Corrosion-Fatigue Testing on Steel Grades with Different Heat and Surface Treatments Used in Rock-Drilling Applications

DEGREE PROJECT IN MATERIALS SCIENCE AND ENGINEERING, SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2016

CORROSION-FATIGUE TESTING ON STEEL GRADES WITH DIFFERENT HEAT AND SURFACE TREATMENTS USED IN ROCK-DRILLING APPLICATIONS

LUIS MIGUEL BÉJAR INFANTE

KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF INDUSTRIAL ENGINEERING AND MANAGEMENT

Abstract Corrosion fatigue is a common failure mechanism in rock drilling components and many other mechanical parts subjected to cyclic loads in corrosive environments. A crucial part in the design of such components resides in the selection of the right materials for the application, which ideally involves testing and comparison of their performance under working conditions.

The present work was performed with the purpose of designing a corrosion-fatigue testing method that would allow the designer to compare the performance of different materials exposed to corrosion fatigue, permitting also the comparison with results from dry fatigue testing. The method was designed for rotating-bending machines. Two different steel grades were used during the work, one through hardened and one case hardened. The effect of these heat treatments and of shot peening over corrosion-fatigue behaviour were studied using the proposed method.

It was proven that the testing speed has a strong impact on the fatigue life of steel. It was found that, at a fixed stress level, the case hardened and shot peened steel reached 3X10^6 cycles at 2300 rpm, while it failed at only 5X10^5 cycles with a testing speed of 500 rpm. A large beneficial influence of the shot peening was demonstrated. It was also observed that, at fixed testing speed, the shot peening on the through hardened steel can increase its fatigue strength from 190 MPa to 600 MPa under corrosion fatigue. Many cracks were found at the surface of the shot peened parts, which are arrested near the surface by the compressive stress layer from the shot peening. It was also found that, for the non-shot peened parts, case hardening had a slightly higher corrosion-fatigue strength than the through hardened. This might be a result of the compressive stresses from carburization, or due to the higher core toughness of this steel grade.

Keywords: Corrosion fatigue, crack, rotating-bending, S-N curve, staircase method, fatigue strength, shot peening, case hardening, through hardening

Preface This work was performed to cover the thesis project in Mechanical Metallurgy as part of the master’s degree in Engineering Materials Science at the Royal Institute of Technology in Stockholm, Sweden. The project was conducted in cooperation with and financed by Atlas Copco Secoroc AB in Fagersta, Sweden.

Very special thanks go to all the people who supported me throughout the whole project, without whom it would not have been possible to complete this work. To Göran Stenberg for the right guidance on every step of the project, and for providing me with scientific criteria to achieve a professional work. To Richard Johanson for the invaluable support in several aspects, including all the laboratory training I received, calibration of the instruments, production processes, metallographic sampling and analyses, and for the interesting discussions that helped to shape the conclusions of this work. To Anders Olsson who shared essential insights to the technical debates that guided the project to success. To Jonas Falkestrom for the support I received within the company throughout the whole timeframe to perform this project without any obstacle.

I also want to thank Alexander Beronius for the important technical contributions and interesting discussions which had a vital impact on the outcome of the project, and Gabriella Brorson for the guidance on essential technical information and valuable criteria.

Special thanks go to my supervisor in KTH, Stefan Jonsson, who followed my project in all technical aspects and provided me with vital insights that helped me take decisive steps. And last but not least, I want to thank my father, who introduced me to the fascinating world of mechanics and metals, for his innumerable teachings in these subjects and in all others.

Luis Miguel Béjar

Fagersta, June 2016

Abbreviations

CH Case hardened

TH Through hardened

SP Shot peened

NSP Non-shot peened

HCF High-cycle fatigue

LCF Low-cycle fatigue

iii

List of Tables Table 1. Different material scenarios used in the present work ...... 28 Table 2. Metallographic samples selection ...... 33 Table 3. Percent Replication ...... 34 Table 4. Average values for number of cycles to failure and time of test for frequency dependence of corrosion fatigue on CH-SP in fresh water ...... 35 Table 5. Average values for number of cycles to failure and time of test for frequency dependence of corrosion fatigue on TH-NSP in fresh water ...... 37 Table 6. Parameters for Proposed Corrosion-Fatigue Testing Method ...... 39

List of Figures Figure 1. Definition of cycles and reversals [3] ...... 4 Figure 2. Load cycles for sinusoidal, square and triangular load paths. Middle figure: Same as above but twice the frequency. Lower figure: A schematic spectrum load found in many applications. [2] ... 4 Figure 3. Definition of components in a stress cycle [2] ...... 5 Figure 4. S-N curve for AISI 4340 alloyed steel [7] ...... 6 Figure 5. Comparison of steel and aluminium fatigue behaviour [6] ...... 6 Figure 6. Illustration of the steps of the fatigue crack evolution on radial and axial cross sections of a cylindrical part [9] ...... 7 Figure 7. Schematic illustration of the variation of fatigue-crack growth rate da/dN with alternating stress intensity ΔK in steel, showing regimes of primary crack-growth mechanisms [8] ...... 8 Figure 8. Development of extrusions and intrusions during fatigue [9] ...... 9 Figure 9. Schematic illustration of crack initiation, stable crack growth and fracture [2] ...... 9 Figure 10. Fatigue crack propagation [6]...... 10 Figure 11. Corrosion-fatigue and its general effect on the behaviour of steel [12] ...... 11 Figure 12. Solubility of oxygen in water at different temperatures ...... 12 Figure 13. Effect of NaCl concentration on the corrosion of Fe [18] ...... 13 Figure 14. Effect of NaCl concentration on crack initiation and crack propagation [19] ...... 13 Figure 15. Effect of MnS inclusions in the film rupture/anodic dissolution process [20] ...... 17 Figure 16. Hydrogen embrittlement ...... 19 Figure 17. The hydrogen embrittlement process [36] ...... 19 Figure 18. Shot peening compressive stress profile [21] ...... 21 Figure 19. Rotating-bending machine diagram ...... 22 Figure 20. Example of staircase fatigue data [3]; suspensions are tests in which the specimens survived ...... 23 Figure 21. S-N testing with a small sample size [3]; suspensions are tests in which the specimens survived ...... 24 Figure 22. Rotating-bending fatigue testing machine - Atlas Copco Secoroc AB materials laboratory 25 Figure 23. Rotating-bending fatigue testing machine, rear side - Atlas Copco Secoroc AB materials laboratory ...... 26 Figure 24. Specimen mounted on the machine and water collector underneath ...... 26 Figure 25. Corrosion chamber ...... 27 Figure 26. Geometry and dimensions of specimen ...... 27 Figure 27. Laser scan micrometer ...... 28 Figure 28. Laser scan micrometer display ...... 28 Figure 29. Specimens used. From left to right: (1) Case hardened, non-shot peened; (2) Case hardened and shot peened; (3) Through hardened, non-shot peened; (4) Through hardening and shot peened ...... 29 Figure 30. Residual stresses on CH-SP ...... 30 Figure 31. Residual stresses on CH-NSP ...... 30 Figure 32. Residual stresses on TH-SP ...... 30 Figure 33. Residual stresses on TH-NSP ...... 30 Figure 34. Corrosion-fatigue frequency dependence on CH-SP ...... 34 Figure 35. Average time spent per test for different testing speeds on CH-SP in fresh water and brine ...... 35 Figure 36. Corrosion-fatigue frequency dependence on TH-NSP ...... 36 Figure 37. Average time spent per test for different testing speeds on TH-NSP in fresh water ...... 36

Figure 38. Corrosion-fatigue frequency dependence comparison between CH-SP and TH-NSP ...... 37 Figure 39. Corrosion-fatigue frequency dependence of CH-SP at two load levels ...... 38 Figure 40. Corrosion-fatigue frequency dependence of TH-NSP at two load levels ...... 38 Figure 41. Testing speed selection ...... 39 Figure 42. S-N curve for CH-SP ...... 40 Figure 43. Corrosion-fatigue strength of CH-SP from staircase method ...... 40 Figure 44. S-N curve for CH-NSP ...... 41 Figure 45. Corrosion-fatigue strength of CH-NSP from staircase method ...... 41 Figure 46. S-N curve for TH-SP ...... 42 Figure 47. Corrosion-fatigue strength of TH-SP from staircase method ...... 42 Figure 48. S-N curve for TH-NSP ...... 43 Figure 49. Corrosion-fatigue strength of TH-NSP from staircase method ...... 43 Figure 50. Comparison of the four conditions (S-N curves and corrosion-fatigue strengths) ...... 44 Figure 51. S-N curve comparison between CH-SP and TH-SP ...... 45 Figure 52. S-N curve comparison between CH-SP, CH-NSP, and dry tests ...... 46 Figure 53. S-N curve comparison between TH-SP, TH-NSP, and dry tests (on non-shot peened) ...... 46 Figure 54. Crack initiation and deformed propagation surface on TH-NSP @200 MPa, Specimen 00947 Figure 55. Crack initiation on TH-NSP @200 MPa, Specimen 009 ...... 47 Figure 56. Surface major crack propagation on TH-NSP @744 MPa, Specimen 087 ...... 47 Figure 57. Several cracks initiating on fracture surface at different planes near on TH-NSP @744 MPa, Specimen 087 ...... 47 Figure 58. Crack initiation morphology on TH-NSP @744 MPa, Specimen 087 ...... 47 Figure 59. CaS inclusion near crack initiation on TH-NSP @744 MPa, Specimen 087 ...... 47 Figure 60. Possible crack initiation from (Ca, Si, Al)S inclusion on CH-NSP @200 MPa, Specimen 043 48 Figure 61. Cracking around initiation spot on CH-NSP @200 MPa, Specimen 043...... 48 Figure 62. Crack initiation on CH-NSP @725 MPa, Specimen 045 ...... 48 Figure 63. Brittle crack propagation on CH-NSP @725 MPa, Specimen 045 ...... 48 Figure 64. Major rack initiation on TH-SP @600 MPa, Specimen 010 ...... 48 Figure 65. Crack propagation surface on TH-SP @600 MPa, Specimen 010 ...... 48 Figure 66. Several crack initiations on TH-SP @750 MPa, Specimen 034 ...... 49 Figure 67. Major crack initiation and sub-cracking on TH-SP @750 MPa, Specimen 034 ...... 49 Figure 68. Crack connecting residual fracture on TH-SP @750 MPa, Specimen 034 ...... 49 Figure 69. Crack propagation surface TH-SP @750 MPa, Specimen 034 ...... 49 Figure 70. Crack propagation at 3mm depth from edge on CH-SP @550 MPa, Specimen 018 ...... 49 Figure 71. Intergranular crack propagation at 4mm from edge on CH-SP @550 MPa, Specimen 018 49 Figure 72. Brittle fracture surface on CH-SP @550 MPa, Specimen 018 ...... 50 Figure 73. Corrosion products at fracture surface on CH-SP @550 MPa, Specimen 018 ...... 50 Figure 74. Main crack initiation and crack meeting facture surface on CH-SP @550 MPa, Specimen 018 ...... 50 Figure 75. Main crack initiation and sub-cracking on CH-SP @550 MPa, Specimen 018 ...... 50 Figure 76. Main crack initiation on CH-SP @725 MPa, Specimen 082 ...... 50 Figure 77. Intergranular crack propagation on CH-SP @725 MPa, Specimen 082 ...... 50 Figure 78. Specimen 091. A) Overall surface view, B) General crack length, C) Cracks showing almost no branching, D) Crack not following the microstructure ...... 51 Figure 79. Specimen 047. A) Overall surface view, B) General crack length, C) Cracks showing much branching, D) Abundant branching of cracks ...... 52 Figure 80. Specimen 045. A) Overall surface view, B) Tiny cracks at the decarburised layer, C) Miniature cracks at the surface and one major crack growing into the material ...... 52

Figure 81. Specimen 095. A) Overall surface view, B) Almost no cracks at the surface, except for a few major cracks that do not propagate to cause failure before 3X10^6 cycles ...... 53 Figure 82. Specimen 012. A) Overall surface view, B) General crack length, C) Cracks showing almost no branching, D) Cracks magnification on etched sample ...... 53 Figure 83. Specimen 013. A) Overall surface view, B) General crack length, C) Cracks showing branching, D) Cracks magnification on etched sample ...... 54 Figure 84. Specimen 036. A) Overall, B) Big crack propagation ...... 54 Figure 85. Specimen 016. A) Overall, B) Few large cracks can cause failure...... 55 Figure 86. Specimen 029. A) Overall cracking on the surface, B) Grown crack ...... 55 Figure 87. Specimen 029. A) Overall surface cracking, B) Crack branching is eaten up by salty water, C) General crack length, D) Few larger cracks that eventually propagate deep enough to cause failure 56

vii

Table of Contents Abstract ...... i Preface ...... ii Abbreviations ...... iii 1. Introduction ...... 1 1.1. Background ...... 1 1.2. Purpose ...... 1 1.3. Delimitations ...... 1 1.4. Methodology ...... 1 2. Theoretical Basis ...... 3 2.1. Fatigue ...... 3 2.2. The S-N Curve ...... 5 2.3. Fatigue Limit and Fatigue Strength ...... 6 2.4. Crack Initiation and Crack Propagation ...... 7 2.4.1. Crack Initiation...... 8 2.4.2. Crack Growth ...... 9 2.5. Corrosion Fatigue ...... 10 2.6. Factors Influencing Corrosion Fatigue ...... 11 2.6.1. Environment ...... 11 2.6.1.1. Concentration of the Corrosion Species ...... 12 2.6.1.2. Temperature ...... 12 2.6.1.3. Effect of NaCl ...... 12 2.6.1.4. Influence of pH ...... 14 2.6.2. Surface Finish...... 14 2.6.3. Stress Ratio ...... 14 2.6.4. Load Frequency ...... 14 2.6.5. Stress Intensity ...... 15 2.6.6. Size Effects and Hardness Gradients ...... 15 2.6.7. Metallurgical Variables ...... 16 2.6.7.1. Inclusions ...... 16 2.6.8. Surface Residual Stresses ...... 18 2.6.9. Crack Closure Effects ...... 18 2.7. Mechanisms of Corrosion Fatigue ...... 18 2.7.1. Pitting Corrosion ...... 18 2.7.2. Corrosion at Preferential Locations ...... 19 2.7.3. Hydrogen Embrittlement...... 19

viii

2.7.4. Rupture of Oxide Protective Film ...... 20 2.7.5. Surface Energy Reduction ...... 20 2.8. Heat and Surface Treatments ...... 20 2.8.1. Case Hardening ...... 20 2.8.2. Through Hardening ...... 20 2.8.3. Shot Peening ...... 21 2.9. Fatigue Testing ...... 21 2.9.1. Rotating-Bending Test ...... 22 2.9.2. The Staircase Method ...... 22 2.9.3. The Median S-N Test Method for Small Sample Size ...... 24 3. Materials and Equipment ...... 25 3.1. Laboratory Equipment ...... 25 3.2. Specimens ...... 27 3.2.1. Geometry and Dimensions ...... 27 3.2.2. Material ...... 28 3.2.3. Heat Treatment ...... 29 3.2.4. Shot Peening ...... 29 3.2.4.1. Residual Stress Analysis ...... 29 4. Experimental Setup and Procedures ...... 31 4.1. Corrosion-Fatigue Testing Method Proposal and Testing Parameters Selection ...... 31 4.2. Testing the Proposed Corrosion-Fatigue Method ...... 32 4.3. Post-Test Analysis ...... 33 4.3.1. Microscope Analysis ...... 33 4.3.2. Metallographic Samples ...... 33 5. Results ...... 34 5.1. Presentation of Corrosion-Fatigue Testing Method ...... 34 5.1.1. Frequency Dependence Results ...... 34 5.1.2. Final Parameters for Proposed Testing Method ...... 39 5.2. Testing the Proposed Corrosion-Fatigue Testing Method ...... 40 5.2.1. S-N curves and fatigue strength results ...... 40 5.2.2. Materials Comparison ...... 44 5.3. Metallographic Results and Observations ...... 47 5.3.1. Fracture Surface Analysis from SEM ...... 47 5.3.2. Cracking Analysis from Optical Microscope ...... 51 6. Discussion ...... 57 6.1. Discussion of Method and Parameters ...... 57

6.2. Discussion of Testing Results ...... 59 6.3. Fracture Surface and Cracking Characteristics in Relation with the Results ...... 60 7. Conclusions ...... 64 8. Further Work ...... 65 9. References ...... 66

1. Introduction In this section, the background, purpose of the project, delimitations of the work, and the methodology followed by the author are presented. 1.1. Background Atlas Copco’s rock drilling components work in a tough environment. The drill string delivers an impact pulse in the 250 MPa range, at a frequency that can reach above 40 Hz, while flushing aqueous media with the purpose of transporting chippings together back to the hole entry. Often the flushing media is water and can differ in pH, salt content and other contaminants. This environment can reduce the service life and the reliability of this machinery.

Corrosion fatigue is found to be a highly recurrent cause of failure in drill string components. To improve the reliability and service life of failing parts, it is vital to analyse the failure mechanisms. For that reason, corrosion fatigue was studied by recreating in a laboratory the conditions that could be encountered in underground applications, to be capable to understand the mechanisms involved in the damage, and furthermore, to be able to predict failure. 1.2. Purpose The scope of this work was to design a testing method to compare the performance under corrosion fatigue of different steel grades with different thermal and mechanical treatments used in drill string components, which implicates cyclic loading and usually the exposure to corrosive environments. The work comprehended proposing the method and characterizing the corrosion-fatigue tests by choosing suitable parameters that enable the design engineer to discriminate between diverse steels performing under cyclic loading and corrosive environments. Once the test was parametrized, the comparison of four different material conditions was performed in order to test the veracity of the method. 1.3. Delimitations The tests were limited to rotating-bending machines. Therefore, limitations regarding the detailed study of crack growth rates must be considered.

The present work is delimitated tests in fresh water and salt water. The effect of different aqueous media is not studied hereby.

The water temperature and oxygen concentration were not varied during the tests. The effects of changes in these variables was not further studied.

Intensity of shot peening performed on the specimens was not varied within the same steel grade. 1.4. Methodology The work started with a thorough literature review to establish the conceptual basis of the project. A literature survey was useful to study similar research that has been performed within the field of corrosion-fatigue testing. Both literature surveys were performed using KTH’s digital library DiVA, ScienceDirect database and Google Scholar. Atlas Copco’s current fatigue-testing methodology was studied and the historical data was reviewed to set up the initial bases for further testing. The corrosion-fatigue tests were performed using rotating-bending machines, and the results were captured in an Excel spreadsheet. The statistical analyses of the gathered data were performed using Excel’s statistical tools, and the calculations were performed within the Excel spreadsheet. The preparation and evaluation of the specimens prior to testing were performed with special equipment

1 supplied by Atlas Copco, and it was all performed within the company’s facilities. Once the tests were completed, metallographic samples were prepared from certain specimens and analysed in the optical microscope. Some specimens’ fracture surfaces were analysed with a Scanning Electron Microscope. The analyses of the specimens after testing, involving fracture surfaces, corrosion, degradation, and metallography, were performed also with equipment supplied by Atlas Copco.

2. Theoretical Basis The purpose of the following section is to present, in general terms, the fundamental concepts involved in the present work. 2.1. Fatigue Mechanical components are exposed to a great variety of stress modes and diverse loads during their function. It is the role of the design engineer to decide the geometry, dimensions and material characteristics of every component to ensure its desired performance and reliability. As it can be supposed, the task of the design engineer when it comes to considering every kind of load the component has to withstand is not a simple one; tension, compression, torsion, bending, or any of the possible combinations of these could be stressing the material in complex ways.

Every material has a stress limit at which it will no longer withstand a static load and it will fail (either fracture or deform permanently, depending on what is considered failure by the engineer), and it can be inferred that under such limit, the static load will not be enough to cause the material to fail. In the past this limit was considered alone for the design of machine components, and engineers started to observe that, surprisingly, this limit was not respected by many of the failing parts. By the middle of the XIX century, this behaviour started to concern engineers, who began to perform tests under cyclic loads below the static load limits and found that the components would eventually fail at such low load values. It was observed that failures happened after the component had been functioning satisfactorily for a period of time, and the general opinion stated that the continuous cyclic load had exhausted the ability of the material to carry load, or in other words, the fatigue of the material occurred [1]. Fatigue is by far the most common cause of failure on mechanical components (about 80% of all the fractures of metals [2]), and it is the most feared one, since it “occurs suddenly without any noticeable plastic deformation, i.e. without warning” [2].

Fatigue starts when a load is high enough to cause plastic deformation on a component, even when it happens only at a small volume of the material, i.e. localized plastic deformation could occur at the highest stress location. When such load is repeated cyclically, the damage accumulates and eventually a crack is formed. If the cyclic load continues, the length of the crack increases until the stress on the component is high enough to cause its sudden fracture [3]. It could be then assumed that the time to fracture by fatigue is a purely cycle-dependent process, but this premise is not always true; when a corrosive environment is present, both fatigue and corrosion have a synergetic effect on the material, which is known as “corrosion fatigue” [2]. Since corrosion is mainly a time-dependent process, crack initiation and crack growth become strongly time-dependent during corrosion fatigue, as it will be explained more deeply in section 2.5.

The phenomenon of fatigue is generally divided into three steps: (1) crack initiation, (2) crack growth/propagation and (3) final fracture. The first two steps of the process are not easy to characterize and a great amount of work on research has been performed to try to understand the mechanisms involved on each one of them; it is even hard to identify a sharp limit to divide when the crack initiation ends and the crack propagation starts [14]. The general principles of such steps will be described in section 2.4.

A load cycle can be applied in many different ways in terms of the load wave shape, magnitudes of the loads, sign of the stress caused by the load and frequency of load application. In general, a load cycle can be characterised by a maximum stress, σmax, and a minimum stress σmin [2]. It should be emphasised that the stresses could be either positive or negative (tensile or compressive, respectively) depending on the particular cycle. The step of going from the minimum to the maximum stress -or vice versa- is

3 called a reversal. A cycle is achieved when the stress is taken from the minimum to the maximum, and then back to the minimum –or vice versa- completing two reversals [3]. Figure 1 shows how a cycle and a reversal are defined graphically.

one one one reversal reversal cycle 300 200 100

0 Stress -100 0 1 2 3 4 5 6 7 -200 -300

Figure 1. Definition of cycles and reversals [3]

Load cycles can be applied in a variety of wave shapes, from the sinusoidal shape shown above to triangular, squared, or even an irregular shape, also called spectrum shape. The frequency of application can also vary. Figure 2 shows the different loading wave shapes and frequencies. All those factors are controlled during a fatigue test.

Figure 2. Load cycles for sinusoidal, square and triangular load paths. Middle figure: Same as above but twice the frequency. Lower figure: A schematic spectrum load found in many applications. [2]

Based on the stresses σmax and σmin explained above, a static component, σ0 (also known as mean stress), and a dynamic component, σa (also known as stress amplitude), can be obtained for any particular loading cycle. The mean stress is defined as the algebraic average of the maximum and minimum stresses in one cycle. The stress amplitude is defined as the difference between the maximum –or minimum- stress and the average stress. Another component that can be naturally defined from the cycle is the stress range, σr, defined as the difference between the maximum and the minimum stresses, i.e. two times the stress amplitude [5]. These components can be graphically seen in Figure 3.

Figure 3. Definition of components in a stress cycle [2]

From the image shown above, it is easy to imagine that the stress wave can be positioned anywhere along the stress axis (without crossing the σUTS limit) depending on the way the load is applied. A fully reversed cycle, for instance, will have the same magnitude for both the maximum and minimum stresses but a different sign, i.e. the maximum stress will be tensile and the minimum will be compressive; this will result in a mean stress with a magnitude of zero. This kind of stress cycle is characteristic of rotating components that are also subjected to bending, like shafts, and it can be simulated with a rotating-bending test [6]. This method will be further explained in section 2.9.1. If for instance both σmax and σmin are positive, the static component of the cycle will also be positive. This concept leads to an important parameter to characterise a fatigue test: the stress ratio. The stress ratio is defined as the algebraic ratio of σmin to σmax (R = σmin/σmax). For a fully reversed cycle, R would be equal to -1. If the cycle is partially reversed, R becomes negative with a value between -1 and 0. If the load goes from 0 to σmax, R will be equal to 0, and if both σmin to σmax are positive, R will be also positive, but less than 1 [6].

Depending on the load levels of the cycle, the stress in the component can be purely elastic or it could jump into the plastic region. When the applied stress is elastic, the fatigue phenomenon is considered to be high-cycle fatigue (HCF). When this happens, the component could reach a really long life (N > 105 cycles) [6]. Although the stress on the component is too low for initiating a general plastic deformation, the crack tip undergoes plastic deformation, and that is the reason why it grows and eventually causes failure. On the other hand, when the stress on the component is high enough to cause plastic deformation, the phenomenon is known as low-cycle fatigue (LCF). Since plastic strain is present in each cycle, the behaviour is more complex than in HCF, as the material undergoes plastic deformation and forms a hysteresis loop on the stress-strain diagram. This phenomenon is well explained elsewhere ([6], [7]). 2.2. The S-N Curve As explained in the previous section, materials will fail as an effect of fatigue when a cyclic load is applied on them. The higher the load, the less number of loading cycles it will endure and vice versa. This Stress-Life relation can be represented by the Wöhler method with the so-called S-N curve, which is usually shown as a plot of the nominal stress amplitude versus the number of cycles to failure on a logarithmic axis [5]. Figure 4 shows an illustration of the S-N curves for notched and unnotched AISI 4340 alloyed steel specimens at room temperature tested with a stress ratio of R = -1 [7].

Figure 4. S-N curve for AISI 4340 alloyed steel [7]

2.3. Fatigue Limit and Fatigue Strength The typical fatigue behaviour of most metals at dry conditions shows a stress level low enough at which the metal will no longer fail under fatigue, and it is said that the material shows infinite life. For instance, in Figure 4 a runout (no failure) is observed for each of the curves at certain stress level; taking the unnotched curve as example, the runout is present at 345 MPa. This maximum stress that the material can endure without failure is called fatigue (endurance) limit. Below this stress level there is a 50% probability that no failure will occur on a material at the same conditions [5].

Most nonferrous metals and ferrous metals under corrosive environments do not exhibit an endurance limit. In such cases, the S-N curve does not even out at certain stress value, but instead it continues falling at a slow rate at high number of cycles [5]. When this happens, fatigue strength, rather than fatigue limit, is reported. Fatigue strength is defined as the stress level that a material can withstand to reach a defined number of cycles, generally enough be considered as ‘infinite life’. In these case, the plot must specify the number of cycles at which the fatigue strength is reported [5]. A typical endurance limit behaviour is shown in Figure 5 for AISI 1045 steel; in the same plot, the constant drop in stress to reach longer life can be seen for aluminium 2024-T6.

Figure 5. Comparison of steel and aluminium fatigue behaviour [6] 6

2.4. Crack Initiation and Crack Propagation It is understood that, in order for the fatigue process to start, there is a requirement for certain plastic strain on the material, as explained in section 2.1. There are three simultaneous conditions that are required for fatigue damage to occur: cyclic stress, tensile stress and plastic strain [14]. The fatigue phenomenon can be either studied as a stress-controlled or as a strain-controlled process. The process of fatigue is generally divided into the following sequential steps [14]:

(1) Cyclic plastic deformation prior to fatigue crack initiation (2) Initiation of one or more microcracks (3) Propagation or coalescence of microcracks to form one or more macrocracks (4) Propagation of one or more macrocracks (5) Final failure

The fatigue process starts with the formation of a crack, which can be either nucleated during the cyclic loading or present in the material from the beginning. As the loading cycles continue, the crack will grow under a stable regime below the fracture toughness limit of the material. When the stress intensity factor of the crack exceeds the fracture toughness of the material, instant fracture occurs [2]. Figure 6 shows a represnetation of the three steps conforming the fatigue process.

Figure 6. Illustration of the steps of the fatigue crack evolution on radial and axial cross sections of a cylindrical part [9]

Although the steps of the fatigue process are generally accepted, there is no general agreement on when the crack initiation process ends and when the crack propagation starts, or at which crack length the crack initiation turns into propagation [14]. In the past it was believed that as much as 95% of the total life of a fatigue-failure was spent in crack initiation [4], however, better methods of crack detection have appeared and nowadays it is possible to observe that cracks nucleate after the first 10% of the total life of the component [7]. In dry conditions, this is mainly dependent on the level of applied stress. In general terms it is considered that “approximately 30 to 40% of the low-cycle fatigue life and approximately 80 to 90% of the high-cycle fatigue life, measured by cycles to failure, involve nucleation of the dominant fatigue crack that eventually causes failure” [5]. Since a large fraction of the life is spent on crack growth on low cycle fatigue, it has become more interesting to study this phenomenon in terms of the mechanical behaviour of the crack growth, i.e. by studying the crack growth and characterizing it by linear elastic fracture mechanics, elastic-plastic fracture mechanics, or fully plastic fracture mechanics. In such cases, the entire crack initiation process is taken as nucleation phase, and the crack propagation is then expressed in terms of crack growth rates [14].

The linear elastic fracture mechanics approach assumes that every mechanical component contains flaws and that cracks grow from an initial size, ao, until reaching a critical size, ac. The rate at which the crack grows is expressed as the change in crack size per each cycle, da/dN. Figure 7 shows the three regimes of crack evolution and the factors that influence each regime. During regime A, the crack growth is very small, but still increases to get to a size large enough to enter regime B, where da/dN can be modelled using the Paris law, which is represented as a straight line along the log ΔK axis. The equation for the Paris law can be expressed as

푑푎 (1) = 퐶(∆퐾)푚 푑푁

where the constants C and m are material coefficients that can be obtained experimentally by plotting log(da/dN) against log(ΔK) [2]. In regime C the crack growth rate grows faster without control, until it reaches final failure [6].

Figure 7. Schematic illustration of the variation of fatigue-crack growth rate da/dN with alternating stress intensity ΔK in steel, showing regimes of primary crack-growth mechanisms [8]

For the present work, linear elastic fracture mechanics was not considered since the aim of this investigation does not deal with crack growth rates or crack nucleation times. In fact, the interesting event for this work was the final failure of the parts and the variable considered was the life reached by the material at the final failure point. Nevertheless, the influence of different factors on each regime shown in Figure 7 is interesting for understanding the behaviour of the materials investigated in the present work. 2.4.1. Crack Initiation Usually cracks start on the surface of the material, where the highest stress is normally found [2]. Moreover, the surface usually contains defects such as machining marks, inclusions, “soft spots” from carburisation, corrosion pits, etc. Below the surface, carbide clusters, inclusions or slag particles can

8 also have the same effect [2]. All those defects serve as crack initiation spots. However, if the material is pure enough to be exempt of such defects and its surface is well polished, the crack nucleation process can start from persistent slip bands. A slip band is a result of the systematic build-up of slip movements on the order of 1 nm [6]. When the material is repeatedly stressed, the surface grains are sheared in near 45° angle with the applied stress and slip bands are formed. Then persistent slip bands will form extrusions and intrusions on the surface of the material, as shown in Figure 8. The intrusions will act as nucleation sites for cracks to form and grow through the first grain following the shear band at about 45° angle to the applied stress [2]. The crack starts as a featureless fracture surface parallel to the slip bands, in what is referred to as Stage I. When the crack is large enough, the stress at the crack tip becomes dominant and the overall crack plane is switched and becomes normal to the principal stress [6].

Figure 8. Development of extrusions and intrusions during fatigue [9]

2.4.2. Crack Growth Once the stress at the crack tip becomes dominant, the crack changes plane and continues to grow perpendicular to the principal stress. This could happen when the crack, growing at 45° relative to the principal stress during the initial stage, reaches the next grain boundary on its way and it deflects. This step could be considered as the start of the stable crack growth stage, referred to as Stage II. Figure 9 shows this phenomenon.

Figure 9. Schematic illustration of crack initiation, stable crack growth and fracture [2]

During Stage II, the crack will propagate as a result of its cyclic opening and closing on each stress cycle. The tensile stress will open the crack, making it grow a tiny bit. The next reversal will relax the stress, closing the crack. As this process is repeated, the crack sharpens and blunts leaving a very small marking on the crack surface on every cycle. These marks, known as striations, leave a characteristic pattern on the fracture surface of the material, with each striation representing one cycle of fatigue.

The steps of stable crack growth are illustrated in Figure 10. During step 1, the crack is “resting” while no tensile stress is applied on it. On stage 2, tensile stress is applied and the crack opens and the crack tip deforms and propagates a bit on step 3. On step 4 the crack starts to close again as the stress is reduced, leaving a permanent deformation at the tip of the crack, i.e. the crack grew longer. On stage 5 the crack is completely closed again as the stress goes back to zero or to negative values, which are harmless to crack propagation. A fatigue cycle covers from step 1 to step 5, so step 6 is the same as step 2 but with a larger crack.

Figure 10. Fatigue crack propagation [6]

In order for a crack to grow, ΔK must exceed a threshold value. When this happens, crack growth can be observed and the crack growth rate can be described by Paris law [2] (equation 1). However, since compression is harmless, the minimum stress used to calculate ΔK is set as zero. 2.5. Corrosion Fatigue Having described the phenomenon of fatigue, it is easy to introduce the concept of corrosion fatigue. Corrosion fatigue is, as its name suggests, the combined action of cyclic stress and a corrosive environment on materials. “Corrosion fatigue is dependent on the interactions among loading, environmental, and metallurgical factors” [4]. In general terms, the effect of corrosion will be detrimental for the fatigue endurance of materials, as corrosion could cause a faster crack initiation, a higher crack propagation rate, or both. The effect of the aggressive media on the fatigue performance can vary widely, and is mainly dependent on the relation between the metal and the environment [5].

One of the main differences between corrosion fatigue and inert-environment fatigue is that there is no “safe stress range” at which metal has infinite life. As the number of cycles to failure is increased, the fatigue strength of the metal in corrosive environment continues to fall [10]. See Figure 11.

Figure 11. Corrosion-fatigue and its general effect on the behaviour of steel [12]

Since corrosion attacks the surface of the part, corrosion fatigue cracks will always originate at the surface, unless the material presents defects underneath the surface, which would act as stress concentration sites and initiate sub-surface cracks [4]. Corrosion-fatigue cracks are usually trans- crystalline, although inter-crystalline cracks can nucleate if the media attacks the grain boundaries preferentially [1]. Fatigue cracks will initiate and propagate in the material depending on the metallurgical properties of each alloy; for carbon steels, for example, cracks often initiate from corrosion pits and usually contain a lot of corrosion products, normally propagate trans-granularly and present branching. However this is not a requisite; cracking in carbon steels can initiate without a corrosion pit and follow grain boundaries [4].

Most of the times corrosion fatigue produces a characteristic failure on which corrosion products can be observed on the fracture surface or at growing cracks. Nevertheless, such corrosion products could be absent in some specific cases, for example, when high-strength steels are attacked by a hydrogen- rich atmosphere. Also, when the frequency is high enough, the fracture surface produced by corrosion fatigue might not differ significantly from fatigue fracture surfaces at non-aggressive environments [4]. 2.6. Factors Influencing Corrosion Fatigue As it has been exposed in previous sections, crack initiation and crack propagation are two separate processes driven by different phenomena. It is then important to notice that the corrosive media will have a different effect on each of them. The following sections explain the factors that are known to affect one or both of these steps. 2.6.1. Environment Corrosion fatigue will be enhanced by an increased chemical activity of the environment. Some of the factors that have a strong influence on corrosion fatigue are temperature, pH, pressure of the gaseous environment, and concentration of the corrosion species. There is no rule of thumb to generalize the behaviour of materials when exposed to diverse environmental conditions since, as mentioned before, the metallurgical characteristics of each material define its behaviour combined with the environment.

However, there are some commonly observed reactions to certain individual conditions. For example, low pH, high pressure of a gaseous environment, high concentration of the corrosive species and high temperature generally have a detrimental effect on the corrosion-fatigue resistance of the material. In high-strength steels, crack growth rates are increased when the water vapour pressure increases until saturation is reached [4].

2.6.1.1. Concentration of the Corrosion Species The presence of oxygen is known to be very detrimental for the corrosion fatigue performance of many metals, since it enhances many aggressive chemical processes. Therefore, if the atmosphere has free access to the metal surface –allowing oxygen to have contact with the metal- the effect of corrosion will be favoured. For this reason, corrosion fatigue strength will be lower if the specimen is sprayed or dripped with the corrosive fluid than if it is totally submerged in it [1]. A similar effect is observed when the corrosive fluid is aerated (contains dissolved oxygen) compared to when it is deaerated, being the aerated solution the most aggressive one.

2.6.1.2. Temperature The effect of temperature is complex and can highly vary depending on the material and the temperature range. For instance, since water temperature decreases its oxygen solubility (see Figure 12), some materials could benefit from a lower oxygen level in the environment and therefore have a better corrosion-fatigue endurance at high water temperatures. On the other hand, by decreasing the oxygen content, the hydrogen evolution mechanism (typical in de-aerated solutions) might take place and cause hydrogen embrittlement [10], for which the process is explained in section 2.7.3.

Figure 12. Solubility of oxygen in water at different temperatures

2.6.1.3. Effect of NaCl It is well known that the presence of NaCl in the corrosive media causes the corrosion fatigue strength to drop drastically. The corrosion fatigue rate has a maximum level when the concentration of NaCl is around 3 – 4 wt% [15], as observed in Figure 13. This maximum is believed to be caused by the combination of effects that the amount of NaCl have in electrical conductivity and on the oxygen solubility, which are opposed. As the amount of salt is increased, the electrical conductivity increases and the oxygen solubility decreases. The shape of the curve in Figure 13 is a result of the two effects acting together in different intensities depending on the concentration of salt.

Figure 13. Effect of NaCl concentration on the corrosion of Fe [18]

Rollins et al [19] studied the influence of NaCl concentration in crack initiation and crack propagation separately for steel. The findings are shown in Figure 14. It can be observed that the number of fatigue cycles to initiate a crack decrease as the NaCl concentration increases from 0 to 4 wt%, has a minimum at 4 wt% and then increases for higher NaCl concentrations. On the other hand, crack propagation seems to be independent of the severity of the corrosive solution (which is maximum at 4 wt% NaCl), otherwise the curve would be expected to have a shape similar to the initiation. The explanations proposed by the authors was that the high conductivity at high NaCl concentrations allows cracks to grow deeper by electrochemical cell action or that the adsorption of chloride ions on the crack tip enhanced the crack propagation.

Figure 14. Effect of NaCl concentration on crack initiation and crack propagation [19]

2.6.1.4. Influence of pH Previous studies [16] have shown that the effect of low pH (1.2 and 5.5) on steel (1.0 wt%C, 1.0 wt%Cr, 0.25 wt%Mo, 0.3 wt%Si, 0.5 wt%Mn) under rotating bending testing have a little influence on fatigue crack initiation, whereas most of the reduction of endurance happened during crack propagation [15].

In other studies performed on the same steel in a 0.4% NaCl solution [16] it was proposed that in a pH range of 4 – 10, the corrosion rate was determined by oxygen diffusion to the cathodes. At pH values below 4, the corrosion rate increases due to hydrogen evolution at the cathodes and an increase in conductivity. At pH values above 12, corrosion did not occur, and it was suggested that the metal was protected by a film of metal hydroxide or adsorbed oxygen, preventing it from developing corrosion- fatigue cracks [15]. 2.6.2. Surface Finish Surface finish have an important effect on fatigue on both inert and corrosive environments. The nucleation if microcracks becomes easier when the material surface presents a rough finish. The effect of, for example, machining marks on the material surface can act as notches where cracks will start forming. Such marks also make the material more vulnerable to corrosion, whereas a polished surface is more resistant to this environmental attack.

Another possible negative effect of manufacturing on the specimen is that, besides the machining marks, the surface could have suffered from hardening by cold-work or softening by decarburization. Also, residual stresses could be introduced into the surface layers as a result of machining and preparation [1]. A positive (tensile) residual stress is highly detrimental for fatigue, while a negative (compressive) residual stress is beneficial, as it will be explained in section 2.8. 2.6.3. Stress Ratio Stress ratio, as described in previous sections, is the ratio between the minimum and the maximum applied stress in the cycle. It indicates the way the part is being loaded in terms of magnitudes of the uniaxial-stress-wave limits. Positive stress ratios indicate that both the maximum and minimum stresses are positive, i.e. tensile, and the higher the stress ratio (closer to 1), the closer the magnitudes are to each other. For dry-fatigue testing, a fully-reversed load (R = -1) is usually the most severe [7]. However, the opposite behaviour can be expected for corrosion fatigue; when the stress is positive, the crack will open, leaving the fresh material in its interior exposed to the aggressive atmosphere. The longer the time of exposure, the higher the corrosive damage suffered at the crack tip. Therefore, it can be deduced that –in general- the rates of crack propagation are increased by high stress ratios [4]. If the stress ratio is increased while the stress intensity (ΔK) is held constant, the crack tip strain and strain rate are increased. As a result, the passive film rupture is enhanced and the crack propagation is therefore increased [5]. 2.6.4. Load Frequency Frequency of the stress cycle has little effect on fatigue behaviour in nonaggressive environments [5]. However, it is the most important factor influencing corrosion fatigue for most material, environment and stress intensity conditions [5]. This happens mainly because corrosion is a time-dependant phenomenon. Then, if corrosion will have effect on the material after a certain fixed time, the frequency will determine the amount of cycles the material will reach before this corrosion time lapse is completed. Therefore, the higher the frequency, the longer the life reached by the component. In fact, “frequencies exist above which corrosion fatigue is eliminated” [4]. On the other hand, at frequencies below 10Hz the detrimental effect on the corrosion-fatigue strength is enhanced [5]. Another effect of frequency is the time the crack is open during each cycle. This is understood under the basis of fresh material exposure by the open crack: when the crack opens due to the positive stress

14 applied, the crack tip exposes fresh material to the corrosive environment, increasing the corrosive attack on each cycle. As a result, the crack growth rate is increased and the total life of the component is reduced.

Load frequency also affects the temperature of the part, depending on its material and on the stress level [1]. Generally, the higher stress levels and higher speeds, the heat generated by the deformation increases and the rate of heat dissipation by the specimen decreases. Specimen overheating is less common on rotating-bending tests than on other static tests since the specimen rotation helps the cooling by forced convection. Moreover, in corrosion fatigue tests, liquid corrosive media can aid the dissipation of the heat by the same principle. 2.6.5. Stress Intensity Although the correlation between the stress intensity and corrosion fatigue cracking varies noticeably, the general tendency is that crack growth rates in corrosion fatigue increase when increasing the stress intensity. However, the behaviour of the crack growth rate is different for each of the three different regimes (near threshold, Paris law region, and final fracture), so it is incorrect to assume that the curve in Figure 7 simply shifts to higher levels. This marked dependence becomes more pronounced for materials that are extremely environment-sensitive, such as ultrahigh-strength steel in distilled water [5]. Also in many cases, the combination of stress intensity with other factors such as stress-wave form, cycle frequency and metallurgical characteristics have diverse and even more drastic effects on crack growth rates.

In general terms, it has been commonly accepted that the fatigue behaviour of metals in aggressive environments at high stress intensities is similar to that in air or inert atmospheres. The reason is that the mechanics-controlled growth occurring in dry fatigue increases when the stress intensity factor is augmented [15]. 2.6.6. Size Effects and Hardness Gradients There are several ways to improve the hardness and toughness of a material, both generally having a beneficial effect on its fatigue resistance. In the present work, the materials studied were subjected to mechanical hardening by shot peening and to heat treatment by case hardening or through hardening. Case hardening creates a layer of harder material on the part compared to its core, whereas through hardening increases the overall hardness of the component. Further explanation on these methods is found in section 2.8 of this document. The harder the material, the more fatigue-resistant it will be, so it can be inferred that the through hardened material will generally perform better under fatigue than a case hardened one.

In non-corrosive environments, it is generally observed that when the specimen diameter increases, the fatigue limit of the material will decrease. This happens for diverse reasons related to the stress distributions throughout the geometry, and it is greatly affected by the kind of test performed, either reversed direct stress or rotating-bending [1]. For rotating-bending specifically, the maximum stress is present at the surface of the specimen. If the surface of the specimen is hardened by case hardening or cold working, surface cracks will not propagate as easy as they would without the treatment. The thickness of this hardened layer depends on the severity of the applied treatment and on the treatment itself, but it will not depend on the specimen size. Thus it can be implied that if the thickness of the hard layer is held constant (produced by the same method), the ratio of the layer thickness to the specimen diameter will decrease when the diameter increases. So a smaller specimen will have a “thicker” layer in proportion to its softer core, increasing its fatigue limit. In general, as the layer/diam. ratio increases, the closer the fatigue life approaches the one of the through hardened material [1].

Besides the general behavior of the induced hardness mentioned above, it is also important to consider the residual stresses present in the material. When the specimen is polished, for example, compressive stresses will be left on the surface, improving the fatigue and corrosion fatigue endurance of the material. The same effect is achieved by shot peening (see section 2.8.3). A thorough explanation about the relation of these variables is presented elsewhere ([1], p. 54-64).

In corrosive environments, on the other hand, the fatigue strength at a given life will increase when the diameter increases [1]. The explanation for this is that fatigue crack growth rates are relatively not affected by the corrosive media -depending on the stress level and crack length- opposite to in-air. However, crack initiation time is dramatically reduced in corrosion fatigue. Therefore, a small part of the total life is spent in nucleating a crack, and a bigger portion is spent on growth until failure. Since the crack growth rate is unaffected, the critical crack will take a longer time to propagate through larger specimens, increasing the total life at the same stress level, or the fatigue strength at the same total-life value [1]. However, by increasing the specimen diameter, the surface area also increases, leaving more material exposed to the aggressive environment, and as a result, making the specimen more vulnerable to corrosion-crack initiations [5]. 2.6.7. Metallurgical Variables Composition of the material also has an effect on fatigue, for example the addition of carbon to steel, since it increases the hardness of the material, and as a result, the fatigue limit is increased. Grain size of the steel has an indirect influence on its behaviour under fatigue since it has a close relation with the material’s strength and fracture toughness. Thus, finer grain sizes have a better fatigue strength than coarse-grained steels [7].

For steels with similar strength levels, microstructure can establish a marked difference in fatigue behaviour. For instance, pearlitic structures have a poor fatigue resistance, whereas a tempered martensitic structure provides the highest fatigue limit [7]. However, a research performed by Novak has shown that, under salty conditions, crack initiation is insensitive to microstructure of a material [10]. However, he concluded that tempered martensite and ferrite-pearlite structures have very poor response to salt containing solutions, and its application under such conditions should be avoided [11]. Some other research has shown that AISI 4130 steel exposed to salty solutions presented crack initiation sites at the grain boundaries, whereas in air environment the initiation sites were inclusions [10].

The tensile strength of a material has a different influence on fatigue and on corrosion fatigue. Generally speaking, the fatigue strength of materials will be higher if their tensile strength is higher; however the corrosion fatigue limit is normally unaffected or even decreased when tensile strength is increased [10].

Some of the factors increasing the brittle corrosion fatigue cracking are: impurities in the steel, such as phosphorus and sulphur segregation at the grain boundaries, solute (e.g. chromium) depletion or sensitization at the grain boundaries, planar deformation associated with precipitates, and large inclusions (mainly MnS). Other important factors include the presence of a martensitic microstructure in steels, carbide formation in stainless steels, and other inhomogeneities, for which a uniform distribution can hardly be assumed [13].

2.6.7.1. Inclusions Inclusions are extremely harmful for a metal exposed to corrosion fatigue conditions, and its impact is markedly higher in corrosive environments than in inert conditions [13]. Most of the literature mentions MnS inclusions as the most hazardous for steels under corrosion fatigue situations. Areas

16 with MnS inclusions create localized anodic regions, which will easily form corrosion pits [20]. A detrimental effect of the presence of these inclusions must be expected on both crack initiation and crack propagation, as explained hereby.

One of the biggest problems with MnS inclusions is that they easily dissolve in the presence of water. When a MnS inclusion is present on the surface of a steel part subjected to corrosion fatigue, it will dissolve, forming a tiny pit on the metal surface. This pit will potentially form a crack since it will act as a stress concentrator. Furthermore, it is easier for this pit to corrode since the MnS will form a highly acidic atmosphere rich in sulphides around the spot. This is associated with the total consumption of oxygen inside the crack, which forms a differential aeration cell. A gradient in corrosion potential from the surface to the crack tip is then formed, causing a shift in the pH within the crack, usually towards acidic values [13].

When a crack intersects a MnS inclusion, it will dissolve and enrich the sulphide within the occluded crack solution. These anions stimulate crack advance by increasing the anodic charge that promotes - film rupture and hydrogen embrittlement; the dissolution of MnS inclusions produces H2S and HS at the crack tip [20]. See Figure 15. In water, the local potential is critical for the hydrogen production since it determines if the process is thermodynamically feasible; hydrogen can separate from water at -0.6V [20]. As a result, a high concentration of hydrogen is present and adsorbed by the fresh metal at the crack tip [20]. Figure 15 also shows a MnS inclusion with hydrogen trapped around it, which will transform into molecular hydrogen at temperatures over 200 °C. This process increases the internal pressure, and if the inclusion cavity is sharp, a crack will form ahead of it. This is common in medium- to-low strength carbon steels [20].

Figure 15. Effect of MnS inclusions in the film rupture/anodic dissolution process [20]

2.6.8. Surface Residual Stresses Although there is no generalization that accurately predicts the effect of residual stresses on the corrosion fatigue performance of metals, it is commonly seen that compressive residual stresses improve fatigue strength, and tensile residual stresses do not [5]. The favourable effect of residual compressive stresses on the fatigue strength is greater in harder materials, whereas softer materials will have a better improvement when work-hardened [5].

The reason for the surface compressive residual stress being beneficial for the fatigue strength is that it will lower –or even cancel- the general stress caused by the loading, slowing down the formation of surface cracks. Plastic deformation can cause a gradual decrease in the compressive stress level [5]. On the contrary, tensile residual stresses will have the opposite effect. Since tensile residual stresses could be produced on the surface by machining, it is of great importance to take special care of this issue, which causes a highly harmful effect on fatigue –and corrosion fatigue- strength.

A typical way to effectively introduce surface compressive residual stresses on steel is by shot peening, carburizing and nitriding. These processes will be further explained in section 2.8. 2.6.9. Crack Closure Effects When the material undergoes the unloading part of the cycle, the crack surfaces make contact with each other, and the material relaxes. When this happens, the stress intensity at the crack tip is reduced, decreasing the rate of crack growth. This phenomenon is particularly relevant for near-threshold crack propagation, at large load applications, and when corrosive embrittlement is present [5]. Environmental embrittlement produces rough, intergranular crack surfaces, which promote crack closure since “uniaxially loaded cracks open in complex three-dimensional mode, allowing for surface interactions and load transfer” [5]. Moreover, the crack surface interactions become more relevant when the crack opening displacement is less than the fractured grain size [5].

Corrosion products forming on the cracked surface also have a great impact on the crack growth rate, and in some cases, it could slow down the growth rates to values even lower than the ones for air or vacuum [5]. This effect depends highly on the stability of the corrosion products during complex tension-compression loading and on the fluid conditions [5]. In a similar way, liquids can also enter the crack cavity and act as a wedge for the crack [6]. 2.7. Mechanisms of Corrosion Fatigue Mainly four mechanisms have been exposed in the attempt of explaining the process of corrosion fatigue. Although every mechanism is able to cover some of the aspects of this complex phenomenon, none of them is able to completely explain the whole process [10]. 2.7.1. Pitting Corrosion One of the first mechanisms proposed as an explanation for the marked reduction in fatigue endurance in corrosive environments was the formation of corrosion pits on the surface of the material, which acted as stress concentration spots where cracks were nucleated easily, as McAdam observed in 1928 [10]. Cyclic-stressing the material accelerates the corrosion pitting and causes transverse extensions of the pits, developing fissures or crevices [1].

Other studies have shown that the constant presence of the corrosive media causes the most damaging effect, compared to situations at which the aggressive environment was only present until the formation of corrosion pits. In the latter cases, the corrosion pits act as mechanical notches on the specimen, presenting considerably higher fatigue strengths at long endurances compared to those having the corrosive media applied continuously throughout the entire test [1].

Decreases in life of components under corrosion fatigue have also been observed on environments and materials that do not exhibit the formation of corrosion pits. In 1971, Laird and Duquette claimed that the formation of pits comes as a result of cracks previously formed [10]. As a result of such discrepancies, other mechanisms have been proposed to explain these behaviours. 2.7.2. Corrosion at Preferential Locations This mechanism proposes that corrosion attacks spots where fresh metal is exposed to the aggressive media, making the material vulnerable to corrosive damage. Such weak spots are created by intrusions and extrusions caused by persistent slip planes (see section 2.4, Figure 8). The attack produces a stress concentrator where the material was already highly strained, causing a decrease in the fatigue strength of the material [10].

Some studies have shown that highly deformed spots on the material are anodic with respect to undeformed areas, enhancing the electrochemical process [1]. Although this phenomenon is not thermodynamically favoured, the fresh metal is believed to present a lower activation energy and therefore is more prone to react with the corrosive environment [10]. 2.7.3. Hydrogen Embrittlement When the steel is strained in the presence of hydrogen, the material microstructure undergoes a dangerous process known as hydrogen embrittlement, which consists on the diffusion of dissolved hydrogen atoms into the metal lattice, causing its dilation and weakening its atomic bonds. See Figure 16.

The process of hydrogen embrittlement is illustrated in Figure 17. (1) First, water molecules or hydrogen ions diffuse between the crack walls to the crack tip. (2) There, electrons are discharged and (3) hydrogen atoms are reduced and adsorb at the crack tip surface (reduction). (4) These atoms will then diffuse to surface locations that are preferential for corrosion activity (surface diffusion). (5) Finally, the hydrogen atoms absorb into the metal and (6) diffuse ahead of the crack tip into critical locations such as grain boundaries (volume diffusion) [15]. This causes an embrittlement of the material surrounding the crack tip, which promotes further propagation of the crack.

Figure 16. Hydrogen embrittlement Figure 17. The hydrogen embrittlement process [36]

Hydrogen embrittlement is typical in high-strength alloys. Since this kind of damage depends on diffusion and adsorption, it is considered to be a time-dependent mechanism. Contrary to the dissolution mechanism, this process does not form any kind of passivating oxide layer, so the hydrogen diffusion can occur during the whole load cycle [15]. The addition of oxygen in small amounts is known to practically eliminate the brittle corrosion-fatigue mechanism in crack growth [13].

2.7.4. Rupture of Oxide Protective Film Some metals, such as aluminium, copper and stainless steel, form a protective oxide layer, which prevents the material underneath from further corrosive attack. Therefore, this mechanism proposes that corrosion fatigue occurs as this protective film is broken and corrosion fatigue cracks form easily. Since the present work does not deal with metals forming protective oxide films, this mechanism will not be considered here. 2.7.5. Surface Energy Reduction The reduction of surface energy due to the adsorption of environmental species is known as the Rebinder mechanism, which initially stated that the surface-active agent adsorbed into the cracks, increasing its internal pressure, resulting in the propagation of the crack. Later on, this theory was modified to state that the adsorbing species actually reduced the surface energy of the material, allowing for an easier formation of protrusions resulting from slipping bands [10]. Although these theories seem reasonable, there is not enough data nowadays to prove its veracity. 2.8. Heat and Surface Treatments There exist some methods to selectively change the properties of a steel at the surface –or of the whole body-, improving its fatigue resistance considerably. These methods can be divided into surface hardening by heat treatment and mechanical working by inducing deformation on the surface. Improvements are due to two different factors: the change in the microstructure and the residual compressive stresses [21]. In surfaces hardened by heat treatments (quenching, for instance), the layer must be deep enough to protect the core from operating forces, but thin enough to maximize the effectiveness of the residual stresses [7]. This is controlled by the heating and cooling parameters, as well as by the carbon content of the steel and its composition. In the case of mechanical working, shot peening and skin rolling are very popular methods. 2.8.1. Case Hardening This thermo-chemical process consists on exposing the finished component to a carburizing atmosphere at a high temperature, ranging between 850-950 °C. As a result, a layer rich in carbon is formed at the surface of the material. After carburising, the component is quenched, which increases its hardness. Quenching causes the carburized layer to transform into martensite, for which the hardness will be in function of its carbon content. The resulting component will have a hard surface and a softer-tougher core [24].

The carbon content and other alloying elements (such as Ni, Mn, Cr and Mo) in the steel will determine its hardenability [24]. The carburised surface usually presents residual compressive stresses as a result of the case hardening, ranging between -150 to -250 MPa [24]. The depth of hardening is chosen depending on the function of the component, and it is achieved by adjusting the parameters of the case hardening process. 2.8.2. Through Hardening Through hardening is a thermal process in which the component is subjected to quenching, i.e. it is heated up and rapidly cooled, to achieve a martensitic or bainitic microstructure throughout its cross section. If the cooling process is not fast enough, the resulting microstructure could result in ferrite, pearlite, or upper-bainite [24]. For large components, the quenching might not be enough to cool down the core of the piece, resulting in an incomplete martensite or bainite transformation in such regions. This will cause a hardness gradient, decreasing from surface to core [24].

2.8.3. Shot Peening Shot peening is a kind of cold-working process that consists on bombarding the surface of the material with small metal hard particles, such as metal, glass or ceramic spheres, or small cylinders made with cut wire. As a result, the surface deforms and changes its properties. The hardness of the deformed material increases, and a residual-compressive-stress layer is formed at the surface. The compressive stress from shot peening forms a gradient with a characteristic shape, as seen in Figure 18, for which the residual stress reaches a minimum (maximum compressive) underneath the surface, and then it relaxes again.

Figure 18. Shot peening compressive stress profile [21]

The improvement in fatigue resistance on the material is caused by the combined effect of cold working and compressive stresses. Cold working involves work hardening and the closure of pores and surface cracks [21]. The results of the shot peening depend on the intensity of the process and on the material being treated. The intensity of the shot peening process is affected by several variables, such as the hardness and size of the balls, speed of the shots, coverage, etc. In general terms, the higher the intensity, the greater the resultant residual compressive stresses and the deeper they go into the material. The result of the shot peening also depends on the material properties. For instance, a higher yield strength of the material will allow higher compressive stresses since any residual compressive stress larger than the yield strength of the material will cause relaxation without any external load applied [22]. 2.9. Fatigue Testing Despite of the enormous amount of money, time and effort spent on studying fatigue and trying to generate valuable data for preventing unpredicted fatigue failure in engineering applications, a scientific basis for reliable estimates of fatigue life for all possible combinations of load and environmental conditions remains intangible. Furthermore, existing approaches to standard test development can result non-conservative in cases where two or more phenomena interact synergistically against the performance of the material. Thus, a designer must rely on experience along with research data to make decisions about the components in service [5].

Laboratory corrosion-fatigue tests can be classified as either cycles-to-failure or crack propagation tests. The former consists on applying a sufficient number of cycles to cause complete fracture of the

21 specimen. With this kind of test it is difficult to identify crack initiation and propagation stages of the fatigue process; nevertheless, the fraction of the total life spent in initiating a crack is generally estimated in the ranges described in section 2.4 [5]. The result of these kinds of tests might not be accurately representative of large components when the tests are performed on small specimens, but the results can be effectively used to compare alloying additions, heat and surface treatments and finishes, corrosion-control methods, etc., since they provide valuable “empirical data on the intrinsic fatigue crack initiation behaviour of a metal or alloy” [5]. Therefore, in the present work, the cycles- to-failure type of test was used for the experimentation.

Fatigue testing methods are firstly characterized by the loading mode of the specimen. Mainly three different loading modes are applied: direct axial loading, plane-bending, and rotating-beam loading [25]. Depending on the loading mode, the design of the machine and the shape of the specimen can widely vary. In general, the testing machines are classified by the basic drive mechanism they use, and the test parameter they allow to control [25]. In the present work, the rotating-bending test is used, and therefore it will alone be described hereon. 2.9.1. Rotating-Bending Test The specimen used for this test method is generally a cylindrical rod with a waist at the centre of the body to concentrate the stress. This kind of test consists on holding the specimen from both extremes using special grips mounted on bearings that allow rotation. One of the extremes is rotated, generally by an electric motor. The specimen is then bent by the application of death-weight loading. Thus, the specimen surface is subjected to a sinusoidal stress loading, as shown in Figure 1, with maximum values at the centre of the waist where the diameter is minimum. Rotating-bending tests generates a fully- reversed loading cycle. Because of the characteristic loading method, it becomes impossible to run mean-stress effect tests with this machine [25]. Since only a small part of the material at the surface of the specimen will be loaded with the maximum stress at a time, the fatigue strengths and lives obtained from rotating-bending tests will be typically higher than those obtained with axial-fatigue tests [25].

Figure 19 shows an illustration of how a cantilever loading rotating-bending machine is designed.

Figure 19. Rotating-bending machine diagram

2.9.2. The Staircase Method One of the most widely used testing methods to determine the statistical distribution of the fatigue strength of a material –usually providing conservative results [3]- is known as the staircase method. It consists on running a set of tests around the estimated fatigue limit, and use the collected data to calculate the fatigue strength statistically. The way to choose the load at which each individual test will be conducted depends on the result of the previous test, as explain further on.

The first step is to select a certain number of cycles to be considered as the infinite life, for instance, 106 cycles. Then, a value for the fatigue limit is estimated from either experience or preliminary S-N data [26], and a fatigue-life test is performed at a stress level just above the estimated value. If the specimen fails before reaching the chosen infinite life, the next test is to be performed at a stress level below the previous one. Else, if the specimen survives, the following test must be performed at a stress level higher than the previous one. This process is then repeated as many times as needed, usually between 20 and 30 [2, 3]. The stress increment –or reduction- is usually decided to be in a range from half to twice the standard deviation of the fatigue limit [3].

The staircase method does not deal with the number of cycles to failure. Instead, it handles the data in a “pass/fail” manner [26]. Once the data is collected, the results can be visualized in a plot as shown in Figure 20. The Dixon-Mood method can then be applied to statistically calculate the mean, µS, and the standard deviation, σe, of the fatigue limit. This method assumes that the fatigue limit follows a normal distribution; if the data does not satisfy this requirement, a transformation (logarithmic, power, squared, cubic, etc.) of the stress values can be applied [26].

Figure 20. Example of staircase fatigue data [3]; suspensions are tests in which the specimens survived

The Dixon-Mood method requires the use of the least frequent event (either survivals or failures) to estimate the mean and standard deviation of the fatigue strength. In the plot shown in Figure 20, for instance, the least frequent event is survivals (denoted as “suspensions”), represented by the white circles. The Dixon-Mood method can be expressed in the following equations [26]:

푖푚푎푥 푖푚푎푥 푖푚푎푥 (2) 2 퐴 = ∑ 푚푖 , 퐵 = ∑ 푖 ∙ 푚푖 , 퐶 = ∑ 푖 ∙ 푚푖 푖=0 푖=0 푖=0

퐵 (3) 휇 = 푆 + 푠 ∙ ( ± 0.5) 푆 0 퐴

퐴 ∙ 퐶 − 퐵2 퐴 ∙ 퐶 − 퐵2 (4) 휎 = 1.62 ∙ 푠 ∙ ( + 0.029) if ≥ 0.3 푒 퐴2 퐴2

퐴 ∙ 퐶 − 퐵2 (5) 휎 = 0.53 ∙ 푠 if < 0.3 푒 퐴2

where i is an integer of the stress level, and the parameter imax is the highest stress level in the staircase.

If the least frequent event is survivals, then S0 will be the lowest stress level at which a specimen survived, and it will correspond to the i = 0 level; contrary, if the least frequent event is failures, S0 will have the value of the lowest stress level at which a failure occurred, and it will correspond to the i = 0 level. The parameter mi is the number of specimens that failed at each stress level i. and s represents the stress step size. In equation (3), the (+) sign is used when the least frequent event is survivals, and the (–) sign is used when the least frequent event is failures [26]. 2.9.3. The Median S-N Test Method for Small Sample Size This method was described by the Japan Society of Mechanical Engineers (1981), and it is used as a guideline to determine the S-N curve with a reliability of 50% and a minimum sample size [3]. It requires at least 14 specimens, 8 for the finite-life region (S-N curve) and 6 for the staircase method. The arrangement and sequence of the tests is suggested as shown in Figure 21. Note that the value of the stress levels and number of cycles in this illustration are mere examples.

Figure 21. S-N testing with a small sample size [3]; suspensions are tests in which the specimens survived

For the generation of the S-N curve, it is recommended that more than one specimen is tested at each stress level in order to produce replicable data, which is necessary to estimate the statistical distribution and variability of the fatigue life [3]. For preliminary and R&D tests, it is recommended to use between 6 – 12 specimens to generate an S-N curve. The percent replication (PR) for R&D testing is recommended to be between 33 – 50. This value indicates the portion of the sample size that can be used to determine an estimate of the variability of replicate tests [3], and it is calculated from equation 6 shown below:

퐿 (6) 푃푅 = 100 ∙ (1 − ) 푛푠 where 푛푠 is the number of samples and 퐿 is the number of stress levels.

3. Materials and Equipment The procedure followed in the present work was adapted to the equipment available in the materials laboratory at Atlas Copco Secoroc AB in Fagersta. The fatigue tests were performed in rotating-bending machines designed to test specimens with certain dimensions and geometries, which are detailed in section 3.2.1. Two materials were tested, each one with different heat treatments. Shot peening was applied on half the specimens of each of the materials, resulting in a total of four different material conditions, as explained in section 3.2.2. 3.1. Laboratory Equipment The materials laboratory at Atlas Copco Secoroc AB has two rotating-bending fatigue testing machines. A picture of these machines is presented in Figure 22. An electric motor drives one of the rotating grips, while the other grip is pulled by a lever connected to dead-weights by pulleys and a cable (see Figure 23), creating the cantilever loading effect on the specimen. The corrosive conditions are created by means of a water supply system, which consists of a stainless-steel water tank attached to the machine, a water pump with maximum flow capacity of 12 l/min installed on one of the sides of the tank, a closed chamber to cover the specimen, a hose connecting the pump with the chamber, and a water collector at the bottom of the chamber which redirects the water into the tank.

Figure 22. Rotating-bending fatigue testing machine - Atlas Copco Secoroc AB materials laboratory

Figure 23. Rotating-bending fatigue testing machine, rear side - Atlas Copco Secoroc AB materials laboratory

Figure 24 shows a specimen mounted in the machine without the chamber housing installed. The two white plastic washer-like parts are put around the specimen to prevent the water from flowing outside the chamber. The water collector can be observed under the specimen. The chamber can be seen in Figure 25.

Figure 24. Specimen mounted on the machine and water collector underneath

Figure 25. Corrosion chamber

The minimum weight that the machine can loaded with is 2.836 kg, which corresponds to the weight of the basket on which the extra load is carried. This would mean that the basket would be left alone with no extra weight added. For these machines, this minimum load corresponds to a stress of 83.6 MPa on a standard specimen with a waist diameter of 15 mm. 3.2. Specimens 3.2.1. Geometry and Dimensions As mentioned in section 2.9.1, the specimens used for the tests were cylindrical bars with a waist in the centre to create a stress concentration. The geometry of the specimens is illustrated in Figure 26. The surface roughness defined in the drawing is for the through hardened material –without shot peening- after its final polishing. Nonetheless, all the other dimensions apply for the case hardened specimens as well. The minimum diameter of the waist is 15 mm and the tolerance for this dimension was always respected, even on the shot peened specimens, which have a rougher surface and generally tend to have a slightly larger diameter.

Figure 26. Geometry and dimensions of specimen

Before testing, the diameter of each specimen was measured using a Mitutoyo laser scan micrometer, which is shown in Figure 27 and Figure 28. With this device, the exact diameter and the ovality of the waist was measured.

Figure 27. Laser scan micrometer

Figure 28. Laser scan micrometer display

3.2.2. Material The two steels used during the present work were a case hardened (which in the present document will be referred to as CH) and a through hardened (which will be referred to as TH) steel grades. Half of the population of each steel was shot peened, resulting in four different material conditions, as shown in Table 1.

Table 1. Different material scenarios used in the present work

Shot peened Non-shot peened Case Hardened (CH) CH-SP CH-NSP Through Hardened (TH) TH-SP TH-NSP

Figure 29 shows a picture of one specimen of each material condition studied in the present work.

Figure 29. Specimens used. From left to right: (1) Case hardened, non-shot peened; (2) Case hardened and shot peened; (3) Through hardened, non-shot peened; (4) Through hardening and shot peened

3.2.3. Heat Treatment Case hardening in CH steel was performed within Atlas Copco’s facilities in Fagersta. After the treatment, the surface hardness was HRC 57, and the core hardness was HRC 43. The hardening depth was 1.4 – 1.2 mm. As a result of the heat treatment, the case hardened parts will have a hard case and a tough core.

The through hardening process was not performed by Atlas Copco. The steel grade was heat treated by the supplier before delivery. The hardness of this material is, by standard, HRC 49 - 51 and is constant throughout the cross section of the parts. 3.2.4. Shot Peening The CH steel (CH-SP) was shot peened with an intensity of 0.58-0.59 mm A. The TH steel (TH-SP) was shot peened with an intensity of 0.66 mm A. The difference in the intensity was due to a maintenance adjustment made on the shot peening machine between the two processes, which were performed in different days. The residual stresses achieved by the shot peening are presented in section 3.2.4.1.

3.2.4.1. Residual Stress Analysis Specimens from the four different conditions were analysed for residual stresses on the surface using the X-Ray diffraction method, which is explained elsewhere ([27]). The measurements were not performed within the Atlas Copco laboratories. The results are shown in the following figures. In Figure 33, Phi=0 corresponds to axial direction and Phi=90 to circumferential direction. The measurement on the axial direction is the one to be taken into consideration, since it is the one affected by the bending moment of the test. The circumferential direction is not relevant for the stress state during the test as long as it has the same sign as for the axial direction. Then, it will not be part of the biggest Mohr’s circle being responsible for the maximum shear stress in the material.

Figure 30. Residual stresses on CH-SP Figure 31. Residual stresses on CH-NSP

Figure 32. Residual stresses on TH-SP Figure 33. Residual stresses on TH-NSP

From Figure 30 and Figure 32, it can be observed that the residual stresses on both shot peened materials is negative with a value of around -400 MPa at the surface, going down to around -900 MPa at 0.07 mm for CH-SP and down to -800 MPa at 0.2 mm for TH-SP.

For CH-NSP, shown in Figure 31, it can be seen that a tensile residual stress is present at the surface, but then it decreases suddenly down to approximately -150 MPa, and it remains negative as the measurement goes deep inside the specimen for more than 1 mm. In the case of TH-NSP, shown in Figure 33, the residual stress is negative on the surface in the axial direction, and then it goes close to zero as the depth increases.

4. Experimental Setup and Procedures The present work was divided into three main steps. First, a corrosion-fatigue testing method was proposed, and the testing parameters were defined. The second step consisted on using the method to test different material conditions and make a comparison to prove the validity of the proposed method. The third step comprised a set of metallographic examinations of the tested specimens to observe the cracking phenomena. The details of these three steps are presented in this section. 4.1. Corrosion-Fatigue Testing Method Proposal and Testing Parameters Selection It was decided that the corrosion-fatigue testing method would cover the materials’ behaviour on both finite and infinite life ranges, i.e. it would involve the S-N curve and the fatigue strength. To be able to compare fatigue results from wet and dry conditions, one key parameter taken from the dry fatigue method is the infinite life value, which is considered to be at 3X106 cycles. The proposed method is shown in section 5.1.

The results obtained with the proposed testing method should permit the visualization of well-defined differences between the materials and scenarios. Since the results from fatigue tests are strongly dependent on the testing parameters (as explained in section 2.6), a right set of testing parameters is required to achieve such a good method. It was crucial to select the main factors influencing corrosion fatigue -based on their relevance on the real-product application and on the capabilities of the laboratory equipment- and perform a series of tests varying the values of such factors to observe their effect on the results. From these observations, the values for these parameters were selected for the final method.

It was decided that frequency was the main parameter defining the new corrosion-fatigue testing method. Therefore, frequency dependence of corrosion fatigue on CH-SP and TH-NSP was studied in order to choose a frequency that would allow a proper comparison of the materials and a good visualization of the data in the proposed testing method. The machines allow a rotating speed in the range of 350 – 3000 rpm. Currently the company performs dry-fatigue tests at 2300 rpm. The chosen range for testing frequency dependence in the present work was therefore 500 – 2300 rpm.

The frequency dependence analysis consisted of a set of tests at fixed load but different rotating speeds: 500 rpm, 1000 rpm, 1600 rpm, and 2300 rpm. The load for these tests was the R95C90 lower bound from the dry fatigue limit, which means that there is a 95% reliability with a 90% confidence interval that the specimens would survive at this load in dry conditions [3]. The one-side lower-bound value can be calculated with equation 7 [3]. The calculations are shown below:

푆푒,푅,퐶 = 휇푆 − 퐾 ∙ 휎푒 (7)

For CH-SP:

Dry fatigue data: µS = 725 MPa ; σe = 26.5 MPa The K factor has a value of 2,894 for R95C90 lower bound and n = 7 [3] From equation 7, R95C90 = 650 MPa

For TH-NSP:

Dry fatigue data: µS = 744 MPa ; σe = 26.5 MPa The K factor has a value of 2,755 for R95C90 lower bound and n = 8 [3] From equation 7, R95C90 = 670 MPa

The results from the frequency dependence tests are presented in section 5.1.1.

Regarding the loading characteristics, the stress ratio R was -1, with a sinusoidal wave shape. These variables are fixed for rotating-bending fatigue testing machines.

In terms of the environment, water conditions were kept constant except for the NaCl content. Water was taken from the tap found in the laboratory, and it was expected to have the same characteristics throughout the whole experimental work timeframe. It is important to mention that the water was renewed every test without exception. Temperature of the water was not varied, and was found to be between 20 and 23 °C, sometimes increasing up to 25.6 °C during long-lasting tests and warmer days. Such variations were neglected since they are considered to be too small to have a significant effect on the results, for instance in the oxygen content, as seen in Figure 12. The pH was also kept constant in a value of 7, as coming from the tap. Oxygen concentration in the water was not controlled with detail, but during the experiments, continuous stirring of the water was ensured to allow constant dissolution of oxygen. The level of oxygen dissolution is not of vital importance, since all the tests – during this work and in the future- are to be performed in the same way, which ensures equal water characteristics on every test, allowing to fairly compare the results. The dependence on NaCl content in the water was tested, performing a set of tests with water containing 4 wt% NaCl.

Further tests of frequency dependence were carried out at a load level corresponding to each material’s dry fatigue limit with the purpose of comparing the results with the ones from R95C90 lower bound stress level. The intention was to prove that the frequency dependence results showed a different trend for each load level, i.e. that the frequency dependence response changes depending on the applied stress. However, these tests were not extensive, and only one specimen was tested at each test speed. The results of these tests are shown in Figure 39 and Figure 40. 4.2. Testing the Proposed Corrosion-Fatigue Method Once the testing method and the parameters were decided, the four material conditions presented in Table 1 were tested using this method in order to prove its functionality. The results for these tests are shown in 5.2. As well as for the frequency dependence tests, the water was renewed for every test.

The approach in the present work was to run corrosion fatigue tests at the same load levels as the dry tests, even though a lower fatigue strength value was expected from corrosion fatigue testing; this way, the S-N curve for corrosion fatigue could be generated in the load range going from the dry fatigue limit down to the corrosion fatigue strength. The separation between the load levels used in the S-N curves was chosen based on the results obtained in the first tests. For instance, if the life of the specimen in the first level was low, the decrement in load for the next level should be large, and vice versa. The fatigue strength for every condition was found using the staircase method.

The first material condition tested was CH-SP. The load for the first test was chosen to be the dry fatigue limit of that material condition, which is 725 MPa. The results for this material condition are shown in Figure 42 and Figure 43. TH-NSP was tested in the same way. For this material condition, the dry fatigue limit is 745 MPa, and the first test was also performed at this load level. The results for TH- NSP are shown in Figure 48 and Figure 49. Then, CH-NSP was tested. Since the steel in CH-NSP is the same as in CH-SP, and only shot peening differentiates these two conditions, the load for CH-NSP was chosen as 725 MPa for the first test, even though the dry fatigue limit is 625 MPa. This would allow a comparison between conditions at high loads. The plots for CH-NSP are presented in Figure 44 and Figure 45. Finally, TH-SP was tested following the same steps. Since this material condition has not been tested in dry fatigue yet, the dry fatigue limit was taken from its partner condition (TH-NSP), so the load for the first test of TH-SP was 750 MPa (745 MPa + 5 MPa to round up the value). The results for TH-SP are presented in Figure 46 and Figure 47.

4.3. Post-Test Analysis After the tests were performed, a set of specimens was selected based on the testing parameters and on the quality of the fracture surface, to be later analysed in SEM and optical microscope. The purpose of these analyses was to characterize the fracture surface, and also to try to understand the cracking process by observing the morphology of the cracks. 4.3.1. Microscope Analysis A set of specimens was selected to be prepared as metallographic samples in order to examine the cracks on a cross-sectional view using a metallographic microscope available in the company’s laboratory in Fagersta. The selection of the specimens is described in section 4.3.2. The pictures obtained in this analysis are shown in section 5.3.2.

Some specimens which failed during the test were also selected in order to analyse the fracture surface on a Scanning Electron Microscope (SEM). The purpose of these observations was to show the crack initiation sites, the morphology of the crack-propagation surface and inclusions on the cracking path. The results of this analysis is presented in section 5.3.1 4.3.2. Metallographic Samples The selection of the specimens was based on the conditions of the test they represent. In general, two specimens of each material condition were chosen: one for the highest load and one for the lowest load tested for the condition. Some other specimens were also selected based on the relevance of the resulting observations, for instance, TH-NSP on low speed values or CH-SP in salty water. Table 2 shows the specimens selected and prepared as metallographic samples.

Table 2. Metallographic samples selection

Stress Level during Test Testing Speed Material Condition Specimen Number [MPa] [rpm] 093 750 1000 016 200 1000 TH-NPS 029 680 1000 036 680 500 012 800 1000 TH-SP 013 550 1000 045 725 1000 CH-NSP 095 200 1000 091 800 1000 CH-SP 047 500 1000 CH-SP in 028 650 500 4.0 wt% NaCl water 032 650 2300

5. Results This section is divided into three parts: the first subsection (section 5.1) presents the frequency dependence plots and the final selection of testing parameters for the proposed method; the second subsection (section 5.2) presents the corrosion-fatigue test results obtained with the proposed method; the third subsection (section 5.3) offers a group of metallographic analyses performed on the already-tested specimens with the intention of understanding the mechanisms involved in cracking and fracture of the materials compared with the proposed method. 5.1. Presentation of Corrosion-Fatigue Testing Method The proposed method for testing corrosion fatigue was the “Median S-N Test Method for Small Sample

Size”, described in section 2.9.3. It consists in building the S-N curve with 8 samples (ns = 8) and 4 stress levels (2 samples on each stress level), and the staircase with 6 samples. The calculated percent replication (PR) from equation 6 is shown in Table 3.

Table 3. Percent Replication

Parameter Value Comment

풏풔 8 OK for preliminary and R&D tests (ASTM, 1998) Stress levels 4 From Nakazawa and Kodama, 1987 PR 50 OK for R&D tests (ASTM, 1998)

5.1.1. Frequency Dependence Results The frequency dependence results for CH-SP and TH-NSP are presented below. Figure 34 shows the corrosion-fatigue frequency dependence of CH-SP in fresh water and brine (water with 4 wt% NaCl). For the values obtained in fresh water, the averages are also shown. Trendlines are included in the plot.

4.0E+06

0% NaCl 3.5E+06 4% NaCl Average 3.0E+06

2.5E+06

2.0E+06

1.5E+06 Number of cycles Numbercycles of failure to 1.0E+06

5.0E+05 y = 128.51x + 525843 R² = 0.5451

0.0E+00 0 500 1000 1500 2000 2500 Test speed [rpm] Figure 34. Corrosion-fatigue frequency dependence on CH-SP

The chart shown in Figure 35 shows the average time spent on the tests performed for the frequency dependence study, so it represents the data shown in Figure 34.

25 0% NaCl 4% NaCl

Time Time ] [hours 10

0 500 1000 1600 2300 Speed [rpm]

Figure 35. Average time spent per test for different testing speeds on CH-SP in fresh water and brine

The values of time for 0% NaCl in Figure 35 are shown in Table 4 below, including the average number of cycles from the fresh water tests plotted in Figure 34.

Table 4. Average values for number of cycles to failure and time of test for frequency dependence of corrosion fatigue on CH- SP in fresh water

Speed Average number Number Average time of [rpm] of cycles of tests test [hrs] 500 382 450 2 12.75 1000 1 165 266 3 19.42 1600 1 878 466 3 19.57 2300 2 988 566 3 21.65

Frequency dependence for TH-NSP was also proven, and the results are shown in Figure 36. The effect of NaCl in the water was not tested on this steel grade. The plot in Figure 37 shows the average time spent by test on each testing speed, and Table 5 shows the corresponding values.

1.4E+05 0% NaCl Average 1.2E+05 y = 3442.4x0.4364 R² = 0.8839 1.0E+05

8.0E+04

6.0E+04

Number of cycles Numbercycles of failure to 4.0E+04

2.0E+04

0.0E+00 0 500 1000 1500 2000 2500 Test speed [rpm]

Figure 36. Corrosion-fatigue frequency dependence on TH-NSP

0% NaCl

1.5

1 Time Time ] [hours

0.5

0 500 1000 1600 2300 Speed [rpm]

Figure 37. Average time spent per test for different testing speeds on TH-NSP in fresh water

Table 5. Average values for number of cycles to failure and time of test for frequency dependence of corrosion fatigue on TH- NSP in fresh water

Speed Average number Number Average time of [rpm] of cycles of tests test [hrs] 500 50 433 3 1.68 1000 73 933 3 1.23 1600 87 366 3 0.91 2300 98 566 3 0.71

A side-to-side comparison of corrosion-fatigue frequency dependence between CH-SP and TH-NSP is shown in Figure 38.

1.0E+07

CH-SP

TH-NSP 1.3015 Potencial (CH-SP) y = 124.28x R² = 0.9292 Potencial (TH- NSP)

1.0E+06 Number of Cycles NumberCycles of tofailure 1.0E+05

y = 3442.4x0.4364 R² = 0.8839

1.0E+04 0 500 1000 1500 2000 2500 Test speed [rpm]

Figure 38. Corrosion-fatigue frequency dependence comparison between CH-SP and TH-NSP

The plots in Figure 39 and Figure 40 show the results of the corrosion fatigue tests performed at a load level corresponding to the dry fatigue limit of each material. For CH-SP, this limit is found at 725 MPa, and the R95C90 value is 650 MPa. For TH-NSP, the dry fatigue limit is 744 MPa and its R95C90 value is 670 MPa.

4.0E+06 R95C90 3.5E+06 Average N Dry fatigue limit y = 124.28x1.3015 3.0E+06 R² = 0.9292

2.5E+06

2.0E+06

1.5E+06

Number of Cycles NumberCycles of toFailure 1.0E+06 y = 470245ln(x) - 2E+06 5.0E+05 R² = 0,9333

0.0E+00 0 500 1000 1500 2000 2500 Test speed [rpm]

Figure 39. Corrosion-fatigue frequency dependence of CH-SP at two load levels

1.4E+05

R95C90 1.2E+05 Average N y = 3442.4x0.4364 Dry fatigue limit R² = 0.8839 1.0E+05

8.0E+04

6.0E+04 y = 2114,4x0,4584 R² = 0,9607

4.0E+04 Number of Cycles NumberCycles of toFailure

2.0E+04

0.0E+00 0 500 1000 1500 2000 2500 Test speed [rpm]

Figure 40. Corrosion-fatigue frequency dependence of TH-NSP at two load levels

5.1.2. Final Parameters for Proposed Testing Method Figure 41 shows the same frequency dependence results for CH-SP in fresh water presented in Figure 34. It can be noticed that a testing speed of 1000 rpm fits the needs of the proposed method in the best way, as it is further discussed in section 6. Thus, this value can be found in the final parameter selection shown in Table 6.

4.0E+06 0% NaCl 3.5E+06 Average

Infinite Life 3.0E+06

2.5E+06

2.0E+06 Marked Difference

1.5E+06 Number of cycles Numbercycles of failure to 1.0E+06 Small Spread 5.0E+05

0.0E+00 0 500 1000 1500 2000 2500 Test speed [rpm] Figure 41. Testing speed selection

The proposed testing parameters are presented in Table 6.

Table 6. Parameters for Proposed Corrosion-Fatigue Testing Method

Parameter Value Comment Loading Characteristics Rotating speed 1000 rpm From frequency dependence Stress ratio -1 For rotating-bending test Stress amplitude variable Dependent on stress level Stress-wave shape sinusoidal For rotating-bending test Water Characteristics pH 7 Tap water NaCl content 0 wt% NaCl Tap water Oxygen content Full Saturation Constant stirring Temperature 20 – 23 °C Room temperature Other Parameters Infinite life 3X10^6 cycles Same as for dry fatigue tests Staircase load increment 50 MPa Same as for dry fatigue tests

5.2. Testing the Proposed Corrosion-Fatigue Testing Method 5.2.1. S-N curves and fatigue strength results The proposed corrosion-fatigue testing method was performed on the four material conditions using the selected parameters presented in Table 6 in order to prove its functionality. With this method, the S-N curves and fatigue strengths for each of the four conditions were generated.

For CH-SP, the data forming the finite life curve adjusted properly to a logarithmic regression, as it can be seen in Figure 42. The highest stress level tested for this condition was 800 MPa, and the mean life expected is 3X10^5 cycles. As the load decreases, this material condition reaches infinite life at a mean stress of 525 MPa, as shown in Figure 43.

850 Failures 800 Run outs 750

700

650 y = -114.2ln(x) + 2240.3 R² = 0.9173

600 Stress Stress [MPa]

550

500

450

400 1.0E+05 1.0E+06 1.0E+07 Number of Cycles Figure 42. S-N curve for CH-SP

650 Runouts

Failures 600 Fatigue strength @ 3X10^6 cycles 550 525 ± 42.8 MPa

500 Stress Stress [MPa]

450

400 0 1 2 3 4 5 6 7 8 9 10 Test number Figure 43. Corrosion-fatigue strength of CH-SP from staircase method

For the case of CH-NSP, the data for the S-N curve, shown in Figure 44, was better adjusted with a power regression. The highest stress tested for this condition was 725 MPa and the expected life at this load is around 4X10^4 cycles, significantly lower than the shot peened version. The corrosion- fatigue strength also dropped considerably without shot peening, reaching a value of 208.3 MPa, shown in the staircase plot in Figure 45.

800.00

Failures 700.00 Run outs

600.00

500.00 y = 18930x-0.308 R² = 0.9438

400.00 Stress Stress [MPa] 300.00

200.00

100.00

0.00 1.0E+04 1.0E+05 1.0E+06 1.0E+07 Number of cycles

Figure 44. S-N curve for CH-NSP

300 Runouts

Failures

250 Fatigue strength @ 3X10^6 cycles

200 208.3 ± 26.5 MPa Stress Stress [MPa]

150

100 0 1 2 3 4 5 6 7 8 9 10 Test number Figure 45. Corrosion-fatigue strength of CH-NSP from staircase method

For the case of TH-SP, the maximum stress level tested was also 800 MPa, and for this material condition, the life expected at this load is around 1.6X10^5 cycles, which is a bit lower than for CH-SP. The S-N curve for this condition is shown in Figure 46. The data for the finite life was also well adjusted to a logarithmic regression. The staircase method, shown in Figure 47, determined a corrosion fatigue strength of 600 MPa.

850

Failures 800 Run outs

750

700

650 Stress Stress [Mpa] y = -80.77ln(x) + 1767.7 R² = 0.9298 600

550

500 1.00E+05 1.00E+06 1.00E+07 Number of cycles

Figure 46. S-N curve for TH-SP

700 Runouts

Failures

650 Fatigue strength @ 3X10^6 cycles

600 ± 26.5 MPa

600 Stress Stress [MPa]

550

500 0 1 2 3 4 5 6 7 8 9 10 Test number Figure 47. Corrosion-fatigue strength of TH-SP from staircase method

Lastly, the S-N curve for and staircase for TH-NSP is presented in Figure 48 and Figure 49, respectively. The highest stress level tested for this condition was 750 MPa, for which the expected life is 5X10^4, which is slightly higher than the CH-NSP. The corrosion fatigue strength was found to be 191.6 MPa.

800

Failures 700

Run outs 600

500

400 y = -165ln(x) + 2523.7

Stress Stress [Mpa] R² = 0.9843 300

200

100

0 1.0E+04 1.0E+05 1.0E+06 1.0E+07 Number of cycles

Figure 48. S-N curve for TH-NSP

300 Runouts

Failures

250 Fatigue strength @ 3X10^6 cycles

200 191.6 ± 26.5 MPa Stress Stress [MPa]

150

100 0 1 2 3 4 5 6 7 8 9 10 Test number

Figure 49. Corrosion-fatigue strength of TH-NSP from staircase method

5.2.2. Materials Comparison The following plots present a comparison between the four material conditions tested. Different combinations of conditions are also shown in an attempt to expose important observations regarding the performance of each condition.

First, a comparison between the four material conditions is presented in Figure 50, containing the finite life predicted by the S-N curves and the corrosion-fatigue strengths of each one of the conditions.

900

800

700

600 MPa 600

525 MPa

500 Stress Stress (Mpa)

400

300 CH-SP TH-SP 208.3 MPa 200 CH-NSP TH-NSP 191.6 MPa Infinite Life - 3X10^6 cycles 100 1.0E+04 1.0E+05 1.0E+06 1.0E+07 Number of Cycles

Figure 50. Comparison of the four conditions (S-N curves and corrosion-fatigue strengths)

Figure 51 shows a magnification of the shot peened conditions, since they appear small in Figure 50.

850

800

750

700

650 Stress Stress [MPa]

600 MPa 600 CH-SP TH-SP 550 CF Strength CH-SP CF Strength TH-SP 525 MPa Infinite Life - 3X10^6 cycles 500 1.0E+05 1.0E+06 Number of cycles

Figure 51. S-N curve comparison between CH-SP and TH-SP

Finally, a comparison between the corrosion fatigue and dry fatigue results is presented. It is important to remember that the latter were obtained only for staircase testing, but they are easily included in these plots since a record of the number of cycles achieved by the fractured specimens is kept by the company.

Figure 52 shows the two versions of CH under corrosion fatigue, and also the results from dry testing, for both shot peened and non-shot peened specimens of the same steel. Figure 53 presents TH-SP and TH-NSP under corrosion fatigue, and the dry fatigue data of the TH-NSP condition generated for the staircase. As mentioned earlier, the TH-SP condition has not been tested in dry conditions yet.

900 CH 800

700

600

500

400 Stress Stress [MPa]

300

Shot Peened 200 Non-Shot Peened 100 Dry tests (shot peened) Dry tests (Non-shot peened) 0 1.0E+04 1.0E+05 1.0E+06 Number of cycles

Figure 52. S-N curve comparison between CH-SP, CH-NSP, and dry tests

900 TH 800

700

600

500

400 Stress Stress [MPa]

300

200 Shot Peened 100 Non-Shot Peened Dry tests (non-shot peened) 0 1.0E+04 1.0E+05 1.0E+06 Number of cycles

Figure 53. S-N curve comparison between TH-SP, TH-NSP, and dry tests (on non-shot peened)

5.3. Metallographic Results and Observations This section presents a series of pictures as part of a post-test analysis in the attempt of explaining the failure behaviour observed in the plots shown in the previous section. 5.3.1. Fracture Surface Analysis from SEM Pictures for TH-NSP are shown below.

Figure 54. Crack initiation and deformed propagation Figure 55. Crack initiation on TH-NSP @200 MPa, Specimen surface on TH-NSP @200 MPa, Specimen 009 009

Figure 56. Surface major crack propagation on TH-NSP Figure 57. Several cracks initiating on fracture surface at @744 MPa, Specimen 087 different planes near on TH-NSP @744 MPa, Specimen 087

Figure 58. Crack initiation morphology on TH-NSP Figure 59. CaS inclusion near crack initiation on TH-NSP @744 MPa, Specimen 087 @744 MPa, Specimen 087

The following pictures show CH-NSP. A large inclusion containing Ca, Si, Al and S was found near a crack initiation of a specimen tested under 200 MPa stress. The fracture surface at low load seems generally deformed after the repetitive cycling of the test. For high loading, the fracture surface shows an intergranular crack propagation, characteristic of brittle fracture.

Figure 60. Possible crack initiation from (Ca, Si, Al)S Figure 61. Cracking around initiation spot on CH-NSP inclusion on CH-NSP @200 MPa, Specimen 043 @200 MPa, Specimen 043

Figure 62. Crack initiation on CH-NSP @725 MPa, Figure 63. Brittle crack propagation on CH-NSP @725 Specimen 045 MPa, Specimen 045

The pictures shown below correspond to TH-SP. Several crack are observed on the fracture surface in shot-peened conditions. Propagation seems rather ductile.

Figure 64. Major rack initiation on TH-SP @600 MPa, Figure 65. Crack propagation surface on TH-SP @600 MPa, Specimen 010 Specimen 010

Figure 66. Several crack initiations on TH-SP @750 MPa, Figure 67. Major crack initiation and sub-cracking on TH-SP Specimen 034 @750 MPa, Specimen 034

Figure 68. Crack connecting residual fracture on TH-SP Figure 69. Crack propagation surface TH-SP @750 MPa, @750 MPa, Specimen 034 Specimen 034

Next, pictures for CH-SP are shown. Several cracks meet the fracture plain near the surface. The crack propagation is intergranular, a possible sign of brittle fracture.

Figure 70. Crack propagation at 3mm depth from edge on Figure 71. Intergranular crack propagation at 4mm from CH-SP @550 MPa, Specimen 018 edge on CH-SP @550 MPa, Specimen 018

Figure 72. Brittle fracture surface on CH-SP @550 MPa, Figure 73. Corrosion products at fracture surface on CH-SP Specimen 018 @550 MPa, Specimen 018

Figure 74. Main crack initiation and crack meeting facture Figure 75. Main crack initiation and sub-cracking on CH-SP surface on CH-SP @550 MPa, Specimen 018 @550 MPa, Specimen 018

Figure 76. Main crack initiation on CH-SP @725 MPa, Figure 77. Intergranular crack propagation on CH-SP Specimen 082 @725 MPa, Specimen 082

5.3.2. Cracking Analysis from Optical Microscope Pictures of samples from the specimens in Table 2 are presented in this section.

 CH-SP in fresh water

Specimen 091 – 800 MPa @ 1000 rpm Many micro cracks at the surface, presenting little branching. Corrosion pits are formed at the crack base.

A B

C D Figure 78. Specimen 091. A) Overall surface view, B) General crack length, C) Cracks showing almost no branching, D) Crack not following the microstructure

Specimen 047 – 500 MPa @ 1000 rpm Many cracks at the surface with profuse branching. Corrosion pit formed at the crack base.

A B

C D Figure 79. Specimen 047. A) Overall surface view, B) General crack length, C) Cracks showing much branching, D) Abundant branching of cracks

 CH-NSP in fresh water

Specimen 045 – 725 MPa @ 1000 rpm

Tiny cracks at the surface of the specimen, growing through the decarburised layer, as can be seen in C. Eventually one or more of those cracks propagate to become catastrophic, as seen in C. The crack in C was not the cause of final fracture, but a similar one made the specimen fail.

A B

Figure 80. Specimen 045. A) Overall surface view, B) Tiny cracks at the decarburised layer, C) Miniature cracks at the surface and one major crack growing into the material

Specimen 095 – 200 MPa @ 1000 rpm

Virtually no cracks form at the surface, only slight cracking at the decarburised layer, as seen in A. A few cracks will nucleate and grow, as seen in B. This specimen survived.

A B Figure 81. Specimen 095. A) Overall surface view, B) Almost no cracks at the surface, except for a few major cracks that do not propagate to cause failure before 3X10^6 cycles

 TH-SP in fresh water

Specimen 012 – 800 MPa @ 1000 rpm

Several cracks form on the surface, and show little or no branching. No corrosion pit at the crack bases.

A B

C D Figure 82. Specimen 012. A) Overall surface view, B) General crack length, C) Cracks showing almost no branching, D) Cracks magnification on etched sample

Specimen 013 – 550 MPa @ 1000 rpm

Plenty of cracks form at the surface, and most of them branch out. Little crack deep propagation was observed. This specimen survived.

A B

C D Figure 83. Specimen 013. A) Overall surface view, B) General crack length, C) Cracks showing branching, D) Cracks magnification on etched sample

 TH-NSP in fresh water

Specimen 036 – 680 MPa @ 500

A few cracks present on the surface, and several of them seem to be able to propagate simultaneously. The cracks in B grew very deep into the specimen, but another one was the cause of failure.

A B

Figure 84. Specimen 036. A) Overall, B) Big crack propagation

Specimen 016 – 200 MPa @ 1000 rpm

At this low load, the surface is almost free of cracks, and only a few will nucleate an grow, in some cases deep enough to cause failure. A corrosion pit can be seen at the base of the cracks.

A B

Figure 85. Specimen 016. A) Overall, B) Few large cracks can cause failure

 CH-SP in brine (4.0 wt% NaCl)

When running in brine, the crack branching seems to be eaten up by the more aggressive environment.

Specimen 028 – 650 MPa @ 500 rpm

A B

Figure 86. Specimen 029. A) Overall cracking on the surface, B) Grown crack

Specimen 032 – 650 MPa @ 2300 rpm

A B

C D

Figure 87. Specimen 029. A) Overall surface cracking, B) Crack branching is eaten up by salty water, C) General crack length, D) Few larger cracks that eventually propagate deep enough to cause failure

6. Discussion This section is divided in three parts. First, a discussion about the frequency dependence results and the testing parameters selection is presented. Then, the validity of the proposed testing method, as well as of the parameters chosen, is evaluated by discussing results obtained through its application. Finally, the fracture surfaces and micrographs are examined to discuss the apparent cause of the observed failure behaviour. 6.1. Discussion of Method and Parameters Before the present work, the company compared materials in terms of fatigue only by means of the fatigue strength calculated with the staircase method. No study about finite life at higher stresses was performed, and the construction of a Wöhler or S-N curve was not made. The comparison of fatigue results between dry and corrosive environments might lack of valuable insight apart from the observable decrease in fatigue resistance that the aqueous media causes. Therefore, the use of fatigue results at stress levels above the corrosion fatigue strength was proposed in the present work. This way, the lives of materials under corrosion fatigue could be compared with the ones in inert environments at high stresses. However, this implicates more extensive tests, which might come along with increased testing times and costs. For that reason, the Median S-N Test Method for Small Sample Size is proposed; since it does not require a large number of tests to build a valid S-N curve and to find the fatigue strength, a decent number of tests can be carried out in a short time and with a relatively low budget.

Regarding the selection of the water characteristics, a recent analysis of the water used with the company’s equipment in several mines around the world was taken into account to select the parameters of the testing method. Most of the parameters were kept as unaffected as possible – meaning that the water was used as it was taken from the tap- to keep the characteristics closest to the product application. The water flushed while drilling usually does not contain any salt, its temperature ranges around 20 °C, and the pH is generally kept around neutral values. In some rare cases, the water contains extreme levels of salt contaminants, but since that goes out of what is considered “normal” for the application, it was not further studied for the method besides the four tests ran for the frequency dependence analysis shown in Figure 34. Other reasons to not continue testing on salty water involve the harm that the testing machines take, and the lack of simplicity in the test preparation. It takes longer time to prepare the water before every test if a precise measure of the salt content is desired. The oxygen level was not controlled in detail, but it was decided to be kept as highest as possible by constant stirring of the water. This way, the method stays simple and with consistent parameters.

As for the selection of a suitable testing speed, the frequency dependence on CH-SP is discussed. From Figure 34, it can be observed that corrosion fatigue in fresh water is markedly frequency dependent, whereas the life of the specimens tested in brine tends to stay almost constant along the tested frequency range. Looking at the results for fresh water conditions in Figure 41, one can appreciate how the number of cycles reached at 2300 rpm averages infinite life. Although the applied load is lower than the dry fatigue limit, two of the three specimens reached more than 3X10^6 cycles, and the average life is similar to the one achieved without corrosion, which means that this speed would not be good for the new corrosion-fatigue testing method. The difference compared to dry fatigue becomes larger as the testing speed decreases. Also the spread of the data is smaller for lower speeds and largest at 2300 rpm. Note that the trend shows how the life tends to a value of zero at a speed of 0 rpm, which would mean that the specimen would not rotate, and the failure phenomenon would be stress corrosion cracking. On the other hand, the values for brine show that life that tends to stay almost constant regardless of the testing speed. This means that corrosion fatigue for this material in

57 salty environments is virtually frequency independent within the tested range. Since a life tending to 0 rpm speed is also expected for this condition, a strong frequency dependence would be expected in the speed range of 0 – 500 rpm.

In Figure 35, a fraction of the time of each test was occupied by crack initiation, and the rest by propagation. In the present work, no effort was made to identify these two separate cracking steps, and thus these fractions were not studied. However, it is known that time plays an important role on corrosion-fatigue cracking, and that both crack initiation and propagation are partly time dependent and partly cycles dependent, as proven by Rollins et.al ([16]) and discussed nu Jonsson ([15]). Therefore, the fraction of time spent on crack initiation is not expected to be the same for all test speeds.

What is desired from the proposed testing method is that a noticeable difference between different materials and conditions can be achieved. Also, that the testing time is not excessive. From the plot in Figure 41 it can be seen that the lower the frequency, the larger the difference is with respect to dry fatigue results, which is desired in the new method. However, the lower the testing frequency, the more time the test takes. For instance, if a test is ran till runout (at 3X106 cycles), it will take double the time at 500 rpm than at 1000 rpm. Therefore, since a marked difference can be observed already at 1000 rpm, this frequency was chosen for the new corrosion-fatigue testing method. Thus, the main reasons for this decision are:

 Frequency independence on dry testing was proven at 1000 rpm  Significant difference in fatigue life exists between wet and dry conditions at this frequency  Testing time is not excessive  Good number of tests can be carried, resulting in solid data generation  Data spread at this frequency is not large  The machine works fine (avoids natural frequencies) at chosen frequency, resulting in reliable results and lower maintenance

For the staircase method, the load increment used corresponded to 50 MPa. The reason for this is that larger load steps result in a fewer number of tests required to find a limit or strength value, reducing the testing time and making the process cheaper. Also, it was chosen since it is the value used on the dry fatigue tests already performed by the company, and its validity has been proven. Furthermore, this value lays between half and twice the standard deviation of the fatigue strength of each of the material conditions, as the method indicates it should be (see section 2.9.2).

It was desired to prove the effect of the load on the frquency dependence response of the materials. From Figure 39 it is clear that the load has a strong effect on the frequency dependence response on CH-SP, while Figure 40 shows that such a load increment does not affect the frequency dependence response significantly. Perhaps the difference is an effect of the shot peening or the lack of it. However, it must be noted that the results from loading with the dry limits were obtained by testing only one specimen at each speed. Thus, the results only show a rough estimation on how the general trend at this load lever would look like.

The proposed method was tested, and valuable results were obtained, proving that the method works fine as designed in the present work. However, it is worth mentioning that, although the Median S-N Test Method for Small Sample Size suggests four load levels with two tests on each load level for the S-N curve, more than two tests were performed at some load levels and in some cases more than four levels were tested for generating the S-N curves. The same happened for the staircase method tests, where more than the suggested number of tests were performed in some cases. The reason for this is

58 that more reliable results can be obtained with a greater sample size, and enough time and specimens were available during the project to carry out more tests. In the case of the staircase method conducted for CH-SP (see Figure 43), it was necessary to conduct more tests than planned, since the spread in the data was rather large. 6.2. Discussion of Testing Results The tests were performed following the methodology of the proposed method. The sequential order of the tests was decided using the schema in Figure 21. The values of the fatigue life reached by every material condition were presented in section 5.2.1, and a description of the testing procedure was presented in section 4.2. As mentioned there, the load level for the first test of the S-N curve corresponded to the dry fatigue limit of each material condition, stepping down and upwards for the other levels. Nevertheless, for CH-NSP and TH-NSP, a load higher than their corresponding dry fatigue limits were tested since the life reached at the first test was already too low (see Figure 44 and Figure 48).

A fifth load level was included in the S-N curve of CH-NSP (Figure 44) since the points on the top level of the staircase (Figure 45) were incorporated as data in the finite life region. For Conditions 1 and 3, a greater amount of tests in the staircase were required (Figure 43 and Figure 47) since the spread around the strength was large, and even four stress levels had to be tested to find the corrosion fatigue strength for CH-SP. On the other hand, the staircase method for TH-NSP (Figure 49) was applied using only the suggested number of tests since the trend was stable enough, compared to the previous conditions.

For TH-SP only two tests per load level were performed, and the results show an even spread along the mean life in plot with a logarithmic x-axis. The staircase method, shown in Figure 47, shows a fairly good distribution of the data around the fatigue strength. This value is substantially higher than the one found for the other shot peened condition, with a difference of 75 MPa.

For the TH-NSP, the life at high stress is slightly higher than the CH-NSP. However, its corrosion fatigue strength also drops enormously compared to its shot peened version, reaching a value of 191.6 MPa, which is even lower than the other non-shot peened condition. The spread of the values conforming the finite life is narrow in almost every stress level, becoming a bit bigger at 250 MPa, where the stress is closer to the strength value. Yet, the spread is smaller than on the other material conditions.

Concerning the comparison shown in Figure 50, the first thing that can be observed is the large gap between the shot peened and the non-shot peened versions of the steel grades. The gap between TH- SP and TH-NSP is 408.4 MPa, and between CH-SP and CH-NSP is 316.7 MPa. It is also noticeable that at high loads (around 800 MPa), the difference in expected life between shot peened and non-shot peened conditions is around 1.7X10^5 cycles. It can be seen how both shot peened and non-shot peened versions cross each other at certain points. For the case of the shot peened conditions, TH-SP has a lower life expectancy than CH-SP at high stresses. However, as the load decreases, the expected lives for these conditions get closer together and intersect at around 640 MPa and 1.2X10^6 cycles. Finally, the corrosion-fatigue strength of TH-SP was found to be higher than the one of CH-SP. A similar behaviour can be observed for the non-shot peened versions. In this case, both material conditions have a close life expectancy of around 4X10^4 – 5X10^4 cycles at 725 – 750 MPa. As the load is reduced, CH-NSP life starts decreasing more drastically than for TH-NSP. Then it evens out, intersecting the other condition at around 7X10^5 cycles and 300 MPa, to end up having a slightly higher corrosion- fatigue strength.

The comparison between fatigue in dry and wet environments (Figure 52) shows a noticeable reduction in fatigue strength in wet environments, even for the shot peened condition. The dry-fatigue strength (shot peened) is 725 MPa, whereas its corrosive equivalent is only 525 MPa. It should be mentioned that, for the shot peened version under corrosion fatigue, a few new data points appear in this chart. The reason why they were not presented before is that they correspond to the data used for the staircase, and the proposed method establishes this separation (see Figure 21). Regarding Figure 53, the drop of fatigue strength is quite large when exposing the non-shot peened version to wet environments, going from 745 MPa in dry conditions down to 191.6 MPa in corrosive media. This plot also shows data that is not contained in the previous graphs, and the reason is the same as for the other steel grade. Note how, in both cases, the dry results show a spread in the data that is considerably larger than the one found on the corrosion-fatigue tests. For some stress levels, the data points are quite far away from each other. This can be explained with the fact that the mechanisms of crack initiation might be different in corrosion fatigue and in dry fatigue. While the time to initiate a crack in dry conditions is variable due to inclusions and surface defects, the corrosive environment will generally enhance the nucleation of a crack within a short period of time. It is also important to consider the fact that the dry data is very close to the fatigue strength, which generally causes a large spread between runouts and failures, a behaviour also observed in some of the corrosion-fatigue tests. 6.3. Fracture Surface and Cracking Characteristics in Relation with the Results The post-testing analysis of the fracture surfaces by the use of SEM, presented in 5.3.1, was performed as an attempt to observe the crack initiation sites and identify the characteristics of the crack propagation process. This task was challenging since the rotating-bending test damages the fracture surfaces in the process, and usually the violent final fracture of the specimen results in scratched surfaces which are practically unusable for the analysis. Nevertheless, a few specimens were collected and successfully observed in the SEM. An important note is that all the cracks were found initiated at the surface of the material, in accordance with the theory presented in the previous parts of this work.

In the case of TH-NSP at 200 MPa, shown in Figure 54 and Figure 55, only one crack was found initiated on the fracture surface, which was the cause of failure. It should be remarked that at this low stress level, some specimens of this material condition survived and some others –like the one analysed in SEM- failed. Some corrosion attack can be observed where the crack meets the surface of the specimen. As discussed in the literature review in this document, it is not feasible to conclude if the crack initiated from the corrosion pit, or if the latter was formed after the crack nucleated. Some tiny cracks can be seen on the surface in Figure 55, and also what seems to be the formation of thin flakes. Propagation is not easily appreciated since the surface appears very flat, perhaps because the crack closure during repetitive cycling had deformed the crack surface. For the same material condition at 744 MPa, presented in Figure 56 to Figure 59, several crack initiations were found on the fracture surface. One major crack was leading the propagation, and the majority of the fracture surface is found on the plane of this crack; however, when the cracking process had advanced enough, the fracture surface was pulled on diverse directions by the other simultaneously growing cracks, since the energy to fracture the material in those directions was lower than the one required to continue propagating in the main plane.

As for the CH-NSP at 200 MPa, shown in Figure 60 and Figure 61, the fracture surface also showed one single crack initiation, with non-metallic inclusion containing Ca, Al and Si sulphides found at this site. Perhaps this was the cause of the crack nucleation in this case. Figure 62 and Figure 63 show the same material condition at 725 MPa. The crack initiation looks severely damaged, perhaps by corrosion, and it could be due to the decarburised layer material being released. The crack propagation seen in Figure 63 is totally intergranular, which indicates brittle fracture that could be caused by hydrogen

60 embrittlement. Apparently the high stress state enhances the brittle propagation since the plastic strain is higher at the crack tip, allowing more hydrogen absorption into the steel.

The TH-SP at 600 MPa presented a ductile crack propagation surface, observable in Figure 65. It was difficult to identify if the crack initiation occurred at the surface of the specimen, as seen in Figure 64, but it seems to be that the propagation lines flow already from the edge of the fracture surface inwards. A few corrosion pits are also appreciable along the edge. Again, it cannot be defined if the corrosion pit is the cause of the crack, or if it is the other way around. The same material condition stressed at 750 MPa, on the other hand, presented many simultaneous crack initiations on the fracture surface, as seen in Figure 66. For that reason, the fracture surface was shaped on different planes. Some cracks that initiated on planes further away from the predominant ones were reached by the shear lip of the residual fracture, as seen in Figure 68. From Figure 67, it is visible that the major crack initiation presented several cracks around it, perhaps initiating from the fracture plane or possibly initiated elsewhere and intersecting it.

In the case of CH-SP at 550 MPa, two different fracture surface characteristics were observed at different depths, as shown in Figure 70 and Figure 71. At 3 mm, the fracture seems more transgranular than at 4 mm, where the propagation looks mainly intergranular. This could be a cause of the crack growth rate, as discussed by Ritchie [8], who stated that when the cyclic plastic-zone size –which increases as the crack grows- is near the grain size, the facets on the surface will form. The crack growth rate is unknown in this study, and therefore this statement is not conclusive. Nevertheless, the intergranular crack propagation is an indicator brittle fracture, possibly caused by the hydrogen- embrittlement mechanism. Some corrosion products were found on the fracture surface, as seen in Figure 73. These could have formed during or after the test, but they seem to be occupying the place of a rounded body, possibly an inclusion that was dissolved by the water. The crack initiation happened also on the surface of the specimen boosted by the corrosive media, as shown in Figure 74. Several cracks were found on the fracture surface surrounding the crack, and equal to the case of TH-SP, these cracks could have either intersected the fracture surface or grown from it. For the same material condition at 725 MPa, multiple crack initiation sites were found around the fracture edge, and the crack propagation was also intergranular, as Figure 77 shows.

The purpose of analysing the cracks on a cross section of the different material conditions, presented in section 5.3.2, was to relate the variables of each condition to its cracking characteristics. This study provided useful information in terms of the ease of crack propagation mainly as a function of the residual stresses present on the materials. It is important to remember that each material condition presented different residual stress scenarios. CH-NSP has a relatively low compressive residual stress due to the carburisation, which reaches -150 MPa in the first 0.02 mm of depth and stays negative further in. On the other hand, TH-NSP has a negative residual stress on the very surface, but quickly reaches a value of zero at 0.05 mm of depth. TH-SP has a highest compressive stress of around -800 MPa at approximately 0.2 mm of depth, being considerably high already at 0.1 mm with a value of around -750 MPa, resulting in a thick layer of compressive stresses below the surface. CH-SP reached a highest compressive stress of around -900 MPa at 0.07 mm, staying high on -840 MPa at 0.14 mm and then abruptly decreasing. It can be highlighted how the compressive stress level in both materials reaches a similar value, but the depth of penetration is higher –approximately double- on the TH-SP. The reason for this could be that the surface hardness of CH is higher than the one of TH, restricting the compressive stress from going deep into the material. However, this hypothesis is not conclusive, since there is a considerably big gap between the measurement at 0.14 mm and the measurement at 0.48 mm, and no data about the compressive stress between these two points is existent. Although

61 the trend seems to go up from the 3rd to the 4th measurement depth, it is not valid to conclude what happens at the gap.

In general terms, the micrographs of shot peened specimens showed abundant cracking at the surface, growing in to reach a certain common depth, before being arrested by the compressive stresses. As this happens, the cracks start forming branches and spreading laterally instead of propagating straight forward radially. Observe for instance the cracks for CH-SP at 500 MPa showed in Figure 79. The cracks grow to a general depth of around 72.4 µm, branching out from around half of their length. The explanation for this phenomenon has not been totally clarified in the present work, but one feasible explanation is that the energy to initiate a crack is lower than the one required to propagate it, and as the cracks are stopped by the compressive stress layer, many cracks nucleate from the initial cracks, fanning out simultaneously. It can be seen from Figure 78-D that the branches do not follow the microstructure. A general behaviour observed in these cases is that, for lower loads, the cracks branch out more than for higher loads. Compare for example CH-SP at 800 MPa and 500 MPa, shown in Figure 78 and Figure 79 respectively. In both cases the cracks reach approximately the same general depth, but at higher load, the cracks have a leading branch with sharp tip that pass the compressive stress barrier and cause failure. This difference is more evident for TH-SP, presented in Figure 82 and Figure 83. The reason for the greater branching on lower stresses could be then that the specimen runs for longer time, allowing for more branches to form, without having enough energy to overcome the compressive stress layer. Additionally, at lower loads, corrosion pits can be seen at the base of the cracks, also explained by the longer run time allowing more exposure to the corrosive media. It is important to mention that the compressive stress around the crack and between the branches is lost, otherwise the branches would close as a result of the compression. It is also suggested that the compressive stress in the material between the cracks is decreased due to the same effect. However, compressive stresses are kept ahead of the cracks front, preventing the cracks to grow forward. The residual stresses were not measured on the specimens post-test, so there are no values to show the extent of this fact.

The shot peening on both conditions arrested the growing cracks at a depth that appears to match the peak of compressive stresses, or close to it. For CH-SP, this value averages 73 µm, for which the highest compressive stress is around 75 µm. For TH-SP, the range is larger since the compressive stress layer is thicker, but already reaches high values at about 100 µm, so the cracks are considerably stopped around that depth. However, the crack front profile is more irregular on this material for this reason, as shown in Figure 82. This information can be used to propose an explanation for the stress-life behaviour seen in the S-N curve in Figure 51. First, it is observed that TH-SP has a higher corrosion- fatigue strength, and its life at low stress levels will be longer, compared to CH-SP. This could be because it results more difficult for the cracks to pass a thicker compressive stress layer. Therefore, TH-SP can hold larger cracks than CH-SP. On the other hand, TH-SP has a lower life at higher stress levels. The reason for this could be that, in both cases the cracks eventually cross the compressive stress layer, and this probably takes longer time in TH-SP since the layer is thicker, but once they grow further in, the propagation is slower in CH-SP since the material core is tougher than in TH-SP.

For the case of the non-shot peened versions, the higher the stress levels, the more cracks that nucleate on the surface of the specimen, as seen for example in TH-NSP at 500 MPa and 200 MPa, presented in Figure 84 and Figure 85, respectively. At high loads, many small cracks were found growing radially in parallel, and some larger cracks were already rather advanced into the material body. At the end, the final fracture surface united several predominant cracks, as discussed in the SEM analysis. Oppositely, at low stresses, only a small number of cracks will nucleate, and as the load approaches the fatigue strength, these cracks might not grow large enough to cause failure before

3X10^6 cycles are reached. It was observed that, for low load cases, CH-NSP presented more cracks than TH-NSP. The reason for this is that, unlike TH-NSP, the compressive stresses from the carburised layer on CH-NSP stop the cracks from propagating, allowing for more cracks to form without becoming critical. Additionally, a corrosion pit is observed at low stress levels in both steels, since the tests run for longer times, allowing more attack from the environment. In the case of CH-NSP at high stresses, the microcracks grow through the decarburised layer, and seem to follow the weaker phases, as seen in Figure 80-B. Same as for TH-NSP, a few predominant cracks grow and cause failure. At low loads, the surface is practically free of critical cracks, and only a few nucleate after many cycles and grow slowly due to the low stress level, as seen in Figure 81-B.

The addition of 4 wt% NaCl in the water for the tests in CH-SP resulted in micrographs similar to the ones from the fresh water tests in the same material condition. Yet, a few differences can be observed. See Figure 86 and Figure 87. Both show specimens tested at 650 MPa, a relatively medium stress for this material condition, but at different test speeds (500 rpm and 2300 rpm). The surface is also covered with branched-out cracks partially arrested by the compressive stress. However, the branching is not clearly defined, but rather eaten up by the extremely aggressive media, as seen in Figure 87-B. Some cracks appreciably crossed through the compressive stress layer evidently easier than in fresh water. It is not possible to know to which extent the salty water affected the cracking process in terms of initiation and propagation, since this information was not generated for any of the situations. Nonetheless, it can be supposed that the aggressiveness of the salty water caused a rather similar crack initiation time regardless of the testing speed, and that the propagation was dominated by the cycles. Then, from Figure 35, it can be deduced that the length of the bands represent the time that the propagation takes at each speed. At 500 rpm, the time is long since less revolutions happen on a given period of time, compared to the opposite extreme, for which the time is shorter since more load cycles are covered in the same amount of time. This explanation encloses the general assumed behaviours, but as detailed by Jonsson ([15]), in reality both initiation and propagation stages have time-dependence and a cycles-dependence components.

7. Conclusions This work investigated the main variables affecting the corrosion fatigue on TH and CH, and offered parameters for a corrosion-fatigue testing method hereon proposed. Rotating-bending machines were used, resulting in a fully-reversed sinusoidal stress wave. Frequency dependence tests were performed to study the response of corrosion-fatigue life as a function of the testing speed, in both fresh water and brine (4.0 wt% NaCl water). The testing method was used on shot peened and non-shot peened specimens of both steel grades, and its validity was proven. The following conclusions were found in relation to the objectives of this project:

1. The Median S-N Test Method for Small Sample Size is proposed as a suitable method for testing corrosion fatigue in martensitic steel grades with different heat and surface treatments, since it requires a relatively small sample size, resulting in relatively cheap studies with low testing times. However, a slightly greater sample size is was used in the present work since the number of tests suggested by the method could not be used to build concrete conclusions. Thus, whenever time and resources are available, a greater sample size should be used to generate more reliable results. 2. The generation of S-N curves allows the study of different behaviours of the materials and their comparison at a broader stress range, opposite to the comparison of the fatigue strengths alone. In the present work, TH-SP was found to have a higher corrosion-fatigue strength than CH-SP, although its expected life is lower at higher loads. 3. A testing speed of 1000 rpm is the best option from all the four speeds tested. The reason for this is that it results in a marked difference between the dry and the wet fatigue results, which is desired for the proposed testing method. The lower the testing speed, the bigger the difference is. However, as the speed is decreased, the testing time increases. The proposed testing speed results in a decent balance between testing time and comparison capabilities. 4. The addition of salt in the water for testing is discarded since it is not a common characteristic in real rock-drilling applications, and the damage mechanisms are different to the ones caused by fresh water. Additionally, the corrosive attack was observed to also affect the test machines, which is not desired by any means. 5. The compressive residual stresses are extremely beneficial for the corrosion-fatigue performance of steel since it arrests the cracks and stops them from growing. Shot peening was found to enhance the resistance of the materials against cracking. A great amount of cracks nucleate on the surface and start growing up to a certain general distance close to the peak of the compressive stress layer. Moreover, the thicker this layer is, the more resistance it builds for the cracks to propagate. 6. The crack growth rate appeared to be smaller in the tough core of CH than in the hard core of TH. In general, it is well accepted that the higher the toughness of a material, the better fatigue properties it has.

The present study provided a useful method to test corrosion fatigue on steel grades used in rock- drilling applications, and generated valuable information about the behaviour of TH and CH. The mechanisms of cracking and failure were discussed and provided explanations for the results obtained from the tests. Nevertheless, it is worth to mention that the phenomenon of corrosion fatigue is not yet fully understood. Considerable discrepancy is occasionally found in literature regarding the failure mechanisms of different materials at different conditions. Additional data generation by testing specific desired conditions results of high importance for R&D activities.

8. Further Work This section presents several recommendations for future work related to the present study.

 Conduct a study to identify the fraction of total life spent in crack initiation and in crack propagation. Since the tests are performed with rotating-bending machines, the use of 2-stage testing method is suggested.  Generate S-N curves for fatigue at dry conditions to be able to compare the behavior of the materials at stresses higher than the fatigue limit.  Apply the proposed testing method to generate data for different combinations of materials and treatments. For instance, test CH steel grade without case hardening treatment, but only with shot peening. If the material performs well, the company could consider removing the heat treatment process for this steel grade.  Evaluate of the effect of different shot peening parameters on the performance of steels subjected to corrosion fatigue. Changing the shot peening intensity could result in different compressive stress profiles, improving the performance of the steel at corrosion fatigue.  Address the problem of the curved specimens after shot peening. The residual stresses from the shot peening usually cause a slight curvature rod-like components. For this reason, the shot peened products generally require a straightening process. The same thing could apply for the testing specimens, which also suffer from this post-peening deformation.

9. References [1]. Frost, N.E., Marsh, K.J., Pook, L.P., (1974), Metal Fatigue, Oxford University Press.

[2]. Jonsson, S., (2010), Mechanical Properties of Metals and Dislocation Theory from an Engineer’s Perspective, Royal Institute of Technology, Sweden.

[3]. Lee, Yung-L., Pan, Y., Hathaway, R.B., Barkey, M.E., (2005), Fatigue Testing and Analysis (Theory and Practice), Elsevier Butterworth-Heinemann.

[4]. Glaeser, W., Wright, I.G., (2003), Forms of Mechanically Assisted Degradation, ASM Handbook, Vol. 13A, Corrosion: Fundamentals, Testing, and Protection, ASM International, p. 322-330.

[5]. Phull, B. (Rev.), (2003), Evaluating Corrosion Fatigue, ASM Handbook, Vol. 13A, Corrosion: Fundamentals, Testing, and Protection, ASM International, p. 625-638.

[6]. ASM International, (2008), Elements of Metallurgy and Engineering Alloys (#05224G), Chapter 14: Fatigue.

[7]. Boardman, B., Deere and Company, Technical Center. (1990). Fatigue Resistance of Steels. ASM Handbook, Volume 1: Properties and Selection: Irons, Steels, and High-Performance Alloys. ASM Handbook Committee, p 673-688.

[8]. Ritchie, R.O., (1979), Near-threshold fatigue-crack propagation in steels, The Metals Society and the American Society for Metals, International Metals Reviews, 1979, Nos. 5 and 6.

[9]. Higgins, R.A., (1993) Engineering Metallurgy - Applied Physical Metallurgy, Arnold, 6th edition.

[10]. Evins, J.L., (2004), Dependence of Strength on Corrosion-Fatigue Resistance of AISI 4130 Steel, Georgia Institute of Technology.

[11]. Novak, S. R., Corrosion-Fatigue Crack Initiation Behavior of Four Structural Steels, Corrosion- fatigue: Mechanics, metallurgy, electrochemistry, and engineering, ASTM STP 801. T. W. Crooker and B.N. Leis (Eds). West Conshohocken: American Society for Testing and Materials, p 26-63.

[12]. Boyer, H. E. (Ed.). (1986). Atlas of fatigue curves, ASM International.

[13]. Katz, Y., Tymiak, N., Gerberich, W.W., (2000), Evaluation of Environmentally Assisted Crack Growth, ASM Handbook, Vol. 8, Mechanical Tenting and Evaluation, ASM International, p. 613-648.

[14]. ASM International, (2000), Fatigue and Fracture Mechanics, ASM Handbook, Vol. 8, Mechanical Tenting and Evaluation, ASM International, p. 681-685.

[15]. Jonsson, S., (1995), A Review on Corrosion-Fatigue of High Carbon Steels, Swedish Institute for Metal Research, Sweden.

[16]. Rollins, V., Arnold, B., Lardner, E., (1970), Br. Corros. J., 5, p. 33-40.

[17]. Scott, P. M., (1983), Chemistry effects in corrosion-fatigue. Corrosion-fatigue: Mechanics, metallurgy, electrochemistry, and engineering, ASTM STP 801. T. W. Crooker, & B. N. Leis (Eds.). West Conshohocken: American Society for Testing and Materials, pp. 319-350.

[18]. Uhlig, H.H., (1963), Corrosion and Corrosion Control, John Wiley & Sons, New York, p 99.

[19] Rollins, V., Arnold, B., Lardner, E., (1967), Metallurgia 75, p. 149, 154.

[20] Milella, P.P., (2012), Fatigue and Corrosion in Metals, Springer.

[21] Freddi, A., Veschi, D., Bandini, M., Giovani, G., (1997), Design of Experiments to Investigate Residual Stresses and Fatigue Life Improvement by Surface Treatment, Fatigue & Fracture of Engineering Materials & Structures Ltd, R&D Center of Norblast Company, Italy.

[22] Sandell, M., Jonsson, S., (2016), Literature Survey – Methods of Inducing Residual Stress and Their Relaxation, Royal Institute of Technology, Sweden.

[23] Yunfeng Cao, Yung C. Shin, and Benxin Wu, "Parametric Study on Single Shot and Overlapping Laser Schock Peening on Various Metals via Modeling and Experiments," Journal of Manufacturing Science and Engineering, vol. 132, 2010.

[24] Andersson, L., Berggren, K., Bodin, J., Ferrari, M.T., Eriksson, A., Eriksson, J., . . . Ölund, P., (2012), Steel and its Heat Treatments, T. Holm, P. Olsson, E. Troell (Eds.), Swerea IVF, Sweden.

[25] Halford, G.R., Lerch, B.A., (2000), Fatigue, Creep Fatigue, and Thermodynamical Fatigue Life Testing, ASM Handbook, Vol. 8, Mechanical Tenting and Evaluation, ASM International, p. 686-716

[26] Pollak, R., Palazotto, A., Nicholas, T., (2006), A Simulation-Based Investigation of the Staircase Method for Fatigue Strength Testing, Mechanics of Materials 38, Elsevier, p 1170-1181.

[27] Fitzpatrick, M.E., Fry, A.T., Holdway, P., Kandil, F.A., Shackleton, J., Suominen, L., (2005), Determination of Residual Stresses by X-ray Diffraction – Issue 2, Measurement Good Practice Guide No. 52, National Physical Laboratory, United Kingdom.

[28] Laird, C., Duquette, D. J., (1971), Mechanisms of Fatigue Crack Nucleation, Corrosion-fatigue: Chemistry, mechanics, and microstructure. National Association of Corrosion Engineers, pp. 88-115.

[29] Ray, G.P., Jarman, R.A., Thomas, J.G.N., (1985), The Influence of Non-Metallic Inclusions on the Corrosion Fatigue of Mild Steel, Corrosion Science, Vol. 25, No. 3, Pergamon Press Ltd., p. 171-184.

[30] Weng, L., Zhang, J., Kalnaus, S., Feng, M., Jiang, Y., (2013), Corrosion Fatigue Crack Growth of AISI 4340 Steel, International Journal of Fatigue 48, Elsevier Ltd., p. 156-164.

[31] Delmotte, E., Micone, N., De Waele, W., Testing Methodologies for Corrosion Fatigue, Ghent University, Belgium.

[32] Nishikawa, H., Oda, Y., Noguchi, H., (2011), Investigation of the Mechanism for Brittle-Striation Formation in Low Carbon Steel Fatigued in Hydrogen Gas, Journal of Solid Mechanics and Materials Engineering, Vol. 5, No. 8, Japan.

[33] Carter, C.S., (1969), Stress Corrosion Crack Branching in Sigh-Strength Steels, Boeing Commercial Airplane Division, Washington, United States.

[34] Rabb, B.R., (2003), Staircase Testing – Confidence and Reliability, Fatigue Damage of Materials: Experiment and Analysis, Transactions on Engineering Sciences, Vol 40, WIT Press.

[35] Roylance, D., (2001), Fatigue, Massachusetts Institute of Technology, United States.

[36] Suresh, S., (1998), Fatigue of Materials, 2nd Ed., Cambridge University Press.

67 www.kth.se