Werkstofftechnische Berichte | Reports of Materials Science and Engineering Mustafa Mamduh Mustafa Awd Machine Learning Algorithm for Fatigue Fields in Additive Manufacturing Werkstofftechnische Berichte | Reports of Materials Science and Engineering Reiheherausgegebenvon FrankWalther,LehrstuhlfürWerkstoffprüftechnik(WPT),TUDortmund, Dortmund,Nordrhein-Westfalen,Deutschland In den Werkstofftechnischen Berichten werden Ergebnisse aus Forschungspro- jekten veröffentlicht, die am Lehrstuhl für Werkstoffprüftechnik (WPT) der Technischen Universität Dortmund in den Bereichen Materialwissenschaft und Werkstofftechnik sowie Mess- und Prüftechnik bearbeitet wurden. Die Forschungsergebnisse bilden eine zuverlässige Datenbasis für die Konstruktion, Fertigung und Überwachung von Hochleistungsprodukten für unterschiedliche wirtschaftliche Branchen. Die Arbeiten geben Einblick in wissenschaftliche und anwendungsorientierte Fragestellungen, mit dem Ziel, strukturelle Integrität durch Werkstoffverständnis unter Berücksichtigung von Ressourceneffizienz zu gewährleisten. Optimierte Analyse-, Auswerte- und Inspektionsverfahren werden als Entschei- dungshilfe bei der Werkstoffauswahl und -charakterisierung, Qualitätskontrolle und Bauteilüberwachung sowie Schadensanalyse genutzt. Neben der Werkstof- fqualifizierung und Fertigungsprozessoptimierung gewinnen Maßnahmen des Structural Health Monitorings und der Lebensdauervorhersage an Bedeutung. Bewährte Techniken der Werkstoff- und Bauteilcharakterisierung werden weit- erentwickelt und ergänzt, um den hohen Ansprüchen neuentwickelter Produk- tionsprozesse und Werkstoffsysteme gerecht zu werden. Reports of Materials Science and Engineering aims at the publication of results ofresearchprojectscarriedoutattheChairofMaterialsTestEngineering(WPT) at TU Dortmund University in the fields of materials science and engineering as wellasmeasurementandtestingtechnologies.Theresearchresultscontributetoa reliable database for the design, production and monitoring of high-performance products for different industries. The findings provide an insight to scientific and applied issues, targeted to achieve structural integrity based on materials understanding while considering resource efficiency. Optimized analysis, evaluation and inspection techniques serve as decision guid- ance for material selection and characterization, quality control and component monitoring, and damage analysis. Apart from material qualification and produc- tion process optimization, activities concerning structural health monitoring and servicelifepredictionareinfocus.Establishedtechniquesformaterial andcom- ponent characterization are aimed to be improved and completed, to match the high demands of novel production processes and material systems. Mustafa Mamduh Mustafa Awd Machine Learning Algorithm for Fatigue Fields in Additive Manufacturing MustafaMamduhMustafaAwd Dortmund,Germany PublicationasDoctoralThesisinFacultyofMechanicalEngineeringofTUDortmund University. Locationofdoctorate:Dortmund Dateoforalpresentation:05.09.2022 Chairman:Prof.Dr.MoritzSchulzeDarup 1.Reviewer:Prof.Dr.-Ing.habil.FrankWalther 2.Reviewer:Prof.Dr.-Ing.SebastianMünstermann Assessor:Prof.Dr.-Ing.Dr.-Ing.E.h.A.ErmanTekkaya ISSN2524-4809 ISSN2524-4817 (electronic) WerkstofftechnischeBerichte|ReportsofMaterialsScienceandEngineering ISBN978-3-658-40236-5 ISBN978-3-658-40237-2 (eBook) https://doi.org/10.1007/978-3-658-40237-2 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringer FachmedienWiesbadenGmbH,partofSpringerNature2022 Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher, whetherthewholeorpartofthematerialisconcerned,specificallytherightsoftranslation,reprint- ing, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physicalway,andtransmissionorinformationstorageandretrieval,electronicadaptation,computer software,orbysimilarordissimilarmethodologynowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthis publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexempt fromtherelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthors,andtheeditorsaresafetoassumethattheadviceandinformationinthis bookarebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernorthe authorsortheeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontained hereinorforanyerrorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwith regardtojurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations. This Springer Vieweg imprint is published by the registered company Springer Fachmedien WiesbadenGmbH,partofSpringerNature. Theregisteredcompanyaddressis:Abraham-Lincoln-Str.46,65189Wiesbaden,Germany Foreword Theimpactofprocess-specificcharacteristicssuchassurfaceroughness,remnant porosity, residual stresses, and microstructure on process-oriented characteristics of engineering materials has an effect on the property profile of emerging struc- tures. TU Dortmund University’s Chair of Materials Test Engineering (WPT) capitalizes on the promise of additive manufacturing methods by contributing tothedesignofnovelstructuresoptimizedforadditivemanufacturing.Structures utilizedintransportation,industry,medicalequipment,andelectroniccomponents aresusceptibletofatiguefailure.Thereisagrowingneedtoconnectcutting-edge experimentalcharacterizationtonumericalandartificiallyintelligenttoolsthatare probabilisticallygrounded.Additivemanufacturinghasdevelopedintoatransfor- mative digital manufacturing technique with a high degree of adaptability. The physics involved in this process chain are computationally too complex to grasp using conventional numerical techniques. The present work focuses on the application of experimental techniques, numerical methods, and digital twinning to intelligently map fatigue strength to selective laser melting process parameters across a given process window. The experiments are designed to get a better knowledge of the relationship between process-induced microstructural characteristics and the ensuing consis- tency under laboratory and arbitrary conditions. A novel technique is developed in which the fatigue cyclic deformation and crack propagation behavior are dig- itally twined using a Bayesian-based machine learning algorithm. The suggested machinelearning(ML)andBayesianstatisticstechniqueswereusedtoconstruct the defect-correlated evaluation of fatigue strength. It facilitated the mapping of v vi Foreword structural and process-induced fatigue features to a geometry-independent load density chart throughout a wide variety of fatigue regimes. Dortmund Prof. Dr.-Ing. Frank Walther October 2022 Preface ThisdissertationistheproductofmyworkasaheadoftheWorkgroupModeling and Simulation at the Chair of Materials Test Engineering (WPT) of TU Dort- mund University. At this point, I want to express my gratitude to everyone who contributed,directlyorindirectly,totherealizationofthiswork.Iwanttoexpress my deepest gratitude to my doctor father, Prof. Dr.-Ing. Frank Walther, head of the Chair of Materials Test Engineering (WPT), for his ever-lasting support and guidance through technical assistance and motivating mindset. I warmly appre- ciate the interest of Prof. Dr.-Ing. Sebastian Münstermann in my work and his helpfulandthoughtfulcomments.IwanttoexpressmygratitudetoProf.Dr.-Ing. A.ErmanTekkayaforhisinterestinthisworkandparticipationinthecommittee, and Prof. Dr. Moritz Schulze Darup for chairing the examination committee. I am eternally grateful to my wife, Lobna Saeed, and my mother, Nahla Abdelkhalek, for their unconditional support throughout a journey spanning so manyyears.Nottobeforgottenisthecontinuoussupportofmyformercolleague Mr. Shafaqat Siddique and current colleague, Mr. Mohamed Merghany. Mr. Shafaqat Siddique impacted my research residence and doctoral studies so posi- tively and cooperated with me so closely, especially in the early phases. Mr. Jan Johannsen at Fraunhofer IAPT provided excellent support in specimen manufac- turing and thermal measurements. I want to express my gratitude to the German Research Foundation (DFG) for funding the project entitled: “Mechanism-based understanding of functional grading focused on fatigue behavior of additively processed Ti-6Al-4V and Al-12Si alloys; project number 336368661.” based on which this dissertation was prepared. This work is dedicated to my wife, parents, and parents-in-law. I want to expressmygratitudetothemandmysonOmarforunderstandingmycontinuous absence over the past several months and for their invaluable personal support. vii viii Preface Finally,Iwanttoexpressmyever-lastingdebttomywife,Lobna,whohasbeen standingfast,loving,anddependablehelpineveryscenario.Theprospectofour future together has always motivated me. Dortmund Mustafa Mamduh Mustafa Awd October 2022 Abstract Fatigue failure occurs in structures used in transportation, industry, medical equipment, and electronic components. There is an increasing need to build a link between cutting-edge experimental characterization and probabilistically grounded numerical and artificially intelligent tools. Additive manufacturing has evolved as game-changing digital manufacturing technology with high modu- larity. Researchers and scientists appear to be maximizing the yield of this technologybycreatingphysics-basedcause-and-effectrelationships.Thephysics involvedinthisprocesschainiscomputationallyprohibitivetocomprehendusing traditional computation methods. The mean surface temperature of the mechan- ical testing specimens exhibits a strong positive gradient up to a certain depth. Theremeltingprocedureusedsuccessfullyreducedthenumberofporesandtotal defectvolumeinTi-6Al-4V.ScansusingX-raymicrocomputedtomographypro- vide pure two-dimensional images in angular increments. X-ray microcomputed tomography was used to assess the efficacy of the remelting technique. Keyhole pores were formed as the scanning speed increased, accompanied by a decrease in metallurgical pores. Using machine learning and Bayesian statistics, a defect- correlated estimate of fatigue strength was developed. AlSi10Mg and Ti-6Al-4V had a tensile strength that was higher than the cast and wire+arc counterparts, respectively. Fatigue failures arose from the surface and sub-surface flaws in Ti- 6Al-4V, whereas clusters of pores caused fatigue failures in most situations in AlSi10Mg. Since load is very low in the VHCF range, the crack propagation pathshowedlinearelasticcharacteristics.Deviationscouldbeduetomicrostruc- tural discontinuities which exist in AlSi10Mg but not Ti-6Al-4V. The extended finiteelementmethodandcontourintegraltechniquesinAbaqusareusedtostudy arbitrarycrackpaths.Whentheenergyreleaserateexceedsathresholdvalue,the cracks will propagate from defects. Crack propagation rate curves are used to ix