1

TABLE OF CONTENTS
Symbol. Abbreviations. CHAPTER -1 Introduction to Fatigue of Welded Structures............................... CHAPTER -2 2.1 Fatigue as a Phenomenon in the Material 2.2 Different phases of the fatigue life CHAPTER- 3 3.1 Why metal parts fail from repeatedly-applied loads 3.1.1 What is fatigue loading 3.1.2 How is the fatigue strength of a metal determined 3.1.3 Is there any relationship between UTS and fatigue strength 3.1.4 Why is the surface so important 3.1.5 Is the endurance limit an exact number 3.1.6 Do real-world components exhibit the "laboratory" EL 3.1.7 Is fatigue loading cumulative. CHAPTER- 4 4.1 Fracture Mechanisms 4.1.1 Ductile Fracture 4.1.1.1 General Macroscopic Appearance of Ductile Fractures 4.1.2 Brittle Fracture 4.1.2.1 Microscopic Aspects of Fracture 4.1.3 Transgranular Ductile 4.1.4 Transgranular Brittle Fracture 4.1.5 Quasi-Cleavage 4.1.6 Mechanisms of Intergranular Fracture. 4.1.6.1 Intergranular brittle cracking. 4.1.6.2 Dimpled intergranular fractures.. 4.1.6.3 Intergranular fracture surfaces with corrosion. 4.1.7 Macroscopic Aspects of Overload Failures 4.1.8 Cracks propagating from a pre-existing stress raiser or notch.. CHAPTER 5 5.1 Fatigue testing Part 1. 5.1.1 S/N curve. 5.1.2 Palmgren-Miners rule 5.2 Fatigue testing Part 2. 5.2.1 Preparations and measurements .. 5.2.2 Test results 5.3 Crack growth tests guidelines for test setup and specimen monitoring 5.4 Welded Components. 5.5 Fatigue testing Part 3 5.5.1 BS 7608:1993 5.6 Potential modes of failure of welds.. 5.7 Tubular joints. 5.8 Weldments . 2 2 4 7 9 15 15 17 18 19 20 21 22 24 25 28 30 32 34 38 40 44 46 47 48 50 61 63 65 66 67 69 74 75 80 85 86 95 104 105

2 CHAPTER 6 Designing against Fatigue of Structures. 8.8 Different types of structural fatigue problems 8.8 Designing against fatigue 8.8 The crack initiation aspect 8.8 Material selection 8.8 Surface treatments 8.8 Detail design for an improved stress distribution 8.8 Large-scale design issues 8.8 Uncertainties, scatter and safety margins 8.8.1 Uncertainties 8.8 Scatter and safety factors 8.8.1 The fatigue limit and the safety factor 8.8 Safety factors for finite fatigue life problems under CA loading 8.8 Safety factors for finite fatigue life problems under VA loading 8.8 Safety factors and fatigue crack growth . 8.8 Safety aspects associated with a corrosive environment and low frequency fatigue.. CHAPTER 7 Methods of revealing fatigue cracks 7.1 Dye-penetrant testing 7.1.1 An example of dye penetrant testing used on bicycle components 7.2 Photoelasticity . CHAPTER- 8 8.1 Causes and recognition of fatigue failures 8.1.1 General Causes of Material Failures: 8.1.2 Recognition of Fatigue Failure 8.2 Design Considerations 8.2.1 Influence of Processing and Metallurgical Factors on Fatigue 8.2.1.1Processing Factors . 8.2.1.1 Metallurgical Factors . 8.3 Experimental Analysis of Fatigue Life Curves 8.6 Fatigue Crack Growth.. 8.7 Real Life-Design and Manufacturing Considerations . 8.8 Recommendations for Designs to Avoid Fatigue Failures . Annexure 1 Additional Scanning Electron Microscope Images Annexure 2 Metallography/Microstructure Evaluation Annexure 3 Microscopic characteristics of fatigue fracture Macroscopic characteristics of fatigue fracture Lack of Deformation Beachmarks Ratchet Marks Similarities between Striations and Beachmarks Differences between Striations and Beachmarks Annexure 4 Samples failure REFERENCES

109 109 111 112 113 113 113 114 114 114 115 115 117 118 118 122 124 125 126 128 129 129 129 131 131 131 133 134 134 135 135 137 139 147 147 148 148 149 149 150 151 154

Symbols
a a0 ac af c C D da/dN dU/da E G K K = KIc N N P r S T a m crack length, or semi-crack length, or depth of part through crack initial crack length final or critical crack depth final crack length (semi) crack length of surface crack constant in Paris equation diameter crack growth rate strain energy release rate Youngs modulus shear modulus stress intensity factor Kmax Kmin fracture toughness number of cycles fatigue life until failure load root radius of notch nominally applied (gross) stress temperature strain Poisson ratio local stress in material stress amplitude mean stress shear stress

Abbreviations
AW BS bcc CA CP CT CTOD FC As-Welded British Standards Abbreviation for body-centered cubic crystal structure. Constant Amplitude Cathodic Protection Compact Tension Crack Tip Opening Displacement Free Corrosion

fcc FEA FEM HAZ HB hcp HS LEFM NDI SAW SEM TEM TIG VA

Abbreviation for face-centered cubic crystal structure Finite Element Analysis Finite Element Method Heat-Affected Zone Hardness Brinell Abbreviation for hexagonal close-packed crystal structure Hot-Spot Linear Elastic Fracture Mechanics Non-Destructive Inspection Submerged-Arc Welding Scanning electron microscopy Transmission electron microscopy Tungsten Inert Gas Variable Amplitude

CHAPTER 1
Introduction to Fatigue of Welded Structures
Fatigue failures in metallic structures are a well-known technical problem. Already in the 19th century several serious fatigue failures were reported and the first laboratory investigations were carried out. Noteworthy research on fatigue was done by August Whler. He recognized that a single load application, far below the static strength of a structure, did not do any damage to the structure. But if the same load was repeated many times it could induce a complete failure. In the 19th century fatigue was thought to be a mysterious phenomenon in the material because fatigue damage could not be seen. Failure apparently occurred without any previous warning. In the 20th century, we have learned that repeated load applications can start a fatigue mechanism in the material leading to nucleation of a small crack, followed by crack growth, and ultimately to complete failure. The history of engineering structures until now has been marked by numerous fatigue failures of machinery, moving vehicles, welded structures, aircraft, etc. From time to time such failures have caused catastrophic accidents, such as an explosion of a pressure vessel, a collapse of a bridge, or another complete failure of a large structure. Many fatigue problems did not reach the headlines of the newspapers but the economic impact of noncatastrophic fatigue failures has been tremendous. Fatigue of structures is now generally recognized as a significant problem. As a result of extensive research and practical experience, much knowledge has been gained about fatigue of structures and the fatigue mechanism in the material. Much has been learned from laboratory research. However, accident investigations have also highly contributed to the present state of the art. Fatigue failures in service can be most instructive and provide convincing evidence that fatigue may be a serious problem. The analysis of failures often reveals various weaknesses contributing to an insufficient fatigue resistance of a structure. This will be illustrated here by a case history. The front wheel of a heavy motorcycle completely collapsed, see Figure 1.1a. Ten spokes of the light alloy casting were broken. Examination of the failure surfaces indicated that fatigue cracks occurred in all spokes, see Figure 1.1b.Why was the fatigue life of this wheel insufficient? A first question of a failure analysis must be: Was the failure a symptomatic failure or was it an incidental case? If it is a symptomatic failure, all motorcycles of the same type are in danger and immediate action is required. However, the failure may be an incidental case for some special reason applicable to that single motorcycle only: for instance, unusual

6 and severe damage of the material surface. In the case of this motorcycle, the same failure had occurred in several wheels in different countries, although predominantly in motorcycles of the police. The wheel shown in Figure 1.1 collapsed when a policeman suddenly had to use the brakes to stop before a railway crossing. He survived after some heavy shocks.

Fig 1.1a Front wheel, broken spikes, axle part with drum

Fatigue fractures brakes Fig. 1.1b Collapse of the front wheel of a motorcycle by fatigue of the spokes.

7 A structure should be designed and produced in such a way that undesirable fatigue failures do not occur during the design life of the structure. A special issue is how to account for environmental effects. Experimental data used in the predictions are generally obtained under laboratory conditions and relatively high testing frequencies. However, in service corrosive environments may be present and the load frequency can be much lower. As an example, think of a welded structure for a drilling platform in the sea. The environment is salt water, and the loading rate of water waves is relatively low [1].

CHAPTER 2
2.1 Fatigue as a Phenomenon in the Material
In a specimen subjected to a cyclic load, a fatigue crack nucleus can be initiated on a microscopically small scale, followed by crack grows to a macroscopic size, and finally to specimen failure in the last cycle of the fatigue life. Microscopic investigations in the beginning of the 20th century have shown that fatigue crack nuclei start as invisible micro cracks in slip bands. After more microscopic information on the growth of small cracks became available, it turned out that nucleation of micro cracks generally occurs very early in the fatigue life. Indications were obtained that it may take place almost immediately if a cyclic stress above the fatigue limit is applied. The fatigue limit is the cyclic stress level below which a fatigue failure does not occur. In spite of early crack nucleation, micro cracks remain invisible for a considerable part of the total fatigue life. Once cracks become visible, the remaining fatigue life of a laboratory specimen is usually a small percentage of the total life. The latter percentage may be much larger for real structures such as ships, aircraft, etc. After a micro crack has been nucleated, crack growth can still be a slow and erratic process, due to effects of the microstructures, e.g. grain boundaries. However, after some micro crack growth has occurred away from the nucleation site, a more regular growth is observed. This is the beginning of the real crack growth period. Various steps in the fatigue life are indicated in Figure 2.1. The important point is that the fatigue life until failure consists of two periods: the crack initiation period and the crack growth period. Corrosive environments can affect initiation and crack growth, but in a different way for the two periods. It should be noted here that fatigue prediction methods are different for the two periods. The stress concentration factor Kt is the important parameter for predictions on crack initiation. The stress intensity factor K is used for predictions on crack growth [1].

Fig. 2.1 Different phases of the fatigue life and relevant factors.

Fig. 2.2. Different scenarios of fatigue crack growth. The crack initiation period includes the initial micro crack growth. Because the growth rate is still low, the initiation period may cover a significant part of the fatigue life. This is illustrated by the generalized picture of crack growth curves presented in Figure 2.2. which schematically shows the crack growth development as a function of the percentage of the fatigue life consumed (= n/N), with n as the number of fatigue cycles and N as the fatigue life until failure. Complete failure corresponds to n/N = 1 = 100%. There are three curves in Figure 2.2, all of them in agreement with crack initiation in the very beginning of the fatigue life, however, with different values of the initial crack length. The lower curve corresponds to micro crack initiation at a perfect surface of the material. The middle curve represents crack initiation from an inclusion. The upper curve is associated with a crack starting from a material defect which should not have been present, such as defects in a welded join. Figure 2.2 illustrates some interesting aspects: The vertical crack length scale is a logarithmic scale, ranging from 0.1

10 nanometer (nm) to 1 meter (1 nanometer = 109 m = 4108 inch). Micro cracks starting from a perfect free surface can have a sub-micron crack length (<1 m = 106 m). However, cracks nucleated at an inclusion will start with a size similar to the size of the inclusion. The size can still be in the submillimeter range. Only cracks starting from macro defects can have a detectable macro crack length immediately. The two lower crack growth curves illustrate that the major part of the fatigue life is spent with a crack size below 1 mm, i.e. with a practically invisible crack size. Dotted lines in Figure 2.2. indicate the possibility that cracks do not always grow until failure. It implies that there must have been barriers in the material which stopped crack growth [1].

2.2 Different phases of the fatigue life

Fatigue fractures have a characteristic appearance which reflects the initiation site and the progressive development of the crack front, culminating in an area of final overload fracture. Initiation site(s). Progressive development of crack front characterised by beach marks. Culminating in an area of final fracture.

Fig. 2.3a illustrates fatigue failure in a circular shaft. The initiation site is shown and the shell-like markings, often referred to as beach markings because of their resemblance to the ridges left in the sand by retreating waves, are caused by arrests in the crack front as it propagates through the section. The hatched region on the opposite side to the initiation site is the final region of ductile fracture. Sometimes there may be more than one initiation point and two or more cracks propagate. This produces features as in Fig. 2.3b with the final area of ductile fracture being a band across the middle. This type of fracture is typical of double bending where a component is cyclically strained in one plane or where a second fatigue crack initiates at the opposite side to a developing crack in a component subject to reverse bending. Some stress-induced fatigue failures may show multiple initiation sites from which separate cracks spread towards a common meeting point within the section [2].

11

Fig. 2.3 a,b,c

12 Fatigue strength is determined by applying different levels of cyclic stress to individual test specimens and measuring the number of cycles to failure. Standard laboratory test use various methods for applying the cyclic load, e.g. rotating bend, cantilever bend, axial push-pull and torsion. The data are plotted in the form of a stress-number of cycles to failure (S-N) curve, fig 2.4. Owing to the statistical nature of the failure, several specimens have to be tested at each stress level. Some materials, notably low-carbon steels, exhibit a flattening off at a particular stress level as at (a) in Fig.2.4 which is referred to as the fatigue limit. As a rough guide, the fatigue limit is usually about 40% of the tensile strength. In principle, components designed so that the applied stresses do not exceed this level should not fail in service. The difficulty is a localised stress concentration may be present or introduced during service which leads to initiation, despite the design stress being normally below the safe limit. Most materials, however, exhibit a continually falling curve as at (b) and the usual indicator of fatigue strength is to quote the stress below which failure will not be expected in less than a given number of cycles which is referred to as the endurance limit.

Fig 2.4

13

Fig

2.5 a,b

Sample pictures of fatigue failure

14

Fig 2.6

Micro crack at grain boundary during fatigue test.

Fig 2.7

Fatigue failure

15 Although fatigue data may be determined for different materials it is the shape of a component and the level of applied stress which dictate whether a fatigue failure is to be expected under particular service conditions. Surface condition is also important. Often complete components or assemblies, e.g. railway bogie frames or aircraft fuselage will be tested by subjecting them to an accelerated loading spectrum reproducing what they are likely to experience over their entire service lifetime.

16

CHAPTER 3
3.1 Loads
Long ago, engineers discovered that if you repeatedly applied and then removed a nominal load to and from a metal part (known as a cyclic load), the part would break after a certain number of load-unload cycles, even when the maximum cyclic stress level applied was much lower than the UTS, and in fact, much lower than the Yield Stress. These relationships were first published by A. Z. Whler in 1858. They discovered that as they reduced the magnitude of the cyclic stress, the part would survive more cycles before breaking. This behaviour became known as FATIGUE because it was originally thought that the metal got tired. When you bend a paper clip back and forth until it breaks, you are demonstrating fatigue behaviour. Some common questions about metal fatigue are: What is fatigue loading? How do you determine the fatigue strength of a material? Does the strength of a material affect its fatigue properties? Why is the surface of a part so important? Is fatigue life an exact number? Do real-world parts behave the same as laboratory tests? Are fatigue cycles cumulative?

Why Metal Parts Fail From Repeatedly-Applied

3.1.1

WHAT IS FATIGUE LOADING?

There are different types of fatigue loading. One type is zero-to-max-to zero, where a part which is carrying no load is then subjected to a load, and later, the load is removed, so the part goes back to the no-load condition. An example of this type of loading is a chain used to haul logs behind a tractor. Another type of fatigue loading is a varying load superimposed on a constant load. The

17 suspension wires in a railroad bridge are an example of this type. The wires have a constant static tensile load from the weight of the bridge, and an additional tensile load when a train is on the bridge. The worst case of fatigue loading is the case known as fully-reversing load. One cycle of this type of fatigue loading occurs when a tensile stress of some value is applied to an unloaded part and then released, then a compressive stress of the same value is applied and released.

Fig 3.1 A rotating shaft with a bending load applied to it is a good example of fully reversing load. In order to visualize the fully-reversing nature of the load, picture the shaft in a fixed position (not rotating) but subjected to an applied bending load (as shown in Fig 3.1). The outermost fibres on the shaft surface on the convex side of the deflection (upper surface in the picture) will be loaded in tension (upper green arrows), and the fibres on the opposite side will be loaded in compression (lower green arrows). Now, rotate the shaft 180 in its bearings, with the loads remaining the same. The shaft stress level is the same, but now the fibres which were loaded in compression before you rotated it are now loaded in tension, and vice-versa. To illustrate how damaging fully-reversing load is, take a paper clip, bend it out

18 straight, then pick a spot in the middle, and bend the clip 90 back and forth at that spot (from straight to L shaped and back). Because you are plastically-deforming the metal, you are, by definition, exceeding its yield stress. When you bend it in one direction, you are applying a high tensile stress to the fibres on one side of the OD, and a high compressive stress on the fibres on the opposite side. When you bend it the other way, stresses are reversed (fully reversing fatigue). It will break in about 25 cycles. The number of cycles that a metal can endure before it breaks is a complex function of the static and cyclic stress values, the alloy, heat-treatment and surface condition of the material, the hardness profile of the material, impurities in the material, the type of load applied, the operating temperature, and several other factors. 3.1.2 HOW IS THE FATIGUE STRENGTH OF A METAL DETERMINED?

The fatigue behaviour of a specific material, heat-treated to a specific strength level, is determined by a series of laboratory tests on a large number of apparently identical samples of that specific material. This picture shows a laboratory fatigue specimen. These laboratory samples are optimized for fatigue life. They are machined with shape characteristics which maximize the fatigue life of a metal, and are highly polished to provide the surface characteristics which enable the best fatigue life.

Fig 3.2 A single test consists of applying a known, constant bending stress to a round sample of the material, and rotating the sample around the bending stress axis until it fails. As the sample rotates, the stress applied to any fibre on the outside surface of the sample varies from maximum-tensile to zero to maximum-compressive and back. The test mechanism counts the number of rotations (cycles) until the specimen fails. A large

19 number of tests is run at each stress level of interest, and the results are statistically massaged to determine the expected number of cycles to failure at that stress level. The cyclic stress level of the first set of tests is some large percentage of the Ultimate Tensile Stress (UTS), which produces failure in a relatively small number of cycles. Subsequent tests are run at lower cyclic stress values until a level is found at which the samples will survive 10 million cycles without failure. The cyclic stress level that the material can sustain for 10 million cycles is called the Endurance Limit (EL). In general, steel alloys which are

subjected to a cyclic stress level below the EL (properly adjusted for the specifics of the application) will not fail in fatigue. That property is commonly known as infinite life. Most steel alloys exhibit the infinite life property, but it is interesting to note that most aluminium alloys as well as steels which have been case-hardened by carburizing, do not exhibit an infinite-life cyclic stress level (Endurance Limit).

3.1.3 IS THERE ANY RELATIONSHIP BETWEEN UTS AND FATIGUE STRENGTH? The endurance limit of steel displays some interesting properties. These are shown, in a general way, in this graph, (Figure 7) and briefly discussed below. It is a simplistic rule of thumb that, for steels having a UTS less than 160,000 psi, the endurance limit for the material will be approximately 45 to 50% of the UTS if the surface of the test specimen is smooth and polished. That relationship is shown by the line titled 50%. A very small number of special case materials can maintain that approximate 50% relationship above the 160,000 psi level. However, the EL of most steels begins to fall away from the 50% line above a UTS of

20 about 160,000 psi, as shown by the line titled Polished. For example, a specimen of SAE-4340 alloy steel, hardened to 32 Rockwell-C (HRc), will exhibit a UTS around 150,000 psi and an EL of about 75,000 psi, or 50% of the UTS. If you change the heat treatment process to achieve a hardness of about 50 HRc, the UTS will be about 260,000 psi, and the EL will be about 85,000 psi, which is only about 32% of the UTS. Several other alloys known as ultra-high-strength steels (D-6AC, HP-9-4-30, AF-1410, and some maraging steels) have been demonstrated to have an EL as high as 45% of UTS at strengths as high as 300,000 psi. Also note that these values are EL numbers for fully-reversing bending fatigue. EL values for hertzian (contact) stress can be substantially higher (over 300 ksi). The line titled Notched shows the dramatic reduction in fatigue strength as a result of the concentration of stress which occurs at sudden changes in cross-sectional area (sharp corners in grooves, fillets, etc.). The highest EL on that curve is about 25% of the UTS (at around 160,000 psi). The surface finish of a material has a dramatic effect on the fatigue life. That fact is clearly illustrated by the curve titled Corroded. It mirrors the shape of the notched curve, but is much lower. That curve shows that, for a badly corroded surface (fretting, oxidation, galvanic, etc.) the endurance limit of the material starts at around 20 ksi for materials of 40 ksi UTS (50%), increases to about 25 ksi for materials between 140 and 200 ksi UTS, then decreases back toward 20 ksi as the material UTS increases above 200 ksi. 3.1.4 WHY IS THE SURFACE SO IMPORTANT?

Fatigue failures almost always begin at the surface of a material. The reasons are that

the most highly-stresses fibres are located at the surface (bending fatigue)

the intergranular flaws which precipitate tension failure are more frequently found at the surface.

Suppose that a particular specimen is being fatigue tested (as described above). Now suppose the fatigue test is halted after 20 to 25% of the expected life of the specimen

21 and a small thickness of material is machined off the outer surface of the specimen, and the surface condition is restored to its original state. Now the fatigue test is resumed at the same stress level as before. The life of the part will be considerably longer than expected. If that process is repeated several times, the life of the part may be extended by several hundred percent, limited only by the available cross section of the specimen. That proves fatigue failures originate at the surface of a component. 3.1.5 IS THE ENDURANCE LIMIT AN EXACT NUMBER?

It is important to remember that the Endurance Limit of a material is not an absolute nor fully repeatable number. In fact, several apparently identical samples, cut from adjacent sections in one bar of steel, will produce different EL values (as well as different UTS and YS) when tested, as illustrated by the S-N diagram below. Each of those three properties (UTS, YS, EL) is determined statistically, calculated from the (varying) results of a large number of apparently identical tests done on a population of apparently identical samples. The plot below shows the results of a battery of fatigue tests on a specific material. The tests at each stress level form statistical clusters, as shown. A curve is fitted through the clusters of points, as shown below. The curve which is fitted through these clusters, known as an S-N Diagram (Stress vs. Number), represents the statistical behaviour of the fatigue properties of that specific material at that specific strength level. The red points in the chart represent the cyclic stress for each test and the number of cycles at which the specimen broke. The blue points represent the stress levels and number of cycles applied to specimens which did not fail. This diagram clearly demonstrates the statistical nature of metal fatigue failure.

22

Fig 3.4 DO REAL-WORLD COMPONENTS EXHIBIT THE LABORATORY EL?

3.1.6

Unfortunate experience has taught engineers that the value of the Endurance Limit found in laboratory tests of polished, optimized samples does not really apply to realworld components. Because the EL values are statistical in nature, and determined on optimized, laboratory samples, good design practice requires the determination of the actual EL will be for each specific application, known as the Application-Specific Endurance Limit (AS EL). In order to design for satisfactory fatigue life (prior to testing actual components), good practice requires that the laboratory Endurance Limit value be reduced by several adjustment factors. These reductions are necessary to account for: (a) (b) the differences between the application and the testing environments, and the known statistical variations of the material.

This procedure is to insure that both the known and the unpredictable factors in the application (including surface condition, actual load, actual temperature, tolerances, impurities, alloy variations, heat-treatment variations, stress concentrations, etc. etc. etc.) will not reduce the life of a part below the required value. An accepted contemporary practice to estimate the maximum fatigue loading which a

23 specific design can survive is the Marin method, in which the laboratory testdetermined EL of the particular material (tested on optimized samples) is adjusted to estimate the maximum cyclic stress a particular part can survive (the ASEL). This adjustment of the EL is the result of six fractional factors. Each of these six factors is calculated from known data which describe the influence of a specific condition on fatigue life. Those factors are: 1. Surface Condition (ka): such as: polished, ground, machined, as-forged,

corroded, etc. Surface is perhaps the most important influence on fatigue life; 2. Size (kb): This factor accounts for changes which occur when the actual size of

the part or the cross-section differs from that of the test specimens; 3. Load (kc): This factor accounts for differences in loading (bending, axial,

torsional) between the actual part and the test specimens; 4. Temperature (kd): This factor accounts for reductions in fatigue life which occur

when the operating temperature of the part differs from room temperature (the testing temperature); 5. Reliability (ke): This factor accounts for the scatter of test data. For example, an

8% standard deviation in the test data requires a ke value of 0.868 for 95% reliability, and 0.753 for 99.9% reliability. 6. Miscellaneous (kf): This factor accounts for reductions from all other effects,

including residual stresses, corrosion, plating, metal spraying, fretting, and others. These six fractional factors are applied to the laboratory value of the material endurance limit to determine the allowable cyclic stress for an actual part: Real-World Allowable Cyclic Stress = ka * kb * kc * kd * ke * kf * EL

3.1.7

IS FATIGUE LOADING CUMULATIVE?

It is important to realize that fatigue cycles are accumulative. Suppose a part which has been in service is removed and tested for cracks by a certified aircraft inspection station, a place where it is more likely that the subtleties of Magnaflux inspection are well-understood. Suppose the part passes the inspection, (i.e., no cracks are found)

24 and the owner of the shaft puts it on the good used parts shelf. Later, someone comes along looking for a bargain on such a part, and purchases this inspected part. The fact that the part has passed the inspection only proves that there are no detectable cracks RIGHT NOW. It gives no indication at all as to how many cycles remain until a crack forms. A part which has just passed a Magnaflux inspection could crack in the next 100 cycles of operation and fail in the next 10000 cycles (which at 2000 RPM, isnt very long) [2].

25

CHAPTER 4 4.1 Fracture Mechanisms

In very general terms, when crack size (a) in a stressed part reaches a critical size (acr), the fracture process occurs almost instantaneously with complete and sudden separation of the part. This stage of final fracture is referred to as overload fracture. However, it must be recognized that many overload failures occur after subcritical (a acr) crack growth from various progressive damage mechanisms such as fatigue or environmentally assisted cracking. Overload cracking can be categorized into three general types of mechanisms:

Brittle overload from cleavage (i.e., transgranular brittle cracking) Ductile overload failures that involve the fracture mechanisms of ductile tearing and/ or micro void formation caused by transgranular slip Stress rupture (sometimes referred to as de-cohesive rupture), when a pair of free surfaces are created from a pre-existing grain boundary or second-phase boundary

Ductile and brittle cracking are the two main types of overload fracture, while stress rupture includes various types of mechanisms that may be either brittle or ductile. For example, a brittle stress-rupture failure may occur from grain boundary embrittlement, while a ductile stress rupture failure may occur from grain-boundary slip (i.e., timedependent creep deformation in polycrystalline metals). When examining fracture surfaces, it is important to obtain an overall perspective from both macroscopic and microscopic study. Examination beyond the fracture surface also provides information. For example, visual inspection of a fractured component may indicate events prior to fracture initiation, such as a shape change indicating prior deformation. Metallographic examination of material removed far from the fracture surface also can provide information regarding the penultimate microstructure, including the presence of cold work (bent annealing twins, deformation bands, and/or grain shape change), evidence of rapid loading and/or low-temperature service (deformation twins), and so forth. These types of investigative methods are also important in the analysis of fractures. In general, identifying the cause and corrective action of a fracture benefits by the careful documentation of various macroscopic and microscopic observations. Typically, observation begins with a visual examination of a fracture surface, where general

26 features and surface roughness can be revealed under favourable lighting. Examination at low magnification (about 15 X or less) can also reveal features regarding the nature of the fracture path. Metallographic and fractographic techniques then can be used to reveal microscopic features. The failed piece may be properly sectioned for preparation of metallographic samples and examination under an optical microscope. Alternatively or in addition, an electron microscopetypically a scanning electron microscope (SEM), or a transmission electron microscope (TEM)with higher magnifications and depth of field may be used to make direct examination of the raw fracture surface.

Table 4.1

Maximum shear stress as a function of states of stress

Both the macro- and micro scale appearances of fracture-surface features can tell a story of how and sometimes why fracture occurred in terms of the following information: Crack initiation site and crack propagation direction Mechanism of cracking and the path of fracture Load conditions (tension, bending, shear, monotonic, or cyclic) Environment Geometric constraints that influenced crack initiation and/or crack propagation Fabrication imperfections that influenced crack initiation and/or crack propagation

4.1.1

Ductile Fracture

The mechanisms and appearances of ductile fracture are best introduced with the simple example of an unnotched bar subjected to conventional tension testing at quasistatic strain rates (0.1 s-1). In general, tension test specimens of a ductile material

27 have a visible region of necking (Fig. 4.1), while a brittle specimen results in fracture with little or no visible evidence of any necking (Fig. 4.2). Necking is a region of strain localization that forms when the increase in stress due to decrease in the crosssectional area of the specimen becomes greater than the increase in the load-carrying ability of the metal due to strain. Necking generally occurs at the point of maximum load on the engineering stress-strain curve. Prior to the onset of necking, strain is uniform along the gage length, while plastic deformation becomes concentrated in the necked region of the tension specimen. The size of the neck and the extent to which it is visible depends primarily on strain hardening and strain-rate hardening (when temperatures are below 0.4 Tm of the metallic material, where Tm is the material melting point on the Kelvin scale). The resulting necked region is, in effect, a mild notch, which introduces a complex state of stress that has a large tensile-hydrostatic component. This tensile-hydrostatic (or triaxial) stress is highest in the centre of the specimen, where micro voids occur from the tensile separation of the ductile matrix from harder second-phase particles or inclusions (which are present in most commercial alloys). This central region of the fracture surface also is typically flat (at the macro scale), which indicates separation from a tension stress state. Thus, even though the ductile fracture involves deformation, the microscopic mechanism of crack initiation involves the brittle like effect of tensile separation in the region of a

Fig. 4.1

Cup-and-cone fracture of a low-carbon steel bar under tension.

28

Fig. 4.2

Brittle fracture of a smooth (unnotched) tensile test specimen.

Deformation-induced notch. The micro voids also grow and connect to ultimately form and fracture near the central region of the specimen. However, as the central crack grows, the material in the outer annulus deforms by stresses along the shear plane (45 to the direction of the tensile load). This results in distinctive shear lips of ductile fracture and the classic cup-and-cone profile of the fracture surface. One piece has a cone like profile, while the other piece (shown in Fig. 4.1) has the surface profile of a cup. The ratio of the area of the flat-face region to the area of the shear lip usually increases with section thickness. The area of the shear lip also depends on the extent of necking. The amount of necking depends on the extent of strain and strain-rate hardening, which in turn are influenced by factors such as temperature and material condition. Lowering the temperature below room temperature generally increases the strainhardening exponent, n. This increases the strain to neck formation. In contrast, an ideal plastic material (in which no strain hardening occurs) becomes unstable in tension and begins to neck as soon as yielding occurs. Most metals and alloys (not heavily cold worked) undergo strain hardening, which tends to increase the load-carrying capacity of the specimen as deformation increases. Commercial engineering materials also typically contain inclusions, second-phase particles, and other constituents. These microscopic constituents can influence the process of fracture nucleation and crack growth. In an ideal material containing neither inclusions nor second phases, ductile fracture would be expected to occur by slip and possibly twinning, resulting in complete

29 reduction in area. Alternately, cleavage across a grain on a single plane would be expected to result in a smooth fracture surface. Such results are sometimes observed in high-purity single-crystal specimens, but are seldom seen in commercial engineering materials. Fracture by uninterrupted plastic deformation is a special type of plastic fracture. For metals that do not work harden much, the metal under tension would be drawn down almost to a chisel edge or a point before breaking apart (see Fig. 4.3a). This special circumstance can hardly be termed fracture in a normal sense, as there is no fracture surface at all. It is usually referred to as rupture because the process arises from prolonged shear on slip planes within the worked region of the crystals, which finally at one point shear apart. It should also be noted that strain rate and adiabatic deformation also play an important role in this type of behaviour. That is, high temperature and very slow strain rates can result in extensive uninterrupted deformation. Another form of uninterrupted plastic deformation, which may or may not result in a chisel point type fracture, is the shearing of a single crystal. Deformation of a single crystal is governed by the critical resolved stress and slip occurs in an active slip plane and slides in a specific direction. Figure 4.3(b) shows a copper-aluminium single crystal that has gone through such a prolonged extension. Final separation of the specimen eventually occurs by shearing off at one of the slip planes. Whether a single crystal will fracture by shearing off or by drawing down to a chisel point depends on the slip system of a particular crystal.

4.1.1.1

General Macroscopic Appearance of Ductile Fractures. For ductile

fracture, macroscopic features include:

A relatively large amount of plastic deformation precedes the fracture. Shear lips are usually observed at the fracture termination areas. The fracture surface may appear to be fibrous or may have a matte or silky texture, depending on the material. The cross section at the fracture is usually reduced by necking. Crack growth is slow.

30

Fig. 4.3

Examples of uninterrupted shear failure.

(a) Polycrystalline aluminium bars pulled at 600 C (1110F). (b) Extended copper-aluminium single crystal.

Ductile fractures often progress as single cracks, without many separated pieces or substantial crack branching at the fracture location. The region of crack initiation typically has a dull fibrous appearance that is indicative of cracking by micro void coalescence. The crack profiles adjacent to the fracture are consistent with tearing. The fracture surface may have radial markings, chevrons, and/or shear lips, depending on the specimen geometry and material condition. Depending on the state of stress and geometric constraints on macroscopic ductility, the fracture of a ductile material may occur by plane strain, plane stress, or mixed mode. In general, these variations in fracture profiles are related to fracture toughness, which depends on section thickness (B) and the crack size (a) of a pre-existing discontinuity such as a crack or notch. Figure 4.4 is a schematic illustration of this for an inherently ductile material with varying section thickness. Plane-strain fracture is characterized by a flat surface perpendicular to the applied load. Plane stress fracture occurs when shear strain becomes the operative mode of deformation and fracture (as maximum stresses occur along the shear plane from the basic principles of continuum mechanics). In plane-stress cracking, the fracture profile is characterized by shear lips, which are at about a 45 oblique angle to the maximum stress direction (although this angle may vary depending on material condition and loading condition). The classic cup-and-cone appearance that results from ductile fractures of unnotched cylindrical tension test specimens is a good example of mixed-mode fracture. In this

31 case, crack initiation near the specimen centre occurs from fracture under triaxial tension and thus has a flat fracture surface normal to the applied load. When fracture reaches the region near the outer surface, deformation by slip (shear) becomes dominant, and the stress state of fracture changes to plane stress. However, shear lips are not necessarily the definitive characteristic of a ductile fracture. The macro scale appearance of a fracture is also influenced by geometric constraints and stress-state conditions. For example, even though a flat centre region of crack initiation is characteristic of tensile (brittle like) separation, the flat fracture region of a ductile fracture also has a dull fibrous appearance with small microscopic dimples. Microscopic dimples are indicative of separation from displacement within a ductile matrix (or localized regions) of a material.

4.1.2

Brittle Fracture

Materials that do not develop a neck before fracture are generally considered brittle. When the specimen lacks ductility (due to low temperature, environment, strain rate, or the material itself), the fracture is brittle and occurs by separation on a plane that is normal to the direction of the applied load. Due to absence of necking, deformation is approximately uniform until fracture occurs from complete separation under tension. However, lack of ductility depends on a number of other factors, such as environment, strain rate, and the internal state of stress created (influenced by part geometry and discontinuities in the material). Therefore, lack of ductility is not just due to the material itself, but is also influenced by complex relationships of stress state, part geometry, localized deformation, and internal discontinuities. Conversely, a fracture surface normal to the applied load also is not necessarily a definitive indication of an inherently brittle material or a brittle mechanism (such as cleavage or intergranular fracture). This important distinction is indicated in the central region of the classic ductile fracture in Fig. 4.1. Initial deformation causes a region of localized strain (i.e., necking), which is essentially a mild notch that results in the development of hydrostatic tensile stresses in the interior of a tension test specimen. This triaxial state of tension then causes crack initiation (void formation) by tensile separation around small inclusions, second-phase particles, or discontinuities in the centre region of the tension bar. In effect, ductile mechanisms lead to a stress state that causes tensile separation around less ductile.

32

Fig. 4.4.

Schematic of variation in fracture behaviour and macro

scale features of fracture surfaces for an inherently ductile material. As section thickness increases, plane strain conditions develop first along the centre line and result in a flat fracture surface. With further increases in section thickness, the flat region spreads to the outside of the specimen, decreasing the widths of the shear lips.

In general, brittle fracture can be distinguished by these characteristics:

Little or no visible plastic deformation precedes the fracture. The fracture surface is generally flat and perpendicular to the loading direction and to the component surface. The fracture may appear granular or crystalline and is often highly light reflective. Facets may also be observed, particularly in coarse-grain steels. Chevron patterns may be present. Rapid crack growth results in catastrophic failure, sometimes accompanied by a loud noise.

Brittle overload failures, in contrast to ductile overload failures, are characterized by little or no macroscopic plastic deformation. Brittle fracture initiates and propagates more readily than ductile fracture or for so-called subcritical crack propagation processes such as fatigue or stress-corrosion cracking (SCC). Because brittle fractures are characterized by relatively rapid crack growth, the cracking process is sometimes referred to as being unstable or critical because the crack propagation leads quickly to final fracture. The macroscopic behaviour of brittle fracture is essentially elastic up to the point of failure. The energy of the failure is principally absorbed by the creation of new

33 surfacesthat is, cracks. For this reason, brittle failures often contain multiple cracks and separated pieces, which are less common in ductile overload failures. All brittle fracture mechanisms can exhibit chevron or herringbone patterns that indicate the fracture origin and direction of rapid fracture progression. Herringbone patterns are unique microscopic features of brittle fractures. Ductile cracking, which occurs by micro void coalescence, does not result in a herringbone pattern. On a microscopic scale, the features and mechanisms of fracture may have components of ductile or brittle crack propagation, but the macroscopic process of fracture is characterized by little or no work expended from deformation.

4.1.2.1

Microscopic Aspects of Fracture

Although the examination of a fracture surface begins logically with macroscopic observations, the microscopic aspects of fracture are also essential in understanding the causes of fracture. In general, fracture on a micro scale can be distinctively defined in terms of transgranular brittle fracture (cleavage), transgranular slip (ductile fracture), or intergranular fractures. These three types of fracture paths result in distinctly different fracture surfaces, as seen in Fig. 4.5. These distinct microscopic appearances provide important information on the underlying mechanisms of fracture. On a microscopic level, most engineering materials (alloys) are polycrystalline solids that consist of many grains (crystals) with grain boundary regions between the crystals. Typically, the grain boundaries are stronger than individual grains in properly processing polycrystalline plastic/elastic solids. The reason for this can be understood in simple terms. The grain boundaries are disruptions between the crystal lattice of individual grains, and this disruption provides a source of strengthening by pinning the movement of dislocations. Thus, a finer-grain alloy imparts more grain-boundary regions for improved strength. Moreover, because a greater number of arbitrarily aligned grains are achieved when grain size is reduced, the stressed material has more opportunity to allow slip and thus improve ductility. However, the grain boundaries are also a region with many faults, dislocations, and voids. This relative atomic disarray of the grain boundaries, as compared to the more regular atomic arrangement of the grain interiors, provides an easy path for diffusionrelated (thermally activated) alterations. For this reason, grain boundaries can be a preferential region for congregation and segregation of impurities, preferential phase precipitation, and/or absorption of environmental species. These thermally activated

34 alterations are potential mechanisms for weakening or embrittlement along the grain boundaries. In addition, grain-boundary regions are weakened when the temperature is high enough to activate diffusion-induced flow deformation along grain boundaries at stresses below the yield strength. This onset of time-dependent flow (i.e., creep deformation) is roughly representative of the viscoelastic deformation that occurs when the temperature is higher than 0.4 Tm. This basic overview of the relative strength of grains and grain boundaries provides a general framework to describe the basic mechanisms of transgranular and intergranular fracture. Transgranular fracture of a crystalline material can occur by either a brittle process of cleavage or by the ductile process of micro void formation. These two mechanisms of transgranular fracture are very distinct in terms of appearance and have clear causes. Brittle transgranular fracture of a crystalline material takes place by cleavage along low-index crystallographic planes within grains, while ductile transgranular fracture occurs when small voids (micro voids) form and coalesce in the region of fracture. These micro voids leave distinctive concave depressions called dimples on both surfaces of the fracture. Intergranular fractures are also very distinctive in appearance (Fig. 4.5a), but the underlying causes or mechanisms can be complex and varied. Grain boundaries are weakened in various ways, and the surface of an intergranular fracture does not necessarily reveal the evidence of the underlying mechanism that leads to fracture along a grain-boundary region. This illustrates the need for careful analysis of the overall circumstances that lead to fracture. In many instances, SEM fractography provides a means to identify the fracture path as intergranular, but it may yield little other information. Additional important information can be obtained by chemical analysis in the SEM and by microstructural examination. Use of the SEM for microstructural examination in addition to optical light examination may be appropriate depending on the scale of micro constituents present. Important clues on the underlying cause of intergranular separation also may be revealed by fractography at assorted magnifications.

35

Fig. 4.5

SEM images of (a) intergranular fracture in ion-nitrided layer of

ductile iron (ASTM 80-55-06), (b) transgranular fracture by cleavage in ductile iron (ASTM 80-55-06), and (c) ductile fracture with equiaxed dimples from micro void coalescence around graphite nodules in a ductile iron (ASTM 65-40-10). Picture widths are approximately 0.2 mm (0.008 in.) from original magnifications of 500X.

4.1.3 Formation).

Transgranular Ductile Fracture (Transgranular Slip and Micro void

In terms of inherent material structure of crystalline materials, the deformation processes of slip and twinning compete with the brittle fracture process of cleavage. Cleavage is a brittle process that occurs on the plane of maximum normal stress, while slip mechanisms are associated with plastic deformation and ductile fracture. At temperatures lower than 0.4 Tm, plastic deformation occurs by transgranular slip and/or

36 twinning in the crystalline lattice. If other events do not intervene, this deformation culminates in fracture first by strain localization (necking or shear band formation), and then final fracture occurs in the volume or region of strain concentration. At temperatures of ~0.4 Tm or higher, however, deformation can occur by slip and viscous grain-boundary flow, as the grain boundary regions become weakened at high temperature. Thus, the predominant fracture path becomes intergranular in the region of creep (time-dependent) deformation at temperatures of _0.4 Tm or higher. The overall process of ductile fracture is illustrated by the preceding example of a ductile fracture in an unnotched tension test specimen (Fig. 4.1). Ductile fractures are uniquely characterized by micro voids that form in the region of high stress. In an unnotched tension test bar, micro voids nucleate and grow in the central region, where the diffuse notch (created by necking) causes separation due to triaxial (hydrostatic) tension. These voids coalesce and join together to form a microscopic crack. At the same time, more small cavities are formed and distributed over the remaining section of the test piece. A typical example is presented in Fig. 4.6, which shows a cross section of a tensile specimen containing numerous voids at a stage between necking and final fracture. The fracture process consists of these voids joining on the plane perpendicular to the loading direction and coalescing into a central crack. This crack grows until it approaches the outer annulus, where a change in fracture path occurs as the process approaches final fracture. At some point (depending on ductility), final fracture occurs along the shear plane and results in shear lip. The amount of shear lip varies, depending on ductility, strain rate, and temperature. The surfaces of plastic fractures are characterized by microscopic dimples, also called cupules. These dimples represent the numerous concave depressions left on the opposite fracture faces of the broken specimen. Dimples on fracture surfaces are observed in many materials, including carbon and alloy steels, austenitic steels, alloys of aluminium, titanium, and copper, and plastics. It has been suggested that dimples represent the coalesced voids, and the voids initiate from inclusions or intermetallic particles. Dimples also can take different shapes, depending on loading condition (Fig. 4.7). Round dimples occur from separation under tension, while the dimples have an elongated parabolic shape when ductile fracture occurs from shear, torsion, and tearing (or bending). In general, the concavity of the parabola is oriented toward the direction of relative displacement of the other half of the specimen, and the axis of symmetry

37

Fig. 4.6

Section through the neck area of a tensile specimen of

copper showing cavities and crack formed at the centre of the specimen as the result of void coalescence.

38

Fig. 4.7

Schematic of plastic fracture. (a) Normal plastic (formation of

round dimples). (b) Shear plastic (formation of elongated dimples pointing in the direction of shear on each fracture surface). (c) Tear plastic (formation of elongated dimples pointing in the direction opposite to the direction of each propagation. (d) Dimple elongation from out-of-plane shear. I Dimple elongation from mixed mode of screw sliding with ductile tear (c _ d).

39 is parallel to the direction of propagation of the rupture front (Fig. 4.7). Thus, when the two opposite surfaces of rupture due to in-plane shear (Fig. 4.7) are examined, the concavity of these dimples is turned in opposite directions on the opposite faces. Parabolic dimples from torsional loading are shown in Fig. 4.8. An example of matching surfaces is shown in Fig. 4.9 for dimple shape and orientation for bending/ ductile tearing. In the case of bending, there may be some question as to how often elongated dimples are seen in bending loading (or seen at all), because dimples on typical compact tension specimens (axial _ bending) appear mostly equiaxed.

4.1.4

Transgranular Brittle Fracture (Cleavage).

Cleavage is a low-energy fracture that propagates along well-defined low-index crystallographic planes known as cleavage planes. Theoretically, a cleavage fracture should have perfectly matching faces and should be completely flat and featureless. However, engineering alloys are polycrystalline and contain grain and sub grain boundaries, inclusions, dislocations, and other imperfections that affect a propagating cleavage fracture so that true, featureless cleavage is seldom observed. These imperfections and changes in crystal lattice orientation, such as possible mismatch of the low-index planes across grain or sub grain boundaries, produce distinct cleavage

Fig. 4.8

Parabolic shear dimples from torsional fracture of cast Ti-6Al-

4V alloy. Original magnification, 2000 X

40 fracture surface features, such as cleavage steps, river patterns, feather markings, herringbone patterns, and tongues. Cleavage fracture also occurs in ceramics, inorganic glasses, and polymeric materials. The cleavage mode of fracture is controlled by tensile stresses acting normal to a cleavage plane and is brittle in nature. Its fracture surface, which is caused by cleavage, appears at low magnification to be bright or granular, owing to reflection of light from the flat cleavage surfaces. It exhibits a river pattern when examined under an electron microscope. It occurs in bcc and hcp metals, particularly in irons and steels, below the ductile-to-brittle transition temperature (DBTT). Cleavage fracture is very seldom found in fcc metals. The fcc metals (e.g., copper, aluminium, nickel, and austenitic steels) have a large number of slip systems (12), which is one reason why they exhibit high ductility. The fcc metals also are more closely packed (i.e., a shorter distance exists between atoms in the crystal cell) than are bcc metals. This partly explains why cleavage fracture does not normally occur in the matrix of fcc metals. However, it is not just a matter of the multiplicity of ways for slip or cleavage to occur. Two other factors also control the inherent ductile brittle behaviour of crystalline materials:

Fig. 4.9

Fracture markings on Plexiglas. TEM, matching fracture

surfaces. Note the matching features A, A and B, B on the two fracture faces. The parabola markings are similar to the plastic dimples observed in a tear ductile fracture of a metallic material.

41 Critical shear stress required to initiate slip Critical normal stress required to propagate a cleavage crack

As long as there is only one type of atom in the lattice, the shear stress for slip is low. The presence of foreign atoms raises this stress and, depending on the location of the foreign atoms (random or periodic), may cause a severe loss in ductility, as is the case for the ordered intermetallic alloys. Thus, the critical stresses required for slip and cleavage also determine whether or not cleavage occurs. The cleavage process in bcc and hcp metals occurs by separation normal to crystallographic planes of high atomic density. Microscopic examination of a fracture surface from cleavage typically reveals distinctive river lines indicative of propagation by fracture along nearly parallel sets of cleavage planes. The direction of crack propagation is indicated by the flow of the river lines as marked by an arrow in Fig. 4.9.

4.1.5

Quasi-Cleavage.

Totally brittle fracture in metals at the microscopic level (ideal cleavage or pure cleavage) occurs only under certain well-defined conditions (primarily when the component is in single-crystal form and has a limited number of slip systems) and is correctly described as cleavage fracture. More commonly in metals, the fracture surface contains varying fractions of transgranular cleavage and evidence of plastic deformation by slip. Grains oriented favourably with respect to the axis of loading may slip and exhibit ductile behaviour, whereas those oriented unfavourably cannot slip and will exhibit transgranular brittle behaviour. When both transgranular fracture processes operate intimately together, the fracture process is termed quasi-cleavage. The dividing line between cleavage and quasicleavage is somewhat arbitrary. The term quasi-cleavage applies when significant dimple rupture and/or tear ridges accompany the cleavage morphology. The fracture surface is typically dominated by cleavage, but there are usually small patches of micro void coalescence present or thin ribbons of micro void coalescence contained in the fracture surface. As the patches increase, the fracture surface is more accurately described as (micro scale) mixed cleavage and micro void coalescence. Another term is cleavage with ductile tear ridges. Quasi-cleavage should not be confused with the decohesion along certain

42 crystallographic planes that can occur by shear, by sliding off (plastic shear), or by separation along weak, still poorly defined interfaces. This type of decohesion has been referred to as glide-plane decohesion. Quasi-cleavage fractures also should not be confused with those in which cleavage appears in brittle second phases with the characteristic dimples of micro void coalescence appearing in the more ductile matrix. In quasicleavage, there is no apparent boundary between a cleavage facet and a dimpled area bordering the cleavage facet (Fig. 4.10).

Figure 4.11 is a schematic representation of quasi-cleavage. The occurrence of quasicleavage is usually distinguished by: Initiation within facet boundariesin contrast to fracture by cleavage, which usually initiates from one edge of the region being cleaved (Fig. 4.12) Cleavage steps appearing to blend directly into tear ridges of the adjacent dimpled areas

Many high-strength engineering metals fracture by quasi-cleavage, which is a mixed mechanism involving both micro void coalescence and cleavage. When tested under embrittling conditions, such as those imposed by corrosive mediums or triaxial stress states, quasi-cleavage can occur in metals that normally are not known to have active cleavage planes (e.g., austenitic stainless steels, and nickel and aluminium alloys). One explanation is that facets that exhibit quasi-cleavage features fracture ahead of the moving crack front; then, as the stress increases, the cleavage facet extends by tearing into the matrix around it by micro void coalescence.

Quasi-cleavage fracture surfaces appear in steels from: Quasi cleavage, or cleavage in complex microstructures, is more difficult to identify than the cleavage found in low-carbon steel made up of ferrite and pearlite. When identification is uncertain, it is essential to relate the fracture features to the microstructure, including the prior austenite grain size, the martensite plate size, and the distribution, size, spacing, and volume fraction of fine carbide particles precipitated sudden or impact loading, low temperature, high levels of constraint (ambient temperature), in heavily cold worked parts (ambient temperature).

43 during tempering. With a few exceptions, intergranular fractures are macroscopically brittle with little or no mechanical work expended as part of the fracture process. However, the micro mechanisms of intergranular fracture may be brittle or ductile, depending on how the grain-boundary regions are weakened or embrittled. For example, grain boundaries may become embrittled by a film of a brittle phase or by the segregation of an impurity in the boundary region. In this case, the mechanism of intergranular cracking may be brittle by cleavage in the brittle phases that congregate in the grain boundaries. Conversely, intergranular fracture may also occur from ductile micro mechanisms (slip) that involve localized formation of ductile micro voids in the region near the grain boundaries. For example, voids along the grain boundaries may form at a particle in

Fig. 4.10

Effect of quasi-cleavagemixed cleavage and micro void

coalescenceon the fracture surface appearance of 17-PH stainless steel. TEM p-c replica, 4900X

44

Fig. 4.11

Fracture model showing a cleavage step blending with a tear

ridge in a quasi-cleavage fracture surface. At top left is the lower surface of a fracture, showing a step at the lower left and a ridge at the upper right. At right and at bottom are sections through the fractured member, showing profiles of both the upper and the lower fracture surfaces.

The grain boundary or in a precipitate free zone (PFZ) adjacent to the grain boundary. This feature is sometimes referred to as dimpled intergranular fracture. The interpretation of intergranular fracture is more complex than the distinct mechanisms of transgranular fracture (i.e., fatigue, cleavage, or ductile with micro void coalescence). However, the appearance of intergranular fracture is relatively easy to recognize, and the causes are fairly limited. The presence of intergranular fracture (especially in the region of crack initiation) also is often helpful in narrowing the potential cause for failure. Some common circumstances that have been known to induce intergranular cracking have been classified into four general categories:

Presence of grain-boundary precipitates Thermal treatment or exposure that causes segregation of certain impurities to the grain boundaries without an observable second phase Stresses applied at elevated (creep-regime) Temperatures

45

Fig. 4.12

Cleavage in a large second-phase particle on a fracture

surface of A-286 steel

Environmental assisted alteration or weakening of the grain boundaries by various mechanisms such as hydrogen embrittlement, liquid-metal embrittlement, solid metal embrittlement, oxidation or reduction potentially in the grain boundaries, radiation embrittlement, and SCC

Large grain size also plays a role in causing a change from transgranular to intergranular cracking and can enhance any of the above mechanisms.

4.1.6

Mechanisms of Intergranular Fracture.

On an atomic scale, crack growth occurs by any one or a combination of the following: Tensile separation of atoms (decohesion) Shear movement of atoms (dislocation egress or insertion) Removal or addition of atoms by dissolution or diffusion

All of these processes can occur preferentially along the grain boundary by various phenomena, such as:

Segregation of embrittling elements to the grain boundaries More rapid diffusion of elements along grain boundaries than along grain interiors More rapid nucleation and growth of precipitates in grain boundaries than in grain

46 interiors Greater adsorption of environmental species in the grain-boundary regions

These basic mechanisms of intergranular cracking are more varied than those of transgranular fracture. However, except for conditions of creep stress rupture at elevated temperatures, intergranular fracture is not common in properly processed materials in a benign environment. There also are a fairly limited number of circumstances of intergranular fracture associated with improper processing of a material and/or some aggressive service environment. Some specific situations include:

High-carbon steels with a pearlitic microstructure Segregated phosphorus and cementite at prior-austenite grain boundaries in the high carbon-case microstructures of carburized steels Stress-relief cracking Grain-boundary carbide films due to eutectoid divorcement in low-carbon steels Grain-boundary hypereutecoid cementite in carburized or hypereutectoid steels Iron nitride grain-boundary films in nitrided steels Temper embrittlement in heat-treated steels due to segregation of phosphorus, antimony, arsenic, or tin Embrittlement of copper due to the precipitation of a high density of cuprous oxide particles at the grain boundaries Embrittlement of steel due to the precipitation of MnS particles at the grain boundaries as a result of overheating

Fig. 4.13

Quasi-cleavage in the surface of an impact fracture in a

47 specimen of 4340 steel. The same area is shown in both SEM fractographs, but at different magnifications. The small cleavage facets in martensite platelets contain river patterns and are separated by tear ridges. Shallow dimples, marked by arrowheads, are also visible. Direction of crack propagation is from bottom to top in each fractograph. The specimen was heat treated at 845C (1550F) for 1 h, oil quenched, and tempered at 425C (800F) for 1 h. Fracture was by Charpy impact at _196C (_321F). (a) 1650X. (b) 4140X

Grain-boundary carbide precipitation in stainless steels (sensitization) Improperly precipitation-hardened alloys, resulting in coarse grain-boundary precipitates and a denuded region (PFZ) Embrittlement of molybdenum by interstitials (carbon, nitrogen, oxygen) Embrittlement of copper by antimony Reduction of Cu2O in tough-pitch copper by hydrogen Hydrogen embrittlement by grain-boundary absorption of hydrogen Stress-corrosion cracking (sometimes intergranular, but also transgranular) Liquid metal induced embrittlement (LMIE), for example, mercury in brass, lithium in type 304 stainless steel Solid metal induced embrittlement (SMIE)

In all these cases, SEM fractography can provide the means to identify the fracture path. However, it cannot yield sufficient information on the underlying causes (or mechanisms) of intergranular fracture. Thus, additional important information may be needed in terms of chemical analysis or fractographic examination at assorted magnifications. The following sections briefly describe appearances for three general categories of intergranular fracture:

Intergranular brittle cracking Dimpled intergranular fracture Intergranular fracture surfaces with corrosion products Intergranular brittle cracking typically has a relatively clean fracture

4.1.6.1

surface with the faceted appearance of cracking along grain contours. The general appearance of intergranular brittle fracture may include:

48

Brittle second-phase particles and/or films in grain boundaries Fracture where no film is visible and, due to impurities, atom segregation at the grain boundary Environmentally induced embrittlement where there is neither a grain-boundary precipitate nor solute segregation

Grain-boundary segregation of elements (such as oxygen, sulphur, phosphorus, selenium, arsenic, tin, antimony, and tellurium) is known to produce intergranular brittle fractures. Studies of the effects of such impurities in pure iron have been greatly aided by the development of Auger electron spectroscopy. In the case of brittle grainboundary films, it is not necessary for the film to cover the grain boundaries completely; discontinuous films are sufficient. Some common examples of intergranular embrittlement by films or segregants include:

Grain-boundary carbide films in steels Iron nitride grain-boundary films in nitrided steels Temper embrittlement of alloy steels by segregation of phosphorus, antimony, arsenic, or tin Grain-boundary carbide precipitation in austenitic stainless steels (sensitization) Embrittlement of molybdenum by oxygen, nitrogen, or carbon Embrittlement of copper by antimony

Grain-boundary strengthening is characteristic of intergranular fractures caused by embrittlement. Intergranular brittle fracture can usually be easily recognized, but determining the primary cause of the fracture may be difficult. Fractographic examinations can readily identify the presence of large fractions of second-phase particles on grain boundaries. Unfortunately, the segregation of a layer a few atoms thick of some element or compound that produces intergranular fracture often cannot be detected by fractography.

4.1.6.2

Dimpled intergranular fractures result in low macroscopic ductility (i.e.,

the grains separate rather than deform) and a fracture surface that reveals microscopic dimples at higher magnifications (typically on the order of 1000 to 5000X). For example, a fracture surface from stress rupture of a nickel-base alloy is shown in Fig.

49 4.14 at two levels of magnification. The higher-magnification image reveals a dimpled topology on the grain facets. Another example is shown in Fig. 4.15 for a high-purity aluminium-copper precipitation hardened alloy with a coarse grain structure. In this example (with the coarse grain size), ductility is limited, and the yield strength in the local region of the grain boundary becomes lower than the matrix; thus, fracture tends to develop first within the grainboundary zone by micro void coalescence. The two levels of magnification provide a useful combination of images, one demonstrating the intergranular fracture path and the other revealing the microscopic mechanism of micro void coalescence. Other circumstances of dimpled intergranular fracture include:

Uniform void nucleation aided by the formation of methane bubbles at the grain boundaries (Ref 4.7) Void nucleation at precipitates in the grain boundaries of precipitation-hardening alloys (such as Al-Mg-Zn alloys) with large precipitates on grain boundaries (Ref 4.8) and wide PFZs

Voids aided by impurities that adsorb strongly on the grain-boundary surface Stress-relief cracking of chromium-molybdenum steels (Ref 4.7)

4.1.6.3

Intergranular fracture surfaces with corrosion products can provide

some evidence of cause. For example, corrosion products are

Fig. 4.14

SEM image of the fracture surface of a nickel-base alloy

50 (Inconel 751, annealed and aged) after stress rupture (730 C, or 1350 F; 380 Mpa, or 55 ksi; 125 h). (a) Low-magnification view, with picture width shown at approximately 0.35 mm (0.0138 in.) from original magnification of 250X. (b) High-magnification view, with picture width shown at approximately 0.1mm (0.004 in.) from original magnification of 1000X.

Fig. 4.15

SEM fractographs of the tensile test fracture surface of a high-

purity, coarse-grained Al-4.2Cu alloy with (a) intergranular facets at low magnification (10X) and (b) uniform dimples on one facet at higher magnification (67X). The microstructure indicated alloy depletion at the grain boundaries.

Frequently observed on the separated grain facets of fractures of intergranular stresscorrosion cracking (IGSCC) (Fig. 4.15). Intergranular fractures from stress rupture may also have oxidation products. For example, Fig. 4.16 is an example of turbine blade failure from a combustion gas environment. The original reference reported it as an IGSCC failure, when in all likelihood the fracture is one of creep rupture with oxidation products on the surface. Oxides are commonly observed on the fracture surface of creep fractures. This illustrates the importance of using all available information sources, such as stress analysis and surface chemical analysis. Depending on the environment, cracks from hydrogen embrittlement may also reveal corrosion products on a fracture surface. Analysis of the cause of fracture in metal parts and components that have been exposed to corrosive environments is often difficult because of interactions of fracture mechanisms or because fractures generated by different mechanisms have similar

51 appearances.

4.1.7

Macroscopic Aspects of Overload Failures

In terms of macroscopic behaviour, overload cracking is either ductile or brittle, but the entire fracture may occur from different combination: 1. Totally ductile 2. Totally brittle

Fig. 4.15

Intergranular fracture of 201-T6 cast aluminium after SCC

testing. (a) Optical micrograph, Kellers etch, approximately 75X. (b) SEM image of fracture surface.

Fig. 4.16

Surface from fractured U-700 turbine blade. (a) Region with

transgranular and intergranular fracture feature. (b) Debris on intergranular

52 facets may be indicative of oxidation at high temperature after creep cracking.

4. Initially ductile, then brittle 5. Mixed mode (ductile and brittle)

In the last two cases (4 and 5), the ductile appearance may not be directly visible at the macro scale. Initially ductile fractures (case 4) are usually associated with rising-load ductile tearing, or the initial ductility may be inferred by transverse strain at the crack tip. The size of the plastic zone may be micro scale in this case. Mixed-mode ductile and brittle cracking (case 5) would be inferred due to the presence of an intimate mixture of cleavage and micro void coalescence at the micro scale (quasi-cleavage) or by the presence of shear lips at the macro scale. On a macroscopic scale, ductile and brittle fractures often are simply determined by visual examination for evidence of whether the part is deformed excessively, not at all, or somewhere in between. In the case of a tensile bar, for example, one might judge the degree of brittleness of that material by its stress-strain curve (i.e., by the amount of elongation or reduction in area before final fracture). More often, however, the macro scale appearance is insufficient to convey the full story about a fracture. The microscopic mechanisms and appearances of fracture also are needed for thorough understanding and testing of hypotheses in a case study of a fracture. Uncovering both the macro- and micro scale mechanisms allows the source and cause of a given failure to be identified, and thus a course of corrective actions can be reached more reliably. Surface roughness and optical reflectivity also provide qualitative clues to events associated with crack propagation. For example, a dull/ matte surface indicates micro scale ductile fracture, while a shiny, highly reflective surface indicates brittle cracking by cleavage or intergranular fracture. In addition, when intergranular fracture occurs in coarse-grain materials, individual equiaxed grains have a distinctive rock-candy appearance that may be visible with a hand lens. In terms of documenting surface conditions, one major problem with optical (light) macroscopic or microscopic examination of fracture surfaces is its inability to obtain favourable focus over the entire surface if the magnification exceeds 5 to 10X. Therefore, SEM also has become a standard metallographic tool in failure analysis. Surface roughness provides clues as to whether the material is high strength (smoother) or low strength (rougher) and whether fracture occurred as a result of cyclic

53 loading. The surfaces from fatigue crack growth are typically smoother than monotonic overload fracture areas. The monotonic overload fracture of a high strength quenched and tempered steel is significantly smoother overall than is the overload fracture of a pearlitic steel or annealed copper. Also, fracture surface roughness increases as a crack propagates, so the roughest area on the fracture surface is usually the last to fail. Fracture surface roughness and the likelihood of crack bifurcation also increase with magnitude of the applied load and depend on the toughness of the material. Brittle failures often contain multiple cracks and separated pieces, while ductile overload failures often progress as single cracks, without many separated pieces or substantial crack branching at the fracture location. Macroscopic features typically help identify the fracture initiation site and crack propagation direction. For example, crack branching and T junctions (Fig. 4.17) can indicate the direction of crack propagation and location of crack initiation. Similar techniques also apply when brittle materials fracture into multiple pieces (Fig. 4.18). The orientation of the fracture surface, the location of crack initiation site(s), and the crack propagation direction should correlate with the internal state of stress created by the external loads and component geometry. In general,

Fig. 4.17

General features to locate origin from crack path branching (a)

and sequencing of cracking (b) by the T-junction procedure, where fracture A

54 precedes and arrests fracture B

Fracture initiates in the region where local stress (as determined by the external loading conditions, part geometry, and/or macroscopic and microscopic regions of stress concentration) exceeds the local strength of the material. This includes micro scale discontinuities (such as an inclusion or forging seam) and macroscopic stress concentrations (such as a geometric notch or other change in cross section). The fracture surface orientation relative to the component geometry may also exclude some loading conditions (axial, bending, torsion, monotonic versus cyclic) as causative factors. For example, crack initiation is not expected along the centre line of a component loaded in bending or torsion, even if a significant material imperfection is present at that location, because

Fig. 4.18

Characteristics of crack direction and branching in fractures of

brittle materials from (a) impact, (b) bending, (c) torsion, and (d) internal pressure

55 a shear stress at this location in bending, but in a homogeneous material, it is too small to initiate fracture. That might not be the case for a laminated structure loaded in bending.) Likewise, the profile of a fracture surface can indicate the direction of crack growth. For example, the region of plane-strain fracture indicates the direction of fracture in a shear overload fracture of annealed iron sheet (Fig. 4.19). Under the right conditions, fracture surfaces may also have radial marks and chevrons, which are macroscopic surface features that indicate the region of crack initiation and propagation direction. They are common and dominant macroscopic features of the fracture of wrought metallic materials, but are often absent or poorly defined in castings. The V of a chevron points back to the initiation site, and a sequence of Vs across the fracture surface indicates the crack propagation direction. The appearance of chevrons or radial marks near the crack origin depends in part on whether the crackgrowth velocity at the surface is greater or less than that below the surface. If crackgrowth velocity is at a maximum at the surface, radial marks have a fan-shape appearance (Fig. 4.20). If crack growth rate is greatest below the surface, the result is chevron patterns (Fig. 4.21). In rectangular sections, specimen dimensions can affect the appearance of radial markings and chevron patterns. For example, the macro scale fracture appearances of unnotched sections are shown schematically in Fig. 4.22 for sections with various width-to-thickness (w/t) ratios. The w/t ratio influences the ability of the sample to maintain a unidirectional state of stress during tension. In a thick section (top), strain in the width direction is constrained and thus tends to a condition of plane-strain (mode I) fracture. In this case, a large portion of the fracture surface

Fig. 4.19

Fracture surface showing a localized zone of plane-strain

fracture (left) from shear overload failure of annealed Armco iron sheet at _196C (_321F). The configuration indicates that the fracture propagated

56 from left to right in this view. Light fractograph, 5X

is comprised of radial markings or chevron patterns indicative of rapid, unstable cracking. At higher w/t ratios, the radial zone is suppressed in favour of a larger shearlip zone. In very thin sections (bottom), plane-stress conditions apply, and the fracture surface is comprised almost entirely of a shear lip outside the fibrous zone of crack initiation. Figure 4.23 is a schematic of radial marks and chevrons when fracture initiates from surface notches. If conditions are right, radial markings associated with rapid or unstable crack propagation may also appear on the fracture surface of bar sections. The extent of radial markings depends on ductility, as seen in Fig. 4.24 from tension testing of unnotched bars after at different temperatures. As temperature decreases, ductility and

Fig. 4.20 Radial marks typical of crack propagation that is fastest at the surface (if propagation is uninfluenced by part or specimen configuration)

Fig. 4.21 Chevron patterns typical when crack propagation is fastest below the surface. It is also observed in fracture of parts having a thickness much smaller than the length or width (see middle illustration in Fig. 4.22).

57

Fig. 4.22 Typical fracture appearances for unnotched prismatic tension test sections

Fig. 4.23 Schematic of typical fracture appearances for edge- and side-notched rectangular tension test sections

58

Fig. 4.24 Fracture surfaces of unnotched AISI 4340 steel specimens (heat treated to a hardness of 35 HRC) after tension testing at three different temperatures. (a) A shear lip surrounding a fibrous region is visible in the specimen tested at 160C (320F). (b) At a lower test temperature (90C, or 195F), a radial fracture zone formed around the fibrous region, which formed first; also, the shear lip here is smaller. (c) No fibrous region formed in the specimen tested at _80C (_110F). Instead, fracture formed a radial zone that extends nearly to the specimen surface and terminates in a narrow shear lip. Original magnifications all at 15X

the extent of shear-lip formation are reduced. The fracture surface of radial marks is visually distinct from the fibrous region near the centre on an unnotched bar (Fig. 4.24b). The radial markings (sometimes called a radial shear, star, or rosette) are perpendicular to the crack front. With an unnotched tension-test bar, radial marks point to crack initiation in the central region of the bar (Fig. 4.25a). This is not true with notched bars (Fig. 4.25b), where crack initiation occurs near the root of the notch, and where the radial zone points to the region of fast final fracture near the central region of the notched bar. The extent of radial marking depends on the degree of ductility, as shown in Fig. 4.26 from tension testing of notched 4340 bar at various temperatures. As an illustration of environmental effects, tension-overload fractures of a notched quenched and tempered 4340 steel at room temperature are presented in Fig. 4.27, which shows fracture surfaces from sustained-load cracking after hydrogen charging. Note that the progressive decrease in fracture stress from Fig. 4.27(a) to (d) is related to a progressive increase in the size of the fibrous zone in the outer region by the notch.

59

Fig. 4.25 General fracture-surface regions from ductile fracture of an unnotched (a) and notched (b) tension test bar. (a) Radial zones on an unnotched point to the region of crack initiation near the centre of the specimen. (b) In the notched tensile specimen, the fibrous zone surrounds the radial zone, because fracture initiates near the root of the notch (and completely around the specimens in this idealized case without additional stress raisers). Fast final fracture occurs in the centre.

60

Fig. 4.26 Overload fracture in notched AISI 4340 steel specimens (35 HRC) from tension testing at three different temperatures. (a) The surface of the specimen tested at _40C (_40F) shows only fibrous marks. (b) The specimen tested at 80C (_110F) has a fibrous zone that surrounds a radial zone, which is off-centre because of non-symmetrical crack propagation. (c) The specimen tested at _155C (_245F) has a small annular zone of fibrous fracture, with prominent radial marks in the central region of final fast fracture. All at ~17X

61

Fig. 4.27

Effect of sustained loading with hydrogen charging on the

fracture-surface characteristics of notched specimens of quenched and tempered AISI 4340 steel tension tested at room temperature. (a) The specimen with a relatively small fibrous zone at the right edge was broken by tension overload with notched tensile strength of 2005 Mpa (291 ksi). The specimen in (a) was not charged with hydrogen, while the three other specimens were charged with hydrogen and then subjected to sustained loading as follows: (b) Broke in 1.65 h under a stress of 1380 Mpa (200 ksi). (c) Broke after 5.35 h under a stress of 1035 Mpa (150 ksi). (d) Broke after 5.5 h under a stress of 690 Mpa (100 ksi). Notch radius was 0.025 mm (0.001 in.). Microstructure of all four specimens was tempered martensite. All at ~8X

62 4.1.8 Cracks propagating from a pre-existing stress raiser or notch may

propagate totally in plane stress with net section yielding, totally in plane strain, or there may be a fracture transition as the crack propagates. In some instances, buckling may also occur. In terms of crack initiation, the likelihood of brittle crack initiation at the free surface is not high, unless the material is inherently brittle. Initial cracking is typically by a ductile mechanism (with three types of appearances: a tear zone, micro void coalescence, or a tear zone followed by micro void coalescence), but further cracking may change to cleavage or quasi-cleavage as the crack reaches some (small) critical length, depending on the temperature, loading rate, and grain size. This is common in bcc ferrous materials.

In steels, three types of crack initiation are found, depending, in part, on the temperature.

In the first case, crack-tip blunting followed by ductile crack propagation by tearing initiates at the free surface. In the second, quasi-cleavage fracture and/or ductile micro void coalescence, fracture initiates at the location of maximum constraint. In the less common third case, the material has sufficient toughness and sufficient crack blunting occurs that ductile fracture occurs on a shear plane at the crack tip. If blunting occurs, the peak stress is reduced, and the stress falls off more gradually behind the notch.

Figure 4.28 shows schematically the differences in cracking behaviour in ductile alloys that exhibit a ductile-brittle transition with temperature. The presence of the stretch zone can be used to quantitatively estimate the magnitude of the nominal stress and the fracture toughness [3].

63

Fig. 4.28 Schematic of the brittle-to-ductile fracture transition. The relative area on the fracture surface of the three micro scale fracture mechanisms (stretch zone, dimple zone, and cleavage zone) are indicated.

64

CHAPTER 5
5.1 Fatigue testing Part 1

Fatigue as a specific failure mechanism has been recognised since the early part of the nineteenth century but it was the development of rail travel that resulted in a major increase of interest in this type of fracture. The premature failure of wagon axles led to Wohler in Germany investigating fatigue failure under rotating loading. This led to the design of the first standardised test a reversing stress rotating specimen, illustrated in Fig. 5.1.

Fig.5.1. Wohler rotating fatigue test

There are many mechanisms that can lead to failure but fatigue is perhaps one of the most devious since it can lead to a catastrophic failure with little or no warning one well known example being the failure of the Comet aircraft in the 1950s. Failure can occur at a fluctuating load well below the yield point of the metal and below the allowable static design stress. The number of cycles at which failure occurs may vary from a couple of hundreds to millions. There will be little or no deformation at failure and the fracture has a characteristic surface, as shown in Fig.5.2.

65

Fig.5.2. Typical fatigue crack fracture surface The surface is smooth and shows concentric rings, known as beach marks, which radiate from the origin; these beach marks becoming coarser as the crack propagation rate increases. Viewing the surface on a scanning electron microscope at high magnification shows each cycle of stress causes a single ripple. The component finally fails by a ductile or brittle overload. Fatigue cracks generally start at changes in section or notches where the stress is raised locally and, as a general rule, the sharper the notch the shorter the fatigue life one reason why cracks are so damaging. An unwelded ferritic steel component exhibits an endurance limit a stress below which fatigue cracking will not initiate and failure will therefore not occur. This is not the case with most non-ferrous metals or with welded joints these have no clearly defined endurance limit. The reason for this is that in arc welded joints there is an intrusion a small defect at the toe of the weld, perhaps only some 0.1mm deep. Provided that the applied stress is sufficiently large a crack will begin to propagate within an extremely short period of time. The endurance limit for a welded joint is therefore dependent on the intrusion size

66 that does not result in crack propagation at the applied stress range. In the case of a welded joint, therefore, a fatigue limit a safe life is specified, often the stress to cause failure at 2x106 or 10 7 cycles. During fatigue the stress may alternate about zero, may vary from zero to a maximum or may vary about some value above or below zero. To quantify the effect of these varying stresses fatigue testing is carried out by applying a particular stress range and this is continued until the test piece fails. The number of cycles to failure is recorded and the test then repeated at a variety of different stress ranges.

5.1.1

S/N curve

This enables an S/N curve, a graph of the applied stress range, S, against N, the number of cycles to failure, to be plotted as illustrated in Fig.5.3. This graph shows the results of testing a plain specimen and a welded component. The endurance limit of the plain specimen is shown as the horizontal line if the stress is below this line the test piece will last for an infinite number of cycles. The curve for the welded sample, however, continues to trend down to a point where the stress range is insufficient to cause a crack to propagate from the intrusion. By testing a series of identical specimens it is possible to develop S/N curves. In service however, there will be variations in stress range and frequency. The direction of the load may vary; the environment and the shape of the component will all affect the fatigue life, as explained later in this article. When designing a test to determine service performance it is therefore necessary to simulate as closely as possible these conditions if an accurate life is to be determined.

67

Fig.5.3. S/N curves for welded and unwelded specimens

5.1.2

Palmgren-Miners rule

In order to enable the fatigue life to be calculated when the stress range varies in this random manner, the Palmgren-Miners cumulative damage rule is used. This rule states that, if the life at a given stress is N and the number of cycles that the component has experienced is a smaller number, n, and then the fatigue life that has been used up is n/N. If the number of cycles at the various stress ranges are then added together n 1 /N 1 + n 2 /N 2 + n 3 /N 3 + n 4 /N 4 etc. the fatigue life is used up when the sum is of all these ratios is 1. Although this does not give a precise estimate of fatigue life, Miners rule was generally regarded as being safe [1]. This method, however, has now been superseded with the far more accurate approach detailed in the British Standard BS 7608. The design of a welded joint has a dominant effect on fatigue life. It is therefore necessary to ensure that a structure that will experience fatigue loading in the individual joints has adequate strength. The commonest method for determining fatigue life is to refer to S/N curves that have been produced for the relevant weld designs.

5.2

Fatigue testing Part 2

Hydraulic digitally-controlled testing machines are used to subject test specimens to repetitive loading. Although it is possible to simulate variable amplitude loading, most tests are carried out using constant amplitude loading. The testing is carried out under constant amplitude loading at various stress levels. The machine is usually in the load control mode, and it is recommended that the specimen stresses are monitored by strain gauges in addition to the information provided by the machines control panel. The fatigue test specimens should usually have geometry and a loading mode that are representative for in-service condition. Smaller specimens are tested in standard servo-hydraulic testing machines, whereas full-scale testing of larger structural parts (e.g. tubular joints, large girders) are tested in specially built frames with adaptable hydraulic cylinders. A cruciform fillet-welded joint subjected to axial loading is shown in Figure 5.4. The specimen has a width of 60 mm and a thickness of 25 mm. The specimen is mounted directly between the two grips of an AMSLER hydraulic testing machine with a performance of 250 kN dynamic loading. It is the upper piston that inflicts the loading by vertical displacement. The main plate of the welded detail is in a vertical position and the specimen is subjected to an axial force perpendicular to the welding direction. The electrical cables seen on the photo pertain to unidirectional strain gauges in the loading direction located 10 mm from the weld toe. Figure 5.5 shows a tubular joint with an X configuration in the same type of testing machine. The diameter of the chord is 320 mm, whereas the diameter of the branches is 250 mm. The wall thicknesses of the tubes are 16 and 12 mm respectively. Due to the size of the joint, it is mounted in a frame on the table of the testing machine. The upper piston of the machine introduces a vertical force on the chord, with associated bending in the two branches through vertical movement of the chord. The lower photo also shows all the strain gauges that are attached to the tube surfaces to control the strain distribution in vicinity of the welded intersection between the cord and the branches. This type of joint geometry gives data for establishing the T-class S-N curve. The geometrical stress range is used as the key parameter to determine the fatigue life for this case.

69

Figure 5.4. Cruciform fillet-welded joint in a hydraulic testing machine

Figure 5.5. Tubular joint with an X geometry subjected to in-plane bending of the braches. Upper part: setup in testing rig. Lower part: detail of branch-chord intersection at the crown point with strain gauges

70

5.2.1

Preparations and measurements

Fatigue testing includes all the phases such as test planning, the preparation of the specimens and the statistical analyses of the results. It is important to keep track of all the parameters that have an influence on the test results. This is necessary to define a homogeny specimen population as a basis for the obtained design curve. If a population involves test specimens with large differences in these important parameters, the compiled test results will exhibit large scatter that can be difficult to explain and cope with. The practical results will be that high-quality joints will be penalized because they have been merged with joints of lower quality. For the populations defining categories in current rules and regulations, this is often the case and improvement in this area will lead to more accurate predictions due to reduced scatter. Thus, high-quality joints will get the fatigue life predictions they deserve.

The following background information should be gathered before the test: - Steel type, chemical composition, mechanical characteristics, - Welding procedure, method, electrodes, number and sequence of passes, heat input, - Manufacturing sequence, - Global specimen geometry and local weld toe geometry, - Axial or angular distortion, - Non-destructive testing (NDT), - Estimate of residual stresses in the specimens, - Rolling direction of plates with regard to applied loading, - Microscopy of the heat-affected zone (HAZ).

Most of this information is straightforward to obtain. The chemical composition and mechanical characteristic for a C-Mn steel is given in Table 1. The data are typical for medium-strength steels with nominal yield stress close to 350 Mpa. This is one of the most widespread steel types used for welding. Both the fillet joint in Figure 5.4 and the tubular joint in Figure 5.5 are made of this steel. The welding has been carried out by shielded metal arc welding (SMAW). Regarding the given information for mechanical parameters, hardness measurements should be carried out after the welding has been completed. This gives important information regarding the base metal, weld deposit, and the HAZ. Figure 5.6 and Table 5.2

71 show typical results from hardness measurements carried out on the material in Table 5.1 after welding had been carried out.

Table 5.1. Chemical composition in % and mechanical characteristics for a C-Mn steel

Table 5.2. Hardness measurement

Figure 5.6. Results for hardness measurements for cruciform plated joint (C-Mn steel in Table 5.1)

72 The local toe geometry parameters are measured by applying replica material on the weld toe and inserting cuts of replica cross sections into a profile projector with a typical magnification of 10. The replica cuts must be made transverse to the weld seam direction. Table 5.3 shows the results from two different test series with specimens. Series 1 has low toe angle (mean value 30 degrees) and larger radius (mean value 2.7 mm). Series 2 has a mean angle of 58 degrees and a toe radius of 0.75 mm only. The fatigue testing actually showed that the mean fatigue life of series 1 was twice as long as for series 2. If local toe geometry had not been measured, this difference in fatigue life would have been difficult to understand and explain.

Table 5.3. Statistical results from local toe geometry measurements of a fillet weld

Non-destructive testing should be carried out at the same level as is typical for details entering into service conditions. Specimens with flaws that would have been rejected for in-service purposes should, of course, also be excluded from the test series. The most frequent method is magnetic particle inspection and no crack detection is the only acceptable result. Estimating the actual residual stresses is probably the most challenging part of the preparation work. In general, small specimens will have lesser residual stresses than larger specimens. If the specimens are properly stress relieved, the residual stresses may be set to zero. If not, they can be measured by boring a hole and measuring the change in stresses in vicinity of the hole by train gauges. This will reveal the residual stress gradient. The problem is that the residual stresses vary over the volume of the test specimen. Finally, metallographic study of the HAZ will reveal any abnormal phases or grain size.

73

During the test the following should be measured and registered: loading mode, load range and nominal stress range, the applied R-ratio for the applied stresses, load frequency, stresses by strain gauges monitoring, crack growth measurements by an electrical method, beach marking of the fatigue surface by temporarily decreasing the stress range, number of cycles to final failure.

The loading mode is usually uni-axial loading or bending. One should be aware of the fact that most S-N data are compiled without taking into account the difference between these two modes; it is only the maximum stress range that is assumed to matter. Most specimens are subjected to a positive load ratio to avoid problems with buckling. Typical values of R are in the range of 0.1 to 0.3. The loading frequency is assumed to play a minor role in air environment, and most tests are accelerated so that the test program will not consume too much time. A typical frequency of 6-10 Hz is often applied. However, special care should be taken for the environmental condition (e.g. corrosive seawater.) In such cases there is an interaction between the fatigue damage process related to the oscillating stresses and the time-dependant electro-chemical reaction at the crack front. Therefore, test frequency becomes important. The load frequency should usually not exceed 1 Hz. Detailed knowledge of the strain distributions can be obtained by the use of electrical resistance strain gauges as was shown in Figure 5.5. These gauges operate on the principle that the resistance of a wire changes when its length changes due to stretching or compressing. Hence, when bonded to the surface of the specimen, the local strain at the surface of the material is captured. Some test specimens should be equipped with strain gauges to reveal the actual strain in vicinity of the weld toe during the test. Based on these measurements, the strain concentration can be determined and also the related stress concentration. The stress should account for the gross stress concentration due to the global geometry of the joint, but not the local effect caused by the weld profile. The gauge distance from the weld toe should be large enough to avoid the local effect of the toe itself, typically 10-15 mm. The strain measurements are normally compared to stresses obtained by FEA analysis. FEA obtained stresses were shown in Figure

74 4.7. Local strain distribution for a tubular joint is shown in Figure 5.8. The distribution is obtained with a rack of uni-axial gauges with 0.3 mm length and 2 mm spacing. As can be seen, the strains measured at the chord are higher than the measurements at the branches.

Figure 5.7 Results from a local FEA model for a cruciform fillet-welded joint

Figure 5.8. Strain measurements at brace and chord at welded intersection (Strain gauges length 0.3 mm spacing 2 mm, see bottom of Figure 5.5)

75

5.2.2

Test results

Traditional fatigue life testing produced one result only: the number of cycles to failure. For small test specimens the failure is defined when a final fracture separates the specimen into two parts. For larger joints the failure state is a question of definition. It can be defined as a fracture separating the joint into two pieces or by a through-thickness crack. For the cruciform joint in Figure 5.4, these two time stages will coincide, whereas for the tubular joint in Figure 5.5, the joint will still have integrity when a through-thickness crack has appeared. Typical life data are shown in Figure 5.9 for small-scale testing with base material. The material is a high strength weld able steel for mooring chains. For these small specimens, the failure occurred when the crack was less than 1 mm. As can be seen, the 18 tests are carried out at three stress levels, six tests at each level. The tests were carried out in seawater without any protection against corrosion and the load frequency was 1 Hz. Some results in air are added for comparison. The mean and the design curves are drawn. When this steel is used in a larger structural item, it must be born in mind that the life data in Figure 5.9 corresponds to early cracking of the item and to not final failure. Therefore, if the curves in Figure 5.9 are used directly as life data for a structural item, it will be overly pessimistic.

Figure 5.9. Results from S-N testing with small specimens of high-strength steel in seawater

76 5.3 Crack growth tests guidelines for test setup and specimen monitoring

Crack growth testing is based on the hypothesis that it is the stress intensity factor range (SIFR, K) that governs the crack rate. The SIFR uniquely determines the local severe stress field ahead of the sharp crack front under linear elastic conditions. Crack growth tests are usually carried out on standard test specimens for which the SIF (stress intensity factor) can be determined with great accuracy. One typical test specimen is the compact tension (CT) specimen shown in Figure 5.10.

Figure 5.10. Compact tension specimen for crack growth measurements

The specimen is fabricated with specified dimensions (W and H) that are given once the thickness T of the specimen is chosen. As can be seen, the specimen has a sharp prefabricated notch from which the fatigue crack will grow. Definition of the loading is shown in Figure 5.11. The load F varies with constant amplitude between its maximum and minimum value. The corresponding SIF can be calculated and the SIFR is defined as the difference between them. The evolution of the SIFR with time is shown in Figure 5.12. As can be seen, there is a slight increase in K for constant F. This increase is due to the increase in crack length.

77

Figure 5.11. Definition of loading and SIFR

Figure 5.12. SIFR as a function of time

The crack depth is measured as a function of applied number of cycles, as shown on the left-hand side of Figure 5.13. The measurements can be made optically or by an ACPD technique. At chosen stages, the crack growth rate a/N in m/cycle is calculated and plotted against the actual range of the stress intensity factor K. This range has to be based on the mean crack size during the observation period

78 N. The procedure is shown more detailed in Figure 5.13 where one point in the log a/N-log K diagram is determined to the right on the figure. By following the same procedure, at several stages along the crack path, the crack growth rate parameters can be obtained as shown in Figure 5.14 for a log scale.

Figure 5.13. Transforming measurements crack growth rates as functions of K

Figure 5.14. Crack growth rate as function of K for a log-log scale

79 As can be seen from Figure 5.14, the results fall into three different regions depending on the magnitude of the K. At low K there is a threshold value K0 below which a crack will stop to propagate. Furthermore, there is an almost linear relation between da/dN and K for a log-log scale in an intermediate region. It is in this region that the results obey the Paris law. The slope m and parameter C defined by the intercept with the da/dN axis are obtained. Again, we see from the figure that the linear relation for a log-log scale is an approximation; deviations do occur. Hence, a linear regression analysis has to be carried out in this region in the same way as for the S-N data. At higher values of K, the maximum value of K is close to the critical value of what the material at the crack front can sustain. The consequence will be unstable fracture. As a result it is the SIFR that governs the growth rate, components with various geometries, but made of same material, may have different a-N curves, but when they are transformed into a da/dN-K diagram, as in Figure 5.14, the curves will coincide. This is true provided the tested components have the same loading ratio R and the same environment. The great benefit is then that the curve in Figure 5.14 can be applied to predict the crack growth rate for various types of crack in any component if we are able to calculate the SIFR for the crack and component geometry in question. Hence, the da/dN-K diagram and related equations are of great importance.

Figure 5.15 shows the test setup and the result for crack growth tests carried out in seawater. The CT specimens with 25 mm thickness are taken from high-strength steel in a mooring chain. Steel bars are forged and flash welded to obtain chain links from this type of steel. The results to the right in the figure pertain to a test in air, but other specimens were submerged in the small seawater basin as shown to the left in the figure. The tests were carried out with an applied SIFR in the range between 10 and 30 Mpam0.5. As can be seen, this gives growth rates from 10 -5 to 10-4 mm/cycle. To reveal the threshold value, the testing has to be carried out at smaller SIFR. Typical values for K0 are found between 2 and 8 Mpam0.5, dependent on the environment. The threshold value can be defined when the rate da/dN is less than 10-7 mm/cycle.

80

Figure 5.15. Test setup and growth rate results with high-strength steel

Figure 4.16 shows the final fracture of the CT specimen. This will occur when the either the net ligament ahead of the crack front is too small (global criterion) or if the crack front has too high local stresses (local criterion). In the latter case, the final fracture will be of the brittle type [4].

Figure 5.16. Final fracture of a CT specimen after fatigue crack growth

81

5.4

Welded Components

A welded joint exhibited no clearly established fatigue limit as in an unwelded component. In service, few structures experience purely static loads and that most will be subjected to some fluctuations in applied stresses and may therefore be regarded as being fatigue loaded. Motorway gantries, for example, are buffeted by the slipstream from large lorries and offshore oilrigs by wave action. Process pressure vessels will experience pressure fluctuations and may also be thermally cycled. If these loads are not accounted for in the design, fatigue failure may occur in as few as a couple of tens of cycles or several million and the result may be catastrophic when it does. Fatigue failures can occur in both welded and unwelded components, the failure usually initiating at any changes in cross section a machined groove, a ring machined onto a bar or at a weld. The sharper the notch the greater will be its effect on fatigue life. The effect of a change in section is illustrated in Fig. 5.17, where it can be seen that the stress is locally raised at the weld toe. The illustration shows a bead-onplate run but a full penetration weld will show the same behaviour.

Fig.5.17. Stress concentrating effect of a change in thickness

82 In addition, misalignment and/or distortion of the joint will cause the applied stress to be further increased, perhaps by introducing bending in the component, further reducing the expected fatigue life. A poorly shaped weld cap with a sharp transition between the weld and the parent metal will also have an adverse effect on fatigue performance. In addition to these geometrical features affecting fatigue life there is also the small intrusion at the weld toe, as illustrated in Fig.5.18. In an unwelded component the bulk of the fatigue life is spent in initiating the fatigue crack with a smaller proportion spent in the crack propagating through the structure. In a welded component the bulk of the fatigue life is spent in propagating a crack. The consequences of this difference in behaviour are illustrated in Fig.5.19.

Fig.5.18. Weld toe intrusion

83

Fig.5.19. Effect of stress concentration on fatigue life This shows that this small intrusion reduces the fatigue life of a fillet welded joint by a factor of perhaps 10 compared with that of an unwelded item and some eight times that of a sample with a machined hole. The other consequence is that fatigue cracks in welded joints almost always initiate at the toe of a weld, either face or root. It may be thought that the use of a higher strength material will be of benefit in increasing fatigue life. The rate of crack propagation, however, is determined by Youngs Modulus a measure of the elastic behaviour of the metal and not simply by tensile strength. Alloying or heat treatment to increase the strength of a metal has very little effect on Youngs Modulus and therefore very little effect on crack propagation rates. Since the bulk of a welded components life is spent in propagating a crack, strength has little or no influence on the fatigue life of a welded item. There is thus no benefit to be gained by using high strength alloys if the design is fatigue limited. This is illustrated in Fig.5.20 which shows the benefits of increasing the ultimate tensile strength of steel if the component is unwelded or only machined but how little effect this has on the life of a welded item.

84

Fig.5.20. Effect of increase in tensile strength on fatigue life

One additional feature in welded joints that set them apart from unwelded or machined items is the presence of residual tensile stress. In a welded component there will be stresses introduced into the structure by, for example, assembly stress. These stresses are long range reaction stresses and from a fatigue point of view have little effect on fatigue life. Of far greater significance with respect to fatigue are the short range stresses introduced into the structure by the expansion and contraction of material close to and within the welded joint. Whilst the actual level of residual stress will be affected by such factors as tensile strength, joint type and size and by run size and sequence, the peak residual stress may be regarded as being of yield point magnitude. The implications of this are that it is the stress range that determines fatigue life and not the magnitude of the nominal applied stress. Even if the applied stress range is wholly compressive and there is apparently no fluctuating tensile stress to cause a crack to form and grow, the effect of welding residual stress is to make the structure susceptible to fatigue failure. This is illustrated in Fig.5.21, where it can be seen that, irrespective of the applied stress, the effective stress range is up to the level of residual stress at the welded joint.

85

Fig.5.21. Effect of residual stress on stress range

It would seem reasonable, therefore, that a post-weld stress relief treatment would be of benefit to the fatigue life by reducing the residual stresses to low levels. This is only true, however, where the applied stress range is partly or wholly compressive. If the applied stress range is all tensile, research has shown that as-welded and stress relieved components have almost identical fatigue performances with only a marginal improvement in the stress relieved joints. This is the result of the bulk of the fatigue life of a welded joint being spent in crack propagation where propagation rates are only marginally affected by mean stress. It may be difficult therefore to justify the cost of stress relief if the only criterion is that of improving fatigue life [1].

86

5.5

Fatigue testing Part 3

Fig.5.22. Examples of joint classification from BS 7608

87 A welded joint behaves in a radically different way from an unwelded item, even if this item contains a significant stress raiser. For both welded and bolted steel structures it has been established that the fatigue life is normally governed by the fatigue behaviour of the joints, including both main and secondary joints. The best fatigue behaviour will be obtained by ensuring that the structure is so detailed and constructed that stress concentrations are kept to a minimum and that where possible the elements are able to deform in their intended ways without introducing secondary deformations and stresses due to local restraints. Stresses may also be reduced by increasing the thickness of parent metal which should improve fatigue life although with some types of joint fatigue strength tends to decrease with increasing thickness. The best joint performance is achieved by avoiding joint eccentricity and misalignment and welds near free edges and by other controls over the quality of the joints. Performance is adversely affected by concentrations of stress at holes, openings and re-entrant corners [1].

5.5.1

BS 7608:1993

This British Standard gives recommendations for methods for the fatigue design and assessment of parts of steel structures which are subject to repeated fluctuations of stress. Table 1 to Table 12 corresponds to the following basic types of details:

plain material; lapped or spliced, riveted or bolted joints; fasteners; continuous longitudinal welded attachments; other welded attachments; transverse butt welds in plates; transverse butt welds in sections and tubes; load-carrying fillet and T-butt joints; slotted connections and penetrations through stressed members; details relating to tubular members; seam welds in vessels; branch connections to vessels.

Each classified detail is illustrated and given a type number. Table 1 to Table 12 also give various associated criteria and the diagrams illustrate the geometrical features and

88 potential crack locations which determine the class of each detail. This information is used to assist with initial selection of the appropriate type number. A detail should only be designated a particular classification if it conforms in every respect to the tabulated criteria appropriate to its type number. Class A is generally inappropriate for structural work and the special inspection standards relevant to classes B and C cannot normally be achieved in the vicinity of welds in structural work. In BS 7608 each joint type is assigned a classification letter. For example, a plate butt weld with cap and root ground flush is class C, an undressed plate butt weld class D and a fillet weld class F ( Fig.5.22). In the case of members or elements connected at their ends by fillet welds or partial penetration butt welds and flanges with shear connectors, the crack initiation may occur either in the parent metal or in the weld throat; both possibilities should be checked by taking into account the appropriate classification and stress range. For other details, the classifications given in Table 1 to Table 12 cover crack initiation at any possible location in the detail [5].

89

90

91

92

93

94

95

96

5.6

Potential modes of failure of welds

Some potential modes of failure are given below:-

97

98

99

100

101

102

Fig.5.23. Effects of joint classification on fatigue life

For each classification a fatigue curve has been developed and from these curves the

103 design life can be predicted as shown in fig 5.23. This is obviously an over-simplification of what can be a very complicated task the forces acting on a joint arising from changes in temperature, changes in internal or external pressure, vibration, externally applied fluctuating loads etc. can be complex and difficult to determine. Whilst the joint design has a major effect on design life and is the basis for calculating service performance, the weld quality also has a decisive effect any fatigue analysis assumes that the welds are of an acceptable quality and comply with the inspection acceptance standards. However, in practice it is not always possible to guarantee a perfect weld and cracks, lack of fusion, slag entrapment and other planar defects may be present, reducing the fatigue life, perhaps catastrophically. Other less obvious features will also have an adverse effect. Excessive cap height or a poorly shaped weld bead will raise the stress locally and reduce the design life; misalignment may cause local bending with a similar effect. Good welding practices, adherence to approved procedures and competent and experienced staff will all help in mitigating these problems. In some applications an as-welded joint will not have a sufficient design life and some method of improving the fatigue performance needs to be found. There are a number of options available. The first and perhaps simplest is to move the weld from the area of highest stress range, the next is to thicken up the component or increase the weld size. Note that, using a higher strength alloy will not improve the fatigue life. Local spot heating to induce compressive stresses at the weld toes will also help, although this needs very accurate positioning of the heated area and very careful control of the temperature if an improvement is to be seen and the strength of the metal is not to be affected. For these reasons, spot heating for fatigue improvements has been virtually discontinued. Hammer peening with a round nosed tool or needle gun peening gives very good results although the noise produced may prevent their use. Shot peening can also be used to introduce compressive stresses at weld toes with equally good results. Compressive stresses can be induced in a component by overstressing a pressure test of a pressure vessel is a good example of this where local plastic deformation at stress raisers induces a compressive stress when the load is released. This technique needs to be approached with some care as it may cause permanent deformation and/or any defects

104 to extend in an unstable manner resulting in failure. Although the next techniques described are not as beneficial as hammer peening of the weld toes they have the advantage of being more consistent and easier to control. The techniques rely upon dressing the weld toes to improve the shape and remove the intrusion. The dressing may be carried out using a TIG or plasma-TIG torch which melts the region of the weld toe, providing a smooth blend between the weld face and the parent metal. Alternatively the toe may be dressed by the careful use of a disc grinder but for best results the toe should be machined with a fine rotary burr as shown in Fig.5.24 and 5.25. Great care needs to be exercised to ensure that the operator does not remove too much metal and reduce the component below its minimum design thickness and that the machining marks are parallel to the axis of the main stress. Ideally the dressing should remove no more than 0.5m depth of material, sufficient to give a smooth blend and remove the toe intrusion [1]. The results of these improvement techniques are summarised in Fig.5.26.

Fig.5.24. Grinding tools

Fig.5.25. Burr machining of weld toes

105

Fig.5.26. Improvement in fillet weld fatigue life

5.7

Tubular joints

One group of joints that needs special attention is welded joints between hollow sections such as cylindrical or rectangular pipes. One simple example is the T-joint between two pipes as depictured in Figure 5.27. It consists of a chord member with diameter D and wall thickness TC and a brace member with diameter d and thickness TB. The principal loading is the axial force and the in-plane bending moment acting on the branch. For these kinds of joints the stress concentration factor can be very high near the intersections mainly due to secondary plate bending in the pipe walls near the welds. Details are shown to the right in Figure 6 with the stress distribution through the chord thickness and the maximum surface stress at various distances from the potential crack site at the weld toe. The section shown is defined as the crown point of the joint. There is a similar phenomenon on the branch side. The problem is that the stress concentration can be very different for different loading modes and thickness ratios, although these loading modes may give the same nominal stresses (elementary beam stresses) away from the critical intersection. Hence, for these types of joints, the nominal stresses are not an appropriate key to the fatigue life given by an S-N curve. It is much more logical to use the stress concentration at the weld toe, ignoring the part of the concentration created by the notch effect of the weld toe itself. The main argument for this choice is that whereas the geometrical stresses may differ significantly from one joint to another due to changes in loading mode and gross geometry, the local stress

106 concentration due to the weld notch will remain the same. Hence, the weld notch effect is not explicitly taken account of, but will be inherent in the S-N curve. The so-called geometrical stress concentration or hot-spot stress has also gained popularity for plated joints, as a strategy to reduce scatter when presenting fatigue results [4].

Fig 5.27. Tubular T-joint with in-plane loading. Geometrical stress (fully-drawn line) and local stress (dotted line) at the weld toe on the chord crown point

5.8

Weldments

Many welding processes are available that can produce satisfactory joints in steels. The selection of a particular process is based on many factors that include the thickness and size of the parts to be joined, the position of the weldment, the desired properties and appearance of the finished weldment, the particular application, and the cost of fabrication as well as other factors. No single process can be used to produce satisfactory weldments for all steels, thicknesses, and positions. The most suitable process is the one that produces the desired properties in the final product at the lowest possible cost.

In arc welding, which is the most widely used welding process for structural steels; filler metal is melted and used to fill a weld groove. The arc-welding process, welding procedure, and joint geometry influence weld penetration and admixture of the filler metal with the base metal. Because of this admixture, the chemical composition of the base metal can have significant influence on the microstructural and mechanical

107 properties of the weld metal. This influence is significant; especially for electro slag and electro gas welds, because they are high heat-input single-weld-pass processes. The final properties of the weld metal depend on many factors, especially the composition of the weld metal and the conditions governing its solidification and subsequent cooling. Because the heat flow in the weld metal is highly directional toward the adjacent cooler metal, the weld metal develops distinctly columnar grains. Furthermore, the rapid cooling of the weld metal may not allow sufficient time for diffusion of the chemical constituents, resulting in microstructural heterogeneities. This segregation and the directional solidification of the weld metal may result in weld-metal properties having pronounced directionality. For the arc welding processes, the maximum temperature of the weld metal is above the melting temperature of the base metal joined. This temperature decreases as the distance from the weld increases. Thus, partial melting of the base metal occurs at the weld-metal-base-metal interface, and microstructural changes occur in the base metal in the immediate vicinity of the weld, forming a heat-affected zone. The size of this zone is determined by the rate of heating, the volume and temperature of the weld metal, and the rate of cooling of the weld metal and surrounding base metal. These factors as well as the composition and microstructure of the base metal determine the grain size, the grain-size gradient, the microstructure, and therefore the fracture toughness of the heat affected zone. Because of the high temperatures and the large variations in temperature gradient and cooling rate, adjacent regions in the heat-affected zone can exhibit large differences in microstructure and properties. In general, for carbon and low-alloy steels, the closer the distance to the weld, the coarser the microstructure. Coarse-grain regions adjacent to the weld interface generally exhibit the poorest toughness. Illustrations of intermediate heat-affected-zone microstructure associated with bead-on-plate deposits on hot-rolled plain-carbon steels and on a low-alloy quenched-and-tempered steel are shown in Figures 5.28 and 5.29, respectively. Both welds were made on plates of the same thickness, using identical heat-input conditions; but, the nominal carbon contents were slightly different. They were 0.20% for the carbon steel and 0.16% for the low-alloy quenched-and-tempered steel. The heataffected zone in a single-pass weld forms under the influence of a single thermal cycle. The temperature and temperature distribution from the weld metal into the base metal in a direction perpendicular to the weld is essentially identical at different locations along the weld groove of a simple butt joint for two constant-thickness plates. Consequently, the various microstructural regions in the heat-affected zone can be continuous.

108 However, the weld metal in a multipass weld is built up by the deposition of successive weld beads. The structure and properties of deposited weld beads and existing heataffected zones are usually altered by the heating effects of subsequent weld-bead deposits. The heat from subsequent weld passes may refine the grain size of the deposited weld metal and existing heat-affected zone, may change the columnar structure of the weld metal to an equiaxed structure, and may temper the microstructure of the existing heat-affected zone. In multipass welds, unlike single-pass welds, the heat-affected zone regions that exhibit relatively low toughness occur intermittently adjacent to the weld interface. In either case, these lower-toughness regions are surrounded by heat-affected zone regions of higher toughness[6].

Figure 5.28 Microstructures typical of the weld metal and the heat-affected base metal in a mild-steel weld.

109

Figure 5.29 Microstructures typical of the weld metal and the heat-affected base metal in a quenched-and-tempered low-alloy steel weld.

110

CHAPTER 6
Designing against Fatigue of Structures
8.8 Different types of structural fatigue problems

The question about how to define problems of designing a structure against fatigue is obviously associated with the goals to be achieved. In principle it implies that satisfactory fatigue properties of a structure should be obtained, but it depends on the type of structures which fatigue properties should be explored. For the present discussion three categories are considered: Rotating blades of turboprop engines, wind turbines and compressors are examples in the first category. Many components of various engines are also in this category with a crankshaft as a well-known case. A fatigue failure in such components would be a kind of a disaster. The fatigue limit of the structure is the important fatigue property and highcycle fatigue is an important issue. However, fatigue failures may also be unacceptable in pressure vessels for which the number of pressurization cycles is not very large, e.g. not exceeding 105. If all cycles have practically the same load range, the relevant fatigue property is the crack initiation life under CA loading. A variety of structures can also occur in the second category. Obviously the crack initiation life is again of interest, and it should be large enough for a satisfactory lifetime in service. If a complete failure is unacceptable, a reliable inspection procedure is indispensable. This applies to aircraft structures, and it can also be applicable to several welded structures. As a consequence both the crack initiation life and the crack growth life are of interest. Moreover, fatigue under VA amplitude loading may also be a relevant condition. The third category includes various utilities for which final failure simply implies that it must be replaced by a new one. Various housekeeping articles are in this category, e.g. washing-machines, vacuum cleaners, but not stairs. Bicycles are another typical

Structures for which fatigue failures are unacceptable. Structures in which fatigue cracks may occur after a sufficient lifetime but without the risk of a complete failure. Structures for which crack initiation and crack growth until a complete failure are acceptable, but for which a reasonable lifetime is still desirable.

111 example in which fatigue failures do occur. The fatigue property is the fatigue life until failure with lifetime as an economical criterion. Data on crack initiation life and crack growth properties are not required, but again both CA and VA load histories can be significant. The three categories of structure have been defined because within each category similar fatigue properties should be predicted. The literature on fatigue prediction problems is quite diverse. The world of building steel bridges and the world of manufacturing wind turbines are two different cultures, but still with similar fatigue problems. As an example, in both worlds load spectra are consisting of a combination of deterministic loads and random loads. In practice the designer who is faced with fatigue endurance problems, must also consider other durability issues, such as: maintenance, inspections, repairs,

replacements, service conditions with implications for corrosion, wear and tear. They are all a matter of concern dealing with the structure as an object that should be in function for a long time. Anyway, the possibility of fatigue crack initiation is a relevant problem because it can have a large economic impact. Designing against fatigue crack initiation is one of the responsibilities of the designer of the structure. Following information is required:

information about the structure, analysis and fatigue data, and last but not least, the load spectrum.

In the literature it is sometimes suggested that our fatigue problems are solved if an accurate prediction model would be available. This is misleading. The present physical understanding about fatigue damage accumulation is reasonably well developed in a qualitative sense. And just because of this understanding, it must be accepted that accurate quantitative predictions on fatigue lives are illusory. Problems of fatigue life and crack growth prediction for notched elements indicate how estimates of the fatigue limit of notched elements could be obtained. Unfortunately, similar prediction procedures are not applicable to fatigue of joints. Empirical data of joints must be available to arrive at estimated of the fatigue limit. Predictions on the fatigue life under VA loading are even more complicated. The Miner rule is physically rather primitive. The rule starts from the idea that damage can be characterized by a single damage parameter which essentially disagrees with the present knowledge about fatigue damage accumulation. At best, the Miner rule gives some weighted indication of

112 the load spectrum, but not of the severity of the load spectrum. The Miner rule fully breaks down in comparisons of load spectra severities. When the Miner rule is used to obtain some rough indication of the fatigue life under VA loading, one should realize that the prediction is an extrapolation of S-N data which by itself have already a limited reliability. The situation appears to be more convenient for predictions of fatigue crack growth. Crack growth prediction for CA loading based on the well-known fracture mechanics methodology can be reasonably reliable. But the situation is less satisfactory for fatigue crack growth under VA loading. A major problem is to account for interaction effects of cycles with different amplitudes. If the interaction effects are ignored, predictions will probably be conservative, but it can lead to significant under-predictions for crack growth under steep load spectra. In general terms, it must be accepted that fatigue predictions are speculative in a way that the order of magnitude may be instructive, but the predictions should be evaluated with appreciable judgement. In cases of doubt, the design variables should be reconsidered to see where weak links are present. Estimates of fatigue properties can be improved by experiments. Whether this is really necessary depends on safety margins and costs involved. Detailed stress analysis, fatigue experiments and load spectrum measurements can improve the significance of predictions. It is possible that a simple fatigue analysis shows that the occurrence of a fatigue failure problem is very unlikely, and no further design improvements are necessary. It is also possible that fatigue failures in service are acceptable because a simple replacement of the failed element is not expensive and safety is not involved. In such cases, a cost-benefit analysis can show that efforts to improve the fatigue prediction are not really worthwhile. But it is also possible that a simple prediction indicates that structural improvements must be considered, i.e. designing against fatigue. It then is useful to have some idea about the accuracy of preliminary fatigue life predictions. Several sources of uncertainties in the prediction technology should be considered, including the strategy of applying safety factors.

8.8

Designing against fatigue

A designer should know whether he is designing against crack initiation, or for an acceptable crack growth behaviour, or for both. Moreover, he also should be aware of the question whether his problem is associated with high-cycle fatigue or low-cycle fatigue.

113 The initiation period is basically a material surface phenomenon, whereas crack growth is a matter of crack growth resistance of the material as a bulk property. As a consequence, fatigue related influences are essentially different for the two periods. The crack initiation period and the fatigue limit are heavily depending on material surface conditions, whereas most of these conditions are practically irrelevant for the crack growth period. Understanding of the effects of these variables is essential for designing against fatigue.

Fig. 6.1 Survey of topics associated with designing against fatigue.

8.8

The crack initiation aspect

It is easily understood from Figure 5.1 that designing against fatigue crack initiation is concerned with the general layout of a structure, detail design, material selection and surface treatments. The layout of a structure depends on the purpose of the structure. But there are various possibilities to obtain an improved load distribution in a structure,

114 e.g. by changing local dimensions such as a locally increased thickness to reduce the stress level around a fatigue critical detail. Another example is associated with eccentricities which are causing unfavourable secondary bending.

8.8

Material selection

The selection of the material depends on many circumstances, such as static properties, workshop properties, corrosion resistance, thermal properties, costs, etc. A material with a higher S0.2 may have a higher Sf for unnotched specimens, but also an increased notch sensitivity. Similarly, welded joints of a higher strength material usually do not necessarily have an improved fatigue strength. If a new material is considered for a structural application, it should be supported by results of service-simulation fatigue tests on specimens which are representative for fatigue critical details of the structure under consideration.

8.8

Surface treatments

The designer can specify the quality of the material surface, and also certain surface treatments. Some typical examples of surface treatments are: fine machining, nitriding of steel, shot peening, surface rolling, prevention of fretting and corrosion protection. It has been pointed out that surface treatments are carried out for various purposes: improvement of fatigue properties, protection against corrosion, improved wear resistance, restoring poor surface quality, and cosmetic reasons. Surface treatments can increase the hardness of a surface layer, and thus hamper cyclic micro plasticity. At the same time, residual compressive stresses restrain the opening of micro cracks in the surface layer and thus will reduce or even arrest the growth of these cracks. As a result, the major benefit of surface treatments is on the crack initiation period. They are important for high-cycle fatigue, and in particular for the fatigue limit.

8.8

Detail design for an improved stress distribution

An essentially different approach is associated with the reduction of stress concentrations. For fatigue critical notches it generally boils down to increasing root radii or applying stress relieving grooves if that is possible. Non-circular fillets are rarely considered. Recently Kt values were calculated for elliptical fillets and results have shown significant reductions of the Kt value. It may be repeated here that various Kt values are not always very accurate as a result of older techniques used to determine

115 the various graphs in the book. By now FE analysis can produce more accurate Kt values as well as stress gradients.

8.8

Large-scale design issues

Detail design is associated with dimensions which are significant for local stress concentration, e.g. a hole diameter or notch root radii. On a larger scale, the designer is considering the general concept of the structure. As an example, although perhaps a somewhat curious one, rather different concepts can be contemplated for designing a bridge. Another noteworthy example, in various structures joints are present, but the variety of different joint concepts is also large. Decisions to be made on the type of structure are generally depending on experience of the industry, and in the industry on economic implications.

8.8

Uncertainties, scatter and safety margins

The purpose of designing against fatigue is to prevent disasters, and also to avoid nonfatal incidents in view of unwanted economic consequences. Unfortunately,

uncertainties about the fatigue performance of a structure cannot be solved by accurate and rational arguments. It implies that some philosophy about safety factors or other measures should be considered. A solid rational frame work to arrive at safety factors cannot be formulated. Statistical distribution function are unknown. Information about scatter of fatigue properties is largely coming from laboratory test series and not from service experience. The choice of reasonable safety factors is a matter of experience and engineering judgement. Both economic and safety consequences of the occurrence of a premature fatigue failure must be considered. In view of limited accuracies of quantitative fatigue predictions, it must be asked how this situation should be carried on. The variety of sources for uncertainties is fairly large.

8.8.1 Uncertainties Three reasons for uncertainties about the prediction of the fatigue performance of a structure are:

Uncertainties about the load spectrum. Uncertainties about the fatigue properties of the structure. Uncertainties about the reliability of predictions.

116 Not all these uncertainties can be associated with scatter of some properties. Variations of the load spectrum include differences between deterministic loads (applied by the operator) and stochastic loads (random type loads depending on environmental conditions). With respect to the deterministic loads, all structures of the same type are not used in exactly the same way. The designer must consider the variability of loads which should be taken into account (functional loads, manoeuvres). But another part of the variability of the load spectrum does not depend so much on the operator. Stochastic loads are relevant to structures operating under a variety of weather conditions (aircraft, boats, drilling platforms) or moving over various roads (passenger cars, trucks, coaches). Statistical distribution functions and power density spectra can be involved for air turbulence, sea-waves and road roughness. Several types of structures will see a combination of deterministic and stochastic loads. Cranes, bridges and buildings offer interesting combinations of both types of loads. The second reason for uncertainties is associated with fatigue properties of the structure. These uncertainties are of an entirely different nature. Statistical variations are related with material properties and production quality. The fatigue properties of a material with a standardized composition may be obtained from data banks, but it cannot be guaranteed that these properties are always the same. Scatter may occur as a result of batch to batch differences, but even in a single plate statistical variations are possible. Moreover, the crack initiation fatigue life is depending on the quality of the production of components. There are sufficient reasons why components produced during a number of years cannot be considered to be samples of the same statistical population. Finally, the third source of uncertainties is associated with the reliability and accuracy of a prediction model. Estimates can be obtained for S-N curves, fatigue limits and crack growth.

8.8

Scatter and safety factors

8.8.1 The fatigue limit and the safety factor An important category of problems of designing against fatigue is associated with highcycle fatigue and a flat load spectrum. If fatigue failures are unacceptable, the criterion is that all load cycles should be below the fatigue limit of the structure. The variables involved are the maximum load cycles occurring in service load spectra and the fatigue limit of the structure. They are both affected by uncertainties. It can be tried to obtain an estimated value of the fatigue limit of the structure, but even if the analysis would be

117 supported by experiments, some unknown scatter must be expected. With respect to the load spectrum uncertainties are involved, not so much as a consequence of scatter, but due to different utilizations. Under these twofold conditions, a safety factor cannot be defined with rational arguments. If a fatigue failure of the structure would cause a fatal accident, relevant experimental efforts should be considered.

Fig. 6.2 Safety margin on load level S1 for required life time N1. A similar margin for the fatigue limit is unrealistic.

A full-scale fatigue test on a representative part of the structure with the step by step increasing load (see Figure 6.3) can give useful information about the fatigue limit.

Fig. 6.3

Load history in a step test to obtain an approximate fatigue limit with

a single specimen. A small sa and a large N-value should be adopted.

118

A full-scale CA load tests at the estimated load level of the load spectrum cannot be recommended. The test should be continued to a very high number of cycles, say > 108. However, if the load level is just below the unknown fatigue limit, see Figure 6.2, then failure will not occur. In view of the scatter band of the fatigue limit, an other similar structure can fail at a fatigue life between 106 and 107 cycles, just above the average fatigue limit. It implies that information about the safety level remains unknown. In the high-cycle fatigue regime and for the fatigue limit, scatter of fatigue lives is not the relevant issue. Scatter of the fatigue strength, and in this case of the fatigue limit, is crucial. For this reason the step by step increasing test of Figure 6.3 should be preferred. Of course the number of cycles in each step (N) should be large enough in order to be in the high-cycle fatigue regime, for instance N = 106 or 2 106 cycles. The fatigue limit Sf obtained with the step-by-step method and also the load spectrum in service are not free from uncertainties. A safety factor should be adopted. Since quantitative indications on scatter are lacking, an intelligent guess must be made. Possible consequences of fatigue failures in service have to be considered. It is believed that a safety factor of 1.5 can be sufficient in many cases. However, if more confidence is desirable, more fatigue tests should be carried out. Another approach is to carry out load history measurements in service to have more information about the load spectrum.

8.8

Safety factors for finite fatigue life problems under CA loading

Crack initiation cannot be avoided if stress amplitudes above the fatigue limit occur in the service load spectrum. As a consequence, fatigue crack initiation is possible and a finite life should be considered. A typical example is represented by a pressure vessel. A safe approximation of the load spectrum is that the pressure vessel is always loaded to the same maximum operational pressure. Load spectra of other structures with a flat load spectrum can be approximated in the same way. A safety factor can now be defined in two different ways. The factor can be applied to the fatigue life or to the fatigue strength. If a finite life is envisaged, the natural approach is to think in terms of endurances which guarantee a sufficient lifetime. If N1 is the required lifetime and N2 the estimated fatigue life, see Figure 6.2, then the safety factor is f N = N2/N1. However, in terms of the fatigue strength, if S1 is the required fatigue strength and S2 is the estimated fatigue strength, then the safety factor f
S

= S2/S1. Adopting the Basquin

119 relation (S
k

N = constant), the relation between the two safety factors is f

= (f S) k. If

loads exceeding S1 should not be expected or even be impossible, and then the safety factor for the fatigue life should be considered. However, if required lifetimes larger that N1 are of little interest then the safety factor for the stress level is more appropriate. The size of these safety factors to be adopted depends on the consequences of a fatigue failure. Obviously larger factors are necessary if fatal accidents are possible; say 1.5 on the stress level or 6.0 on lifetimes. In such a case, a realistic experimental verification test must be advised. If the consequences of a final failure are not serious, a smaller safety factor can be adopted, say 1.2 on the stress level, or 2.5 on the fatigue life. If the quality of the stress raisers is poor (e.g. in low-quality welds), larger values may be worthwhile. Engineering judgement and experience from previous structures should be practiced.

8.8

Safety factors for finite fatigue life problems under VA loading

The VA load case offers an additional uncertainty if compared to the CA load case. Predictions for a VA load history are affected by the unreliability of the Miner rule. It is difficult to understand how this might be accounted for by a safety factor. When using the Miner rule, it appears to be wise to extrapolate S-N data below the fatigue limit. In cases of doubt, some exploratory service-simulation fatigue tests are much recommended.

8.8

Safety factors and fatigue crack growth

Safety factors on fatigue crack growth have to be considered if the crack growth period covers an essential part of the lifetime in service. This can occur when cracks are initiated at material defects, corrosion pits, or sharp corners with a high stress concentration. It can also start from unintentional surface damage caused in-service (nicks, dents, scratches, impact damage, etc.). In welded structures, crack initiation is possible from weld defects, but also at the edge of the weld toe due to a locally unfavourable profile. All these situations are undesirable, but they cannot always be avoided. In view of safety, it may be necessary to consider fatigue lives with a practically zero crack initiation period. It is kind of a worst case analysis which should be made if complete failure is unacceptable. Two different cases can be defined: Crack growth is accepted, but the occurrence of a complete failure must be prevented by periodic inspections.

120 The crack growth period until failure should be larger than the design lifetime of the structure because inspections for cracks in service are undesirable or not feasible.

The first case is well-known for aircraft structures for which so-called damage tolerance requirements are laid down in official airworthiness regulations. It can also be applicable to nuclear pressure vessels or other structures if fatigue failures are inadmissible and periodic inspections must be done to detect fatigue cracks before failure occurs. The problem setting is illustrated in Figure 6.4 by a schematic crack growth curve and a corresponding curve of the decreasing static strength of the structure caused by the growing fatigue crack. Failure of the structure is supposed to occur at a critical crack length, ac. Cracks can be detected at the crack length denoted

Fig. 6.4 Principle of safe crack growth by period inspections.

As ad . The period for crack detection covers crack growth from ad to ac, see Figure 6.4. The number of uncertainties is fairly large:

the initial crack length a0, the final crack length ac, the crack growth data of the material, the load spectrum,

121 the crack growth prediction model, the probability of detecting a fatigue crack.

The probability of crack detection depends on the non-destructive techniques adopted. Questions can be raised whether a surface crack with a length of a few illimetres can

be detected. In general, very small cracks, say 1 mm (0.04 inches) cannot be detected reliably. Crack detection of invisible cracks, e.g. in joints, must be done with special inspection techniques. Secondly, it must also be established how far the crack may grow before the risk of a large failure is present. The crack must be found within the crack growth range between the detectable crack size (ad) and the critical crack size (ac). A safety factor should then be applied to this period to assess the inspection period. In the past, a factor 3 has been used for transport aircraft, but more recently, the tendency is to use a factor 2. Obviously, the choice of the safety factor is a matter of judgement, which requires that all sources of uncertainties are recognized and understood. It should also include the human factor of the inspection procedure. If a large number of structures must be inspected, most of which will be free from cracks, an occasionally occurring small crack might escape detection. Situations of finding cracks in order to prevent dangerous situations are not confined to aircraft. It also applies to other types of structures if a fatigue failure cannot be accepted, e.g. for pressure vessels. Operators of large structures try to combine inspections with periodic maintenance for economic reasons. Actually, operators prefer structures which do not require inspections. The size of the initial crack length (a0) must be associated with the size of some initial defect. This is a difficult issue because the crack growth rate of initially small cracks is very low. As a consequence, the predicted crack growth life will significantly increase for a smaller value of a0 (see Table 6.1).

Table 6.1

Illustrative crack growth life predictions for a carbon steel.

122 It is more conservative to select a larger a0-value, but which size? The final crack length, ac, is reached at the moment of failure. It requires that the reduction of the residual strength of the structure is calculated as a function of the increasing crack length, which is not a simple calculation because macro plasticity will occur. However, the crack growth rate in the last part of the crack growth period is relatively high, and assuming a lower ac will have a small effect on the crack growth period, see again Table 6.1. The crack growth prediction model is less problematic for a CA load spectrum than for a VA load spectrum. In case of CA loading, predictions may give reasonably reliable results provided that K solutions are available. Quite often, K solutions are not available, even for structural elements with a simple geometry. Small cracks are usually part through cracks at the material surface. If K-values are not available, they can be calculated with FE techniques, but it requires expertise on this topic. Predictions on crack growth during VA loading offer problems due to interaction. Ignoring these effects should be expected to give a conservative prediction for most load spectra. The basic CA crack growth data used in the prediction are also subjected to uncertainties. Variations can occur between nominally similar materials from different producers. Even differences between batches from the same producer have been found, see Figure 6.5.

Fig. 6.5 Comparison between crack growth lives of sheet specimen of different producers and different batches

123 Small cycles with K < Kth can still contribute to crack growth. It was proposed to extrapolate the da/dNK function in the Paris regime to low K < Kth.

8.8

Safety aspects associated with a corrosive environment and low frequency fatigue

The effect of corrosion on fatigue depends on the material/environment system. Unfortunately, most types of steel and aluminium alloys are sensitive to corrosion. It can imply that these materials are also sensitive to the frequency and wave shape of the load cycles. Unfortunately, the effect of corrosion fatigue cannot simply be described by a quantitative model. Experience should indicate how to deal with safety issues introduced by a corrosive environment. Corrosion can affect both crack initiation and crack growth. Obviously, the application of safety factors does not preclude the occurrence of corrosion. Pitting and other local corrosion phenomena can occur in a corrosive environment, and subsequent crack growth will be activated. It might be hoped that cracks should not grow at low stress amplitudes, but it would require a high safety factor (see Figure 5.6 for mild steel).

Fig. 6.6 The effect of environment and load frequency on the S-N curve of unnotched mild steel specimens

124 The best solution is to prevent corrosion at the material surface. Sometimes this is done by preventing the access of the aggressive environment to a fatigue critical element of a structure. Corrosion resistant surface layers can be considered also, but experience should indicate whether this will be successful. Another solution is shot peening of the material surface. This would not prevent corrosion at the material surface, but the residual compressive stresses may prevent crack opening and further crack growth. An example of this application is shot peening of springs used in cars. If water is trapped in the structure, the consequences of a stagnant water environment may be disastrous. Trapping of water should be avoided, either by design or sealing of critical locations. Corrosion fatigue can be problematic for structures used in the open air or in the sea, e.g. for bridges, cranes, ships, offshore structures, but also for many other structures. In the open air, rain and fog are causing a moist environment of usually polluted water, which is an aggressive environment. After fatigue cracks have been initiated, the corrosive environment can enhance crack growth. On welded joints, accelerated crack growth has been observed in comparative tests carried out in air and salt water. In salt water, crack growth could be about three times faster. A safety factor of three applied on the crack growth life may be reasonable. If fatigue failures in the environment of the structure would have serious consequences, it might be necessary to support the fatigue analysis by relevant experimental work. The problem is how a service-simulation fatigue test should then be carried out in view of corrosion being a time dependent phenomenon. The frequency of the cyclic loads in service may be low and an exact simulation can imply an unacceptably long duration of the test. A compromise should be considered. Certain parts of the load-time history can be simulated faster than the history in service, while the more damaging load cycles can be applied with the loading rates relevant for the service load-history. It then should be recognized that the increasing load part of a cycle is the most important part for fatigue crack increments. Another interesting alternative to service-simulation tests is to build a few prototypes of the structure and to test these prototypes in a realistic but severe application in service. This has been done for cars and trucks, which were tested by severe driving along selected tracks with rough road conditions. Actually, such tests are not done for fatigue only. It should show a satisfactory functioning of all parts of a structure under severe conditions. However, it also can reveal insufficient fatigue properties[1].

125

CHAPTER 6

Methods of revealing fatigue cracks

126
7.1 Dye-penetrant testing

Dye-penetrant

testing;

method

of

examining components to detect surfacebreaking flaws, such as cracks. The technique is based on the ability of a liquid to be drawn into a "clean" surface-breaking flaw by

capillary action. It is essential that the component to be inspected is thoroughly cleaned to remove all traces of dirt and grease. It is then sprayed with a penetrating liquid, usually a brightly coloured liquid or a

fluorescent dye, which penetrates any surfacebreaking cracks or cavities. The liquid is allowed to soak into the components surface. (Fig1 right) After soaking, the excess liquid penetrant is wiped from the surface and a developer applied. The developer is usually a dry white powder, which draws the penetrant out of any cracks by reverse capillary action to produce indications on the surface.(Fig2 right) These (coloured) indications are broader than the actual flaw and are therefore more easily visible. Fluorescent penetrants are normally used with a UV lamp to enhance sensitivity. These systems are often used to check weld quality during fabrication.

7.1.1 An example of dye penetrant testing used on bicycle components.

This cycle crank arm was returned to the supplier after a very short time in use. The owner had seen a crack coming from the square taper axle attachment and suspected a smaller crack close to the pedal thread

The suspect areas were sprayed with red penetrant dye and left to soak. The square hole location was clearly cracked but the minor region may be a surface scratch. Such distinctions are very important in the performance of engineering components.

After

the

dye was

was

cleaned with

off

the chalk

component

sprayed

developer. The crack running from the square axle drive hole gave a very distinct red indication at its precise location,

indicating it was clearly cracked. The other feature showed no red line on development indicating a surface scratch not an

embyronic crack.

The crank spider arm of this chainset fractured and unseated the cyclist in heavy city traffic. It was old but well looked after and cleaned regularly The growing fatigue crack was undetected until the dangerous failure. The cyclist was concerned if a similar crack had been nucleated in the matching plain crank arm shown alongside. The equivalent area was sprayed with red penetrating dye which was then left to soak into any cracks or fissures in the component. After several minutes of soaking the dye was cleaned off the components surface. The pre-soaked and cleaned area was then sprayed with developer spray which is basically chalk powder in a volatile carrier. Any defects present show up as the red dye is pulled out of any cracks or fissures in the component. It was concluded that none were present. The only red marks were from dye that had been retained in the stamped product identification marks.

129

7.2 PHOTOELASTICITY
A method of examining transparent polymer models of structures etc. to isolate stress concentrations and other weak zones. The model is placed between crossed circular polarizing filters (eg Polaroid sheets) and a force applied. The technique also enables residual stress to be shown in transparent articles. Stress fields (applied and residual) can be exposed using models of structures in photosensitive material placed between polarising filters in the crossed polar position. Here the stresses in a 7 member model bridge truss, centrally loaded and simply supported are shown. These injection moulded safety spectacles contain residual moulding stresses shown here using photo elastic viewing techniques [2].

130

CHAPTER 8
8.1 CAUSES AND RECOGNITION OF FATIGUE FAILURES

8.1.1 General Causes of Material Failures: Design deficiencies Manufacturing deficiencies Improper and insufficient maintenance Operational overstressing Environmental factors (i.e. heat, corrosion, etc.) Secondary stresses not considered in the normal operating conditions Fatigue failures

Improper and insufficient maintenance seems to be one of the most contributing factors influenced by some improper designs such as areas that are hard to inspect and maintain and the need for better maintenance procedures. In many circumstances the true load is difficult to predict resulting in a structure being stressed beyond its normal capabilities and structural limitations. When a structure is subject to cyclic loads, areas subject to fatigue failure must be accurately identified. This is often very hard to analyse, especially in a highly composite structure for which analysis has a high degree of uncertainty. Thus, in general, experimental structural fatigue testing is frequently resorted to. 8.1.2 Recognition of Fatigue Failure Two fatigue zones are evident when investigating a fracture surface due to fatigue, the fatigue zone and the rupture zone. The fatigue zone is the area of the crack propagation. The area of final failure is called the rupture or instantaneous zone. In investigation of a failed specimen, the rupture zone yields the ductility of the material, the type of loading, and the direction of loading. The relative size of the rupture zone compared with the fatigue zone relates the degree of overstress applied to the structure. The amount of overstressing can be determined from the fatigue zone as follows: highly overstressed if the area of the fatigue zone is very small compared with

131 the area of the rupture zone; medium overstress if the size or area of both zones are nearly equal; low overstress if the area of rupture zone is very small. Figure 8.1

Figure 8.1

Figure 8.2 Typical fatigue zone with identifying marks. Describes these relations between the fatigue and rupture zones. The fatigue zone can be described as follows: a smooth rubbed, and velvety appearance, the presence of waves known as clam-shells or oyster-shells, stop marks and beach marks, and the herringbone pattern or granular trace which shows the origin of the crack.

132 In general, stop marks indicate the variations in the rate of crack propagation due to variations in stress amplitude in a cyclic application varying with time. Figure 7.2 is a schematic representation of the fatigue zone.

8.2

Design Considerations

Even if careful attention to good design practices is constantly the goal of design engineers, fatigue problems are sometimes introduced into the structure. Fatigue failures are often the result of geometrical or strain discontinuities, poor workmanship or improper manufacture techniques, material defects, and the introduction of residual stresses that may add to existing service stresses. Typical factors affecting fatigue include the following: Stress raisers, usually in the form of a notch or inclusion; most fatigue fractures may be attributed to notch effects, inclusion fatigue specimens are rare. High strength materials are much more notchsensitive than softer alloys. Corrosion is another factor that affects fatigue. Corroded parts form pits that act like notches. Corrosion also reduces the amount of material which effectively reduces the strength and increases the actual stress. Decarburization, the loss of carbon from the surface of the material, is the next factor. Due to bending and torsion, stresses are highest at the surface; decarburization weakens the surface by making it softer. Finally, residual stresses which add to the design stress; the combined effect may easily exceed the limit stress as imposed in the initial design. 8.2.1 Influence of Processing and Metallurgical Factors on Fatigue A myriad of factors affect the behaviour of a material under fatigue loading. Obvious factors include the sign, magnitude, and frequency of loading, the geometry and material strength level of the structure and the ambient service temperature. However, processing and metallurgical factors are not often considered, but these factors determine the homogeneity of materials, the sign and distribution of residual stresses, and the surface finish. Thus, processing and metallurgical factors have an overriding influence on the performance of a structure. 8.2.1.1 Processing Factors

Stresses are normally highest at the surface of a structure, so it follows that fatigue usually initiates at the surface. Stress raisers are more likely to be present as a result of

133 surface irregularities introduced by the design of the structure or produced in service or resulting from processing. Processing factors can introduce a detrimental or beneficial effect into a structure, usually in the form of effect on strength level or residual stress condition of the surface material. Therefore, the effect of processing on the mechanical properties of a material, especially the surface of the material, directly affects fatigue properties. Processing factors that influence the fatigue life of a structure include the following: the process by which a part is formed, such as die casting; the heat treatment of a material, such as quenching, which builds up residual stresses and annealing, which relieves internal stress (see Figure 7.3); case hardening, such as carburization or nitriding, which increases surface hardness and strength (see Figure 7.4); surface finish, such as polished smooth by electro polishing; cold working, which increases strength; also, cladding, plating, chemical conversion coatings, and anodizing.

Figure 8.3

Effect of hardness on the fatigue life of threads rolled before and

after heat treatment.

134

Figure 8.4 Bending fatigue test results on sections from crankshafts: endurance limit versus surface treatment.

8.2.1.1

Metallurgical Factors

Metallurgical factors refers to areas within the material, wither on the surface or in the core, which adversely affect fatigue properties. These areas may arise from melting practices or primary or secondary working of the material or may be characteristic of a particular alloy system. In virtually all instances the detriment to fatigue properties results from a local stress-raising effect. Therefore, metallurgical factors affecting fatigue include the following:

surface defects sub-surface and core defects inhomogeneity, anisotropy improper heat treatment localized overheating corrosion fatigue fretting corrosion.

135 8.3 Experimental Analysis of Fatigue Life Curves

Failure due to repeated loading is known as fatigue. A small crack, a scratch, or some other such minor defect causes localized deformation. This deformation leads to a small crack if one was not initially present. After cyclic loading, that is, loading in the same way multiple times, the crack grows, and eventually the material fails. A fatigue life curve is a graphical representation of the cyclic loading. Simply, a fatigue life curve, also known as an S-N curve is a plot of the stress amplitude versus the number of cycles the material goes through before it fails. That is, for a certain stress, the material will fail within a certain number of cycles. Figure 8.5 is an example of a typical fatigue life curve.

Figure 8.5

Typical Fatigue Life Curve.

8.6

Fatigue Crack Growth

If an engineering component contains a crack, and if a cyclic or repeated load is applied, the crack is likely to grow slowly with increasing number of load cycles. This process is known as fatigue crack growth. In a fatigue crack growth experiment, the progress of a crack growing under a cyclic load is measured, and the results are plotted as a fatigue crack growth rate curve, da/dN versus K (that is, change in crack length divided by change in number of cycles to failure versus change in fracture toughness). A typical fatigue crack growth curve is shown in Figure 8.6.

136

Figure 8.6 points.

Crack growth rates obtained from adjacent pairs of a vs. N data

In the simplest form of a fatigue crack growth rate test, a cyclic load is applied that has fixed maximum and minimum loading levels. The test specimen is usually a plate of material in which a crack has already been started at the end of a V-bottom machined slot. In a typical fatigue crack growth experiment, the sample is loaded in a closed-loop servo hydraulic testing machine and data for crack length, number of cycles to failure, and fracture toughness is recorded. From this data the mechanical behaviour for a certain material can be described under fatigue crack growth loading by the fatigue crack growth rate curve. This sort of experiment is useful for materials that would undergo high cyclic loading stresses such as an airplane wing or a helicopter rotor [2].

8.7

Real Life-Design and Manufacturing Considerations

The following describes a relationship between factors that shape the S-N curves as they are influenced by design and manufacturing conditions and the effects of such conditions on the fatigue properties of materials, components, and structures.

8.8

Recommendations for Designs to Avoid Fatigue Failures

A designer can help to minimize the possibility of fatigue failure by proper design of structural components. Many fatigue failures may be attributed to lack of sufficient consideration of design details or a lack of appreciation of engineering principles. These principles, which are an integral part of good design of structures subject to

137 fatigue, are well reported in literature, but this information has been scattered throughout sources and may be inaccessible to a designer who needs to understand and utilize the principles. It is good design practice to seek out sources of this information and to utilize the principles before, during and after the design process.

138 Annexure 1

Additional Scanning Electron Microscope Images [2]

Scanning Electron Microscopy

Scanning Electron Microscope (SEM) image of intergranular fracture indicative of hydrogen embrittlement. (Mag: 400X)

Scanning Electron Microscope (SEM) image of fatigue striations indicative of cyclic crack propagation. (Mag: 700X)

Scanning Electron Microscope (SEM) image of beach marks indicative of a progressive fatigue failure. The area of fatigue initiation is noted at the arrow. (Mag: 180X)

139

Scanning Electron Microscope (SEM) image of the surface of a casting void in the fracture surface of an aluminium casting. (Mag: 1,000X)

Scanning Electron Microscope (SEM) image of a brittle fracture surface in an aluminium casting. The angular particles in the surface are silicon particles that contribute to the brittleness of the material. (Mag: 1,000X)

Scanning Electron Microscope (SEM) image of microbiological activity in a fire protection system piping. (Mag: 700X)

140 Annexure 2

Metallography/Microstructure Evaluation
The properties of a material and its performance in a specific application depend on its microstructure. Our metallographs (light microscopes) are capable of examinations at magnifications from 15X to 1,000X. Analyses of microstructure and material defects in cross-sectioned samples determine material properties, flaw characteristics, and defect mechanisms. Metallurgical Technologies, Inc. (Mti) has full metallographic preparation capabilities from sectioning and mounting the specimen through the grinding and polishing stages to proper selection and etching techniques of the tested material.

View of intergranular stress corrosion cracking (IGSCC) in an Inconel heat exchanger tube. Note that the crack follows the grain boundaries. (Mag: 500X)

View of chloride stress corrosion cracking in a 316 stainless steel chemical processing piping system. Chloride stress corrosion cracking in austenitic stainless steel is characterized by the multi-branched lightning bolt transgranular crack pattern. (Mag: 300X)

Microstructure evaluation of the heat-affected zone of a welded stainless steel piping flange etched to reveal the carbide distribution. Fine carbide particles outline the grain boundaries, indicating a "sensitized" condition resulting in susceptibility to intergranular corrosion. (Mag: 600X)

A cross-section through a seam weld in a 400 series ferritic stainless steel tube. The seam exhibits a wide fusion zone and a large grain size contributing to brittleness of the weldment. (Mag: 25X) Cyclic Fatigue Cracks Propagated by a Rust Pit (stress corrosion) Again, many of the high strength steel alloys are susceptible to stress corrosion. The photos illustrate such a failure. The first picture is a digital photo with an arrow pointing to the double origin of the fatigue cracks. The second photograph at 30X magnification shows a third arrow pointing to the juncture of the cracks propagating from the rust pits. L- 19, H-11, 300M and Aeromet 100, are particularly susceptible to stress corrosion and must be kept well-oiled.

142

143

144

Slide 1: Typical fatigue failures in steel components.

Slide 2: Striations in an aluminium alloy.

145

Slide 3 : Fatigue failures in the Alexander L Kielland platform.

Slide 4 : Fatigue crack initiation at an inclusion in a high strength steel alloy.

146

Fractures are analysed using the latest scanning electron microscopy (SEM) and other metal testing techniques. You receive a comprehensive written report with photographic documentation showing each stage of the laboratory analysis. We identify the cause of failure and recommend correction of material processing such as heat treatment, plating, machining, and/or design to prevent recurrence of the problem [2].

147

148 Annexure 3 Microscopic characteristics of fatigue fracture

Striations are the most characteristic microscopic evidence f fatigue fracture, o although striations are not always resent on fatigue fracture surfaces, as will be seen. p However, each time the crack is opened by a tensile stress f sufficient magnitude, o creating a tiny ridge, or striation, on ach of the mating fracture surfaces. If the e maximum cyclic load remains constant, the striations ear the fatigue origin are n extremely small and closely paced; the crack grows at a slow rate because the part is s s till quite strong. However, as the crack gradually ropagates, the spacing between p striations increases and he crack grows at an increasingly rapid rate because the rack t c greatly weakens the section. Eventually, complete inal fracture stage 3) and f ( separation occur. Unfortunately, striations are not always visible on fatigue fracture surfaces for a ariety of reasons: On very hard or very soft metals. Artifacts caused by v rubbing or other post fracture damage may produce parallel idges that resemble r striations. Certain lamellar microstructures in metals esemble fatigue striations. r However, careful study in the electron microscope ill reveal that the orientation of the w platelets varies randomly from one location o another, whereas true striations are t generally concentric around the origin.

Macroscopic characteristics of fatigue fracture Information can be learned about a fatigue fracture with only macroscopic examination. That is, study with the unaided eye and relatively low magnification up to perhaps 25 to 50 times magnification is usually the most important single way to study and analyse fatigue fractures.

149

Lack of Deformation
Since initiation of fatigue fracture does not require a high stress, there is usually little or no deformation in a part or specimen that has fractured by fatigue. If the maximum stress did not exceed the yield strength (actually the elastic limit), there can be no gross plastic, or permanent, deformation, although the final rupture region may have some obvious macroscopic deformation. The typical fatigue fracture that occurs in most loadbearing parts, which have relatively low-stress, high-cycle loading. Not only the fracture surface but the entire part should be examined for deformation. For example, if a unidirectional (one-way) bending fracture is observed, it is useful to carefully reassemble the pieces to determine if there was gross deformation in the part prior to fracture. Of course, the origin of the fracture would be on the convex side, which is the tension side in bending. As pointed out at the beginning of this section, in a true high cycle fatigue fracture, there will be no deformation in the fatigue region, provided that there has been no post fracture damage to the fracture surface. If the final rupture region (stage 3) is ductile, the resulting deformation will prevent close realignment of the fractured pieces; however, if the final rupture region is a truly brittle fracture, there should be no gross deformation, except for post fracture damage. A partially ductile/brittle final rupture region probably will show some degree of deformation.

Beachmarks Beachmarks are a unique feature found in many fatigue fractures, and their presence is a positive means of identifying fatigue fractures. Beachmarks also have been called stop marks, arrest marks, clamshell marks, and conchoidal marks, all in an attempt to describe their origin or characteristic appearance. The term beachmarks is the most commonly used term but is not really as descriptive as some of the others. At any rate, this term is used to describe macroscopically visible marks or ridges that are characteristic of interruptions in the propagation periods (stage 2) of fatigue fractures in relatively ductile metals.

150 Beachmarks must not be confused with striations, although they frequently are present on the same fracture surface; there may be many thousands of microscopic striations between each pair of macroscopic beachmarks.

Ratchet Marks

The term ratchet marks is used to describe features that are very useful in identification of fatigue fractures and in locating and counting the number of fatigue origins. These marks are essentially perpendicular to the surface from which fatigue fractures originate. Therefore, in circular, shaft-like parts, the ratchet marks are essentially radial, pointing toward the centre; in flat parts, such as leaf springs, they initially are perpendicular to the surface but may curve if the bending is unidirectional. The ratchet marks are not the origins themselves; each ratchet mark separates two adjacent fatigue fractures. As the cracks become deeper, the cracks from each origin tend to grow together and become essentially one fatigue fracture that has numerous origins. The number of ratchet marks equals or is one less than the number of origins; thus recognition of the number of ratchet marks is important in determining the number of origins.

Similarities between Striations and Beachmarks Both striations and beachmarks identify the position of the tip of the fatigue crack at a given point in time. Both striations and beachmarks expand from the fatigue origin or origins, often in a circular or semicircular fashion.

151 Both striations and beachmarks are relatively parallel ridges which do not cross similar features from another origin. Some fatigue fracture surfaces have neither striations nor beachmarks. Artifacts, or false features, can confuse observation of both striations and, beachmarks.

Differences between Striations and Beachmarks The most obvious difference between striations and beachmarks is size. Striations are extremely small ridges, visible only with an electron microscope. Beachmarks are much larger than striations. If they are present, they are normally visible to the unaided eye. The other difference between striations and beachmarks, as previously mentioned, is the factors that cause them. Striations represent the advance of the crack front by one load application in many ductile metals, whereas beachmarks locate the position of the crack front when repetitive, fluctuating loading was stopped for a period of time [2].

152

Annexure 4

SAMPLES FAILURE

Stub Axle failure This is the classic reverse bending fatigue of a steel stub axle from a road vehicle. Notice cracks have grown from 8 oclock upwards and to a lesser extent from 2 oclock downwards. The rough central region is the final ductile rupture.

Bending fatigue fracture This 100 mm diameter steel shaft failed after a long period of service on a large dumper truck. The keyway terminated in a circumferential groove approximately half the depth of the keyway.

153

Fatigue cracks in steel cycle frame Although steel has a fatigue endurance limit certain parts of the frame are stressed above this limit and are prone to fatigue cracking. The rear triangle, comprising the chainstays and seatstays are particularly vulnerable.

On the opposite side, the associated cracked point had caused surface corrosion. Notice that the cracks had occurred close to the brazed-on cross member. The heat affected zone associated with brazing and welding usually reduces the endurance limit and raises the chance of fatigue crack initiation.

154

A metallurgical investigation of the extent of the heat affected zone is in progress. The manufacturers did not fulfil their lifetime guarantee in this instance by suggesting that damaged paintwork caused corrosion, which in turn initiated fatigue.

155 REFERENCES [1] Jaap Schijve, Fatigue of Structures and Materials, Second Edition, Springer 2009, ISBN-13: 978-1-4020-6807-2 [2] An Introduction of mechanical Testing Pictorial Basic CMM NDT Services (www.cmmok.xinwen520.com ) [3] Alan F. Liu, Mechanics and Mechanisms of Fracture: An Introduction, ASM International, 2005, ISBN: 0-87170-802-7 [4] Tom Lassen, Fatigue Life Analyses of Welded Structures, ISTE UK 2006, ISBN-10: 1-905209-54-1 [5] BRITISH STANDARD BS 7608:1993, Fatigue design and assessment of steel structures, ISBN 0 580 21281 5 [6] John M. Barsom, Fracture and Fatigue Control in Structures: Applications of Fracture Mechanics, ASTM 1999, ISBN 0-8031-2082-6