DTIC ADA527719: Classification of Corrosion Defects in NiAl Bronze Through Image Analysis

1.3 MB·English

by  Defense Technical Information Center

CorrosionSciencexxx(2010)xxx–xxx
ContentslistsavailableatScienceDirect
Corrosion Science
journal homepage: www.elsevier.com/locate/corsci
Classification of corrosion defects in NiAl bronze through image analysis
Ramana M. Pidapartia,*, Babak Seyed Aghazadeha, Angela Whitfieldb, A.S. Raob,1, Gerald P. Mercierb
aDepartmentofMechanicalEngineering,VirginiaCommonwealthUniversity,401WestMainStreet,Richmond,VA23284,USA
bDepartmentofAlloyDevelopmentandMechanics,CarderockDivision,NavalSurfaceWarfareCenter,9500MacArthurBlvd,WestBethesda,MD20817,USA
a r t i c l e i n f o a b s t r a c t
Articlehistory: Corrosionsurfacedamageintheformofpittingandmicro-cracksisobservedinmanymetals.Cracksusu-
Received22March2010 allyinitiatefromthepitsandgrowundercyclicstressesandeventuallyleadtomaterialfailure.Animage
Accepted7July2010 analysisbasedonwavelettransformsandfractalswasusedtostudythecorrosionmorphologyofnickel
Availableonlinexxxx
aluminumbronzemetalundervaryingcorrosionconditionsandappliedstresses.Imagefeatureparam-
eterswereextractedandanalyzedtoclassifythepits/cracksinthemetalsamples.Theresultsobtained
Keywords: indicatethatclassificationofpits/cracksispossiblewithimageanalysisandmaybeusedforcorrelating
A.Nickel
service/failureconditionsbasedoncorrosionmorphology.
B.SEM
(cid:2)2010ElsevierLtd.Allrightsreserved.
B.ModelingStudies
C.PittingCorrosion
C.AcidCorrosion
C.CorrosionFatigue
1.Introduction Micro-cracksusuallyinitiatefromthecorrosionpitsitesinthe
material. Under the interaction of cyclic load and the corrosive
Engineering components made from nickel aluminum bronze environment, cyclic loading facilitates the pitting process, and
(NAB)metalarebeingusedintheNavy’sshipandpropellercom- corrosionpits,actingasgeometricaldiscontinuities,leadtocrack
ponentsduetotheirhightoughness.However, thismetalissub- initiation and propagation and then final failure [6–9]. Corrosion
jected to pits and stress corrosion cracking due to the operating canleadtoacceleratedfailureofstructuralcomponentsunderfa-
conditionsofsaltwaterenvironment.Inordertoreducethemain- tigue loading conditions. Understanding and predicting corrosion
tenancerequirements,expediterepair,andextendthelifeofnaval damage morphology in the form of pits/cracks is therefore very
nickel aluminum bronze components, engineering studies are importantforthestructuralintegrityofmaterialsandstructures.
beingcarriedouttoinvestigatethemechanismofstresscorrosion ElectrochemicalprobetechniquessuchasScanningReferenceElec-
crackinitiationthroughexperimentsandmodeling. trode Technique (SRET), Scanning Vibrating Electrode Technique
Pittingandstresscorrosioncrackingareknowntobesomeof (SVET),andLocalizedImpedanceSpectroscopy(LEIS)[10–13]have
the major damage mechanisms affecting the integrity of NAB beenusedforelectrochemicalmeasurementsatthemetalsurface.
materialsandstructuresinnuclearandnavalengineering.Corro- There are no probe related techniques that quantitatively assess
sionpitsgenerallyinitiateduetosomechemicalorphysicalheter- thecorrosionmorphologyintheformofpitsandstresscorrosion
ogeneityatthesurface,suchasinclusions,secondphaseparticles, cracking in metals. Recently, Pidaparti et al. [14,15] developed a
flaws,mechanicaldamage,ordislocations.ManyNABalloyscon- computational methodology to predict stresses around pits and
tainnumerousconstituentparticles,whichplayanimportantrole foundthatpitinducedstressesreachesaplateauaftercertaincor-
incorrosionpitformation[1–5].Tobetterunderstandparticle-in- rosion time and may lead to crack initiation after that, and this
duced pitting corrosion, optical microscopy, Scanning Electron phenomenachangeswithcorrosionenvironment.
Microscopy (SEM) and Transmission Electron Microscopy (TEM) Image analysis has been used to characterize corrosion mor-
techniques have been used. Due to operating and special service phology in materials subjected to a variety of environmental
environments (e.g. saltwater), corrosion pits and stress corrosion conditions.Kapsalasetal.[16]proposedamethodtodetectcorro-
cracking are readily formed between the constituent particles sionsizeandtopologyinstoneworksurfacesbytestingandeval-
andthesurroundingmatrixinthesealloys[5]. uating image segmentation schemes. They have showed that
their results are in good agreement with assessments based on
chemical analyses performed on the same surfaces. Choi et al.
[17] have analyzed the corrosion surface damage using digital
* Correspondingauthor.Tel.:[phone];fax:[phone].
imageprocessingtechniques.TheyemployedHISmodelinterpre-
E-mailaddress:[email protected](R.M.Pidaparti).
1 Presentaddress:NuclearRegulatoryCommission,Washington,DC,USA. tation,co-occurrencematrix,andmultidimensionalscalingproce-
0010-938X/$-seefrontmatter(cid:2)2010ElsevierLtd.Allrightsreserved.
doi:10.1016/j.corsci.2010.07.017
Please cite thisarticle inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis,Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017
Report Documentation Page Form Approved
OMB No. 0704-0188
Public reporting burden for the collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and
maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information,
including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington
VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to a penalty for failing to comply with a collection of information if it

does not display a currently valid OMB control number.
1. REPORT DATE 3. DATES COVERED
MAR 2010 2. REPORT TYPE 00-00-2010 to 00-00-2010
4. TITLE AND SUBTITLE 5a. CONTRACT NUMBER
Classification of corrosion defects in NiAl bronze through image analysis
5b. GRANT NUMBER
5c. PROGRAM ELEMENT NUMBER
6. AUTHOR(S) 5d. PROJECT NUMBER
5e. TASK NUMBER
5f. WORK UNIT NUMBER
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Naval Surface Warfare Center Carderock Division,Department of Alloy REPORT NUMBER
Development and Mechanics,9500 Mac Arthur Blvd,West
Bethesda,MD,20817
9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSOR/MONITOR’S ACRONYM(S)
11. SPONSOR/MONITOR’S REPORT
NUMBER(S)
12. DISTRIBUTION/AVAILABILITY STATEMENT
Approved for public release; distribution unlimited
13. SUPPLEMENTARY NOTES
14. ABSTRACT
Corrosion surface damage in the form of pitting and micro-cracks is observed in many metals. Cracks
usually initiate from the pits and grow under cyclic stresses and eventually lead to material failure. An
image analysis based on wavelet transforms and fractals was used to study the corrosion morphology of
nickel aluminum bronze metal under varying corrosion conditions and applied stresses. Image feature
parameters were extracted and analyzed to classify the pits/cracks in the metal samples. The results
obtained indicate that classification of pits/cracks is possible with image analysis and may be used for
correlating service/failure conditions based on corrosion morphology.
15. SUBJECT TERMS
16. SECURITY CLASSIFICATION OF: 17. LIMITATION OF 18. NUMBER 19a. NAME OF
ABSTRACT OF PAGES RESPONSIBLE PERSON
a. REPORT b. ABSTRACT c. THIS PAGE Same as 6
unclassified unclassified unclassified Report (SAR)
Standard Form 298 (Rev. 8-98)
Prescribed by ANSI Std Z39-18
2 R.M.Pidapartietal./CorrosionSciencexxx(2010)xxx–xxx
duretocharacterizeimagesbythreecategoriesofcolour,texture, Cpþ1 ¼XXgðmÞhðnÞCp
4kþ2;ði;jÞ k;ðmþ2i;nþ2jÞ
andshapefeatures.Wangetal.[18]usedwaveletpacketdecom- m n
positionenergiesofimagesatdifferentwaveletsub-bandsasfea- Cpþ1 ¼XXgðmÞgðnÞCp ð1Þ
4kþ3;ði;jÞ k;ðmþ2i;nþ2jÞ
turestostudytheatmosphericcorrosionbehaviorofzincsamples. m n
Theyacquiredarelationshipbetweentheselectedimagefeatures
whereC0istheimageAandkisanindexofthenodesinthewave-
and the corrosionweightloss. Tao etal. [19]analyzed theatmo- 0
letpackettreedenotingeachsub-band.handgarethefiltercoef-
sphericcorrosionoffieldexposurehighstrengthaluminumalloys.
ficients of low-pass and high-pass filters, respectively. Supposing
They used wavelet packed decomposition energies at different
that Haar basis has been used, h¼f(cid:3)0:7071;0:7071g; and g¼
sub-bandstoestimatethecorrosionlossoffivetypesofaluminum
f0:7071;0:7071g: At each step, the image Cp is decomposed into
alloys. It was found that the values of sub-image energies de- k
fourquarter-sizeimagesCpþ1,Cpþ1 ,Cpþ1 ,andCpþ1 :Thecapitallet-
creased with increasing exposure time. There are no studies in 4k 4kþ1 4kþ2 4kþ3
ters (N or M) are maximum constants defined by the image size.
theliteratureonclassificationofpits/cracksinNABmetalsunder
However,smallletters(morn)aredefinedateachstep.Forexam-
combinedcorrosionandcyclicstresses.
ple,whenp=5,pcanbe1,2,3,or4andsoon.
In order to qualitatively and quantitatively characterize the
TheShannonentropyindifferentsub-bandsiscomputedfrom
early stage damage mechanisms in NAB metals under corrosion
thesub-bandcoefficientmatrixas:
and cyclic stresses, image analysis of corrosion morphology may
beusedasatooltopredictthenucleationofcracksresultingfrom Entropy ðkÞ¼(cid:3)XXjCp j2logjCp j2 ð2Þ
p k;ði;jÞ k;ði;jÞ
initiationandgrowthofpits.Inthisstudy,imageanalysisbasedon i j
wavelet transforms and fractals was used to study the corrosion
morphologyofNickelAluminumBronzemetalundervaryingcor- Energy ðkÞ¼XXjCp j2 ð3Þ
rosionconditionsandappliedcyclicstresses.Imagefeatureparam- p k;ði;jÞ
i j
eterswereextractedandanalyzedtoclassifythepits/cracksinthe
NABmetalsamples. whereEnergypðkÞandEntropypðkÞaretheenergyandentropyofthe
image projected onto the subspace at node ðp;kÞ. The entropy of
each sub-band providesa measure of the image characteristics in
2.Imageanalysismethods
that sub-band. The energy distribution has important discrimina-
tory properties for images and as such can be used as a feature
Severalimageanalysismethodsareavailableforcharacterizing
fortextureclassification.Fromtheequationsabove,itfollowsthat
the corrosion surface morphology. In this study, wavelet trans-
thewaveletentropyisminimumwhentheimagerepresentsanor-
formsandfractalswereusedfortheimageanalysisbasedonour
dered activity characterized by a narrow frequency distribution,
experience,andaredescribedbelowbriefly.
whereastheentropyishighwhenanimagecontainsabroadspec-
trumoffrequencydistribution.
2.1.Waveletpackettransform
Inthisstudy,firstlytheSEMimagesofthesurfaceshavebeen
transformedfromTIFtoJPGtoexporttoMATLABsoftware.Haar
In order to improve the images obtained from microscopy
motherwaveletwasemployedtodecomposethegrayscalecorro-
techniques such as SEM and others, multi-resolution wavelet
sion images to the second level of decomposition and Shannon
transformscanbeused.Thedetailsofstandardwavelettransfor-
entropiesofsub-imagesateachsub-bandhavebeenusedasimage
mationscanbefoundintheliterature[20].Asanextensionofthe
features.Theprocesshasbeendoneusingthewavelettoolboxof
standardwavelets,waveletpackets(WPs)whichrepresentagener-
MATLABsoftwareonaPentiumDual-Coreand1.66GHzprocessor
alization of the multi-resolution analysis use the entire family of
speedPCcomputer.
sub-band decompositions to generate a complete representation
ofimages.Therearemanyoutstandingpropertiesofwaveletpack-
2.2.Fractalanalysis
ets, which encourage researchers to employ them in widespread
fields.Ithasbeenshownthatsparsityofcoefficients’matrix,com-
Thefractaldimension(FD)isamathematicalconcepttoclassify
putationalefficiency,multiresolutionandtime–frequencyanalysis
certain sets in more detail than the topological dimension can.
canbeusefulindealingwithmanyengineeringproblems.Thehier-
There are many interesting and aesthetic examples of such sets
archicalWPtransformusesafamilyofwaveletfunctionsandtheir
(knownas‘‘fractalsets”)thatcanbecreated.Manyscientistsinlife
associatedscalingfunctionstodecomposetheoriginaltimeseries/
and material sciences and engineering use the fractal dimension
signal into subsequent sub-bands. The decomposition process is
(FD)asaparametertocharacterizeroughlinesorsurfaces.Inother
recursivelyappliedtoboththelowandhighfrequencysub-bands
words,thefractaldimensionisameasureofthemorphology,tex-
togeneratethenextlevelofthehierarchy.
ture,androughnessinthesurface(orimages).FDcannumerically
Two-dimensional discrete wavelet packet transform (DWPT)
characterizethevariationinsurfacestructurecausedbycorrosion,
decomposition allows us to analyze an image simultaneously at
which corresponds to morphology changes in grey value images
differentresolutionlevelsandorientations.In2-DDWPTmethod,
capturedbymicroscopytechniquessuchasSEMorAFM.Various
animageisdecomposedintooneapproximationandthreedetail
methodshavebeenproposedtoestimatefractalvalues[21]:Fou-
images.Theapproximationandthedetailimagesarethendecom-
rier,Kolmogorov,Korcak,Minkowski,rootmeansquare,slitisland,
posed into a second-level approximation and detail images, and
etc. These methods differ in computational efficiency, numerical
the process is repeated. The standard 2-D DWPT can be imple-
precisionandestimationboundary.However,astrongcorrelation
mentedwithalow-passfilterandahigh-passfilter[20].The2-D
was reported between the relative ranking of fractal values ob-
DWPT of an N(cid:2)M discrete image A up to level Pþ1 ðP6
tainedfromdifferentfractalmeasuringtechniques[21,22].Among
minðlog N;log MÞÞ is recursively defined in terms of the coeffi-
2 2 thesemethodstheFourieranalyticaltechniqueisthemostprom-
cientsatlevelpasfollows:
isingoneinthatithasseveraladvantages.First,theFouriermeth-
Cpþ1 ¼XXhðmÞhðnÞCp od is relatively insensitive to the presence of noise in images;
4k;ði;jÞ k;ðmþ2i;nþ2jÞ
m n second,thereexistsafastalgorithm,theFFT,whichprovideseffi-
Cpþ1 ¼XXhðmÞgðnÞCp cientimplementationparticularlyforonlineapplication;andthird,
4kþ1;ði;jÞ k;ðmþ2i;nþ2jÞ
m n the computationof fractal values is based on an explicit formula
Please cite this article inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis, Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017
R.M.Pidapartietal./CorrosionSciencexxx(2010)xxx–xxx 3
and is potentially more accurate computationally. For these rea-
sonsFourieranalysiswasadoptedtoestimatefractalvaluesinthis
study.
Forasurfaceimagefðx;yÞ,thepowerspectraldensity(PSD)is
calculatedas:
Sðu;vÞ¼jFðu;vÞj2; ð4Þ
whereF(u,v)istheFouriertransformoff(x,y),anduandvarethe
spatialfrequencies(numberofwavesperunitwavelength)inthex
and y directions, respectively. The PSD Sðu;vÞ is converted to the
polarcoordinatesystemS(f)suchthatf ¼pffiuffiffi2ffiffiffiþffiffiffiffiffivffiffiffi2ffiffi.Thevaluesof
SðfÞ,ateachradialfrequencyf,areaveragedoverangulardistribu-
tions.Forafractalsurface,thepowerspectrumshowsalinearvar-
iation between the logarithm of SðfÞ and the logarithm of the
frequencyf[23].Theslopeofthelinearregressionlinebisrelated
totheFDbyequation[24]:
FD¼ð7(cid:3)bÞ=2 ð5Þ
Inthisstudy,FDanalysiswascarriedoutonSEMimagesofcor-
rodedspecimensundervaryingconditionswithagoalofdifferen-
tiatingpitsandcracksinthem.
3.Experimentaldata
Nickel aluminum bronze alloy metal samples were cut into
50mm(cid:2)5mm(cid:2)1mm dimensions from a plate. The chemical
composition of the NAB plate (wt.%) [1] is 8.7–9.5% Al; 4.0–4.8%
Ni; 3.5–4.3% Fe; 1.2–2.0% Mn; 0.1% Si; 0.02% Pb and the rest is
Cu with an average tensile strength of 731MPa and Rockwell B
Hardness of 203 (HRB). The NAB plate’s typical microstructure
consistsofcontinuousequiaxedcrystalswithsmallareasofmeta-
stablebphases.kPhaseprecipitatesarefoundatgrainboundaries
in a phase and b phases areas [1]. Initially, the specimens were
firstinspectedvisuallyanddiamondpolisheduptoasurfacefinish
of 3lm before the corrosion testing. These specimens were
immersed in 10% ammonia (NH OH) (by vol.) – 90% seawater
4
(by vol.) prepared as per ASTM D1141 standard and the cyclic
stress was applied under three point bending according to ASTM
G39.Thechemicalcompositionofthesimulatedseawaterisasfol-
lows: initially salt solution was prepared using salts NaCl
0.413moldm(cid:3)3; MgCl 0.0546moldm(cid:3)3; Na SO 0.0288mol
2 2 4
dm(cid:3)3;CaCl 0.0105moldm(cid:3)3;andKCl0.0009moldm(cid:3)3,anddis-
2
solving them in 1000mL of de-ionized distilled water. Concen-
tratedammoniasolutionof100mLwasthenaddedto900mLof
thesaltsolutiontoobtainthesimulated.
Afterexposingthespecimensforaspecificcorrosiontime(1–
37days) under specified cyclic stress (0–310MPa), the samples
wereexaminedunderSEMtocharacterizethesurfacemorphology.
TheexperimentswereconductedbytheresearchersattheDepart-
ment of Alloy Development and Mechanics, Carderock Division,
NavalSurfaceWarfareCenter,Bethesda,Maryland.Theyprovided
the SEMimage data to VCU for further analysis. An investigation Fig.1. SurfacemorphologyofcorrodedNABspecimens.
of the SEM data for defects classification was carried out in this
study. The corrosion morphology of typical NAB specimens with
defects,pits,andcracksisshowninFig.1.TheseSEMimageswere
werethencalculatedateachsub-band.Fractaldimensionsofthe
usedintheimageanalysistodevelopfeaturestocharacterizethe
imagesaswellastheShannonentropiesofwaveletpacketdecom-
pits/cracksinNABmetal.
positionsof imageswere usedas potentialfeatures. Fig. 2 shows
the Shannon entropies of four sub-images of wavelet packet
4.Resultsanddiscussion decomposition plotted against cyclic stress for samples after one
timestep.ItcanbeseenfromFig.2thatthemetalsamplesunder
4.1.Defectdetection the stresses of 170, 240, and 310MPa have higher entropies in
comparisonwithothersamplesandaremostlikelycandidatesto
Inordertostudytheeffectofcyclicstressonthecorrosionrate havedefects.Therefore,byobservingtheSEMimagescorrespond-
of samples, the images of samples were decomposed using 2D ingtothesestresses,defects,pits,andcrackscanberecognizedas
wavelet packet decomposition. Shannon entropies of sub-images shown in Fig. 1. Interestingly, these findings of no cracks, and
Please cite thisarticle inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis,Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017
4 R.M.Pidapartietal./CorrosionSciencexxx(2010)xxx–xxx
Fig.2. Shannonentropiesoffoursub-imagescalculatedfromwaveletimageanalysiswithvaryingappliedcyclicstressesofNABspecimens.
Cyclicstresslevelsof50,and240MPawerechosentoshowthe
effectofcorrosiontime onsurfacemorphologyofthe NABmetal
samples. Fractal Dimensions of images were calculated using Eq.
(5)andplottedagainstcorrosiontimeinFig.3.Itcanbeseenfrom
Fig.3thatthefractaldimensionsofcorrodedimagesofNABmetal
specimenshavearoughlyincreasingtrendwithtime.Also,when
SEM images with high fractal dimensions were observed, defects
in the samples were clearly visible, whereas SEM images with
low fractal dimensions did not show any visible damage on the
surfacesofmetalsamples.
4.2.Crack/pitclassification
Theclassificationofcracks/pitswasinvestigatedthroughimage
analysisofcorrodedspecimenSEMimages.Theprocedureusedto
establishthedefectclassification(pitsversuscracks)isshownin
Fig.4.Twelvesub-images,sixcontainingcracks,andsixcontaining
pits,wereanalyzed.Eachimage’ssizewas256(cid:2)256pixels.Fig.5
Fig.3. Fractaldimensioncalculatedfromfractalanalysiswithvaryingcorrosion
shows the typical images containing normal, pits, crack-pit, and
time(timestep–numberofcorrosiondays)underappliedcyclicstresses.
cracksextractedfromtheentirecorrodedspecimenundervarying
timeandcyclicstresses.Fractaldimensionsontheseimageswere
crackshasbeenobservedbyVasanthandHays[1]intheirstudies calculatedandpresentedinFig.6.ItcanbeseenfromFig.6(a)that
onNABmetalundervaryingenvironmentalconditions. fractal dimension increases for cracks and pits in comparison to
Fig.4. OverviewofthedefectclassificationprocedureforNABspecimens.
Please cite this article inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis, Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017
R.M.Pidapartietal./CorrosionSciencexxx(2010)xxx–xxx 5
Fig.6. (a)Fractaldimensionand(b)Shannonentropycalculatedfromtheimagesof
Fig.5classifyingvariousdefects.
Fig.5. Typicaldefects(normal(nodefect),pits,cracks-pits,andcracks)observedin Fig. 7. Two-dimensional projection of image features (fractal dimension and
NABmetalspecimensundercombinedcorrosionandcyclicstresses. Shannonentropy)forclassifyingpits/cracksinNABmetalsamples.
Please cite thisarticle inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis,Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017
6 R.M.Pidapartietal./CorrosionSciencexxx(2010)xxx–xxx
normal(nodefect)images.Also,theimageswithpitshavehigher References
fractaldimensionsincomparisontoimageswithcracks.Thesam-
plesarelinearlyseparableinthisfeaturespaceandathresholdcan [1] K.L. Vasanth, R.A. Hays, Corrosion assessment of nickel–aluminum bronze
(NAB)inSeawater,Corrosion2004,PaperNo.04294.
be chosen to distinguish cracks from pits and its transition. This
[2] F.M.Dersch,A.Cohn,UltrasonicInspectionofNickelAluminumBronzefor
showstheeffectivenessofthefractaldimensionfeatureintheclas- StressCorrosionCracking,NSWCCD-61-TR-2001-21,2001.
sificationofcorrosioncracksfrompitsandtheirtransition. [3] M.D. Fuller, S. Swaminathan, A.P. Zhilyaev, T.R. McNelley, Microstructural
Inaddition tofractaldimension,the Shannonentropies ofthe transformations and mechanical properties of cast nial bronze: effects of
fusionweldingandfrictionstirprocessing,Mater.Sci.Eng.463(1-2)(2007)
wavelet packet decompositions of images containing normal (no 128–137.
defects), pits, cracks, and combination of pit-crack cases (Fig. 5) [4] J.A. Wharton, R.C. Barik, G. Kear, R.J.K. Wood, K.R. Stokes, F.C. Walsh, The
werealsocalculated.ItcanbeseenfromFig.6(b)thatthecracks corrosionofnickel–aluminumbronzeinseawater,Corros.Sci.47(2005)3336–
3367.
havehigherentropiesincomparisontopitsincomparisontonor-
[5] K. Oh-Ishi, T.R.T.R. McNelley, Microstructural modification of as-cast NiAl
mal(nodefect)intheirimages.Ithasbeenobservedthatwithonly bronzebyfrictionstirprocessing,Metall.Mater.Trans.35(2004)2951–2961.
increasingcorrosion,therearevisiblepitsinSEMimages.Incon- [6] G.T.Burstein,C.Liu,R.M.Souto,S.P.Vines,Originsofpittingcorrosion,Corros.
Eng.Sci.Technol.39(2004)25.
trast, when the cyclic stress is increased, there are visible cracks
[7] S.I.Rokhlin,J.Y.Kim,H.Nagy,B.Zoofan,Effectofpittingcorrosiononfatigue
and pits in SEM images. To provide a better illustration, a two crackinitiationandfatiguelife,Eng.Fract.Mech.2(1999)425–444.
dimensionalprojectionofimages(normal(nodefects),pits,cracks, [8] S.Ishihara,S.Saka,Z.Y.Nan,T.Goshima,S.Sunada,Predictionofcorrosion
fatigue lives of aluminum alloy on the basis of corrosion pit growth law,
andcombinationofpit-crackcases)intothefeaturespaceispre-
FatigueFract.Eng.MaterStruct.29(2006)472–480.
sentedinFig.7fortheimagesconsideredinthisstudy.Itcanbe [9] M.R. Sriraman and R.M. Pidaparti, Life prediction of aircraft aluminum
seen from Fig. 7 that using the features of fractal dimension and subjected to pitting corrosion under fatigue conditions, J. Aircraft (7–8)
(2009)1253–1259.
Shannon entropy of the wavelet packet decomposition, cracks
[10] H.S. Isaacs, Initiation of stress corrosion cracking of sensitized type 304
andpitsarelinearlyseparableand,thus,easilyclassifiable. stainlesssteelinductilethiosulfatesolution,J.Electrochem.Soc.135(1988)
2180–2183.
[11] T.Suter,H.Bohni,Anewmicroelectrochemicalmethodtostudypitinitiation
5.Conclusions
onstainlesssteels,Electrochim.Acta42(1997)3275–3280.
[12] R. Oltra, V. Vignal, Recent advances in local probe techniques in corrosion
Animageanalysisbasedonwavelettransformsandfractalswas research–analysisoftheroleofstressonpittingsensitivity,Corros.Sci.49
(2007)158–165.
used to study the corrosion morphology of nickel aluminum
[13] C.H.Paik,R.C.Alkire,RoleofsulfideinclusionsonlocalizedcorrosionofNi200
bronzemetalundervaryingcorrosionconditionsandappliedcyc- inNaClsolutions,J.Electrochem.Soc.148(2001)276–281.
lic stresses. Image feature parameters were extracted from the [14] R.M.Pidaparti,R.R.Patel,CorrelationbetweencorrosionpitsandstressesinAl
alloys,J.Mater.Lett.62(2008)4497–4499.
experimental data obtained from SEM, and analyzed to classify
[15] R.M.Pidaparti,K.Koombua,A.S.Rao,Corrosionpitinducedstressesprediction
thepits/cracksinthemetalsamples.Theresultsobtainedfromim- fromSEMandfiniteelementanalysis,Int.J.Comput.MethodsEng.Sci.Mech.
ageanalysisindicatethatclassificationofpits/cracksispossibleas 10(2009)117–123.
[16] K.Y.Choi,S.S.Kim,Morphologicalanalysisandclassificationoftypesofsurface
theyarelinearlyseparablebasedonthefractaldimensionanden-
corrosiondamagebydigitalimageprocessing,Corros.Sci.47(2005)1–15.
tropyfeatures.Theapproachpresentedinthispapermaybeused [17] P. Kapsalas, M. Zervakis, P. Maravelaki-Kalaitzaki, Evaluation of image
forcorrelatingservice/failure conditionsbasedoncorrosionmor- segmentationapproachesfornon-destructivedetectionandquantificationof
phologyinmetals.However,moreresearchisneededtostudyfur- corrosiondamageonstonework,Corros.Sci.49(2007)4415–4442.
[18] S.Wang,S.Song,Imageanalysisofatmosphericcorrosionexposureofzinc,
ther the proposed methods/analysis based on images obtained Mater.Sci.Eng.A385(2004)377–381.
fromnon-destructiveinspectiondevicesaspartofroutineinspec- [19] L. Tao, S. Song, X. Zhang, Z. Zhang, F. Lu, Image analysis of atmospheric
tion and correlating the fractal dimension to service/failure corrosionoffieldexposurehighstrengthaluminumalloys,Appl.Surf.Sci.254
(2008)6870–6874.
conditions.
[20] S.Mallat,AWaveletTourofSignalProcessing,Academic,NewYork,1999.
[21] J.Zhang,P.Regtian,M.Korsten,Monitoringofdryslidingwearusingfractal
Acknowledgements analysis, in: 10th TC-10 IMEKO Conference on Technical Diagnostics, 9–10
June2005,Budapest,Hungary.
[22] J.Rawers,J.Tylczak,Fractalcharacterizationofwear-erosionsurfaces,Mater.
ThefirsttwoauthorsthankUSNationalScienceFoundationfor EngPerformance8(6)(1999)669–676.
supportingthisworkthroughaGrantDMR-0505496.Theauthors [23] J.C.Russ.FractalSurfaces.PlenumPress,NewYork,1994.
[24] H.Peitgen,D.Saupe,M.F.Barnsley,TheScienceofFractalImages,Springer-
alsoacknowledgethefinancialsupportfromDr.AiranPerez,Pro-
Verlag,NewYork,1988.
gramOfficer,OfficeofNavalResearch,Virginia.
Please cite this article inpress as: R.M.Pidaparti et al.,Classification of corrosion defects inNiAl bronze through image analysis, Corros.Sci. (2010),
doi:10.1016/j.corsci.2010.07.017