SECTION 3.2 Designing for Fatigue Loads 1. ENDURANCE LIMIT When the load on a member is constantly vary- g in value, or is repeated at relatively high fre~ quency, or constitutes a complete reversal of stresses with each operating cycle, the material's endurance limit must be substituted for the ultimate strength where called for by the design formulas. Under high load values, the variable or fatigue mode of loading reduces the material’s effective ul- timate strength as the number of cycles increases. At agivenhigh stress value, the material has a defi- nite Service or fatigue life, expressed as Neycles of operations. Conversely, at a given number of ser- vice cycles the material has a definite allowable fatigue strength. The endurance limit is the maximum stress to which the material canbe subjected for agiven ser- vice life. 2. NATURE OF FATIGUE LOADING Fatigue failure is a progressive failure over a eriod of time which is started by a plastic move- sent within a localized region. Although the average unit stresses across the entire cross-section may be below the yield point, anon-uniform distribution of these stresses may cause them to exceed the yield point within a small area and cause plastic movement. This eventually produces a minute crack. The localized plastic movement further ag- gravates the non-uniform stress distribution, and further plastic movement causes the crack to pro- gress. The stress is important only inthat it causos the plastic movement. Any fatigue test usually shows considerable scatter in the results obtained. This results from the wide range of time required before the initial crack develops in the specimen. Once this has oc- FIGURE 1 Sires fo) curred, the subsequent time to ultimate failure is fairly well confined and proceeds ina rather uniform ‘The designer when first encountering a fatigue loading problem will often use the material's en- durance limit or fatigue strength value given inhis engineering handbook, without fully considering what this value represents and how it was obtained. This procedure could lead to serious trouble. There are many types of fatigue tests, types of loading, and types of specimens. Theoretically the fatigue value used by the designer should bedeter- mined in a test that exactly duplicates the actual service conditions. The sample used should pref- erably be identical to the machine member, the test- ing machine should reproduce the actual service load, and the fatigue eycle and frequeney should be the same as would be encountered in actual service. For example, if the actual problem isabutt weld in tension, the allowable fatigue strength used in the design must come from data obtained from loading butt weld in axial tension on a pulsating type of fa~ tigue testing machine, with the same range of stress. 3. ANALYZING THE FATIGUE LOAD Figure 1 illustrates a typical fatigue load pat- tern, the curve representing the applied stress at any given moment of time. There are two ways to represent this fatigue load: 1. As a mean or average stress (on) with a super imposed variable stress (o.). 2. As a stress varying from a maximum value (o=s) to a minimum (om). Here, the cycle can be represented by the ratio—3.2-2 / Special Design Conditions One approach to this problem is to let the variable stress (o.) be the ordinate and the steady or mean stress fon) be the abscissa, When the mean stress (om) is zero, see Figure 2, the variable stress (o:) becomes the value for fa complete reversal of stress (0,). This value would have to be determined by experimental testing, and becomes point b in the diagram. When there is no variation in Eiress, ke. a steady application of stress, a. becomes zero, and the maximum resulting mean stress (on) is equal to the ultimate stress for a.steady load (o.); this becomes point a. ‘ . « ‘, FIGURE 2 where: fatigue strength for a complete reversal of stress which is superimposed upon o. = variable stre steady stress ultimate strength under steady load (Gome set ou equal to the yield strength, oy.) on = mean stress (average stress) A line connecting points b and a will indicate the relationship between the variable stress (o.) and the mean stress (on) for any type of fatigue cycle, for a given fatigue life (N). This straight line will yield conservative values; almostall of the testdata Will lie just outside of this line. = min sess 0 Mean stress FIGURE 3 From similar triangles it is found that-— ees It can be shown by similar triangles that the same relationship holds: omy FIGURE 4 min stressDesigning for Fatigue Loads / 3.2-3 Minimum stress, ksi ae - é Ultimate é ks strength 0, eof > ‘% 2 fi 60,000 He S =|9 sog00., /& 0 etl x 2|e 2 37,800 e* : Butt Welds — A7 (373) Steel Dependable valves: © "100,000 cycles 2,000,000 cycles Allowable velyes— = — == — 16,000, T=172 but not to exceed ed 16,000. Snae = 18,000 9,000 12 8/10k -%0 20 -10 0 +10 +2 +3 +4 +50 +4460 +70 FIGURE 5 The Goodman diagram of Figure 3 may be mo- dified so that the ordinate becomes the maximum and the abscissa becomes the minimum stress (=); see Figure 4. Itcanbe proved that all three diagrams yield the same results. The Am- erican Welding Society (Bridge Specification) uses this last type of diagram to illustrate their fatigue data test results. If the maximum stress (om) lies online a-b, this value is found to be-- Qo oor — Kou co) where K = ‘The next diagram, Figure 5, is constructed with the values for complete reversal(ox) and the ultimate strength (o.) for butt welds in tension. The fatigue data from test results are also plotted. Notice the values lie on or slightly above these straight lines for service life (N) of 100,000 cycles and that of 2 million cycles. ‘These “dependable values” have been reduced to some extent below the minimum values obtained in the test. A factor of safety is applied to obtain al- lowable values; these are shown by dotted lines. This is expressed as a formula along with a value which should not be exceeded. In this case, the maximum allowable is 18,000 psi. This formula re~ presents the slanting line, but a maximum value must be indicated so that it is not carried too far. Figure 6 illustrates several types of fatigue eyeles, with corresponding K values to be used in the fatigue strength formulas. 4. ALLOWABLE MAXIMUM STRESS Fatigue strength formulas, for determining the allowable maximum stress for a given service life of N cycles, are presented in Table 1 for A-7 mild steel and in Table 2for T-1, quenched and tempered high yield strength steel. Required fatigue life or number of cycles will vary but usually starts at several hundred thousand cycles. It is assumed that by the time the value of several million cycles is reached, the fatigue strength has leveled off and further stress cycles would not produce failure, For any particular spec- imen and stress cycle there is a relationship be- tween the fatigue strength (a) and fatigue life (N) in number of cycles before failure. The following3.2-4 / Special Design Condition empirical formula may be used to convert fatigue strengths from one fatigue life to another: (a) fatigue strength for fatigue life Ne fatigue strength for fatigue life Ne fatigue life for fatigue strength o« fatigue life for fatigue strength ov ‘The constant (k) will vary slightly with the spec imen; however, 0.13 has been widely used for butt welds and 0.18 for plate in axial loading (tension and/or compression). ‘The curve in Figure 7 illustrates the general increase in fatigue life when the applied fatigue stress is reduced. As an example, in this case, reducing the fatigue stress to 75% of its normal value will in general increase the fatigue life about nine times. TABLE 1 - ALLOWABLE FATIGUE STRESS for A7, A373 And A36 Steels And their Welds 2, 000, 000 600, 000 100, 000 But Not to eycles cycles cycles Exceed Base Metal @__s500 10, 500 15,000 In Tension T-2/sK Pst T- 2/3 K Pst 1-273 K Pst Connected 2P, By Fillet sxe Welds But not to exceed =} BR PB BR ® @ Base Metal Compression pst Connccted by Fillet Wolds Butt Weld 6 18,000 In Tension on ABO? pot | Best 2 a weld © 5.000 : Compression pat + > ps 1g Butt Weld ® 9,000 S 000, att We ve 1 In Shear pee PEG pot | 29,000 pot 2 2 ir P29 tin, [toon Welds ae “2 = Leg Size Adapted from AWS Bridge Specifications. K = min/max P: = Allowable unit compressive stress for member. P, = Allowable unit tensile stress for member.Designing for Fatigue Loads /3.2—§, z (steody) i| ° min = +Y4mox K= +i a i] cs in k=0 Time —> 7 . reversal) Time —~ FIGURE 6 TABLE 2 - ALLOWABLE FATIGUE STRESS for Quenched & Tempered Steels of High Yield Strength And their Welds 2,000, 000 600, 000 100,000 But Not to ‘eyeles cycles cycles Exceed Base Metal @ @ In Tension-Not|,~ 29,000 33,000 y= 29,500 : Adjacent to 1-.65K Pst i-t60x Pst 1-.50K Pst 6, 000 pel Welds mutt waa | 26,500 © n,000 21,000 : In Tension: 1-.80K Pst i-:75K Pst {60K Pst 9 = 54,000 psi aner steel EAR! ths/in, | F= 2:908-*tps/in, | f= 12-5002 hs/in, Jt = 26,1606 Ibs/An, Above values adapted from "The Fabrication and Design of Structures of T-1 Steel" by Gilligan and England, United States Steel Corporation,3.2-6 / Special Design Conditions Problem 1 | ‘Test data indicates a fatigue life of N. = 1,550, boo cycles when the member is stressed to 0» = 30,000 psi, What would be the fatigue strength at a life of 2,000,000 cycles? (For butt welds, k= 0.13) as =(% azo” (So Using logarithms® for the right hand side: = 0,13 (log 0.775) = 0.13(9.88930 10) 1.285609 - 1.3 (add 8.7 toleft side and sub~ +87 28.7 tract 8.7 from wight side) 9.985609 -10.0 . J Tag-log slide rule could be used to find the value of rrength ‘The anti-log of this is 0, 96740; hence —%_ = 096740 30,000 ‘on = 30,000 X 0.96740 29,020 psi (at Ne The nomograph, Figure 8, further facilitates such conversion and permits quickly finding the rel~ ative allowable stress for any required fatigue life provided the fatigue strength at some one fatigue fife is known and that the constant k value has been established. Conversely, the relative fatigue life can be readily found for any given stress and any constant (k). 5. RELATIVE SEVERITY OF FATIGUE PROBLEM In Figure 9, the allowable fatigue stress is the vertical axis (ordinate) and the type of fatigue stress Cycle (K=min/max) isthe horizontal axis (abscissa). ‘The extreme right-hand vertical line (K = +1) represents a steady stress. As we proceed to the eft. the severity of the fatigue cycle increases; finally at the extreme left-hand axis (K = -1) FIGURE 7 For butt welds, k = .13 llowoble Town fatigue required fatigue life = fatigue life for which . is 5 ‘6 Increase in fatigue lifeDesigning for Fatigue Loads / 3.2~7 -99 97 96 95 ms - 94 93 33 : oi 3 vio $ ee vos 86 FIGURE 8 8 FATIGUE NOMOGRAPH 196 e 104 Given: Test data indicates o butt-weld fatigue life 70% of N 1,580,000 eycles when the member is stressed woe Yo o2 = 30,000 pi Find: The welds ftigue srength (6) ot 2,000,000 cycles (Ny Ny _ 7,000,000 _ 29 N= 1580,000 101 ‘and since the butt weld’s k factor is .13, the nomograph indicates Be nea ere is a complete reversal of stress. This is ast one method of illustrating fatigue stress con- itions. The important thing to be noticed here is tat actual fatigue strength or allowable fatigue alues are not reduced below the steady stress con- ition until the type of cycle (K = min/max) has rogressed well into the fatigue type of loading. In the case of 2 million cycles, the minimum ress must drop down to 4 of the maximum stress efore there is any reduction of allowable strength. 1 the case of 100,000 cycles, the minimum stress an drop to zero before any reduction of allowable trength takes place, Even at these levels, the tember and welds would be designed as though ley were subjected to a steady load. The stress yele must extend into a wider range of fluctu- 30,000 x 96.8% = 29,000 psi ation before it becomes necessary to use lower fatigue allowables. In other words, a fatigue problem occurs only if- 1. Stress is very high, 2, Anticipated service number of cycles, 3. Stress fluctuates over a wide range. extends for a great And it generally requires all three of these situations occuring simultaneously to produce a critical fatigue condition worthy of consideration. ‘The allowable fatigue strength values obtained from the formulas in Table 1 take all three of these into consideration, and it is believed they will result in a conservative design.3.2-8 / Special Design Conditions Fatigue Values for But! Welds in Tension from AWS-Brdge Formulae 7,11, 15 Siess cycle, K Fig. 9 Severity of fatigue depends on stress value and range of fluctuation, as well as service life. TABLE 3 - FATIGUE STRENGTH OF BUTT WELDS Summary of Results, Using 7/8-In. Carbon-Steel Plates Fog Sversh in 10005 tp Description Tension 10 on 5 Tesion Equal Compresion | 019 Tension ension oe 1/2 or Great We CRE R= = N= 100,000 2,000,000 2,000,000, ms | 53.3 | 36. 1s Welded Reinforcement On Stes Relieved Reinforcement Machined OFF [Not Sires Relieved Reinforcement Machined Off Stress Relieved Reinforcement Ground OFF Not Stes Relieved Plain Plow Mill 'Seale On Plain Pave Mill Scale Machined OFF fond Surface Polished ur Wold, Reinforcement cond Mill Scale Machined Off fond Surface PolishedDesigning for Fatigue Loads / 3.2—9 TABLE 4 - EFFECT OF TRANSVERSE ATTACHMENTS ON FATIGUE STRENGTH ae || es zs 704000 ERLES | _ 25 800 psi 25,400 Psi 22,900 ps Zo00000cnes | 22,800 psi 18, 900 psi 13,100 psi 6. COMBINED FATIGUE STRESSES Several formulas are available for this consid~ eration but very little actual testing has been done on this. In many cases there is not very good agreement between the actual test and the formulas. 1, Principalstress theory — HB yore 2, Maximum shear-stress theory — on = Vor =o +4 tt 3. Shear-stress.invariant theory — a = Yor oa FoF bom 4. Combined bending and torsion. Findley corrected shear-stress theory for anistrophy — a= Joe +(7) rot y 7 where o1/r is the ratio of fatigue strength in pure bend. ing to that in pure tension, 5. Combined tensile stresses. Gough suggests — ou: = fatigue strength in (x) direction om = fatigue strength in (y) direction o. and oy = applied stresses, 7. INFLUENCE OF JOINT DESIGN Any abrupt change of section along the path of stress flow will reduce the fatigue strength. It is not welding that effects a reducing of the fatigue strength but the resultant shape or geometry of the section. It is for this reason that fillet welds have lower fatigue strength, simply because they are used in lap joints and all lap joints including ri- veted joints have lower fatigue strength. By means of Table 3, we cansee that removing the reinforcement of a’ butt weld increases its fatigue strength to that of unwelded plate, also that stress relieving the weld has no appreciable effect on its fatigue strength, Table 4 illustrates the effect of transverse fillet welds upon the fatigue strength of plate; this is °s" plate. Recommended Try to avoid 7 ul omer Lown | FIGURE 103.2-10 / Special Design Conditions The attachment causes an abrupt change in sec- tion, and this reduces the fatigue strength of the plate, It is believed these results could be dupli- cated by machining these joints out of solid plate, without any welding. Figure 10 presents some general recommenda- tions on joint design when fatigue loading is a problem. 8. GUIDES TO DESIGNING FOR FATIGUE LOADING 1, In general, a machine is stressed to the full maximum value for only a portion of its fatigue life or cycles. For most of its fatigue life, the machine is stressed to a much lower value, and not to its full rated capacity; hence, most fatigue loading is not as severe as it may first appear. “Sitstion of not sire Pe nie firection of hot rolling of sheets wr ateel walla Ceconiend at Least on bottom half or third, or whole tank, sheets be run lengthwise with tonk Fig. 11 Groin direction of sheet orplote should be in line with force, for greater fatigue strength Consider actual stress rather than average stress. Reduce if possible the range of stress without increasing the maximum or average stress. 2, Fatigue loading requires careful fabrication, smooth transition of sections. ‘Avoid attachments and openings at locations of high stress. Avoid sharp corners. Use simple butt weld instead of lap or T fillet weld. Grinding the reinforcement off of butt welds will increase the fatigue strength. This weld will have about the same fatigue strength as unwelded plate. Grinding, however, should not be specified unless essential, since it does add to the final unit cost. Avoid excessive reinforcement, undercut, over- lap, lack of penetration, roughness of weld. Avoid placing weld in an area which flexes. Stress relieving the weld has no appreciable effect upon fatigue strength, Difficulties are sometimes caused by the welds being too small, or the members too thin. 3. Undereritical loading, place material sothat the direction of rolling (of sheet in steel mill) is in line with force, because the fatigue strength may be higher in this direction than if placed at right angles with the direction of rolling. See Figure 11. 4, Where possible, form member into shape that it tends to assume under load, and hence prevent the resulting flexial movement. 5. Avoid operating in the critical or resonant frequency of individual member or whole structure to avoid excessive amplitude. 6. Perhaps consider prestressing a beam in axial edmpression. This will reduce the tensile bending stress and lessen chance for fatigue failure even though the compressive bending stress is in- creased to some extent. 7. Avoid eccentric application of loads which may eause additional flexing with each application of load. 8, Stiffeners decrease flexibility of member and result in better fatigue strength, unless it causes more abrupt change of section. If the latter should happen, the stiffeners may do more harm than good. 9. A rigid frame type of structure or statically indeterminate type of structure may be better than a simple structure since the load is shared by other members; hence, the structure is less likely to collapse immediately if a fatigue failure starts in one member. This will result ina more gradual failure of one part, then another, and this wouldDesigning for Fatigue Loads / 3.2—11 L Baffle plote — Fotigue crack FIGURE 12 provide a better opportunity to notice that a fatigue failure is in progress. 10. Avoid biaxial and triaxial stresses, avoid restrained internal sections. Problem 2 . Which type of steel should be selected for the following fatigue loads for minimum weight of material? case A ‘A built-up beam subjected to a complete re- versal of « stress, The flange plates are subject to alternating tension and compression. «-=2 N= 2,000,000 eycles From Tables 1 and 2, formulas for butt welds Steel TA Steel 16500 _ 77-30) pai =10,570p0i| 6 psi = 9,170 pa In this case, A-7 (mild steel) would be sel- ected because it has about the same fatigue al- lowable as T-1 steel. Case B ‘The web plates of a pump crankcase are con— nected to the face plate supporting the cylinders and bearings of the crankshaft. Compression is on the forward stroke only. Each web is subjected to axial tension only, and under the worse condi- tions drops to about 75% of the maximum tension, hence: min =+.75 N= 2,000,000 cycles From Tables 1 and 2, formulas for butt welds in tension: Ta Steel 16,500 6 = 21500 pj = 41,200 psi ata P (but not to exceed 18,000 psi — AWS.Bridge Spec) In this case, T-1 steel would be selected be~ cause it has 2.29 times the allowable fatigue stress as A-T (mild steel), and would require just 44% of the plate thickness and weight. Problem 3 After three years service a tank trailer is developing fatigue:cracks In its 12-ga bottom shell, ‘Adjacent to an internal baffle plate. Itis necessary to improve this so it will last for the expected life of the unit which is 20 years. See Figure 12, Although the tank trailer is subject to bending, the bottom shell is mainly in tension, Fatigue data for butt welds in tension will be used. force, Ibs/in “thickness: Syme Hoy (20 yrs) tu te _ (12 Ga = -1046 Te" Tee So i and t 32 338 or use 10.ga steel (t ‘To avoid using the above fatigue formula, the curve in Figure 7 could be used. An extension in 1345") life of about 6% times is desired. The curve indicates this can be obtained by reducing the fatigue stress to about 78% of the original. This reduction in stress could be obtained with a sheet thickness 1.28 times the original, or 10-gage.3.2-12 / Special Design Conditions bending stress Direct Pull Bending Action fl {= [w =F ‘Moment * relation of fotigue life to fotigue stress FIGURE 13 A bracket (Fig. 18) ona farm implement machine is failing in fatigue after about 1 year in service. If we should increase the thickness of the bracket from %e” to %+" , what extension in fatigue life can we expect? From the above analysis of loading, it is ap- parent that fatigue failure is the result of the ten- sile stresses from both bending moments. For lack of fatigue data for plate in bending we shall fatigue data for bult welds in tension: Me_6M ail aoe then, by substituting the bending moments into the Side of the above fatigue formula i = reer om = Ge ws (42) 1 year Geass = 164 yrs