A computational approach to thermomechanical fatigue life predictions of dissimilarly welded superheater tubes

Contents

Acknowledgement

Zusammenfassung

Abstract

1 Introduction
1.1 Background
1.2 Thesis outline

2 Thermal effects on power plant steels
2.1 Typical cracking phenomena in weldments
2.2 Weld metallurgy
2.3 High temperature behavior of power plant steels
2.3.1 Fatigue
2.3.2 Creep
2.3.3 Creep-fatigue interaction

3 Viscoplastic deformation and TMF life prediction model
3.1 Deformation models
3.1.1 Simple material law
3.1.2 Viscoplastic material law
3.2 Mechanism based model for fatigue lifetime prediction

4 Welding and heat treatment of boiler tubes
4.1 Materials
4.2 Multipass welding of dissimilar boiler tubes
4.3 Post weld heat treatment of the welded tubes
4.4 Hardness distribution of welded and post weld heat treated tubes
4.5 Component test
4.6 Damage assessment of welded superheater tube
4.6.1 Detection of cracks
4.6.2 Metallographic analyses
4.7 Summary of the chapter

5 Experimental program for material characterization
5.1 Introduction
5.2 Relaxation experiment
5.3 Low cycle fatigue experiment (LCF)
5.4 Thermomechanical fatigue experiments on VM12 base material
5.5 Thermomechanical fatigue experiments on VM12/Alloy617 cross weld spec- imens

6 Parameter identification for deformation model
6.1 Parameter identification for viscoplasticity model
6.2 Validation through thermomechanical experiments
6.3 Parameter identification for stress relaxation model
6.4 Parameter calibration for heat affected zone
6.5 Summary of the chapter

7 Parameter identification for lifetime model
7.1 Parameter identification for mechanism-based lifetime model
7.1.1 Lifetime prediction of T91 base material and HAZ .
7.1.2 DTMF parameter identification for VM12
7.2 Manson-Coffin, Ostergen and SWT models
7.3 Summary of the chapter

8 Numerical simulation of welding and post weld heat treatment
8.1 Overview
8.2 Modelling of multipass welding process
8.2.1 Thermal analysis
8.2.2 Mechanical Analysis
8.3 Simulation of post weld heat treatment

9 Results of welding and PWHT simulations
9.1 Results of thermal simulation
9.2 Effects of PWHT on residual stresses
9.3 Summary of the chapter

10 Fatigue life prediction of welded superheater tubes
10.1 Influence of heat affected zone
10.1.1 Influence of heat affected zone on crossweld specimen model
10.1.2 Influence of heat affected zone on welded component
10.2 Influence of weld angle
10.2.1 Influence of weld angle on crossweld specimen model
10.2.2 Influence of weld angle on welded component
10.3 Influence of residual stress and distortion
10.3.1 Influence of residual stresses on TMF lifetime prediction . .
10.3.2 Influence of distortion on the thermomechanical fatigue lifetime
10.4 Summary

11 Discussion
11.1 Mechanism based TMF lifetime prediction
11.2 Applicability of mechanism based lifetime models .
11.3 Residual stresses after welding and PWHT
11.4 Failure assessment of welded component

12 Conclusion

Bibliography
A Classical life prediction models
A.1 Manson-Coffin and Basquin relationships
A.2 Ostergen damage function
A.3 Smith-Watson-Topper model
B Welding mechanics
B.1 Thermal analysis
B.2 Solid state phase transformation
B.3 Mechanical analysis
C Stresses in internally pressurized thick-walled tubes

Acknowledgement

This dissertation was written during my employment as Research Assistant at the Fraun- hofer Institute for Mechanics of Materials IWM in Freiburg, Germany. It would not have been possible to write this dissertation without the cooperation and constant support of the people around me

Firstly, it is an honor for me to express my deepest gratitude to my doctoral thesis advisor Prof. Dr. Hermann Riedel. Special thanks is also reserved to Prof. Dr. Volker Schulze from the Institute of Production Sciences at Karlsruhe Institute of Technology KIT, for co- advising this thesis

I would like to thank Dr. Dieter Siegele for giving me an opportunity to be a part of his Structural Integrity and Fracture Mechanics team. It was a great pleasure working with him and he made available his support in many ways

It is a pleasure to thank my office collegue Mrs.Valerie Friedman. Exchanging knowledge and experience with her was always constructive. I would also like to show my gratitude to Dr. Jörg Hohe whose door I always knocked to get hints and tips on troubleshooting numerical problems

I am indebted to many of my colleagues who supported me during the various phases of my work. Especially, I would like to thank Dr. Thomas Seifert, Martin Tandler, Markus Brand, Christoph Schweizer, Gerhard Maier, Yves Sguaizer, Michael Burdack, Sandra Meyerhoff, Lutz Reissig and Aruna Prakash

I would really like to express my deepest gratitude to Melanie Bloss, Dominik Suwito and Luciana Meinking as they created a home for me in Freiburg. Most importantly I would like to thank Iyas Khader for being a great friend with continuous motivation and support

Thanks to my teachers, God and the Universe for giving me a good family, relatives, true friends and showing me the right path throughout the journey of my life

Lastly, I express my regards to all of those who supported me in any respect during the completion of the dissertation

Zusammenfassung

Primäres Ziel dieser Arbeit ist die Lebensdauervorhersage von Schweißverbindungen unter thermomechanischer Ermüdung (TMF). Es wurden sowohl experimentelle als auch nu- merische Methoden eingesetzt, um die Lebensdauer von Mischverbindungen von Über- hitzerrohren für den Kraftwerkeinsatz vorherzusagen. Das Ermüdungsverhalten und die Lebensdauervorhersage derartiger Komponenten sind insbesondere bei hohen Tempera- turen sehr komplex, da hier eine Wechselwirkung zwischen thermisch aktivierten und zeitab- hängigen Prozessen auftritt. Außerdem unterliegen Komponenten bei hohen Temperaturen oft dem Einfluss transienter Temperaturen, die durch thermisch induzierte zyklische Span- nungen bei An- und Abfahrvorgängen auftreten. Die Überlagerung von thermischen tran- sienten mit mechanischen Lastzyklen, in Kombination mit einer ungleichmäßigen Erwär- mung der Komponententeile führt zu einer komplexen Schädigungsentwicklung und zu einem Versagen der Komponente durch thermomechanische Ermüdung. Darüber hinaus können Degradationsmechanismen in geschwei ßten Kraftwerkkomponenten durch mögliche Eigenspannungen noch beschleunigt werden

In dieser Arbeit wurde ein computergestützter Ansatz auf Basis von mehreren, aufeinan- der folgenden Simulationen entwickelt und für Lebensdauervorhersagen angewendet. Mit dieser Technik wurde das Verhalten der Mischverbindung aus dem neuen 9 − 12% Cr- Stahls (VM12) und der Nickelbasislegierung Alloy617, die für zukünftige Anwendungen in Kraftwerken vorgesehen ist, analysiert. Die einzelnen Simulationsschritte umfassen die Berechnung der Eigenspannungen von Mehrlagen-Schweißverbindungen, die Charakter- isierung der Materialeigenschaften aus Modellexperimenten, die Modellierung der Wärmee- influsszonen (HAZ) des VM12 Materials, die Berechnung der Eigenspannungsrelaxation durch die Simulation der Wärmebehandlung (PWHT) sowie die Lebensdauervorhersage bei thermomechanischer Ermüdung unter Berücksichtigung der relaxierten Eigenspannun- gen und des Schweißverzugs

Für die Berechnung der transienten Spannungen und Dehnungen bei thermozyklischer Belastung wurde ein zeit- und temperaturabhängiges zyklisches, viskoplastisches Mod- ell eingesetzt. Für eine gute Anpassung der Modellparameter wurde eine Reihe von Ex- perimenten mittels eines komplexen low cycle fatigue (CLCF) Programms durchgeführt. Ebenso wurden TMF Experimente mit zyklischen Temperaturwechseln durchgeführt, die die Situationen bei Kalt- und Warmstart in Kraftwerken repräsentieren. Schließlich wurden zur Parameteridentifikation des Wärmebehandlungsprozesses Relaxationsversuche bei ho- hen Temperaturen, die dem PWHT-Prozess entsprechen, durchgeführt

Die sequenziell gekoppelten Schweißsimulationen der Mehrlagen-Mischverbindungen zwis- chen VM12 und Alloy617 wurden entsprechend der Schweißreihenfolge der einzelnen Lagen durchgeführt. Mit den Schweißeigenspannungen als Anfangsbedingung erfolgte die Simulation des Wärmebehandlungsprozesses. Das Relaxationsmodell reproduziert das Spannungsrelaxationsverhalten von VM12 und Alloy617 sehr gut. Das Modell sagt ebenso vorher, dass nach PWTH die Eigenspannungen umgelagert aber nicht vollständig abgebaut sind

In dieser Arbeit wurde eine neue Methode für die Modellierung der HAZ auf der Ba- sis der Härteverteilung vorgestellt. Der Einfluss spezifischer Annahmen im HAZ Modell sowie des Schweißwinkels wurde an 3D Modellen der Schweißnaht und der HAZ sowie an 2D axialsymmetrischem Modell der Schweißnaht im Rohr analysiert. Die Ergebnisse der TMF Simulationen zeigten, dass die Einbeziehung der HAZ im Modell die vorhergesagten Lebensdauern nicht wesentlich beeinflusst, aber Unterschiede in der örtlichen Verteilung der größten Schädigung vorliegen. Der Schweißwinkel zeigt ähnlich wenig Einfluss auf die vorhergesagten Lebensdauern der beiden Modelle. Die Bauteilsimulationen zeigten inter- essante Ergebnisse. Die Analysen ergaben, dass die nach PWHT verbliebenen Eigenspan- nungen durch die thermomechnische Wechselbelastung in den ersten Lastzyklen weiter relaxieren. Die vorhergesagten Lebensdauern hängen daher davon ab, in welchem Lastzyk- lus der Schädigungsparameter ausgewertet wird. Bei der Verwendung der Spannungs- Dehnungs-Hysterese nach dem ersten Lastzyklus werden niedrigere Ermüdungslebensdaue- rn vorhergesagt als bei Verwendung späterer Lastzyklen, bei denen die Eigenspannungen weiter abgebaut sind. Ein weiteres wesentliches Ergebnis ist, dass die Analysen ohne Berücksichtigung des schweißbedingten Verzugs eine um den Faktor 3.5 höhere Zykle- nanzahl bis Schädigungsbeginn vorhersagen als die Analysen mit Berücksichtigung des Verzugs nach dem Schweißen

Basierend auf den Ergebnissen dieser Arbeit kann die vorgestellte Simulationskette erfol- greich für folgende Problemstellungen eingesetzt werden: zur Vorhersage von Schweiß- eigenspannungen, zur Abschätzung der Umlagerung von Eigenspannungen nach PWHT und zu Lebensdauervorhersagen unter TMF von geschweißten Komponenten, wobei die Entstehung der Eigenspannungen durch den Schweißprozess und der Abbau bzw. die Umverteilung der Eigenspannungen durch die Wärmebehandlung und die Belastung von besonderer Bedeutung sind. Die vorgestellte Berechnungsmethodik ermöglicht eine Lebens- dauerbewertung von Kraftwerkkomponenten und trägt damit zu einer erheblichen Reduzieru- ng von kosten- und zeitintensiven Experimenten und Komponententests bei

Abstract

The primary focus of this work is the thermomechanical fatigue (TMF) life predictions of dissimilarly welded joints. Both experimental and numerical methods have been adapted in order to predict the lifetimes of dissimilarly welded superheater tubing in power plants. The fatigue behavior and lifetime predictions of such components are complicated at high temperatures since a complex interaction between thermally activated, time-dependent pro- cesses are involved. At the same time, components that operate at elevated temperatures are often subjected to transient temperatures due to start-up and shut-down where thermally induced cyclic stresses can occur. The superposition of thermal transients with mechanical load cycles along with unequal heating of parts of a component result in a complex evolu- tion of damage, leading to TMF failure of the component. To add to the complexity, the degradation mechanisms relevant to welded power plant components can be accelerated by the presence of residual stresses

Hence, in this work a computational approach comprising a chain of simulations has been developed and presented for lifetime predictions. Herein, the behavior of the new 9-12% Cr steel (VM12) has been investigated when it is welded with the nickel based Al- loy617 for future power plant applications. The computational framework comprehends the calculation of residual stresses with multipass welding simulations, characterization of VM12 base material properties from model experiments, modeling of heat affected zone (HAZ) of VM12 material, computation of residual stress relaxation with post weld heat treatment (PWHT) simulation, and fatigue lifetime predictions taking into account the re- laxed residual stresses and welding distortions

A time and temperature dependent cyclic viscoplasticity model is used to calculate the transient stresses and strains under thermocyclic service loading. Series of experiments are performed to obtain a good model adjustment for the cyclic viscoplasticity model by making use of a complex low cycle fatigue (CLCF) program. Likewise, TMF experiments are carried out in order to take into account the temperatures due to cold and warm start in power plants. Additionally, relaxation tests are carried out at high temperatures considering the temperature spectra of the PWHT process

A sequentially coupled multipass welding simulations of dissimilar superheater tubes between the VM12 and Alloy617 are performed in two steps. Thereafter, PWHT simula- tions are performed using the welding residual stresses as initial condition. The relaxation model reproduced the stress relaxation behavior of VM12 and Alloy617 quite well. The model also predicted that after the PWHT simulation, the residual stresses are redistributed but not relaxed completely

A novel method for modeling the HAZ, based on hardness distribution has been pro- posed in this work. The influence of specific assumptions in the HAZ model and the influ- ence of weld angle are investigated on crossweld gauge length models and 2D axisymmetric welded tube models. Preliminary results of the TMF simulations showed that the incorpo- ration of HAZ do not influence the predicted lifetimes significantly but difference on the location of highest damage within the HAZ of VM12 tube. Similarly, the variation of weld angle has little significance for the predicted lifetimes of both the models Component simulations have revealed interesting outcomes in this work. To start with, the simulations show that the residual stresses that remain after PWHT futher relax in the first loading cycles. The predicted lifetimes depend on the residual stresses and thus, on the choice of loading cycle in which the damage parameter is evaluated. If the first loading cycle, where residual stresses are still present, is considered, lower fatigue lifetimes are predicted compared to predictions considering loading cycles with relaxed residual stresses. Another prominent result here is that calculations with undistorted geometry predicts 3.5 times more cycles to failure than those with distortions

Based on the knowledge acquired within this work, the chain of simulations presented can be ascertained as a successful method to predict residual stresses due to welding, to estimate residual stress redistribution due to PWHT and to predict TMF lifetime of dissimi- lary welded superheater tubes by laying emphasis on the residual stresses generated during welding and redistributed during PWHT. It allows the lifetime assessment of power plant components with a significant reduction of expensive and time-consuming experiments and component tests.

Chapter 1 Introduction

1.1 Background

The 9-12% Cr martensite superheater tubes are commonly used in fossil power plants due to the combination of high creep and fatigue strength and its moderate cost. The instal- lation and repair of such tubes involve the process of welding, which introduces changes of microstructures and residual stressses, specifically in the heat affected zones (HAZ), which can lead to the initiation of failure. Furthermore, dissimilar metal welds are used in power station boilers for joining such ferritic/martensitic steel tubes to austenitic steel tubes in the superheater/rehetaer sections. The austenitic steel tubing is used in the final stages of the superheater/rehetaer, where increased resistance to creep and oxidation is needed. The thermal cycles during dissimilar welding cause non-uniform heating and cooling in the material, thus generating inhomogeneous plastic deformation and residual stresses in the weldment.

Such welded power plant components must endure severe cyclic mechanical and ther- mal loads throughout their service life. Inhomogenous temperature distributions during thermal cycling and the interaction with adjacent components inhibit the thermal expansion and contraction of the components, resulting in local stresses and inelastic strains. The superposition of thermal transients with mechanical load cycles results in a complex evo- lution of damage, leading to thermomechanical fatigue (TMF) of the material. To add to the complexity, the degradation mechanisms relevant to power plant components can be accelerated by the presence of residual stresses. Tensile welding residual stresses in par- ticular can contribute to fatigue crack development in a structure even under compressive cyclic loading (CHENG 2003), (FRICKE 2005). Residual stresses also are known to affect fracture processes, (LIDBURY 1984), (READ 1989) and have been shown to accelerate the onset of creep damage (TURSKI ET AL. 2004), (BOUCHARD ET AL. 2004). Therefore, some researches claim that the inclusion of residual stresses is significant in power plant integrity assessment particulary in the prediction of fatigue lifetimes (AINSWORTH 2006) and (LEI ET AL. 2000).

The present work is written based on its contribution to the European COST536 project for "Alloy Development for Critical Components of Environmentally Friendly Power Plants". The main aim of the project was to investigate the behavior of dissimilarly welded new 9- 12% Cr steel (VM12) and nickel based alloy (Alloy617) for future power plant applications operating between 600 ◦ C and 650 ◦ C with a perspective of 700 ◦ C.

A computational approach is developed and presented for fatigue life prediction of dis- similarly welded superheater tubes. This systematic approach is then applied to a dissimilar weld between tubes made from the nickel based Alloy617 and the 9-12 % chromium steel. The computational framework comprising a chain of simulations is shown in Figure 1.1. The approach comprises the calculation of the residual stresses in the welded tubes with a multipass dissimilar welding simulation, the relaxation of the residual stresses in a post weld heat treatment (PWHT) simulation and the fatigue life prediction using the remaining residual stresses as initial condition. A cyclic viscoplasticity model after Chaboche is used to calculate the transient stresses and strains under thermocyclic service loadings. The fa- tigue life is predicted with a damage parameter DTMF which is based on fracture mechanics considerations. The adjustable parameters of the model are determined based on LCF and TMF experiments. The computational approach comprises the following steps, which are described thoroughly in this work:

1. Computation of residual stresses in a multipass welding simulation
2. Characterization of base material properties based on model experiments (LCF, TMF, relaxation)
3. Modeling of HAZ properties based on hardness measurements
4. Computation of the residual stress relaxation in a post weld heat treatment simulation
5. Fatigue life predictions taking relaxed residual stresses and welding distortions into account

The material data for VM12 are determined within this work. The material data for the nickel based Alloy617 have been taken from an existing database at Fraunhofer IWM. The Chaboche model and the DTMF fatigue lifetime model were adjusted to LCF and TMF experiments and thus applied in this work.

1.2 Thesis outline

This thesis is divided into twelve chapters and additional appendices. In Chapter 2, the literature review that follows this introduction, addresses the temperature effects on power plant steels and its influence on the weld metallurgy. The chapter is then completed by reviewing the high temperature behavior of power plant steels under operating conditions.

In Chapter 3, the constitutive equations of cyclic viscoplastic deformations to describe behavior of metals at elevated temperatures and mechanism-based model for TMF life predictions are presented.

Chapter 4 is devoted to the information concerning the material being investigated as well as the details about these processes. Details about the component test are presented

illustration not visible in this excerpt

Figure 1.1: Chain of simulations for multipass welding, post weld heat treatment and thermomechanical fatigue lifetime predictions

later in this chapter. The outcome of the component test concludes this chapter.

Chapter 5 deals with the experimental determination of VM12 material behavior. This chapter describes the basic experimental techniques in order to trace the relaxation, low cycle fatigue and the thermomechanical fatigue behavior of the VM12 base material. It also describes the thermomechanical fatigue experiment of crossweld specimens.

In Chapter 6, emphasis is placed on the characterization of the cyclic viscoplastic deformation model and the stress relaxation model for the VM12 base material. In addition, modeling methods for the VM12 HAZs are also presented.

Chapter 7 addresses the parameter identification procedure for the mechanism-based lifetime model used in this work. Applications to other classical models are also presented here.

The numerical simulation of welding and post weld heat treatment of the materials are outlined in Chapter 8. The results of these computations are presented and briefly summarized in Chapter 9.

The fatigue lifetime predictions are addressed in Chapter 10. Influence of HAZ model properties and weld angle on the lifetime predictions are first addressed here. Investigations on the influence of residual stresses and distortions are presented in this chapter. The outcome of these investigations are summarized at the end of this chapter.

In Chapter 11, the main points that emerged from the results out of this work are discussed. Besides the interpretation of the results, the reliability, validity and possible limitimations of the methods applied are described in detail.

Chapter 12 presents the conclusion of this work addressing the significance and practical applications of the methodology proposed in this work. A perspective for possible extension and future research are given at the end of this thesis.

Chapter 2 Thermal effects on power plant steels

2.1 Typical cracking phenomena in weldments

The process of fusion welding is necessary for the construction and repair of steel pipelines in power plants. Welding involves high thermal cycles which are responsible for the mi- crostructural evolution of the different metallurgical zones within the weld and heat affected zone (HAZ) and the mechanical properties which are related to its temperature history dur- ing welding. Before looking closely at the microstructural zones within the weld region and HAZ, it is convenient to consider possible failure modes in welded pipes under typical operating conditions. The cracking of welded joints is usually classified according to the position of the crack. Type I and type II modes occur within the weld metal, the former confined to the weld metal whereas the latter may grow out of the weld into the plate. Type III cracking occurs in the coarse grained region of the HAZ. Type IV is a pernicious form of cracking where there is an enhanced rate of creep void formation in the fine grained and in- tercritically annealed HAZ of the weld, leading to early failure when compared with creep tests on the unwelded steel (FRANCIS ET AL. 2006). A schematic representation of the long term failure modes in ferritic welds is shown in Figure 2.1 for CrMoV welds; similar results would be expected for other ferritic steels (BECKER ET AL. 2006).

illustration not visible in this excerpt

Figure 2.1: Classifications of cracking in weldment, (FRANCIS ET AL. 2006) 8

The microstructure at each location of the heat affected zone is closely related to the temperature-time characteristics of the welding process (KOHNO & JONES 1978). In addition, geometrical changes occur due to the difference in temperature and the gener- ated residual stresses during welding. Furthermore, residual stresses and microstructural changes influence the reduction in corrosion resistance of the material. Various chemical and physical processes also occur especially in the melted zone region during welding.

2.2 Weld metallurgy

The microstructure of a fusion weld may comprise seven distinct zones as shown in Fig- ure 2.2. In such a weld, the metal adjacent to the fusion zone is rapidly heated to tempera- tures approaching the melting point of the weld metal and then is cooled at rates determined by the conductivity and hence the mass of the surrounding metal. The microstructure that develops in the heat affected zone is determined by the thermal cycle, the kinetics of austen- ite formation, the grain growth kinetics, and the relevant continuous cooling transformation reaction. The welding thermal cycle produces peak temperatures and cooling rates that are highest at the fusion boundary. In a single pass weld in ferritic steel, four distinct regions can be identified in the HAZ alone (FRANCIS ET AL. 2006) (see Figure 2.2)

- Coarse grain region (CGHAZ) - material near the fusion boundary that reaches a temperature well above AC 3 during welding. Any carbides, which constitute the main obstacle to growth of the austenite grains, dissolve resulting in coarse grains of austenite. In 9-12% Cr steels, this austenite transforms into martensite on cooling.
- Fine grain region (FGHAZ) - away from the fusion boundary where the peak temperature TP is lower, but still above AC 3. Austenite grain growth is limited by the incomplete dissolution of carbides. Fine grained austenite is produced, which subsequently transforms into martensite in the 9-12% Cr steels.
- Intercritical region (ICHAZ) - Here AC 1 < TP < AC 3 , resulting in partial rever- sion to austenite on heating. The new austenite nucleates at the prior austenite grain boundaries, whereas the remainder of the microstructure is simply tempered. The austenite (in 9-12% Cr steels) tranforms into untempered martensite during cooling.
- Over tempered region - with TP below AC 1 the original microstructure of the material undergoes further tempering The HAZ of a multipass weld is even more complex, and, as a result of multiple heat treatments, may contain more than four regions. Mechanical properties can vary from zone to zone in an unpredictable manner, and modeling of the mechanical behavior of the com- posite configuration can be tedious for the stress analyst. In addition, the fusion line itself, being the last area to solidify, provides an ideal location for collection of impurities, and other debris.

As a result of the variable creep resistance of the base metal, weld metal, and HAZ, re- distribution of the initial elastic stresses can take place in complex ways. These redistribu- tions can be different for girth welds and for seam welds. It has been shown (CANE 1982)

illustration not visible in this excerpt

Figure 2.2: Schematic diagram of the unique heat-affected zone expected in a low alloy steel, (VISWANATHAN 1989)

that in girth welds the pressure stresses in the axial and hoop directions peak at the most creep-resistant location (i.e., the HAZ) and are the lowest at the least creep-resistant location (i.e., weld metal). In contrast, the work of WELLS ET AL. (1987) suggests that in seam welds the peak stresses after stress redistribution occur at the weakest location (i.e. the weld metal) whereas the lowest stress are encountered at the strongest location (i.e., the HAZ). Furthermore, the near-midwall cusp regions (at weld-bead intersections) at the fusion line of a double-V seam weld experiences high hydrostatic stresses and weld-metal creep resistance that is lower than that of the base metal.

Material property data pertaining to welds are very scanty. The limited data that are available relate only to pure base metal, only to weld metal or only to simulated HAZ ma- terial. What is needed is generation of data in the laboratory under conditions that simulate those encountered in service and yet are sufficiently accelerated to produce results in short periods of time. This is far from easy to do. Attempts to grow cracks in laboratory samples at locations similar to those of field failures have often been unsuccessful. Thus, neither the kind of data that needs to be generated nor the manner of obtaining it is readily apparent.

2.3 High temperature behavior of power plant steels

2.3.1 Fatigue

Isothermal fatigue

A metal subjected to repetetive or fluctuating stress will fail at a stress much lower than that required for failure on a singe application of load. Failures occurring under cyclic loading are termed fatigue failures. Fluctuating thermal stresses during start-stop cycles and due to power changes are examples of cyclic loading that can occur in a plant. For convenience, two types of fatigue are distinguished: high cycle fatigue (HCF) and low cycle fatigue (LCF). Although phenomenologically there is no distinction and clear-cut border between two types of fatigue, the traditional approach is to classify failures occuring above 10⁴ cycles as HCF and those occuring below that values as LCF.

Low-cycle fatigue, or high strain fatigue is typically of concern when there is significant cyclic plasticity. However, an important distinction between HCF and LCF is that in HCF most of the fatigue life is spent in crack initiation, whereas in LCF most of the life is spent in crack propagation, because cracks are found to initiate within 3 to 10% of the fatigue life. Traditionally, LCF tests are carried out in laboratory either with axial loading or bending, thus producing only tensile and compressive stresses where the strain is held constant and the stresses are allowed to vary. This procedure results in a plot of σ vs ϵ.

Material response to cyclic deformation

When a smooth solid specimen is subjected to one of the loading programs as shown in Figure 2.3, the response will differ depending on the mode of the controlled variable. The essential characteristics include cyclic hardening or softening, cyclic mean stress relaxation, cyclic creep, rate dependency and memory effects.

Thermomechanical Fatigue

Most low-cycle-fatigue problems in high temperature machinery involve thermal as well as mechanical loadings. In other words, the material is subjected to cyclic temperature simultaneously with cyclic stress. Analysis of these loadings and consideration of the at- tendance of fatigue damage process become very complex, and gross simplifications are often introduced. In the past, thermal fatigue traditionally has been treated as being syn- onymous with isothermal low-cycle fatigue at the maximum temperature of the thermal cycle (VISWANATHAN 1989). Life-prediction techniques also have evolved from the low- cycle fatigue literature. More recently, advances in finite-element analysis and in servohy- draulic test systems have made it possible to analyze complex thermal cycles and to conduct thermomechanical (TMF) tests under controlled conditions. The assumed equivalence of isothermal LCF tests and TMF tests has been brought into question as a result of a number of studies. It has been shown that for the same mechanical strain range, the TMF test can be more damaging under certain conditions than the pure LCF tests (VISWANATHAN 1989).

Thermal fatigue is defined as "the gradual deterioration and eventual cracking of a material by alternate heating and cooling during which thermal expansion is partially or fully constrained" (SPERA & MOWBRAY 1976). Constraint of thermal expansion causes thermal stresses which may eventually initiate and propagate fatigue cracks.

In TMF tests, the temperature and strain are varied independently. Two simple wave- forms in TMF tests are shown in Figure 2.4. If the maximum mechanical strain is applied at the lowest temperature, then it is called the out-of-phase (OP) TMF testing as shown Figure 2.4(a). On the contrary, if the maximum mechanical strain is applied at the highest temperature, it is then called the in-phase (IP) TMF testing as shown Figure 2.4(b). In components operated at high temperatures, constraints to expansion and contraction of an element arise from adjacent material elements which either are at a different temperature (e.g., surface vs interior of a rotor) or a made of different material (e.g., dissimilar welds in boiler tubes, and austenitic stainless steel cladding on ferritic steel in pressure vessels).

2.3.2 Creep

Due to thermal activation, materials can slowly and continuously deform under constant load (stress) and eventually fail. The time-dependent, thermally assisted deformation of components under load (stress) is known as creep. As a consequence of such deformation, unacceptable dimensional changes and distortions as well as final rupture of the component can occur. Depending on the component, the final failure may be limited either by deformation or by fracture. Local creep processes at the tip of a pre-existing defect or stress concentration can also lead to crack growth and eventual failure. Numerous texts deal in detail with the extensive body of literature pertaining to the phenomenology, mechanisms, and constitutive relationships for creep and creep fracture (GAROFALO ET AL. 1965), (BRESSERS 1981), (BERNASCONI & PIATTI 1979).

illustration not visible in this excerpt

Figure 2.3: Schematic material response to various modes of cyclic input variables, (ELLYIN 1997)

illustration not visible in this excerpt

Figure 2.4: Schematic diagram showing waveforms of temperature and strain in thermomechanical fatigue tests, (SEHITOGLU & MORROW 1982)

Creep properties are generally determined by means of a test in which a constant uniax- ial load or stress is applied to the specimen and the resulting strain is recorded as a function of time. The typical shape of creep curve is shown in Figure 2.5. After the instantaneous strain ϵ 0, a decelerating strain-rate stage (primary creep) leads to a steady minimum creep rate, ϵ (secondary creep), which is finally followed by an accelerating stage (tertiary creep) that ends in fracture at a rupture time, tr. The strain at rupture ϵ r, represents the rupture ductility.

illustration not visible in this excerpt

Figure 2.5: Schematic illustration of creep curve shapes

The shape of the creep curve is determined by several competing reactions, including (1) strain hardening, (2) softening processes such as recovery, recrystallization and strain soft- ening (3) microstructural changes, which may cause hardening or softening and (4) damage processes such as cavitation, cracking, and specimen necking (VISWANATHAN 1989). Of these factors, strain hardening tends to decrease creep rate whereas the other factors tend to increase the creep rate. The balance among these factors determines the shape of the creep curve. Secondary-stage creep is explained in terms of a balance between strain hard- ening and the strain softening and damage processes, resulting in a nearly constant creep rate. The tertiary stage marks the onset of internal or external damage processes (item 4), which result in a decrease in the resistance to load or a significant increase in the net section stresses. Coupled with softening processes (item 2), the balance achieved in stage 2 is now offset, and a rapidly increasing tertiary stage of creep is reached. Of all the parameters pertaining to the creep curve, the most important for engineering applications are ϵ and tr. Specifically, their dependence on temperature and applied stress are utmost interest to the designer. This dependence varies with the prevailing creep mecha- nism. A variety of mechanisms and equations have been proposed in the literature and have been reviewed elsewhere (GAROFALO ET AL. 1965), (BRESSERS 1981), (BERNASCONI

& PIATTI 1979). Fortunately, all these mechanisms can be fitted into two basic categories:

(1) diffusional creep and (2) dislocation creep.

Because creep is a thermally activated process, its temperature sensitivity would be expected to obey an Arrhenius-type expression, with an activation energy ” Q” for the rate- controlling mechanism. When secondary creep dominates, the secondary creep strain rate

illustration not visible in this excerpt

where Ao and n are constants and R is the universal gas constant. The values of n and the activation energy Q are sensitive to the process controlling creep. The region in which an individual process dominates can be obtained from a deformation-mechanism map.

Deformation-mechanism Maps

Deformation maps are diagram in the stress-vs.-temperature plane in which the areas are indicated where a particular creep-mechanism predominates (ASHBY 1972). The idea is that different mechanisms operate independently and the fastest one determines the defor- mation behavior and occupies the respective regime in the stress-temperature plane. The boundaries between the different regimes are usually calculated by comparing the strain rate equations for the different mechanims (RIEDEL 1987). Figure 2.6 schematically shows such a deformation map.

The greatest practical advantage of the deformation maps probably lies in the guidance which they provide in the extrapolation of creep data. Most creep tests in the laboratory are done within a year or less, sometimes within a few years, whereas engineering structures like electric power generating plants are designed to last for at least 25 years. The labo- ratory data can be extrapolated to the in-service conditions with some confidence only if both lie in the same regime of the deformation map. The maps are also useful tools for the designer to choose proper constitutive law for the stress analysis of a given part.

It has been observed that, in boiler piping, cavitation is the principal damage mecha- nisms at brittle zones, weld heat affected zones, and high stress regions in the base material (ELLIS 1988).

The first published attempt to relate creep-life consumption of plant components to cavitation was that of Neubauer and Wedel (NEUBAUER & WEDEL 1983). They charac- terized cavity evolution in steels at four stages i.e, isolated cavities, oriented cavities, linked cavities (microcracks) and macrocracks as shown in Figure 2.7. Based on extensive ob- servations on steam pipes in German power plants, they estimated the approximate time intervals required for the damage to evolve from one stage to the next under typical plant conditions. Using this experience, they formulated recommendations corresponsing to the four stages of cavitation also shown in Figure 2.7. Wedel and Neubauer built considerable conservatism into their recommendations and viewed their technique more as a monitoring technique than as a life prediction technique. Their procedure, because of its simplicity, has found worldwide support with power plant operators.

illustration not visible in this excerpt

Figure 2.6: Deformation mechanism map, (RIEDEL 1987)

2.3.3 Creep-fatigue interaction

In components which operate at high temperatures, changes in conditions at the beginning and end of operation or during operation result in transient temperature gradients. If these transients are repeated, the differential thermal expansion during each transient results in thermally induced cyclic stresses, the extent of the resulting fatigue damage depends on the nature and frequency of the transient , the termal gradient in the component, and the material properties. Components which are subject to thermally induced stresses generally operate within the creep range so that damage due to both creep and fatigue have to be taken into account.

In view of the importance of combined creep and fatigue damage with respect to com- ponent reliability, many attempts have been made to develop damage rules as well as com- ponent life prediction under creep-fatigue conditions. The principal method of studying creep-fatigue interaction has been to conduct strain-controlled fatigue tests with variable frequencies with and without holding periods (hold times) during some portion of the test. The lower frequencies and the hold times allow creep to take place. In pure fatigue tests, at higher frequencies and short hold times, the fatigue mode dominates and failures start near the surface and propagate transgranularly. As the hold time is increased, or the frequency decreases, the creep component begins to play a role with increasing creep-fatigue interac- tion. In this region, fractures are of mixed mode involving both fatigue cracking and creep

illustration not visible in this excerpt

Figure 2.7: Classification of creep damage in steam boilers, (VISWANATHAN 1989)

cavitation. With prolonged hold times with occasional interspersed cycles, creep processes completely dominate and can be treated almost as pure creep. In instances where oxidation effects contribute significantly to the creep-fatigue interaction, the situation is more complex than described above (JASKE 1988).

Chapter 3 Viscoplastic deformation and TMF life prediction model

In this chapter, the constitutive equations to describe the cyclic viscoplastic deformation of metals at elevated temperatures are presented. To begin with, the development of such equations is given from its elementary form. The first part of the chapter introduces a basic material law for plasticity, followed by an extension to a phemonelogical model for viscoplasticity. Subsequently, a mechanism-based lifetime model based on a fracture mechanics approach is presented. Both the viscoplastic deformation model and mechanism-based lifetime model are incorporated in this work to investigate the thermomechanical fatigue lifetime (TMF) of dissimilarly welded superheater tubes.

3.1 Deformation models

3.1.1 Simple material law

In classical models which are used to describe the high temperature behavior of metallic materials, the inelastic strain rate is divided into time-independent plastic strain rate and time-dependent creep strain rate. The time-independent, elastic-plastic deformation is described by the Ramberg-Osgood equation (RAMBERG & OSGOOD 1943):

illustration not visible in this excerpt

where K and N are Ramberg-Osgood parameters.

The time dependent creep strain rate is described by the simple Norton creep law for dislocation creep (NORTON 1929):

illustration not visible in this excerpt

where A is the adjustable parameter and n is the Norton exponent. The Norton material law is simple and can be used to describe the secondary creep stage quite well (NORTON 1929).

3.1.2 Viscoplastic material law

More flexible in describing the response to temperature and loading histories than the above models are viscoplastic material models with internal state variables. A well-known representative of such models is the model developed by CHABOCHE (1989). The constitutive equations of the model are given here for the case of uniaxial stress. Isotropic elastic properties are assumed so that the stress σ is computed from

illustration not visible in this excerpt

with Young’s modulus E. The thermal induced strains are also assumed to be isotropic, thus

illustration not visible in this excerpt

where α th is the differential coefficient of thermal expansion corresponding to the slope of the thermal strain-temperature curve and T is the temperature.

The viscoplastic strain rate is described by a viscous overstress formulation with a power law of the form

illustration not visible in this excerpt

where K and n are parameters to control the viscous effect. The backstress, α, captures the directional dependent effects such as the Bauschinger effect due to the plastic flow under cyclic loading.

The backstress α is introduced as the sum of several backstresses ∑

illustration not visible in this excerpt

where each evolves according to

illustration not visible in this excerpt

The first term is the strain hardening term which gives an increasing evolution of α with viscoplastic plastic strain. The second term called the dynamic recovery is collinear with backstress, α and is proportional to the accumulated plastic strain rate. The third term is called the static recovery or time recovery, since it can act independently of any plastic strain. Such term is important since hardening recovery with time occurs at high tempera- tures. The last term includes the temperature rate dependence in the back stress evolution which takes into consideration the non-isothermal effect. The first two terms in the equation were originally introduced by ARMSTRONG & FREDERICK (1966).

Ci and γ i are the hardening modulus and the transition constant, respectively. The quotient Ci/ γ i is the saturation value of the corresponding backstress. Ri is the static recovery constant. The function

illustration not visible in this excerpt

is introduced to model cyclic hardening or softening, where φ ss,i is the saturation value, ω i the transition constant and p the accumulated plastic strain.

In general, all parameters of the model depend on temperature and they are determined from a suitable set of experimental data, such as tension, relaxation, creep and fatigue tests.

In our work, we have taken two back stress variables, α = α 1 + α 2.

In the dynamic recovery term for the two back stressses we let the first evolve with the amount of viscoplastic strain, and the second with the accumulated plastic strain:

illustration not visible in this excerpt

The adjustable parameters in the Equation 3.5 to Equation 3.11 namely K, n, C, γ, R, φ ss, ω in the 1st and 2nd backstresses are temperature dependent. For non-isothermal loading the model parameters are obtained by linear interpolation in temperature.

3.2 Mechanism based model for fatigue lifetime prediction

The lifetime of a component exposed to thermocyclic loading is usually determined by the growth and the propagation of microcracks. Under LCF and TMF conditions, these microc- racks are usually observed to nucleate during the first few percent of the lifetime in metallic materials. The thermomechanical fatigue lifetime is determined by the growth of these mi- crocracks to an unacceptable size or to fracture (SURESH 1998), (ANDERSON 2004).

On the basis of elastic-plastic fracture mechnanics, Heitmann (HEITMANN ET AL. 1984) developed a damage parameter ZD for room temperature applications.

illustration not visible in this excerpt

The first term on the right hand side of the equation represents the elastic part of the ZD parameter while the second term represents the plastic part. N is the Ramberg-Osgood hardening exponent of the cyclic stress-strain curve and E is the Young’s modulus. These parameters are determined from the LCF hysteresis loop at half lifetime. Δ σ I and Δ ϵ vp are twice the stress and viscoplastic strain amplitudes of a stabilized stress-strain loop. The subscripts I and e denote the maximum principal and the von Mises equivalent of the re- spective variable. The subscript ef f indicates that the effect of crack closure is taken into account in the elastic part of ZD.

For high temperature applications, RIEDEL (1987) proposed the DCF parameters (CF for creep-fatigue). The damage parameter DTMF is an extension of DCF for non-isothermal cycles of arbitrary shape (SCHMITT ET AL. 2002). The DTMF parameter is a measure of the damaging effect of a given non-isothernmal loading cycle, provided that failure occurs by fatigue crack growth. Its functional form is:

illustration not visible in this excerpt

σ CY, is the cyclic yield stress measured from the point of strain rate reversal. The function F is introduced to take into account the temperature and time dependent crack growth and increased damage due to creep and holding times at higher temperatures.

illustration not visible in this excerpt

where Q is the activation energy for creep and R is the gas constant. In addition, F comprises the Norton creep exponent n and an adjustable parameter α. For time independent material behavior, F = 1.

In this application, it is assumed that for small degrees of non-proportionality of stress his- tories, the stress-strain hysteresis loops have well defined points of load reversals. Here the solutions of HE & HUTCHINSON (1981) for a penny-shaped crack under multiaxial but proportional loading are used. The solution is futher modified with a factor to account for the fact that a semi-circular surface crack is considered in ZD (SEIFERT 2008).

The equivalent stress, maximum principle stress and the equivalent plastic strain ranges are

defined as: Δ σ e =

illustration not visible in this excerpt

σ and ϵ vp are the stress and viscoplastic strain tensor. The subscripts 0 and 1 denote the lower and upper points of load reversal. σ H is the largest principal stress referred to its absolute value, i.e. σ H =max[ | σ I |, | σ III | ].

The case differentiation is introduced to distinguish whether the loading or unloading curve of the hysteresis loop is considered. The effective stress range is computed from :

illustration not visible in this excerpt

Such modification is taken into account to include the influence of the crack opening in the elastic part in DTMF. R is the stress ratio in this equation.

The cycles to failure Nf are related to the damage parameter DTMF through the power law:

illustration not visible in this excerpt

with adjustable parameters A and B. According to the damage mechanism, B = 1 and

illustration not visible in this excerpt

where a 0 and af are the initial and final length of the crack. β is the constant of proportionality between the crack propagation per cycle and the crack tip opening displacement. Experimental work done by CHAUVOT (2001) and HOFFMEYER ET AL. (2006) has shown that ln (af /a 0) ≈ 4. Additionally with 1 / 3 ≤ β ≤ 1, one obtains 8 ≤ A ≤ 24.

illustration not visible in this excerpt

Chapter 4 Welding and heat treatment of boiler tubes

4.1 Materials

Creep resistant steel (VM12) - Nr 1.4992

The material being investigated in this work namely X20CrMoV12-1 (VM12) is a creep resistant martensitic steel (body centered cubic lattice) specially developed for high tem- perature applications applied in the power plant industries. The pre-welded superheater tubes of VM12 were fabricated by the COST536 project partner Vallourec & Mannesmann under operational conditions from an industrial forging technique. The tubes were solution annealed at 1020 ◦ C for 20 minutes, quenched and tempered at 790 ◦ C for 60 minutes.

The chemical composition of VM12 is shown in Table 4.1

Table 4.1: Chemical composition of VM12 superheater tubes

illustration not visible in this excerpt

The pre-welded superheater tubes of Alloy617 were fabricated by the VDM Stainless Steel firm. These tubes were investigated mainly in the BMWi and AVIF MARCKO 700 Project (MAILE & SEELIGER 2004)¹. The chemical composition of the Alloy617 tubes is given in Table 4.2

Table 4.2: Chemical composition of Alloy617 superheater tubes

illustration not visible in this excerpt

4.2 Multipass welding of dissimilar boiler tubes

The boiler tubes being investigated were welded by using the TIG (Tungsten-Inert-Gas) welding process. The tubes were welded manually at ALSTOM Power in Stuttgart in Germany. The welded tubes have an outer diameter of 38 mm with wall thickness of 6.3 mm and length of 100 mm.

Figure 4.1 shows the setup for the dissimilar welding of the tubes. The austenitic nickel based Alloy617 side of the tube was clamped. Before the welding process, the region close to the weld of the Alloy617 and martesitic creep resistant VM12 steel tubes were heated up to 200 ◦ C. Two weld points were made between the tubes to hold the tubes together before the hand welding was performed. As illustrated by Figure 4.2, the welds were divided into 3 parts. First four roots welds (WL), followed by four intermediate welds (ZL) and finally 8 cover welds (DL) completed the welding process. The welding was performed in such a way that the first root, intermediate and cover welds were started from the austenitic (Alloy617) side and the consequent root, intermediate and cover welds were started on the martensitic (VM12) side. Thermocouples were located on both sides of the tubes for the measurement of the thermal cycles.

4.3 Post weld heat treatment of the welded tubes

Temperature differences within the component being welded induce local plastic deforma- tion and with that residual stress state which leads to negative influence on the lifetime of the component. In the case of dissimilarly welded boiler tubes, in addition to the different properties of the weld and parent material the welding residual stresses may influence the initiation and growth of cracks. Therefore, it is necessary to reduce the residual stresses generated during welding by post weld heat treatment (PWHT) before the component is assembled in the power plant .

The post weld heat treatment of the welded tubes were also conducted by ALSTOM Power Stuttgart. The welded tubes were heated from room temperature for 1 hour up to 760 ◦ C and then heat treated for 30 minutes at 760 ◦ C. The heat treated tubes were then allowed to cool down to room temperature for about 5 hours in free cooling condition.

illustration not visible in this excerpt

Figure 4.1: Welding setup done in ALSTOM Power Stuttgart

illustration not visible in this excerpt

Figure 4.2: Cross-section of multipass welding sequence

The welded and post weld heat treated Alloy617/VM12 tubes were then further investigated at Fraunhofer IWM.

4.4 Hardness distribution of welded and post weld heat treated tubes

The Vickers hardness of the weldment was measured after welding and after PWHT². The measurements were made at mid wall position in the section plane at 4 circumferential po- sitions. The results were comparable for all circumferential positions and the values for one circumferential position are given as an example in Figure 4.3. After welding, the mean hardness level of the Alloy617 base metal is HV1 195 whereas the Alloy617 weld metal is HV1 210 . The maximum hardness in the heat affected zone of VM12 is in the coarse grained heat affected zone (HAZ) with HV1 450 while the minimum value occurs in the intercritically/subcritically reheated HAZ with HV1 220 . A mean hardness of HV1 240 was measured in the VM12.

After PWHT the hardness level of the Alloy617 base metal and weld metal are comparable, with values about HV1 190. The hardness values in the VM12 HAZ and base metal are lower than in the as-welded state. The maximum value in the coarse-grained HAZ is HV1 250 , the minimium value in the intercritically/subcritically reheated HAZ is HV1 200. A mean hardness of HV1 205 was measured in the VM12 base metal.

4.5 Component test

Component and testing device

The dissimilarly welded and post weld heat treated VM12/Alloy617 component was tested under cyclic thermal loading (between 300 ◦ C and 625 ◦ C) with internal pressure.³ The loading conditions were selected in such a way that the componenet fails at about 5000 hours. The component was presurized with a constant internal pressure of 520 bar with its corresponding hoop stress of about 130 MPa.

The component comprised two dissimilar girth welds on a central VM12 tube, as shown in Figure 4.4a. The testing rig was developed at Fraunhofer IWM (DIETSCHE 2007), which was designed for this project for a maximum pressure of 600 bar. Because of the high pressure levels, the heating cartridge was enclosed in a tube of 1.5 mm wall thickness, in order to ensure the operation of the heating cartridge over the expected testing times. The mean heating cartridge power was 2 kW. Narrower electric coils in the weld area (Fig- ure 4.4b) should enable a more homogeneous temperature distribution in the tube. The tube was sealed by a bolted flange and a sealing ring. The flanges, the extension bolt and the sealing ring are made of high temperature resistant nickel based alloys.

illustration not visible in this excerpt

Figure 4.3: Measured distribution of Vickers hardness at the mid-wall thickness

The component was placed in an insulation box which is accompanied by a burst protection system. Argon gas was used as pressurization medium which enabled the heating of the component to the expected maximum temperature of 625 ◦ C.

Testing parameters

A first test was performed with an internal pressure of 520 bar which corresponds to hoop stress of 130 MPa. The thermal cyclic loading is achieved by using a heating cartridge which generates a cyclic temperature change between 300 ◦ C and 625 ◦ C with a dwell time of 1 h at 625 ◦ C.

Test results

The testing times in terms of numbers of cycles, dwell time and total time are tabulated in Table 1. The first test on a dissimilar weld between Alloy617 and VM12 was ended after heating cartridge failure. Upon the cartridge failure, the component had already experi- enced around 3000 cycles. A planned second test at lower stress was not performed due to the risks of cartridge failure. Instead, cross weld specimens were tested (described in Chapter 5).

4.6 Damage assessment of welded superheater tube

Destructive evaluation of components provide the most direct and accurate assessment of their current condition. For this reason, after failure of the heating cartridge, extensive examinations were made on the dissimilarly welded component to identify possible fail-

illustration not visible in this excerpt

Figure 4.4: a) Tube specimen, b) sketch of the tube specimen and the heating cartridge

Table 4.3: Results of the Component Test

illustration not visible in this excerpt

ure mechanisms. This section deals specifically with the fractographic and metallographic analyses of possible crack initiation and growth on the welded component.

4.6.1 Detection of cracks

Figure 4.5 (a) and (b) show the outer and the inner surfaces of the specimen (P2) extracted from the dissimilarly welded tubes after the component test.⁴ A crack was detected through the dye-penetrant inspection and was found to have propagated along the thickness of the tubes as shown in Figure 4.6.

In a subsequent step, the fractographic analysis was extended by inspecting the specimens through scanning electron microscopy (SEM). Figure 4.7 and Figure 4.8 show SEM images of a crack on P2 specimen initially detected through dye-penetrant inspection. Figure 4.7 shows that the crack has initiated at the weld root and propagated along the VM12/Alloy617 material interface through the wall thickness. Similarly, Figure 4.8 shows that the crack has also travelled from the weld root and propagated along the VM12/Alloy617 material interface on the inner surface of the tube.

illustration not visible in this excerpt

Figure 4.5: Dye-penetrant inspection on (a) outer surface,and b) inner surface of the spec- imen (P2) containing the crack, after the termination of the test (approximately after 6800 hours)

illustration not visible in this excerpt

Figure 4.6: The region pointed by the arrow shows the location of crack detected by dye- penetrant inspection on specimen P2, after the termination of the test (approximately after 6800 hours)

[...]

---

¹ Special thanks to Mr. Phillip von Hartrott at Fraunhofer IWM for enabling the usage of experimental data of Alloy617 for the completion of this dissertation.

² The hardness measurement was conducted at Fraunhofer IWM by COST 536 Project team member Mrs. Valerie Friedmann.

³ The experimental setup and component test were implemented with the technical assistance from colleagues at Fraunhofer IWM. Special thanks to Mr. Yves Squaizer.

⁴ The author would like to thank colleagues at Metallographic Department at Fraunhofer IWM for their assistance in conducting the analyses.