponding to an impactpressure of about 810 MPa and 1550 MPa, respectively. While thefirst is chosen to be slightly lower than the materials’ yield strengthvalues, the latter is significantly higher. As a (presumably possible)consequence, the test series at an impact speed of 580 m/s has to berestricted to a total duration of less than 50 h, as the X5 base mate-rial specimen is already significantly eroded after this time intervalpreventing a meaningful continuation of the test.

Fig. 13 shows a graphic representation of the results. Here,

droplet impact pressure according to Eq. (1). The volume loss val-ues at different impact pressures (speeds) are made comparable byadjusting them for the equivalent droplets pulses by incorporating

Fig. 13. Normalised volume loss of steam turbine blade materials in dependence ofdroplet impact pressure according to Eq. (1).

ar 267

tcecvbaai

milbtapprtl1tfectotet

3

aaIlohd7pd(id

F(

M. Ahmad et al. / We

he droplets pulse ratio N/Nstandard, which will counterbalance thehanged rotor speeds effect. The volume loss values are again ref-renced to the corresponding value of X20Cr13 at standard testingonditions. Similar to the previous case of varying surface hardnessalues, the effect of impact speed on erosion rate may be expressedy a power law equation of the form Re∼vn

impact where values of nre found to vary from about 3.3 to 4.5 for high yield strength steellloys to about 5 for Ti6Al4V. These exponents are found to be welln agreement with previously reported results (see e.g. [1,38]).

When being compared to Fig. 10, the absolute ranking of theaterials does not change over the investigated range of relevant

mpact speeds. Ti6Al4V exhibits the highest erosion resistance, fol-owed by the laser treated X5CrNiCu 16-4, X5-equiv and the X5ase material. However, when considering a relative ranking ofhe materials based on how the erosion of a material acceleratest higher impact speeds, the assessment changes slightly. Com-aring the ratios of cumulative volume loss of each material at aarticular impact speed to the cumulative volume loss of the cor-esponding material at 366 m/s, it is observed that the erosion ofhe two base materials, X5 base and X5-equiv, accelerates slightlyess at higher impact pressures than that of laser treated X5CrNiCu6-4 and Ti6Al4V which might possibly be attributed to the loweroughness values of the two latter materials. In other words, byurther increasing the droplet impact speed, a material in a hard-ned condition may eventually be superseded by its non-hardenedondition in terms of absolute volume loss by brittle failure dueo local mechanical overload. Even if this observation is of extrap-lated nature and hence of rather academic value, it again pointsowards the necessity to treat and investigate the material prop-rty “erosion resistance” separately for each material—at least forhe case of liquid droplet impact.

.4. Dependence of erosion on droplet impact angle

A question that is closely related to the impact pressure gener-ted by an impinging droplet is whether or not a varying impactngle has an additional influence on a blade’s material loss [1,39].n other words, whether or not the material fatigue by perpendicu-ar droplet impact may be significantly superimposed by some sortf chipping effect. To address this aspect, four X20Cr13 specimensave been tested simultaneously at otherwise standard testing con-itions at droplet impact angles of 90◦ (perpendicular impact),0◦, 60◦ and 50◦. Unfortunately, more shallow impact angles arehysically not possible due to test set-up specific limitations. The

ifferent impact angles are adjusted by tilting the specimen holdersee Figs. 2 and 3). Referring Fig. 14, it is confirmed that the max-mum volume loss occurs at perpendicular impact and that itecreases as the impact angle decreases. The obtained data is ref-

ig. 14. Normalised volume loss of X20Cr13 as a function of droplet impact angleimpact speed: 488 m/s).

(2009) 1605–1618 1613

erenced to the volume loss of X20Cr13 at 90◦ impact and mirroredwith respect to the 90◦ impact axis. For orientation, the previouslydetermined standard deviation of 2.6% is depicted as well.

At first view, the macroscopic impact angle consideration maybe thought to apply to a pristine specimen surface only. As soonas a surface disturbance is generated by erosion, this disturbancemay tend to amplify itself in a direction perpendicular to the maindroplet impact angle (like drilling inclined holes into a flat plate),hence locally producing erosion conditions as in the datum per-pendicular impact case (90◦ impact angle). In order to avoid thisconceivable effect, the specimens are tilted randomly with respectto their individual rotational axes any time they are reinstalled afterscheduled interim weight loss assessments. A dedicated discussionof such a “self-amplification hypothesis” shall be left to anotherpaper.

As a second step, a model curve (see solid line in Fig. 14) is com-puted based on the combined hypothesis that (a) erosion is relatedto the droplets’ perpendicular impact velocity component only and(b) that the erosion behaviour of X20Cr13 follows a power law rela-tion with impact pressure, similar to that of the other materialstested. The resulting model curve featuring a standard deviation of2.3% to the data is judged to be a surprisingly excellent representa-tion of the observation made. Even though – or maybe because – theexperiments are not extended to more glancing impact angles, theresults do not supply enough evidence that the chipping makes asignificant contribution to the detrimental effect of shallow dropletimpact erosion. Moreover, as the volume loss seems to be rapidlyreduced when the impact angle gets more glancing, any chippingeffect, if possible at all, may be deemed comparatively insignificantfrom a practical standpoint.

3.5. Analysis of eroded specimen surfaces

The topography of the eroded surface might be thought of toplay a vital role in determining the processes leading to erosionsaturation. Initially with no erosion on the surface, the water filmon the surface tends to reduce the intensity of impact pressure upto one half of its original value. As the surface gets eroded withtime, the assessment of erosion process becomes more complicatedwhere water entrained in the cavities cushions the impact whereasthe surface peaks exhibit less erosion due to a reduced durationof impact pressure. In any case the scale of surface roughness issuggested to be related to the impacting droplets sizes.

In the preceding chapter, the texture of the eroded surfacehas been mentioned to be a possible driver for the characteris-tics and the development of the erosion process. This detail maybe especially true if (a) the impact angle dependency is as previ-ously suggested and (b) if the erosion leads to macroscopic surfacetextures whose scales are significantly larger than the impingingdroplet sizes. Then, as the erosive potential of a possible perpendic-ular droplet impact is minimized by the jagged surface, erosion maybe thought to be integrally self-inhibiting in a longer time perspec-tive within a locally increased total surface area of the target. Thisleads, in turn, to the classical explanation for erosion rate saturation.

In addition, for a specific volume loss or erosion resistanceaccording to Eq. (2) of a given material at fix droplet impact condi-tions, a variety of different surface textures are conceivable. Also, agiven surface structure may be attributed to a large range of integralvolume losses. This virtually constitutes independence of surfacetexture and volume loss. However, a distinct interrelationship shallbe postulated that may be further utilised. Consequentially, it is

worthwhile to analyse the eroded surfaces in a greater detail.

A first step is the visual inspection and documentation of theeroded materials by naked eye and microphotography. Fig. 15 showsthree different eroded specimens after 50 h of simultaneous test-ing at an impact speed of 366 m/s. The specimens are made from

1614 M. Ahmad et al. / Wear 267 (2009) 1605–1618

hroug

T5dTciisnalt

t4doalot

tbnswtii

Fd

Fig. 15. Metallographic analysis of specimen #1 t

i6Al4V, the laser treated X5CrNiCu 16-4 and X5CrNiMoCuNb 14--equiv. They are enumerated from #1 through #3 for furtheristinction. To the left, a general top view impression is given.he main pictures in the middle show microphotographs of theorresponding cutting surfaces. Here, only a minor magnifications utilised for an assessment of the surface texture by engineer-ng judgement. Clearly visible is an increasing jaggedness of theurfaces from top to bottom. While the titanium specimen showsearly a negligible macroscopic change of the surface texture andppears merely sand-blasted, the X5-equiv shows distinct craterike structures with a diameter of 300–400 �m uniformly dis-ributed all over the surface.

In order to make the foregoing surface perception more objec-ive, stylus based pertometer measurements according to EN ISO287 [40] are employed. According to the standardised proce-ure, the surface texture is analysed on a linear testing lengthf 5.6 mm along the eroded specimen surface which representspproximately half of a specimen’s diameter. Hence, the testingength is approximately one order of magnitude larger than thebserved macroscopic surface structures’ main scale derived fromhe preceding microphotography.

Given the large variety of possible evaluation criteria accordingo EN ISO 4287 [40], a combined vertical surface texture criterionased on the overall primary profile height Pt and the overall rough-ess amplitude Rt is employed for the assessment. This inferredurface property shall be designated as “jaggedness”. A cross-check

ith microphotography shows that this vertical jaggedness (being

he sum of Pt and Rt) represents the vertical as well as the hor-zontal macroscopic surface disturbances, generated by dropletmpact erosion, quite well. The jaggedness values for the three

ig. 16. Jaggedness of eroded specimen surfaces after 50 h of testing correlated toroplet impact speed.

h #3 for extended erosion resistance assessment.

depicted visually inspected specimen cutting surfaces are given inFig. 15.

From a statistical viewpoint, any surface texture may be treatedas scatter or variance of the integral volume loss. The latter may beinterpreted as the mean value of the area specific material degra-dation (see schematic drawing in the top left of Fig. 16). Havingdefined the jaggedness, this allows for further surface analyseswith respect to impact speed. Fig. 16 shows that the mean jagged-ness increases from about 320 �m to about 410 �m with increasingimpact speed within the investigated speed range for all testedmaterials. It is interesting to note that the jaggedness increases byapproximately 30% on average while the droplets are, if not con-stant, slightly decreasing in size (Fig. 8). Relative to the droplets’Sauter mean diameter, the main surface features are four to fivetimes larger which in turn supports the initially stated erosion sat-uration hypothesis.

4. Erosion assessment and prediction

4.1. Erosion assessment

Accepting that a pristine specimen features neither volume lossnor a significant surface texture eventually leads to an erosionassessment criterion chart as depicted in Fig. 17. In addition, itshall be tacitly presupposed that the jaggedness itself saturateswith increasing impact speeds while keeping the droplet size dis-tribution constant. In the resulting chart, for the four materials(X5CrNiCuNb 16-4, X5-laser treated, X5CrNiMoCuNb 14-5-equivand Ti6Al4V) the surface jaggedness is depicted as a function ofnormalised area specific volume loss after 50 h of testing at differ-ent impact speeds with the impacting droplets’ mean diameter keptconstant. Again, the specimen #1 through #3 as shown in Fig. 15are marked for illustration.

While contemplating the chart, note the relative ranking in theregion of low volume loss values (abscissa). Compared to the X5base material, the other shown materials feature a higher level ofjaggedness while exhibiting the same integral volume loss. Thismay be interpreted such that the fraction of droplets impactinglocally perpendicular to the surface is reduced for the latter threematerials by surface texture modification. This in turn can be inter-preted as a preferable “self-shielding” characteristic. Note also theapproximately 40% higher final jaggedness level of Ti6Al4V com-pared to the laser treated X5CrNiCuNb 16-4. This is when thenormalised volume loss after 50 h approaches 1.0 which virtuallymeans that the droplet impact speeds are approaching higher val-

ues. This somewhat artificial observation may be verified sometimein the future by long-term low-pressure steam turbine field expe-rience.

Having investigated the surface textures of the different spec-imens, a “desirable surface condition” may be defined arbitrarily

M. Ahmad et al. / Wear 267 (2009) 1605–1618 1615

F n acco5

batbisana

cdhljtiifiemr

asciIrgstbibmaaa

jbdi

ig. 17. Surface primary profile height (Pt) and roughness height (Rt) characterisatio0 h of testing for the different blade materials.

y engineering judgement. In this example case, a maximum pass-ble surface jaggedness of 310 �m is chosen. In other words, theitanium specimen surface #1 depicted in Fig. 15 featuring a “sandlasted” jaggedness would be acceptable. The laser treated X5 spec-

men surface #2 would be just passable and the X5-equiv specimenurface #3 featuring distinct craters would just not be acceptableny more. Any higher jaggedness would be, in the present example,ot acceptable, either. A corresponding horizontal line designateds “erosion appearance criterion” is depicted in Fig. 17.

Changing to a turbine blade designer’s point of view, such ariterion would be chiefly motivated by aerodynamic efficiencyeliberations: the more pronounced the surface features are, theigher the associated efficiency loss would be estimated. The abso-

ute level of such an aerodynamic efficiency loss, though, may beudged to be small as the local Reynold’s number levels are low inhe vicinity of the affected blade leading edges. Krzyzanowski, fornstance, specifies a last row blade efficiency drop by heavy dropletmpact erosion including a noticeable shortening of the blade pro-le chord length to be less than 1%-point [41]. However, a detailedvaluation on a case-by-case basis is indispensable for final judge-ent whether or not a blade replacement is indicated for efficiency

easons.The second inferred criterion is the “volume loss criterion” that

ddresses the integral material loss due to droplet impact ero-ion. Besides having an influence on a turbine blade efficiency byhord length reduction, this criterion is related to the structuralntegrity and the dynamic characteristics of a turbine blade [41].t eventually leads to a retirement-for-cause design concept withespect to droplet erosion for the affected machine components. Ineneral, a low volume loss, for instance promoted by a high ero-ion resistance, is desirable. In the present example, the inferredhreshold, marked as a vertical line in Fig. 17, is arbitrarily set toe eight times the erosion resistance of X20Cr13 at standard test-

ng conditions. Any volume loss lower than this value is judged toe even more advantageous. Mapping this limit value to the testedaterials, the titanium specimen #1 depicted in Fig. 15 would be

cceptable. The laser treated X5 specimen #2 would be just pass-ble and the X5-equiv specimen #3 would just not be acceptableny more.

Of course, the area specific volume loss and, especially, theaggedness of an eroded specimen will most likely be influencedy the impinging water droplet size distribution. A shift of theroplet population to larger Sauter mean diameters may cause

ncreased values for the jaggedness of a given eroded material. Yet,

rding to EN ISO 4287 [42] in dependence of the gross area specific volume loss after

as we accept the droplet impact erosion resistance to be an inde-pendent macroscopic material property, we also commit ourselvesto well-defined and controlled assessment conditions, the dropletsize distribution being one of these. Hence, questions concerningdroplet sizes and numbers or, for example, metallurgic modifica-tions and heat treatments to a given material degenerate to a matterof transferability.

Relative to droplet impact erosion, the two assessment cri-teria introduced above together may define a preferred designspace for the combined engineering design task “materialselection + flow path geometry + thermodynamic operating condi-tions + countermeasures against erosion” for steam turbine rotorblades operating in wet steam. It is again important to note thatthis design task is not limited to material science only. However,material science can greatly contribute to the overall performanceof a specific turbine blade design.

4.2. Erosion resistance prediction approach

In the preceding chapters, different turbine blade steel materialsand a relevant blade titanium alloy has been investigated at a vari-ety of droplet impact conditions. On this basis, a preferable erosioncharacteristic with respect to volume loss and surface texture con-dition has been tentatively inferred. Now, a next logical step is toclassify the obtained results with respect to material design allow-ing for extrapolation of the erosion behaviour to other materialsor material conditions. It is intended that this deliberation finallyforms the basis for a material, design and operating condition basedsteam turbine blade erosion protection concept.

The present approach again takes advantage of the similaritytheory in order to reduce possible dependencies. Table 3 sum-marises a list of influencing parameters that affect the dropletimpact erosion resistance. This list is already significantly reducedin length by utilising known coherences and interrelationships like,for example, the common definition of a fluids bulk modulus orYoung’s elastic modulus, both incorporate the acoustic speed ofsound for the fluid and the target material, respectively. On theother hand, metallurgical influences may only be accounted bysensible grouping of materials.

The wetness dispersion, the droplets’ surface tension and theambient conditions (gaseous phase) are taken into account via theimpinging droplets’ Sauter mean diameter according to Eq. (3). Theliquid phase’s type, viscosity, temperature and impact velocity areincluded in the definition of the impact pressure pimpact according

1616 M. Ahmad et al. / Wear 267 (2009) 1605–1618

Table 3List of the parameters affecting the erosion resistance.

Parameter Dimension Description of parameter Classification

RE,50 [s/m] Erosion resistance according to Eq. (2) Research objective

pimpacta [kg/(m s2)] Impact pressure

Droplet ballisticsDSauter

a [m] Droplet Sauter mean diameter according to Eq. (3)˛ [degree] Impact angle

�la [kg/m3] Density of liquid

Liquid propertiesK [kg/(m s2)] Bulk modulus of impacting liquid

Atarget [m2] (Pristine) target area

Target properties

�s [kg/m3] Density of solidHV [kg/(m s2)] (Vickers) hardness of the target surfaceE [kg/(m s2)] Young’s elastic modulusU∗ [kg/(m s2)] Elastic resilience according to Eq. (5)UP ture

ti

airl

,�s

�l︸on

saala

Fi

r

t [kg/(m s2)] Toughnesst + Rt [m] Surface tex

a Parameter chosen for non-dimensioning.

o Eq. (1). The number of droplet impacts per unit area and time isncorporated into the definition of the erosion resistance.

The droplet impact pressure, the droplet Sauter diameter DSauterccording to Eq. (3), and the liquid density �l are taken as referenc-ng quantities for non-dimensionalisation. Identifying the erosionesistance RE,50 according to Eq. (2) to be the key quantity of interesteads to the following set of similarity parameters:

RE,50

√pimpact

�l︸ ︷︷ ︸reducederosion resistance

= f

⎛⎜⎜⎜⎝ ˛,

Atarget

D2Sauter

,pimpact

K︸ ︷︷ ︸liquid droplet characteristics

,Pt + Rt

DSauter,

pimpact

HV,

E

K︸ ︷︷liquid solid interacti

From an experimental point of view the situation now reads

uch that, concerning the liquid droplet characteristics, the impactngle ˛ and the droplet size influence (Atarget/D2

Sauter) can be reli-bly controlled to stay within a confined level of variance. Asong as groups of materials of similar metallurgical structuresre investigated, like in the present case the two groups (a) high

ig. 18. Droplet impact erosion resistance map of steam turbine blade materials; R∗E,50 c

ndices #1 through #3 denote specimen chosen for further surface assessment.

,HV

U∗r

,Ut

U∗r︸ ︷︷ ︸

solid characteristics

⎞⎟⎟⎟⎠ (6)

yield strength steels and (b) titanium, the same applies to thesolid-to-liquid elasticity ratio (E/K) and the density ratio (�s/�l).The jaggedness term (Pt + Rt/DSauter), chiefly relates to the tar-get’s surface texture that may only initially be controlled. As soonas the eroded surface evolves, this parameter is out of activeexperimental control and may be treated as being part of the solu-tion.

Though being greatly reduced in number, the remaining four

influencing similarity parameters constitute a standard problemin similarity theory. It is virtually impossible to separate theinfluences, meaning to keep three of them constant while inves-tigating the fourth one. However, in the present set-up, the fourparameters are of different significance to the reduced erosion

urves denote lines of constant normalised erosion resistance according to Eq. (2);

ar 267

riadtouttmrtr

tiotist(ttncimt

stsenmtre(itmhp

5

hXstiatf

•

M. Ahmad et al. / We

esistance. As water of a certain temperature will always be thempinging liquid of choice, the bulk modulus K will not vary to

significant extent. Hence, the pressure wave term (pimpact/K) iseemed to be sufficiently represented by the liquid–solid interac-ion term (pimpact/HV). By re-examining the employed definitionf the rated elastic resilience according to Eq. (5) and the tab-lated material data according to Table 2, it may be concludedhat X20Cr13 and titanium stay alone, to begin with, whilehe three copper precipitation hardening materials form a com-

on group. For the latter, slightly different hardness-to-resilienceatios are interpreted to contribute to controllable natural scat-er. A similar consideration applies to the toughness-to-resilienceatio.

Having performed this deliberation, the resulting Fig. 18 showshe reduced erosion resistance as a function of the main liquid–solidnteraction term (pimpact/HV) for all tested specimen keeping thether similarity parameters as constant as possible. It is interestingo note that all steel data, including X20Cr13, may be approx-mated by a single curve. Titanium may be approximated by aimilar but shifted curve featuring higher reduced erosion resis-ances at a specific “impact pressure to target hardness” ratiopimpact/HV). Whether or not the solid-to-liquid elasticity ratio (E/K),entatively suggested in the figure is the appropriate parametero describe a conceivable flock of different material curves stilleeds to be verified. However, the two depicted curves may alreadyonstitute a simplified but functional model to estimate dropletmpact erosion resistances of alternative steel and titanium blade

aterials in reference to the materials discussed in the presentext.

Finally, three lines of constant normalised erosion resistance areketched into Fig. 18. In addition, the specimens denominated #1hrough #3, that have been discussed in the preceding chapters arehown again for illustration. If now, like in the previously discussedxample, a normalised erosion resistance of R∗

E,50 = 8 is deemedecessary to fulfil a specific design requirement, the proposed chartay supply an answer to the specification for an “impact pressure to

arget hardness” ratio (pimpact/HV), smaller than 0.2 for steel mate-ials or smaller than 0.3 for titanium materials. This condition mayither be fulfilled by increasing the surface hardness of the targete.g. by local material exchange or surface treatments), or by reduc-ng the relative droplet impact speed being chiefly responsible forhe impact pressure (e.g. by adjustment of thermodynamic or geo-

etric boundary conditions). Hence, the suggested approach mayelp to evaluate different options during a machine specific designrocess.

. Conclusions

The erosion resistance characteristic of five different butighly relevant steam turbine blade materials, including X20Cr13,5CrNiCuNb 16-4, a laser-treated condition of X5CrNiCuNb 16-4, ateel similar to X5CrNiMoCuNb 14-5 and Ti6Al4V, has been inves-igated in an erosion test rig at different erosive conditions. Thesencluded a variation of droplet impact speed and droplet impactngle. Based on repeated volume loss measurements, surface tex-ure analyses and the given macroscopic material properties theollowing conclusions are drawn:

Liquid droplet impact erosion resistance is recommended tobe treated as an independent macroscopic material property.

Though being of niche character, this property is deemed nec-essary to supply a basis for an erosion endurable design ofmachine components operating in gaseous fluids carrying liquiddroplets that exhibit a significant velocity relative to the com-ponents’ structural boundaries. Prominent members of this class

(2009) 1605–1618 1617

of machine components are rotating blades of turbo-machines,like compressor fan blades or low-pressure steam turbine rotorblades.

• The necessity to accept an independent macroscopic materialproperty “droplet impact erosion resistance” leads to the neces-sity of a standard property assessment procedure. An efficientand functional assessment procedure, including the descriptionof a corresponding testing machine, is proposed and describedin a preceding chapter of the present text. The procedure as wellas the testing machine concept may easily be adopted by othermembers of the community.

• The full knowledge of the droplet size and number spectrum isvital for the comprehension and assessment of the droplet impacterosion process. In the case of the proposed testing procedure, theexact droplet spectrum necessarily needs to be known, well deter-mined and fully controlled. Relative to real machine applications,the feasible transfer of the findings to machine components willbe largely affected by the assumed or experimentally acquiredwetness dispersion.

• The droplet impact pressure as well as the droplet impact angleare confirmed to be driving parameters affecting a material’svolume loss by droplet impact erosion. Here, the volume lossincreases exponentially with increasing perpendicular compo-nent of the droplets’ impact velocity.

• The experimental findings suggest that the droplet impact ero-sion resistance of a material is partly predetermined by itshardness and resilience. This finding is used to derive a functionalerosion resistance prediction model with the help of a similarityapproach. The model is deemed sensible for being used duringpreliminary and comparative design considerations including thechoice of materials, necessary countermeasures against erosionand the thermodynamic design of power plant processes.

• Ti6Al4V is confirmed to exhibit a higher erosion resistance thanthe tested high yield strength blade steels within the investigatedmachine relevant droplet impact speed range.

• X5CrNiCuNb 16-4 in a laser-treated condition as being used incontemporary steam turbine bladings shows the most favourableerosion behaviour among the tested high yield strength bladesteels within the investigated machine relevant droplet impactspeed range.

Acknowledgements

The help of the Mr. Timo Steinhilber at Fraunhofer Gesellschaft,Stuttgart, in utilising the Malvern Particle Sizer is gratefullyacknowledged. This research work has been financed by the fed-eral state Baden-Wuerttemberg and the Siemens AG, Energy Sector,within a project KW21 “Research Network Power Plants for the 21stCentury”. The authors wish to thank Siemens AG, Energy Sector,Mülheim an der Ruhr, for the permission to publish this paper.

References

[1] F.J. Heymann, Liquid Impingement Erosion ASM Handbook, vol. 18, ASM Inter-national Material Park, OH, 1992.

[2] G.C. Gardner, Events leading to erosion in the steam turbine, Proc. Inst. Mech.Eng. 178 (1/23) (1963–1964) 593–623.

[3] R.I. Crane, Droplet deposition in steam turbines, Proc. Inst. Mech. Eng. C: J. Mech.Eng. Sci. 218 (August (8)) (2004) 859–870.

[4] M.J. Moore, R.W. Langford, J.C. Tipping, Research at C.E.R.L on turbine bladeerosion, Proc. Inst. Mech. Eng. 182 (Pt 3H) (1967–1968).

[5] J.A. Hesketh, P.J. Walker, Effects of wetness in steam turbines, Proc. IMechE 219(C) (2005) 1301–1314.

[6] O.A.v. Schwerdtner, H.-G. Hosenfeld, Developments for the prevention of bladeerosion in low pressure end stages, VGB PowerTech. 57 (4) (1977) 217–226.

[7] M.S. Akhtar, J. Black, M.J.C. Swainston, Prevention of steam turbine blade erosionusing stator blade heating, Proc. Inst. Mech. Eng. 191 (1977) 17/77.

[8] T.C.-T. Lam, R. Dewey, A study of erosion on L-0 turbine stages, IJPGC2003-40082, in: International Joint Power Generation Conference, 2003.

1 ar 267

[

[

[

[

[

[

[

[

[

[

[

[

[

[[

[

[

[

[

[

[

[

[[

[[

[

[

[

[

[

618 M. Ahmad et al. / We

[9] E. Honegger, Über den Verschleiß von Dampfturbinenschaufeln, BBC Mitteilun-gen, 1927.

10] J.E. Field, The physics of liquid impact, shock wave interactions with cavities,and the implications to shock wave lithotripsy, Phys. Med. B 36 (11) (1991)1475–1484.

11] S.S. Cook, Erosion by water hammer, Proc. R. Soc. Lond. 119 (July (783)) (1928)481–488.

12] F.J. Heymann, On the shock wave velocity and impact pressure in high speedliquid–solid impact, Trans. ASME J. Basic Eng. 90 (1968) 400.

13] G.P. Thomas, J.H. Brunton, Drop Impingement erosion of metals, Proc. R. Soc.Lond. 314 (January (1519)) (1970) 549–565.

14] M.B. Lesser, Analytic solutions of liquid drop impact problems, Proc. R. Soc.Lond. 377 (July (1730)) (1981) 289–308.

15] M.B. Lesser, J.E. Field, The impact of compressible liquids, Ann. Rev. Fluid Mech.15 (1983) 97–122.

16] J.E. Field, M.B. Lesser, J.P. Dear, Studies of two-dimensional liquid-wedge impactand their relevance to liquid-drop impact problems 401 (October (1821)) (1985)225–249.

17] N.L. Hancox, J.H. Brunton, The Erosion of Solids by the Repeated Impact of Liq-uid Droplets, Surface physics Cavendish Laboratory, University of Cambridge,1966.

18] F.P. Bowden, J.H. Brunton, The deformation of solids by liquid impact by Super-sonic speeds, Proc. R. Soc. Lond. 263 (October (1315)) (1961) 433–450.

19] Y.C. Huang, F.G. Hammitt, T.M. Mitchell, Note on shock-wave velocity in high-speed liquid-solid impact, J. Appl. Phys. 44 (April) (1973).

20] J.E. Field, J.P. Dear, J.E. Ogren, Effects of target compliance on liquid drop impact,J. Appl. Phys. 65 (2) (1989).

21] A. Smith, J. Caldwell, D. Pearson, D.H. McAllister, D.G. Christie, Physical aspectsof blade erosion by wet steam in turbines, Philos. Trans. R. Soc. Lond. 260 (July(1110)) (1966) 209–219.

22] E. Somm, Beurteilung der Erosionsgefahr in Niederdruckdampfturbinen, Brown

Boveri Mitt. 10–71 (1971) 458–472.

23] G.S. Springer, Erosion by Liquid Impact, John Wiley & Sons, 1976.24] Z. Ruml, F. Straka, A new model for steam turbine blade materials erosion, Wear

186–187 (1995) 421–424.25] J.A. Krzyzanowski, A.E. Kowalski, A.L. Shubenko, Some aspects of erosion pre-

diction of steam turbine balding, Trans. ASME 116 (April) (1994) 442.[

(2009) 1605–1618

26] J. Krzyzanowski, Z. Szprengiel, B. Weigle, On the erosion prediction of steamturbine blading, in: Proceedings of the 5th International Conference on Erosionby Solid and Liquid Impact, 1979.

27] J. Krzyzanowski, S. Marcinkowski, Z. Szprengiel, B. Weigle, Some new problemsof predicting steam turbine blading erosion, 1981.

28] J. Krzyzanowski, B. Weigle, H. Severin, Semiempirical criterion of erosion threatin modern steam turbines, J. Eng. Power (January) (1971).

29] J. Krzyzanowski, The correlation between droplet steam structure and steamturbine blading erosion, Trans. ASME (July) (1979).

30] B.-E. Lee, K.-J. Riu, S.-H. Shin, S.-B. Kwon, Development of a water droplet erosionmodel for large steam turbine blades, KSME Int. (September) (2002), Rev. 1.

31] R. Mack, P. Drtina, E. Lang, Numerical prediction of erosion on guide vanesand in labyrinth seal in hydraulic turbines, Wear 223–235 (1999) 685–691.

32] A.A. Torrance, Modelling abrasive wear, Wear 258 (2005) 281–293.33] N. Sürken, Systematic analysis of condensing flows in low pressure steam tur-

bines (in German), Doctoral thesis, RWTH Aachen, VDI Verlag, 2007.34] ANSYS CFX-Solver, Release 11.0:Theory.35] C. Baumgarten, H. Lettmann, G.P. Merker, Modelling of primary and secondary

break-up processes in high pressure diesel sprays, in: Proceedings of the 24thCIMAC World Congress, Kyoto, 2004.

36] L. Wang, X. Cai, X. Ouyang, J. Wei, Coarse water measurement in lp turbinesusing a novel integrated probe, in: Proc. ISROMAC-10, 2004.

37] J. Klaestrup Kristensen, I. Hansson, K.A. Mørch, A simple model for cavitationerosion of metals, J. Phys. D: Appl. Phys. 11 (1978).

38] F.J. Heymann, Toward Quantitative Prediction of Liquid Impact Erosion, STP474,ASTM, 1970, pp. 212–248.

39] B. Stanisa, O.A. Povarov, V.A. Rizenkov, Einfluss des Wassertropfenauftref-fwinkels auf den Erosionsvorgang beim Dampfturbinenschaufelmaterial, 1992(in German), BWK, Bd. 44, pp. 93–97.

40] EN ISO 4287, Geometrical product specifications–surface texture: Profile

method–Terms, definitions and surface texture parameters (ISO 4287: 1997);German version EN ISO 4278: 1998.

[41] J. Krzyzanowski, Tropfenschlagerosion und Erosionsschutzmaßnahmen inDampfturbinen, 1986 (in German) BWK Bd. 38, pp. 527–533.

42] de P. Haller, Unsuchungen über die durch Kavitation hervorgerufnen Korrosio-nen, Schweiz. Bauztg. 101 (1933), pp. 243–246, 260–264.