“T= SauNDONELS ONY SIVESIVN JO ANSI SHLFOREWORD The whole subject of the fatigue failure of metals has to do with the single fact that metals will fracture, without elongation or plastic flow, upon the repeated application of stresses below the limit of pro- portionality of the material in question. That this can happen has been recognized for a least 125 years. Railroad axles were among the first mechanical devices to experience failure in service for this reason, With the passage of the years, through the last century of development of mechanical things, one phase of industry after another has been forced to recognize the importance of this pheno- menon. This occurred as such products were developed to the point that they worked hard enough and long enough to become susceptible to fatigue failure before being discarded for some other reason, Airplane engines and propellers reached this stage some years ago. The last few years have seen airframes themselves become sui- ficiently developed devices that it was physically possible and economi~ cally desirable to use them long enough and hard enough for failure in fatigue to become a matter of real concern, There have been some serious actual occurrences. Fortunately materials suitable for the construction of machines have good enough resistance to failure in this fashion so that actual occurrences are usually the result of local stress raisers that shorten life far below the life of which the material is inherently capable. Such stress raisers can be the result of tool marks or other inadvertencies but are far more likely to be the result of detail design. This means that, from the standpoint of airframe structural design, the degree of likelihood of fatigue failure is almost entirely in the hands of the designer (assuming proper operation of equipment). If this fact is recognized and every detail of design is carried out with due regard to fatigue life, it is doubtful if a serious fatigue failure would ever be seen in service. Unfortunately, the state of the art is not yet infallible as to what separates good practice from bad. A very great deal is known however, as the data collected in these notes will show. Interest in this subject with respect to airframes is so recent that misunderstanding and near hysteria are easily created. It is hoped that students taking this course will not only learn a great deal about fatigue but will also develop a sense of proportion as to its place in the scheme of things - remembering always that "a little knowledge is a dangerous thing". C, R, STRANG CHIEF PROJECT ENGINEER XC~-132Up until ten, perhaps even five years ago, the design of airplane structures was based solely on the maximum loads the airplane could get once in its lifetime. During the past five years a gradual revolution has taken place, and today the design of structural members, and especially structural joints, is no longer based on one maximum load, Now the air- plane structure must also have a reasonable life under the many small repeated loads it will be subjected to in service, This makes a knowledge of the fatigue of metals an essential for every airplane designer. HAROLD W, ADAMS CHIEF DESIGN ENGINEER DOUGLAS AIRCRAFT CO. , INC, We are spending a tremendous amount of effort in this country to avoid catastrophic fatigue failures and thus far we have been very success- ful. Contrary to the impression that some people have gotten, we do have a standard or code of practice and we are monitoring what we do, Our standard is very simply stated: 1, Parts should normally have a life which exceeds the use- ful life of the airplane, As we gain-knowledge, this is becoming more and more generally feasible. A highly-periected transport airplane accumulates flying hours at the rate of 3000 to 4000 hours per year. Assuming a fifteen-year useful life before obso~ lescence, such an airplane might reach 50,000 to 60, 000 hours before being retired. DC-3s in service about this long have in some cases reached over 60,000 hours. DC-3s are still operating, however, and will no doubt continue to do so for some years. While this may be a special case, so far as obsolescence is concerned, other aircraft may also exceed the rough fifteen year average I have taken. On the other hand, if a part is good for 100,000 to 200,000 hours, (and this is an mtirely practicable figure to work to on basic structure), that part is surely adequate from the standpoint of fatigue. It is common practice to design fittings even more conservatively than basic structure and normal for them to outlast the structure they are attached to when re~ peated-load tests are run in the laboratory. 2, Parts which cannot feasibly or with assurance be de- signed for such long life should be designed in such a way that signs of distress, such as cracks, wear, elongated holes, bending or deformation, will become visible long before they become critical. The aircraft should be metho- dically inspected for parts with such signs of distress and these should be immediately repaired, reinforced, or re~ placed.iii 3, The number of critical parts or elements should be kept to a minimum by multiplying the paths through. which loads travel. The failure of any one part should not seriously affect the ability of the remaining structure to carry load. 4. Parts which are both critical and of short life, and these are rare under this system, must be replaced at or before the end of their safe life, conservatively figured, Replacement of parts is a costly process. It is doubly so, when they must be replaced prematutely, long before their actual life is up, in order to be safe. Our whole system is based on the idea of safely lengthening the service life of parts. To this end we design as many as we can to outlast the airplane and as many more as we can to remain in place until they signal some sign of deterioration, cutting down to an absolute minimum the ones that have to be replaced on an arbitrary schedule, This requires a good knowledge of how to design for long life and adherence to those design practices which we know to be good. It requires good shop practice and quality control. It requires continuing and com- prehensive inspection in service. And it requires disciplined self~policing all along the line. Research is needed in order to improve our performance under this system, We need better correlation between laboratory tests and service life. One important factor in this is the compilation of gust load histories now being carried on. Another is the establishment of a meas- uring stick better than simply a record of flight hours. Devices of this sort are being developed and their use will replace generalizations with specific load histories on particular airplanes. The aircraft manufacturers, both civil and military, have an Aircraft Structural Fatigue Panel, operating within the A.I.A., divided into Project Groups under the headings of Environmental Conditions, Design and Analysis, Materials, and Operation, which is concerned with facilitating the interchange of information and the establishment of industry programs of research. A, E, RAYMOND VICE-PRESIDENT, ENGINEERING DOUGLAS AIRCRAFT CO,, INC,iv A SERIES OF EDUCATIONAL LECTURES ON THE FATIGUE STRENGTH OF STRUCTURES, PRESENTED TO ENGINEERING DESIGNERS OF THE DOUGLAS AIRCRAFT COMPANY INC. AT SANTA MONICA CITY COLLEGE 1954 Cover. At top edge - fatigue crack in 75 S aluminum alloy 500X - RHC= Center - Fatigue crack in CM steel at 150X unetched. Photo by A. Phillips, DACo, LL THE MECHANISM OF FATIGUE pod 7 STRESS CONCENTRATION FACTORS IN FATIGUE AND OBSERVATIONS FROM FATIGUE FRACTURED SURFACES bow 3. CORROSION FATIGUE, ELEVATED TEMPERATURE AND SURFACE CONDITION EFFECTS +. 40 4, EFFECT OF RANGE OF STRESS, COMBINED STRESS AND RESIDUAL STRESS IN FATIGUE + 59 5. METHODS TO IMPROVE THE FATIGUE RESISTANCE OF MATERIALS 1.8 6 FACTORS AFFECTING FATIGUE STRENGTH + + 90,01 ih LABORATORY TESTING OF STRUCTURES AND COMPARATIVE TESTING go) 8. DETAIL DESIGN FOR FATIGUE IN AIRCRAFT STRUCTURAL JOINTS: + + 100 %. FATIGUE LOADING AND ANALYSIS OF AIRCRAFT STRUCTURES . . 105 10, TRENDS IN COMMERCIAL AIRCRAFT OPERATION AND DESIGN. . 119 1. VARIOUS FATIGUE ANALYSIS METHODS a e138) 12, DESIGN PRINCIPLES 5 tee) 13, SERVICE FATIGUE FAILURES + + 168 14, FATIGUE MONITORS ela Back, Fatigued designer, Douglas Proprietary rights are included in the information disclosed herein. Neither this document nor the information disclosed herein shall be reproduced or transferred to other documents or used or disclosed to others for manufacturing or for any other purpose except as specifically authorized in writing by Douglas Aircraft Co,, Inc.FATIGUE OF MATERIALS AND STRUCTURES Introduction Someone once said, "The best way to begin a course on any subject is to define the subject briefly in a short, concise manner", Surely this person did not have the subject "Fatigue of Metals" in mind when this statement was made. It is not that the subject is so complicated - be- cause it isn't; it's just that there is so little that is really known about the basic nature of the mechanism of fatigue. There is no textbook on the sub- ject and it is doubtful if one could ever be written. To date, the whole sub- ject of fatigue could at best be compiled by thousands of unrelated investi. gations, studies and tests of the fatigue of metals and structures, It will be the purpose of this series of lectures on the "Life and Death of Metals" to put down in some orderly and chronological fashion the steps to be taken by design and stress engineers in their analyses of structures. It is not to be implied that upon completion of these lectures the engineer will, as Mr. Honey in Neville Schute's "No Highway", be able to predict within the hour or to the cycle when a particular structural member will fail in ser- vice. A rather absurd thought along this vein would be to imagine that you could actually analyze a part to fail in some exact number of prede~ termined loading cycles. What a temptation or challenge it would be to de~ sign your structural member to this ultimate degree of efficiency - and just think what this would do for part sales! During the course of these lectures, there may be times when we touch very lightly upon a particular subject but may ata later date return to the subject and discuss its effects in greater detail, Often it will be necessary to spend what seems to be an undue amount of time on a rather simple point - those of you more familiar with the subject will please bear with us. How, then, might we define our subject? At the risk of making a rather naive explanation, let's consider the loading of a bar of metal. We know if we pull on this bar with sufficient force we will eventually break it, However, let us pull on this bar with say only half the force that fractured it and then relax the loading and re- peat this cycling many hundreds of times. We now examine this bar, measure if, weigh it, X-ray it, and there has been no visible change. How- ever, upon thousands of additional loading cycles there is still no visible change or evidence of plastic deformation and then all of a sudden the bar breaks! What has caused this mysterious phenomenon to occur? Pre- sently, we don't know, Eventually, this phenomenon may be satisfactorily explained but until that time we will have to live with the subject and con- sider it in our designs to the best of our ability. In our particular field, however, it is neither practical nor always possible to design members for an infinite life, since the aspect of weight-saving is all important. Neither are we to be too alarmed that this insidious phenomenon will al- ways or ever result in a catastrophic event. We should take the point ofview that we will meet the problems as they arise and solve them to the best of our ability - in the meantime endeavoring to find out as much as possible about the subject, Fatigue failures if defined as cracks charac- teristically develop slowly and finding them is a matter of inspection. Here, then, is the essence of concern for the eventual structural integrity of any structure, Itis, inevitably, a matter of inspection and maintenance. It will always be so, no matter how our structures are designed and originally built, 2s Hy Christensen * clock ¥ 32861 105542Chapter 1 THE MECHANISM OF FATIGUE Bead, I think the interpreter is the hardest 0 be understood of the tuo! = 2B. B. Sheridon 1. Theories of the basic nature of fatigue failure. ‘The earliest known fatigue tests of materials were those conducted by Alberts in 1829 on welded mine hoist chains although there was no attempt to arrange the test data systematically or explain his results. The most important series of tests that followed Alberts were those of Wohler in 1852, His investigations, to determine the endurance limits of wrought and cast steel bars for railroad axles, required approximately 17 years to complete. If for no other reason Wohler should be admired for his perseverance in conducting these tests for so long a period of time. Wohler was the first to classify his data and arrange it in graphical form, plotting stress against cycles to failure. Hence the term Wohler curve" that was and has been used for many years, Wobler was the first to formulate two fundamental laws of FATIGUE: namely, La Iron and steel may fracture under a unit stress not merely less than the static rputure stress, but also less than the elastic limit, if the stress is repeated a sufficient number of times. 1.2 However many times the stress-cycle is repeated, rupture will not take place if the range of stress between the maxi- mum and minimum stresses is less than a certain limiting value. It was not until 1870 that various experimenters began further studies into the mechanism of fatigue and its causes. A brief chronological dis- cussion of (some of) the better known studies and theories follows: 1884 - Bauschinger theorized that the primitive elastic ranges (tension-compression) were representative of the safe ranges +. in fatigue, The Bauschinger effect states that for nearly equal stresses the elastic limit in tension can only be raised by a corresponding drop in the compression limit. But there was a limit as to how far this range could be stretched. This, however, is not true since we have long since proved that parts and members may fail in fatigue at stresses far below the elastic limits of the material. 1902 - Ewing and Humphrey microscopically observed the slip bands in the fatigueing of Swedish Iron. At first they thought that the fatigueing might loosen the intercrystalline cement rather than damage the individual grains. However their experiments showed an entirely different phenomenon. Micro-photographs1904 1906 1923 at 1000x and at varying numbers of stress cycles showed the growth and broadening of the slip lines and finally a crack within the crystals which ultimately interconnected and formed a continuous fissure originating near or at the surface, See figure 1, Their so called "Attrition Theory" was thus named because they believed a process of attrition occurred along the! slip planes (a wearing away of the slipped surfaces) resulting in a collection of debris which was squeezed out from ‘the crystals as the crack was being formed. Beilbys theory of the mechanism of fatigue agreed in general with those that preceeded his, however, he believed the process of "slip" involved a transition from the soft and crystalline phase to a hard and amorphous phase in the re- gions immediately adjacent to the plane of slipping, the term "amorphous" being used to indicate a state of heterogeneity of molecular arrangement rather than that of a crystalline phase, The Mobility of the amorphous phase involved during slip he believed to be extremely transitory and therefore slip was not reversible, However if the process of repreated strain- ing were continued an increasing amount of the crystalline ma- terial was converted to the amorphous phase and the increasing tensions finally becomming sufficient to initiate a crack. Rosenhain revised Beilbys Amorphous theory by claiming that slip produced a "temporary mobile phase" which required time for setting. Such a phase therefore permitted a reversible strain if applied immediately. He theorized that a state of re- peated slipping will result at each stress reversal and the limi- ting stress at which slip ceases to occur within the metal is the safe range of stress. Gflough and Hanson conducted extensive tests to more or less prove that there were safe ranges of stress that exceeded the static yield stresses. There were many who believed at this time that this condition was not possible. Carrying on with Bauschingers theories and experiments of Bairstow (1909) they measured and observed the hysteresis loops at various stages in the stressing. At stresses below the safe range they ob- served the hysteresis loop to actually diminish with increas- ing cycles of loading after it had reached a maximum growth in width. See figure 2. Very near the safe range of stress they observed The Width of the loop to increase ead eventually vemain constant with increasing reversals. At stresses above the safe range the hysteresis loop gradually increased in width but at a decreasing rate until failure finally occured. How- ever during periods of rest the loop diminished but on recurring cyclic loading the loop widened again until ultimate failure. Gough was one of the first to mention and associate the mechan- ism of fatigue with work or strain-hardening, Hardening, how- ever, can occur under all ranges of stress near at or beneaththe safe range, indicating that slip is not necessarily a weakening process, According to Gough the initiation of a fatigue crack was ascribed to the capacity of the metal for strain-hardening, and when this capacity had been locally exceeded the crack occurred, 1939 - Orowan proposed the following picture of the mechanism of fatigue. "The metal contains weak spots, which act as sources of stress magnification, These may be assumed to be the same as the spots that are responsible for the weakness to- ward rupture in static tests. During the earlier stress cycles, the stress near one or more of these weak spots would rise above the value required for rupture even for stresses some- what below the safe limit if the material were not plastic. Actually, slip occurs in the regions of high stress during the carly stages of the fatigue test; and, as a result, the stress becomes distributed in such a way that the value for rupture is not exceeded, Work hardening accompanies the plastic flow, however, so that the strain producfed by the stress de- creases monotonically during successive cycles. Thus the maximum stress near the weak spot rises with the number of cycles since it is less effectively distributed by plastic flow. In the safe range, the limiting value of the peak stress is less than the critical stress for rupture, whereas it is larger in the unsafe range. According to this picture, the ability of the material to flow raises the static value of the rupture stress. The specimen breaks after successive stress cycles for stresses lower than the static value because its ability to flow is lost as a result of work-hardening, " 1947 - Shear Dislocation Theory. This theory assumes the pre- ‘existence of sub-microscopic cracks. The fatigue failure of a metal will be the result of the growth of cracks caused by dislocation occurring at the edges of these sub-microscopic pre-existing cracks under the influence of cyclic stress. A dislocation consists of a stable arrangement of atoms such that in the region of a few atomic distances, n+ 1 atoms in the slip plane face (n) atoms across the slip plane. The most probable sources of the generation d dislocations are at stress con- centrations such as the above-mentioned pre-existing sub- microscopic cracks. The dislocations are therefore generated at these points of stress concentration upon the application of shear stresses. The small cracks forming in this manner eventually interconnect and the fatigue crack in general occurs and grows along the slip plane. See figure 3. A rather discouraging state of affairs to realize that 100 years of re- search has failed to explain the mechanism of fatigue more satisfactorily, nor does it appear that the original concepts have been enlarged upon to any great extent,It remains to be seen how the understanding of this phenomenon will eventually help us - however fatigue still remains to be a serious threat and challenge to the designer who certainly cannot ignore it, While our long-haired scientists are busily examining crystallographic planes - those of us, not so learned, can attempt to correlate the phenomenon with more tangible and everyday physical properties. 2. Attempts to Correlate Fatigue Resistance With Other Physical Properties. There has not been much success in attempts to correlate the fatigue strength of materials with other basic properties. It can be said that fatigue resistance is not an intrinsic property to be measured in absolute units nor does it exist apart from the conditions of service. One of the most successful correlations is the fatigue endurance limit : of steels as a function of the tensile strength of the material. This rela- a tionship appears to hold true for par bar specimens (no stress concentra- —~" ! tors) up to tensile strengths of 180,000 psi as shown in figure 4.” The rule of thumb for this relationship is that the endurance limit in rotating bending is 1/2 the tensile strength of the particular material, This relationship however breaks down when we introduce a notch, fillet or hole in our speci- men, Attempts to correlate fatigue strength and endurance limit with such properties as static impact strength, yield strength, creep strength, etc., have similarly not been to successful, Non-destruction tests and investigations such as X-Ray diffraction studies, magnetic and electrical resistivity properties, etc., have also been tried and not found entirely satisfactory as a means of determining the de- gree of fatigue damage. Magnaflux, Zyglo, Dy Check, etc., and even the Reflectoscope have been used in attempts to detect fatigue crack fissures. These methods are all very useful in detecting fine cracks but none are sensitive enough to de- tect sub-microscopic cracking or crystalline fragmentations in the process of continuing fatigue damage. 3. Effect of Type of Loading and Range of Stress. Notwithstanding this seeming lack of correlation of fatigue properties with other properties of which we are so much more familiar we are still not to be discouraged, Experimental testing has provided some rather in- teresting evidence as to the behavior of metals under all possible config- urations of loading. In 1870, Gerber proposed a parabolic law that he believed to represent the behavior of metals in fatigue. Briefly, his diagram is described as follows:Onin. TEST DATA ON , DIFFERENT MATUS FALLS WITHIN THESE LIMITS DIAG. A Onax. PARABOLA ST. LINE 9 4 ° a RANGE = T. TOT. ALL_TENSION ° Omax.+ Omi Ovean= OM x MIN, This curve may be plotted in many ways; for instance, as absolute stress values or as a fraction of the static ultimate strength. The endurance limit curve only is shown here, however similar parabolas may be con- structed representing curves of constant life. Also superimposed on this chart may be constructed constant "R" curves which are more common and more frequently used, See Appendix A. Years ago about the only determined endurance limits were for com- pletely reversed bending stresses in rotating parts - today we are not only interested in the upper part of the SN curve but also in the fatigue strengths of materials and parts at some predetermined life other than infinity and for range of stress other than a mean stress of zero, + Type of plot for HALF machine parts under RANGE | rotating bending OF se5. STRESS = RANGE=Omax, ~ oy) CYCLES DIAG. BType of plot for | axially loaded ie members and also R VALUE NOTED for stress ranges other than complete =OMIN. (tens. -comp.) ~ OMAX. reversal, MAX. TENSION STRESS CYCLES DIAG. C It might be well at this time to establish our nomenclature and defi- nitions for a few of the more common terms we will use throughout the dis- cussions, Endurance Limit or Fatigue Limit 1 The highest stress whose repeated application may be applied to a material indefinitely. Fatigue Strength ©: ‘The repeated stress that a material can withstand for a finite number of cycles or life. Stress Ratio(R) |: The ratio of minimum to maximum stress in a load~ ing cycle. — OMAX: Omax. | STRESS RANGE =|— =MAX. T f OMIN. MEAN STRESS = hax. Chan, AN ANGE & R= OMIN. 241 —V——own Omax, COMPLETELY REVERSED UNIDIRECTIONAL‘The SN curves we determine and use today thus yield much more in- formation than the older type plots. CONST. MEAN es STRESS CURVE HAS THIS SLOPE. N So far we have been mainly considering the effect of range of stress for bending loads. However, there is another observed phenomenon in the fatigue strengths of metals in that the characteristics are altered by the type of basic loading. For instance the relative strengths of a material in fatigue, under the same ranges of stress but for different types of load, may in gen- eral be pictured as shown in figure 5. (¢y lito 3.1 ‘The fatigue strengths in Plate Bending are superior to Rot. Beam (this-mey-be-e-meduits-ofmupuine-etfecthy foe pg. 94, 3.2 Fatigue strengths under direct stresses appear to be approxi- mately 70 - 80% of those in Rotating Beam, (A statistical theory of the amount of cross-sectional area stressed has been suggested as an explation for this difference.) 3.3 Strengths in torsion appear to be approximately 50 - 60% of those in rotating bending. (This relationship is usually ex- plained as being the same ratio as the distriotion energy theory indicates 0.87 = Torsion Bending or applying Hooke's law oye = 0.63.) E The above relationships have been observed between the different types of loading at a given range of stress (that of a complete reversal), There are of course an infinite number of configurations of range of stress and it is assumed that the same relationship holds for all cases.10 There is sufficient data available to bear out this relationship al- though there are undoubtedly exceptions to the rule, Returning to Gerber's Parabolic Law we now might briefly expand upon its use as a means of predicting fatigue characteristics for other ranges of stress for which we possibly have no data. Reference Appendix A, page I7. Goodman proposed a diagram which has been used more extensively than Gerber's and which has been revised many times to conform with our present concepts. The form of the diagram widely used today is as shown in the following diagram: For a more detailed explanation of the construc- tion of this and earlier type diagrams refer to Appendix B on page 9 . UTS, o OMIN. 3.4 In addition to these diagrams there have been many attempts to calculate the equations of SN curves - not only to extrapolate to expected lifetimes but possibly to shed some light on a greater understanding of the mechanism of fatigue. Ordinarily SN data is plotted on so-called semi-log paper (no such thing as a semi-log). This method has been found to be the most suitable for the purposes of clarity as it stretches out the region of the curve in the low life range which is the region we are the most interested in, Imagine an SN curve plotted on Cartesian coordinates and how ridiculous it would be to extract design information from the upper part of sucha curve. See figure 6, This presentation shows the test data falling on a straight line which is readily amenable to a straight-line equation,a DIAG. F N 3.5 However the log-log plot often obscures the test data ina par- ticular region of interest, just as the data would be on Cartesian coordinates. A few of the more successful equations for defining a nearly complete SN curve are as follows:(gS) 1, S = CN™ (logarithmic) 2 (S-C,) (N-C,) = C (hyperbolic) 1 . Ss Sy a U 7 x) * No (trigonometric tangent) The log equation (1) fits the data fairly well but it will not bend sharply enough at the "Knee" of the curve, The hyperbolic equation also fits the data well but rises too sharply at high stresses and characteristic of a hyperbola it will never reverse curvature and bend over to the apparent static ultimate stress on the vertical ordinate. The trig-tangent curve developed at DACo appears to match all test data tried with the greatest success. See figure 7 for a comparison of the three methods in matching the test data. Attempts have been made to determine a relationship between the endurance limits for various families of "R" curves as well as constant mean stress curves. However no significant relation- ship appears to exist, See figure 7a. R=.8 no NO SIGNIFICANT PATTERN3.6 Alternating shearing stress of any stress range in &Ty terms of endurance mit for complete torsional reversal, 12 Unlike other types of loading, torsion is unique in that the endurance limit can be predicted for any range of stress if only one range is known, A plot of torsional fatigue data of a wide variety of materials will best illustrate this phenomenon, 1.0] oto—to-A-o-t.b-0-+—9-A-o-Ho| ol os) O Max. shearing stress of BT range expressed as a frac- aa Hon of ty (yleld stress in shear) This relationship appears to hold true for stresses up to 0.8 of the yield strength in shear, However we should realize that these values are endurance limits only and do not apply to the upper region of the SN curve which we are mainly interested in, 7 A= Endurance range (A) of stress constant for all levels of max, stress up to 0.8 ty - DIAG. 1 4, Combined Stresses in Fatigue. It might be well to briefly consider the effects of combined stresses at this time and later go into them in more detail. A few of the more common combinations are as follows: 4.1 The influence of static tension of the torsional fatigue Umit, The torsional fatigue mit has been found to decrease atan in- creasing rate with an increase in the longitudinal tensile stress. The equation suggested for this condition is given by the equation of an ellipse.4.3 13 oe. Q + TF = 1.0 (equation of an ellipse, P & Q semi-axes) 2 2 AF Q = torsional fatigue limit under any given stress combination. Q, = fatigue limit for completely reversed torsion, Ti = applied static tensile stress. T, = ultimate tensile strength, The influence of static torsion stress on the flexural fatigue limit. ee Tigure U-— Tests have indicated that the static torsional stress does not diminish the flexural fatigue limit appreciably unless the torsional stress exceeds the half range of the flexural stress, Above a certain value of the torsional stress however there is an abrupt decrease in the flexural fatigue limit. The equation for this combination may also be represented by an equation for an ellipse as follows: a? , s+ +l 2 2 Ly 1 zg Ey g w . the flexural fatigue limit (semi-range) for any stress combination, the fatigue limit in bending for a complete stress reversal. the applied statical torsion stress. the static ultimate torsional strength. Effect of combining cyclic tension or bending stress with cyclic torsional stress. Similarly this configuration may be represented by the equation for an ellipse: Also see figure 9. = 1.0 Bending and Torsion oF = 1.0 Tension and Torsion where B, T, and Q are the half-ranges of bending, tension and torsion for any given stress combination and By, Ty, andQ, are the reversal fatigue limits for bending, tension and torsion,14 5. So far we have been considering rather idealized cases and depicting an SN curve as a well-defined, smooth curve much like the ordinary stress- strain diagram, Let us now be a bit more realistic about the subject and, as an example, inspect the test data of an ordinary riveted lap joint. Refer- ence figure 10, Plotting peak stress of the loading cycles versus cycles to failure, we find there is considerable scatter in our test results. This is typical of all life testing, whether it is the result of laboratory tested parts or parts loaded in service, No more than there are two hairs or two grains alike are there two specimens with identical lifes. It is true that the apparent scatter in testing may be minimized by careful testing techniques and the following factors all have their effect: (a) Accuracy of maintaining the loading throughout the testing period, (b) Methods of preparing the test samples. (c) Human errors in setting up the test loading. (a) Mechanical characteristics of different fatigue machines of the same type. (ec) Variables within the material itself. Probably the greatest single factor in the scatter of the lifetimes of parts in service is due to the wide variation in the overall service loading of the members, It will therefore have to be left up to the designer, as to the importance of his part, to decide whether he is to base his design on the lower or upper Limit or on the mean of this scatter in life. If he chooses the average or mean life - he may further refine his analysis by bracketing his predictions and calculations by a probability factor - for example 90% of the tested parts will fall within prescribed mits. A few observations have been made from a vast amount of data and the following relationships appear to hold true, Reference figures 10 and 11. (f) Scatter or life variability increases as the magnitude of the re- peated stressing is decreased. (g) Scatter or life variability increases as the stress range is de- creased. (2) It has also been observed that the scatter in the lifes of plain specimens is greater than the scatter for notched specimens - although there is no explanation given for this phenomenon, 5.1 In our brief initiation into the presentation of fatigue test data, we’ should not omit the effects of rate of loading or "periods of rest", The "rate of loading" can be divided into two parts and may best be represented by the sinusoidal load cycle sketch below. Impact type loading 6 o =| +.01 SECS. 02 SECS. ONE CPM. 3000 CPM. DIAG. JIt has been observed (neglecting the effects of corrosion for the present) that it makes no difference whether the loading rate varies from one to 3000 cpm the fatigue life in total cycles is the same. Similarly, it makes no appreciable difference if the in- dividual loading cycle requires one minute to complete or if it occurs in .01 sec, DIAG. K -OISECS, -k-—— 60 SECS. ——] Figure 12 shows the effect of very high speed loading. To compli- cate the picture still more, it has been observed that in the very high speed range the effect may work both ways ~ increasing the fatigue limits and strengths in the lower stress regions as well as decreasing the fatigue characteristics in the higher stress ranges. Among various investigations there is some disagrec- ment on this apparent effect as evidenced in the laboratory - and as long as the effect is small and the frequencies we normally deal with are in the order of 3000 cpm, it will probably be best to ignore it, 5.2 Periods of Rest, Here again, various investigators have disagreed on the effect of “periods of rest", In gencral however, the majority has shown that periods of rest have no effect on a metal where stressed be- low the elastic limit stress levels, and only a very slight improve- ment effect when stressed at the higher levels. Pointing out again that the metal, almost like a human, hasn't forgotten what you have previously done to it. 6. Summary Fatigue data are mortality data and it would be just as inaccurate to expect reliable actuary tables to be derived from data on five people as to ex- pect reliable data from fatigue tests on five specimens. See figure 13. There is a striking similarity between the human mortality curve and those for laboratory specimens. However they differ in detail, Note that the "Life and Death of Metals" curves do not extend to zero life as is the case with humans. Infant mortality is avoided in machine parts and the like be- cause these parts often are rejected in the inspection stage and never reach the period of testing or servicing; naturally this practice is not followed for humans.16 By now we should all realize what an intangible subject we are dealing with. Unfortunately, there are some agencies in this country, as well as in Europe, that are clamoring now for fatigue analyses, which has mainly been brought about by a few who decided they should write a "paper" on the subject. These same people tell us in a roundabout way that they can predict lifetimes and they show us long and rigorous calculations - coming up with fatigue factors of safety of 1.25 or 1.35. Bunk! These factors are no more than factors of ignorance and we would have more respect for them if they admitted it, From what little we have just discussed of the variables and scatter normally encountered in testing - and we haven't as yet included the scatter present due to service loading - the factor for fatigue of 1.25, or what have you, is not only superfluous but absurd.16.1 SLIP_PLANES IN CRYSTAL OF SWEDISH IRON AT 25000 PSI. MAX. STRESS. MAGNIF. [000 X _ AFTER [000 CYC. AFTER 2000 Cyc, AFTER 10.000 CYC. AFTER 40,000 CYC. FIG.1i OQ, MAX. FIBRE STRESS °o 1 {yt 1 STRAIN READINGS FIG. 2 (6.20FIG.3ENDURANCE LIMIT R=-Lo FIG.5 16.40 18Q000 U.T.S. PLATE BENDING ROTATING BENDINGsTRESS - 16.50 IRovATing Beam Teet oF, 1887, 2457 ant Pil PRowD pas a bt) aSRS gee ee WR opt st | LL 20}\— | i! | ol eee “iz 123 a8 67 9 io stress Evcles mn tutions|6.60 ayuntva OL SaATDAD - N a ol eo 40 a -s A = SSauLs WAWIXVW DIWHLIa9907 2IN0eeadAH —— S R ALNZONUL ‘OL — ——— TSd- Wivd isal 0 x oootonon' = (on-nyorer-#), T]\[ SAY Ss 3. ol “SWHOs YBHLO OL dativdWwos sy vive SNOLLVA ONIHDLYW 04 NOLLWNbS LVAONVL “D121. 2 biale71 ENDURANCE | LIMITS UNDER COMBINED STRESS -50000 _O +50000 P- PRINCIPAL STRESS FIG. 816.80 6914 ANAWOW SNIGN3€ : _LNAWOW SNILSIML Oe G Ol Kr ‘006 (aNIMNWY) SS3YLS WdIDNIdd *XVW “LSNOD ; NIVIS (N34) a Wale “XW “LSNOD ADYANZ IWLOL “LSNOD ASYANA NIVULS YVSHS LSNOD 6 gliSas) SSAALS UYVSHS *XVW “LSNOD =OlLVY © NVL O 130 S [3D-IN Xx SD 010 Ef is Ls bk 4 ig E 4.Ol ; Ol D3 vv oas / elem s00Ger Maen GAS) *a@O0Ud NE =z -OlE co _ n PES sla s og eek oe a gat 8) -o2" ee Neco BS 7 SN sinior 98 9 1aaHs! 234 8 oN 8 ‘ = dv7 pogppBp 8 —oGZ Nid KY\ aalaai \ fool eR \ oP oo So ~ N 2 ; \ ; \ S%0| 0 —OF o1 ‘914 \\ > \ p08 \ \° | ®16.91 LIFE VARIABILITY. OF BOLTS vs STRESS RANGE 3S 5 @ oO RANGE = SURVIVORS - PERCENT o 100 200 300 400 500 LIFE -THOUSANDS OF CYCLES FIG.I16.92 00002! 2 O14 4b ONISSSYLS ASIA _ ‘Wd'D — ONILSSL 4O Gaads \ G0009 oooos OO O ONISSSEIS| MOT v ells TBSLS “8D-IN-@ ¥3dd09-0 NOul OWYY-V _ 7VaaLS dW-o LIN -3NSLv4_NI asvauosq LNADY3d AIWIT =SNSlivsa NI IN3DY3d ASV3SYDNI16.93 21°94 ASVYSAV dQ LNaDYAad- SDNVENGNS OOF OS¢ OO0¢ OSI OO! OS O 7 ' —~E. p QO NAN O¢ fg NES, 5 | Ov |p r | a ADNVLOAdXKA | 3411 NVWOH 09 |v ™ DD S108 O SONIYVAE TIVE 08 o SdNV1 vaZWW SM 7 | | NM OOiv APPENDIX A Gerbers' Parabolic Relationship Having test data from one particular "R" curve itis possible to cal- culate the equation of a parabolic curve passing through the test data points as follows. : ' 1 t ° 195 fo |. OMEAN , From the general equation: *O expressed as Fraction of UTS. y= ax? +b 10 or Range = a*-Opgan2 + b when Range = 0, Oyagy = £1.0 and Range = .26, Cygq, = 0.195 a ° [2] 0.26 = a(.1%)* +b Solving [1] and [2] a = 2705 b = 0.2705 or Range = -.2705 (og)® + .2705 for 107 eycles. This process may be repeated for other test data resulting in a family of curves, By superimposing lines of constant #'R" (R= mn)13 3 the corresponding coordinates of stress may now be taken from the pre- dicted curves and SN curves plotted for other ranges of stress. Greater accuracy will undoubtedly be obtained if more than one set of test data are available. In this respect the equation can be calculated that satisfies more than one set of data or points on the curve.19 APPENDIX B Goodman Diagrams The Goodman Diagram and various modifications of it is the most widely used diagram for determining safe ranges of cyclical stressing. Reference figure 15. This example diagram is constructed for mild steel with an F,,1 of 50,000 p.s.i. The diagram is representative of "par bar" specimens t.e., having no stress concentrators. However the upper and lower safe range limits presented here are for infinite lifes for the material in bending, tension and torsion (not combined). We are not particuarly in- terested in infinite lifes for our design parts however, but may use this source type of diagram for any predetermined finite life. A more convenient form of the diagram is suggested as shown in figure 16, In this particular plot actual test data has been used as an aid in constructing the upper stress limit lives of constant life. Naturally, the larger the amount of test data that is used the more accurate the shape of the curves become. In the absence of more than one test point to plot the constant life lives have to be constructed as straight lives, Potentially the amount of information that can be determined from this type of plot is tremendous. However as in all processes of extra~ polation the designer should use care and good old common sense in judging to what extent he may sketch his determinations.(2) (3) Re.2 Omax, 10% 1o* 10% ~—«d10® No UTS. ot OC x if TEST DATA AT_R=.2 x PLOTTED AT OMAx. st —— 6 ; ONST. RATIO TO MAX. =70, 0 Omin. ; UTS. tF LINES OF , 4 Lo CONE LP ~ Bete Ses - a MORE THAN S10 Sy ONE SET OF DATA Oo INDICATES OTHERWISE (4)50 _UTS. -40 -30. -20 - 7 oa Za 7 ae OEY 53 PS.LxlO Sitastic um’ &//N BENDING | & B = 67 exastic_umtr’ = 30 ee Js TENSION; = S a G 2 ; So A Z / BENDING =) a = - Zz = DIRECT ELASTIC LIMIP i ead badese 8 We 2d “ov n , we OFF TORSION - A ‘Vi YP ’ 1 Wa i. / Zz c A i fs AO iio 23040 Vf 7 /nemy Ie ALTER . STRESS a SI ALTERNATING Hes | Ll Zo 40 RESS PSI.x 10% FIG. 15 50 02°6120 APPENDIX C It is the purpose of this section to show how fatigue test data may be expressed by mathematical equations. Also, it will be shown that this pro- cedure is useful in relating the curves for a family of data involving several wR"(1) values, Further the method is presented in the hope that it might assist those who seek to develop mathematical analyses for the phenomenon of fatigue. A vast amount of fatigue test data may be closely matched by a curve whose general equation is: N= No: [TAN(S2=8 )3 Suv S_! 3 where N = Number of cycles No Constant Ss Stress, psi Su Ultimate tensile strength of the material, psi Se Endurance limit stress, psi h = Constant ‘This equation is particularly suited for fitting data from the aluminum alloys such as 24ST and 75ST as will be demonstrated, Figure 17 shows a representative set of fatigue data on 24STAL. As the first step in evaluating the constants of the equation, it is useful to fair in a curve which matches the data as close as possible and which approximately establishes the endurance limit, The values of N, can now be obtained from the plot. No, is the number of cycles at the point where the curve cross the stress The remaining constant, h, is evaluated through the use of the following equation: Ni LOG No LOG - TAN(Se=S+) 1 ($=3)3 "R" ratio of minimum to maximum stress ina cycle. Ratio held con- stant for set of several specimens where Ny and S, are the coordinates of any point on the curve other than at N,. It will be seen that the mathematical curve is constrained to approach the ultimate stress at a small number of cycles and the endurance limit stress ata larger number of cycles. Also, the curve will phas through the points whose coordinates are N,, Su+Se za1 and Ny, Sj. Therefore it is usually best to choose Ny, Sy in the region of the "knee of the curve. Quite often a value of Ny = 10 N, is very con- venient since log 10 = Once the constants have been established, it is useful to write the equation in the form 1 Su~ Se -I7N.\h (eS ) TAN Ge and to solve it in tabular form as illustrated in Table 1, Substituting 90° for makes it possible to use common trigonometric tables in the computation, S=S ‘The mathematical method may be used as an aid in developing a family of curves, In figure 18 is shown data on 75STAL for five different R values. It is reasonable to assume the variables of these data are related in some manner, which could be plotted as a smooth curve. Following this line of thought, an equation for each R curve was derived, The constants obtained are Usted in Table 2. This set of values for h was plotted as shown in figure 19, and a curve was faired through the points to obtain corrected and related values. The values of N, were also arranged to plot on a smooth curve as is shown in figure 18, TheSe revised constants are listed in Table 2 along with slight changes inS,, which aided in matching the data, The curves drawn on figure 18 are the fial results of this process. It is felt that these curves ex- press the family of data more accurately than had each R curve been considered separately. A word might be said about the accuracy with which the mathematical curves fit the test data, It will be noted that while the points are generally matched verly closely at medium and low stress values, some discrepancy occurs at the high stress values, It is believed that part of this deviation is due to practical difficulty in maintaining the proper test load at these levels, and that part is due to a more accelerated increase in the overall ultimate tensile strength of the material from work-hardening in the plastic range. Tests have indicated that after a few cycles of extreme high stress, 24ST will support a load in excess of the normal ultimate tensile strength. Figure 16 shows the data of figure 18 plotted in another form, In our experience, this diagram is one of the most useful ways of presenting a family of fatigue data. From it can be obtained curves for any R value, or for any constant mean stress. It will be found especially useful in aircraft wing fatigue analysis based on the theory of Cumulative Damage. Figure 20 shows the agreement of some basic data on unloaded hole specimens to the calculated equations for the curves. To date, it has been possible to fit all fatigue data that has been attempted with the trigonometric tangent equation, Further, this equation has been applicable where the well known logS = k.logN or hyperbolic (S-k,) (N-k2) = K equations have failed to fit the data throughout the whole range of stress levels, see figure 7 for example.22 TABLE I - CONVENIENT FORM FOR SOLVING EQUATION GENERAL EQUATION: § (8, - 5.) taal whe % N SPECIFIC EQUATION: S = 69,000 - [ cco.0e 20, 500) waa! Gos, 000) »] oo 1 ae = 1/b lay N (N/N,) (N/N,) 1" ta qo dB S,-S, at nh S fa 5 90 Ny 10 00953 +. 00953 0.5° 300 68, 700 3x10° 0286 1.6 900 68, 100 rot +0953 ete, 5.4 2900 66, 100 3x10* +286 16.0 8600 60, 400 10° 953 43.0 23500 45,500 3xc10> 2, 860 70.7 38100 30, 900 10° 9.530 84,0 45300 23, 700 : 107 95.300 89.4 48100 20, 900 TABLE Il - TABULATION OF CONSTANTS FOR FAMILY OF 75STAL CURVES FIRST APPROXIMATIONS FINAL VALUES R Ny 5, b bh +75 2,5x10° 54,500. 7782.30 53, 000 .778 +50 69x10 29,000 1,015 = 7,0x10 28,700 1.050 +20 3,6x10 20,500 1,225 = 3, 6x10* ~— 20,0001, 225. --20. 19x10 15,500 1,340 = L.gxt0* = 15,000 1.340 1.00 9, 35e103 11,500 1,510 9.0107 = 11,000 1.51022.10 O1* ‘I'Sd - SSBYLS WOWINIW do SANIT nw o of op, OO ol i] 201 OF Pp! x OS 2 Z x ou loz = 4p iNwisNo> 40 SaNi1] § ad ; of Xe | ¥ oe 4 $312A234I7_ LNVLSNOD 40 S3NIT~ ; \ a aN orks 4 ‘> =a] oy * «KI _I/ 4 ' h's¢ 00699 «dA NJ 7 \ “Te a ‘184 006%Le'Sin J f ZI log = 4 * eae IN 5 # Sh axl” N =e —T IN SO oL| +4 = 0L = se ale = qi s > \ 2 Nhe y 1 Yo LNVLSNOD / f (990°) LSaHS “WwisSl do SaILWad0ud HIONSYIS ANSILVS “9I'91422.20 BYwNMWd OL SATDAD 2h 0} 50! A 01 TT Ty fet —~ lo IDB A Nao Xe y ‘O00¢o1 4 - of N tt Gongy ~=S NI oN. 4 NolLwioa — —— Wid DISYE 0 &N byy O2'+ =u (igo) 133HS “Wisye wod “WVEsyIC -NS t | Ly "9ld oe SSaals WNWIXvW Sa = «Ol *Sans ol 201 22.30 SL =y Nolynba —— — VINE DISVd -avv080 dgaLON-y (yao) LaaHS TW1ssl wos SWVYOVIC-NS 81 ‘914 OL SATDAD ] 701 “SSIULS "XVWW > SsaaLs ‘NIW =y SBawn> ol- anv t-=y Bos Gasp saain9 B.LITI0 }d> COG] LW ATIVIXY aBInddy avoT NIVED SSOUD GACvVOT SNAWIDads adm | Nollj3as tsa, HAONBaLS O73), “Isa 00699 Of 0) = a ssayis WNAhWixyw22.40 Fig. 19 Pugt oF EXPONENT nh} versus R VALUE FROM DATA OF FIGURE 2. 8 -6 -4 -2 o +2 +4 46 48 +10 Re ee stRess. MIN, STRESS_ MAX. STRESS22.50 10 0 01 AYN OL SFIDAD } ol weil resoeens “uey]ogo9 ='N oor ly ZEN [9 6 iscsi Uue3 Joood ='N cosa zailee ‘S-oogzs Wt )ocer =! ors owl 6 "S-0o9ls wey} 00019 = 'N 5 exalt ears veyfomooce='N A ~wive Isai Weed LWwANDW> Noliwney + aaun> [02 fc Ss a m a oe ® t x be ov “wine Jout+ © x Ysa ons 8 3s GL42e og 4 Notuwniby — —— “daLvoot IAVWYLNID BIH “VIC 109 TWNAC EE sN HLIM La3HS 790° TWLStZ Od SWVYOVIT -NS ot ‘ola ou23 Chapter 2 STRESS CONCENTRATION FACTORS IN FATIGUE AND OBSERVATIONS FROM FATIGUE FRACTURED SURFACES ‘Appearances to the MIBD are of four hinds. things either are See tity appear tober or. they ne ther sre, nor appear fo bez tr ty dres"and do nob sppaar’s0 bas or they are aoty and ped Spr carbs. ighely to Gn th oll hase ches ta th ise ania tosh = Spicteeus 1, Stress or strain concentration, as the name implies, is the condition that exists at a discontinuity in an otherwise uniformly strained member. A discontinuity may be present in many forms such as: mechanical holes, notches and change of cross-section, The metallurgists tell us it may also exist in the form of inclusions and foreign particles in the metal and finer yet, we have evidenced that it may occur as a discontinuous molecular arrange- ment within the crystals or grains of the metal, In evaluating the effect of a discontinuity or stress raiser, we must xemember that there is no part of any designed structure that is entirely free of stress concentrators, Under repeated loading, a fatigue failure will even- tually initiate at the point of maximum stress concentration unless the con- centration is mitigated or the working stresses decreased below the safe range in fatigue. It is possible to calculate mathematically the stress concentration factor at the edge of a hole; it is also possible to measure the strain magni- fication by very short gauge length extensometers, photogrid and photo-elastic methods. However, the picture is not quite as simple as this, For instance, the stress concentration at the edge of a hole in the elastic range is approxi- mately 3.0, However, 3 ELASTIC I~ DISTRIBUTION k-ONoM. = OMax, he} —euasric24 the concentration at the edge of this same hole in the plastic region is some- what less, say 1.2, Plastic yielding of the material in this region redis~ tributes the straining in such a manner that the peak is reduced. If we now put a loaded pin or bolt through this same hole we have an entirely different set of stress concentration factors, In the elastic range, the factor can go as high as 10 or more ~ plastically it is again reduced to possibly 2, To further complicate the picture, we have a stress concentration factor in fatigue. This value, however, is different than either the elastic or plastic concentration factors under static stressing. As a matter of fact, itis a constantly changing factor. Figure 21 shows this phenomenon, 2, In the usual sense the concentration factor in fatigue or fatigue strength reduction facor (K,) is considered as single valued since most calculations are based on the endurance limits of notched specimens to plain specimens. End. Limit Notched End. Limit Plain For the time being, we will assume this is our condition; although we mentioned earlier that in aircraft design we are more often interested in the fatigue strengths of members in the high stress-low life region. It has long been observed that the theoretical (calculated) stress concentration factors (K) are greater than fatigue tests indicate (K > K,). Recently, investigations and studies have been made to explain this discrepancy - so far without much success, The studies have been aimed at the effect of the grain size of the material since it has been established that coarser grain materials will accelerate localized yielding and thus lessen the factor of stress concentra- tion. Briefly, attempts have been made to correlate the macroscopic and microscopic stress states that exist and the term "notch sensitivity index" has been devised to incorporate this effect. Ky - 1 mo -Q notch sensitivity index = fatigue failure stress concentration factor (actual by test) x theoretical (calculated) factor Fortunately, the value of (K,) is generally smaller than (K) and in the absence of substantial test data our calculations would be conservative, This discussion, as previously mentioned, assumes only one value of K, for our particular design but experimental evidence shows us otherwise. For the present we are obliged to use this value in the absence of test data, However, when data exists it should be possible to use a lesser value as obtained by test for our chosen life under repeated stressing.25 From all this we can see that the problem of stress concentration factors in design is going to be a complicated one. The various factors in- volved may be for one design in one particular material but they do not necessarily hold true for different materials. The term "Notch Sensitivity" hhas been used both statically and in fatigue to compare one material with another. For instance, the same geometric notch produces less notch sensi- tivity in cast iron than it does in steel. In the aluminum alloys, for instance, 24ST is more notch sensitive statically yet the reverse holds true for the two materials under fatigue loading. Such a condition can be represented by the following graph. From this graph we can also see that just a few points in our fatigue tests might be misleading if the more complete curve were not established - since the curves actually cross each other, An interesting point in this respect is to imagine a built-up structural member fabricated from two materials possessing this relasionhip with each other. It could be conceived that atthe higher stress levels some components of our structure would fail in the one material while at the lower stress levels the parts first to fail would be in the other material, In a sense this is ex- actly the condition that occurs. Fatigue cracking of structural test com- ponents will often fail in a different manner and in a different location de- pending upon the stress level to which they are tested, Similarly fatigue fail- ures will not occur in the same place as a static ultimate fracture. Right about now might be a good place to pause and review our under- standing of the whole situation, ‘The use of tabulated or calculated stress concentration factors are of some use in designing for fatigue but at the present time the designer is cautioned as to their-frequent use. It is emphasized that calculations for absolute values is not possible. In this respect a calculated (S, C.F.) stress concentration factor in conjunction with an SN curve of the basic material of the structure should not be used in designing a specific part for a predetermined life but rather an improvement in life on a qualitative basis. Otherwise the predicted results obtained in this mann.r may often be misleading and result in a costly project.26 3. If we cannot readily use the classical equations * for stress concentra- tion factors then what are we to base our decisions upon? Qualitatively the designer can at least evaluate his design from geomet- trically similar notches, fillet radii and holes for which there are numerous references and actual test data in the literature, Such a diagram is shown in figures 22 and 23, Common sense, in many cases, and a ca rtoon-type pictorialization of the notch, will often keep the designer out of serious trouble, FATIGUE CRACK OLD IMPROVED UR JgR ¥* See H. Neuber, "Theory of Notch Stresses", 1946, also see "Stress & Strength of Manufactured Parts", Lipson, Noll and Clock.27 Pictorially it can be visualized that the smoother or more evenly dis- tributed the "stress lines" become, the less concentration and "piling" up occurs, which results ina lower peaked s tressed member and, hence, greater fatigue resistance. Once the crack occurs, however, this distribution is changed even on the smoothest and most gradually changing cross-sections, Pes Fatigue crack "A" corresponds to the path of least resistance and is usually normal to the field of stress flow lines, Fatigue cracks progress along this path. As the crack grows there is a constant redistribution of stress flow lines and a denser concentrations of them at the advancing root of the crack, This undoubtedly accounts for the increase in the rate of growth of the crack, It should not be implied here that we are ever to make a fatigue strength calculation or analysis based on the inception of a crack as well as during the growth of the crack. So far, we have in essence admitted that no designed part is free of stress concentrators; this horrible state of affairs exists and we have to live with it, We generally agree that the combined use of stress concentration factors and fatigue properties of the basic material are a means of making a rough prediction but it should not be used unless it is the only means avail- able - even then our calculations should only be considered on a qualitative basis as a percentage improvement over our original design. A typical ex- ample of a fatique* strength analysis based on stress concentration factors is given in Appendix D, 5. The appearance of fatigue fractured surfaces is a subject that is not only interesting but somewhat revealing. In the past 10 years more emphasise has been put into the study of service fatigue fractures. A knowledge of the characteristic features of various kinds of fractured surfaces has often aided in the planning of fatigue testing program and has often helped the designer in judging approximately the percentage of improvement he needs. See figure 24. The subject, however, is by no means so well established that hard and fast rules have been set down, Experience has shown that serious mistakes have been made. An otherwise obvious appearing fatigue fractured surface may possibly have turned out to be cracks developed during the heating and cooling of a forging. 5.1 A few of the simpler observations that have been made and that, in general, hold true may be briefly pictorialized as follows: * A "fatique"' is a fatigue analysis based on gestimations** and is highly subject to disintegration. r An estimated guess.28 If the flexure of a bar is unilateral and the repeated applications of the flexure are defined in the same sense, the fracture usually appears as shown in figure A. The propagation of the fatigue crack seems to originate at a point "A" forming arcs of a circle until it reaches the N.A, and static rupture soon follows. LOAD t (UNIDIRECTIONAL) 5.2 If the loading is bilateral the final (static) rupture is evidenced as a concave meniscus in the location of the middle of the section, LOADING Ht. : oA SZ) SS cs 5.3 In bilateral flexure of bars having circular notches with sharp profiles, the zone of the final fracture forms a convex meniscus as shown, A { — (~—==—NS AD ——\ REVERSAL Cstatic = i SEY A5.4 If, on the other hand, the notch is radiused and not so severe at the root the fractures are of different characteristics and are classified according to the material condition, on 5 2 t DUCTILE MATL BRITTLE 5.5 Characteristics of fatigue fractures for bars in rotating bending: ‘The propagation of the cracks along circular ares is very nearly the same as seen in bilateral flexure. The percentage of frac- tured area due to fatigue loads and final static rupture, however, is a qualitative measure of the magnitude of the load, The following figure shows the faces of the fractured specimens under heavy and light overloading. SPECIMENS WITHOUT NOTCHES oO. HEAVY LIGHT OVERLOADING OVERLOADING30 5.6 On the other hand, bars having circular notches and tested in rotating bending show slightly different fatigue fracture charac- teristics according to the degree of loading and the type of material, Nd eee bucTiLe art ROTATING ie tee WITH See RAPID HI-SPEED OVERLOADING 5.7 A further refinement can be observed on bars with high or low stress concentration and as to a high or low overloading as shown below, LARGE FILLET LO, STRESS CONC. LOW HIGH SMALL_ FILLET STRESSING STRESSING HL_STRESS CONC. | 6 Low HIGH STRESSING STRESSING31 5.8 For bars with internal holes under rotating bending the fracture 5.9 tae Da + propagates from within the bar at the edge of the hole as shown, a3 @ Cie 3 ms An Under axial loading we find the fracture characteristics to follow, in general, in the same manner as under bending loads. In square bars, for instance, the nucleus of the fracture is usually found near or at a corner of the bar. In rare cases, the fracture may be found to start away from the corner as in the case of a flat bar bent and re-straightened such that high residual tension. stresses occurred on the surface. + - HIGH-RESIDUAL_TENS. STRESS ON THIS FACE It might be well to state at this time that the fracture charac- teristics that have been discussed are for service failed parts and not necessarily for laboratory specimens. Laboratory speci- men failures are in general similar to service failed parts, but being continuous and accelerated tests they lack some of the each marks" observed on service fractured faces. The start- ing and stopping of service loaded parts; that is, the intermittent service loading and then period of resting, seems to account for more pronounced "beach marks", Oxidation and discoloration is another factor that makes service and laboratory failures slightly different in appearance.32 5.10 Itis not unusual to have many fatigue cracks all occurring and growing at the same time ina part. Usually these finer cracks will grow in size until they finally interconnect during the later stages of the loading cycles that ultimately statically fracture the part. Z “RATCHET MARKS” ‘The term "Ratchet Marks" is usually applied, however, to the interconnection of the finer fatigue cracks externally, whereas the term "Hackle Marks" describes this same fracture inter~ connection internally. BEACH OR RIB MARKS: “HACKLE MARKS’ An example of these markings is shown in the following photo- graph. The rib and hackle marks are clearly defined on this magnesium impellor subject to combined stressing.33 5.11 Normally, fatigue fractures originate at the surface of a part. However, there,are exceptions to this rule and fractures have been observed to originate beneath the surface, An example of this may be observed in a case carbWwized or nitrided part, ‘The outer layer being so much stronger than the core material and having a greater fatigue resistance, the fracture therefore initiates in the layer between the core and the outer skin. “FISH EYE’ Similar "fish eye" fatigue fractures may also originate in parts wherein the residual tension stress is greatest beneath the sur~ face of the part. 5.12 Characteristics of Fatigue Fractures due to Torsional Loads. Fatigue fractures of solid bars under reversed torsion generally follow a 45° fracture plane. The fracture of tubes, however, are usually divided into two types depending upon the magnitude of the loading. If the loads are below the elastic Limit of the material, the nature of the fissures usually conform to a 45° fracture plane with respect to the main axis of the tube. If the loads are above the elastic Limit, the fissures are usually orthogonal. Shafts and tubes having transverse holes loaded in cylical torsion usually fail at the edge of the hole, and the failure is seen as a crack starting from the edge of the hole; the growth of which pro- gresses ina 45° angle with the main axis of the shaft. Many times one finds two such cracks progressing in two directions but generally at a 45° angle.34 SUMMARY Stress concentration factors are seldom considered indesign, In the past the critical statr that have designed ‘were only used in conjunction with a fitting factor 9! 725 and laboratory tests on these parts confirmed the analystsy this factor was sufficient for ultimate strengt % have pointed out that plastic ytelding has miti- gated the effect of our stress raiser. However, for many members the critical conditions for which they are analyzed are not based on their maximum load carrying capacity; but rather their capacity to carry frequently applied working loads. These are much less than their ultimate strengths but are repeated thousands of times. It is this condition, the fatigue of structures, that is rapidly becoming as im- portant and as critical in our design as the design ultimate static strength, The time will come when general fatigue design requirements will be imposed on our structures, Most designers accept this; the only difference in their opinions is as to when this time will come. At present? Nol Paradoxically, we are now at the d evelopment stage where we know enough about the subject to realize we know next to nothing about it. (What would be the enjoyment of making a research if it proved anything but the need for further research?) We are certainly in no position to set down any specific requirements our- selves or to help anybody else do it, In this chapter, we have considered stress concentration, its effect on the fatigue strength of structures, and the means of incorporating it in our design, We have stated that the stress concentration calculations based on the geometry of our design is not an altogether valid approach; yet have given examples that appear to confirm their usefulness. However, for every fatigue life calculation, made by these methods, that is later proven by ex- periment, we can find one that doesn't work. The acid test for any of our designs is its proof by a carefully planned and conducted test and the methods of stress concentration calculations should only be used to plan the test. The study of fatigue fractured surfaces may also be used in the design of our part but nothing absolute should be derived from our observations. If the impression of hopelessness has been given with all this discussion the implication has been an erroneous one. The apparent discouragement has been aimed at those of you who may believe thatan equation, a graph, a little incense and crystal ball are all we need to make fantastic predictions with which to impress our fellow designers. The main objective in these notes is an honest evaluation of the present "State of the Art", The greater the number of designers that become familiar with this subject, ‘the shorter the time will be when unknowns become knowns.35 no — noo APPENDIX 5 Example 1 Let the tension lug in figure 23 have a width H = 1.75", D = 0.81" and t= 0,83", If the repated tensile load P as applied by a pin or bolt varies from 2300 to 11,700 Ibs., what would be the fatigue life for this 75ST plate lug". To get the stress concentration factor use 0.81 1.75 D/# = = 0,464 From figure 23 the value of K = 2.6, The maximum direct stress is from P/A = 15,000 psi., and the peak stress is 15,000 x 2.60 = 39,000 psi. En- tering the SN curve (figure 25) for this material and for this stress cycle ratio 2300 11, 700 cy = .2) at the maximum peak stress we could predict 500,000 cycles of loading for this tension lug. (Later tests indicated an average of 700, 000 cycles for this lug.) Example 2: In the previous example we have considered a symmetrically loaded tension lug. Let us now consider an unsymmetrically loaded lug as shown in the sketch below. This condition involves a three-dimensional stress distribution which becomes an important consideration when the thickness of the lug is appreciable. A common method of calculating this distribution is based on a triangular distribution of load thru the thickness of the lug. From this distribution and the geometry of the lug the stress calculation is made on a direct stress plus a bending stress basis as sown. 2P Ne36 For this same lug in example 1 except for an unsymmetrical loading we have: = Py + Me = TO, (12,700) (-14) (-415) < 30, 100 prsvi, 1 0.78 045 ‘The peak stress therefore is K = (2.6) (30,100) = 78,000 p.s.i. Using the SN curve for this material as in example 1 we could predict, on the average, 8,000 cycles for this part. (Later tests indicated lifes of 6, 100 to 6, 700 cycles.)36.10 SATDAD Ol gO SIOH_ ds0voIND SNWDAdS ALVId Lsbe TWLSbe NI_SLNIOF dV1 G3AL3aAld || sOl le"ls pOl ——_L » = FOS SNSIWA =o — Ww36.20 OF 02, 03 04 5 + FIG. 22 G —-SMOOTH SPECIMENS— [7% ropa ROOTH Seve! = l2 Oo 4 d- DIAMETER - IN.36.30 1 PIN LOADED H OLES Ol O02 03 O04 O5 O6 D H 3 FIG. 2.NOM. STRESS END. STRESS or 36.40 oO Ol O2 O38 O4 O5 d_. DISP OF RUPT. AREA Ro RADIUS FIG. 2436.50 War O~ Mey ol Oz > ee /' Or Ov OS LS 5S (eu) {SM- 3IDAD NI SS3YLS LOaYIG “XVW37 APPENDIX 6 In the discussions about the effect of mechanically formed stress con- centrators on fatigue strength we have neglected to mention the greatest of all stress concentrations, a fatigue crack, and its effect on the static strength of our member. Stress concentration factors of 1000 or more may be calculated for this sharp a notch but, as we have mentioned before, plastic yielding at the root of the notch reduces the effect, A series of tests which may be of some interest to the designer is given in figure 26, From this chart he should realize the importance of considering fatigue in the de- sign of structures, particularly if they are the main load-carrying members. In secondary structures, the presence of fatigue cracks is naturally not de- sirable but their existance should not be so alarming. There is usually sufficient time to accomplish the necessary rework since the crack growth period is slow. Figure 26: (a) A single hole drilled in the edge of the member reduced its strength 12% (the order of magnitude of our fitting factors). (b) A very fine saw cut reduced the strength to an even greater amount, However, (c) A fatigue crack amounting to only 14% of the total cross-sectional area reduced the load carrying capacity of the member over 330%,37.4 AXIAL LOAD TESTS , | -—4F STATIC SPECIME! Remarks | FAILING | Kr | PLAIN I6,000Les, OO . | 4 RA, EOE 144,000 | I.l2 DIA. HOLE WITH SAW CUT FROM 134,500 pols (© ECs 19 CUT _WITH (mm JEWELERS =| 67250 | 239 E SAW STRESSED + {TO 16300 PSI [ain FoR RS009~| 87000 | 1.85 yD [CRACK Big Sass 28) , |TO 16300 PSI ee CRACK INTO | 48,250 | 334 iD’ BODY OF CAP38 APPENDIX 7 Example 1: To incorporate the first and second weeks notes, let us assume the following (and rather simple) design problem confronts us: What is the fatigue life of 24ST6 clad sheet containing 3/16" holes and being stressed from 25, 000 psi tension to 5,000 psi compression? Let us also assume that the only data available is an SN curve for the plain unnotched material atanR value of + .2. Solution: In using the SN curve of figure 17, for the plain material ata stress ratio of R = 40.2, the first step is to plot the data in the form of a Goodman diagram. By this method it is possible to determine the SN curve for other "RN ratios and for the one described in the example problem, Having this data and therefore calculated SN curve; the next step is the calculation of the stress concentration factor for the holed shect (Ref. figure 23). This factor times the nominal maximum stress (25,000 psi) is then the stress level for which to enter the newly calculated R = - .2 SN curve. The D/H ratio for our holed 24ST cl. sheet is 0, 187 0, 187 1,00 and from figure 23 the value of K is 2.5, Peak stress calc. = KO = (2.5) (25,000) = 62,500 psi and the life in cycles at this maximum stress level is approxi- mately 20,000 cycles (from the above determined SN curve ment e082)" Figure'20 shows the life for this part to be 40,000 cycles as obtained by test, At first hand this discrepancy might appear to be great. However, the calculated life is on the conservative side of our estimate and as many of you may realize, a scatter in life of 2 to 1 is exceptionally good agreement. (Ref, figure 10 on riveted joints. ) By using a more realistic value of (K) - as obtained by test. See figure 21, Peak stress calc. © = (2.2) (25,000) = 55,000 psi and the life in cycles at this maximum stress level is approxi- mately 30,000 cycles. (From figure 17) Figure 20 shows the life for this part, as above, to be 40,000 cycles as obtained by test.APPENDIX 8 The following SN curves (figures 27 to 32) are self-explanatory and show the effect of stress concentration on the fatigue strengths of some materials in various forms.Be002, 39.1 FATIGUE OF J5STALC SHEET 2) MAX, STRESS, = lb 60777 (LoADED_¢ UNLOADED HOLES ). | : alae | 0 th “ reer! g a ko SS LED HeLes M0 NGG ISP| § ! INK) 50 Se I | a hie ML ko SR tt mF UA Ne | a oles 4 Seat J = aa ae \ voles 70, faltuRe tN FIG. 27 SHOT PEENED AND PLAIN. 75ST BARS WITH & WEHOUT 079 DAHOLE (Ramoons er rye ‘Srcomes) II iL. II H+ | 135-145 | Woke 0 wo EXCLES To FALURE FIG.2839.2 x TNC 2487 & 75ST EXTRUSIONS AXIAL LOAD TESTS i ll ‘STD, POLISHED j Srecens | || s i rasticee co! STRESSED. Te MINAL. YIELD nn a TENSION STRESS - 1000 PSI. & ‘L Hl 7 LLL 498. EYCLES WS FAILURE FIG.29 FIG.30 _ 24ST vs 75ST EXTRUDED | MATERIAL (AXIALLY | Ht LOADED) Cd NY STANDARD, NSSeEeMens | rf hee Px TENSION STRESS - 1000 Psi NoToHeD 248 rt NOTCHED 755 LLL. ips Ue to! eXcLes To FAILURE. LVRete 60) 8 30 WK, TENSION STRES-KSI é 8 215-066 eee coat, Par { a STALE g 0 “ASTALEDN te Snes | : Tarte cea concen tanta coe ‘ S sn t 2a - TENSION STRESS S_, AXIAL LOAD TESTS ON 75STALC, 24STALC™ SHEET Un CIRCULAR HOLE. R= ine + | {il opstale _ | #Hstate CXCLES__ TO FAILURE FIG.31 FATIGUE, OF 75$T COUPONS SWMULATING —$TA. 187 (KE) WING SPLICE PLATE RADIUS. 39.3 1 cveces "To ranure | FIG.3239.4 ert paHio ssa i 5 [EN Ol Oz Or Ov OS40 Chapter 3 CORROSION FATIGUE, ELEVATED TEMPERATURE AND SURFACE CONDITION EFFECTS Groat fleas haye Little fleas upon their docks to bite en, and Title fleas have lesser fleas, and'so od infinitu. = De Norgan ‘The last section of notes was concerned with the surface condition of structural members ina broad sense. Notches, holes and changes of shape may be considered surface boundary conditions that unfavorably alter the stress distribution in our members and thus account for reductions in fatigue strength, Ona slightly smaller macroscopic scale corrosion pits, surface material irregularities, and the finest of machining tool marks are also surface conditions of stress concentration that change fatigue properties. It may seem that an undue amount of time is being spent on some of the finer details of the subject and not enough time on the more general rules of designing for fatigue. Itis felt however, that this is the crux of the whole subject - an awareness and understanding of these finer details. Since most, if not all, the fatigue failures that concern us are surface* phenomenon, it will be well to briefly discuss the surface condition of our parts. It is with this knowledge that we are able to improve our design in fatigue. We will discuss some of these effects in the order of a more or less inverse importance to us as aircraft designers. 1, Corrosion 2, Elevated Temperature 3. Surface Finish, 1, Corrosion ‘The subject of corrosion-fatigue may be subdivided into (a) corrosion of the surface before repeated stressing; i.e., corrosion precedes the ser- vice loading and (b) the simultaneous effect of corrosion and repeated stres- sing; i.e., corrosion accompanies the service loading. Primarily corrosion will roughen the surface of our part or structural member. Whether this roughening be accomplished by chemical means or mechanically formed notches the results are similar: that is, the fatigue resistance is decreased, A typical example of this effect occurred in the shop a few years ago as a result of a salvage procedure. Some sheets of 75STAL had become discolored during heat treatment and to remove the dis- coloration the sheets were immersed in an acid bath, The sheets were sup- posedly washed, then placed in a rack to dry. However, in contact with the metal rack, there occurred an electrolytic action which pitted the 75STAL. * Subcutancous fractures are the exception to the rule and are usually brought about by high subsurface residual stresses.41 Inspection revealed this condition which then resulted in a spot laboratory test of the pitted material, Fortunately so, for the corroded material fell far below the uncorroded material in the fatigue test. The indicated stress concentration factor for this condition was from 1,3 to 1.4, Need- less to say, these sheets were not used onan airplane, it should aot be implied that a part that has become corroded is always rejected on this basis alone before it is put into service, Very often the effect is so slight or the removal of the corroded surface so simple that the part may be salvaged, It may be inferred that a corroded part whose corrosive environment is once arrested may then have a definite ¢ stablished fatigue limit or fatigue strength, Contrary to this, a part that is in service and subject to repeated stressing as well as being in a continuous corrosive environment has NO en- durance limit, 1,1 To complicate the effect further, the corrosion-fatigue strength is a function of time rather than the rate of repeated stressing. (For example, the part should not be defined as so many cycles to failure at a given stress excursion unless the corrosive media and rate of loading are also defined. Slow speed tests give much lower fatigue-strength results in cycles to failure.) See figure 33, 1,2 Minimizing of corrosion fatigue can sometimes be accomplished by the substitution of material: ¢.g., stainless steels for car- bon steels, Cladding and surface coating materials are also often used. However, coatings that do not give electrolytic pro- tection must be free of cracks and pin holes, otherwise subcoating corrosion may occur. Fortunately we are not too concerned with this phenomenon in aircraft designing. Submergence oFexposure to salt water is not our particular problem, Exhaust gases and general weather conditions are about the most damaging media and their effects are largely reduced by the careful cleaning and polishing procedures of the airline operators in their esthetic feeling for their property. Some protection coatings worthy of mentioning are hot galvanizing, electrolytic galvanizing, cadmium and chromium plating as well as synthetic resins, There are also surface coatings derived from chemical action that are used, such as nitriding and anodizing. In summary, we have determined the following: 1.3. Corrosion that precedes the service loading of our members has a small effect on the fatigue strength; and is comparable to the effects of surface irregularities such as holes and notches. We may further classify this type of corrosion into further subdivisions as to types that (a) may attack the surface of the part in an evenly distributed (v) may attack certain points locally on the part, (c) intercrystalline corrosion originating at the surface for some materials only - particularly for those materials where the fatigue cracking in itself is always intercrystalline, This is not in general true for the majority of materials.1.4 15 1.6 az Corrosion that accompanies service loading of parts has an enormous effect on their fatigue strengths. Also, this effect is further accelerated with an increase in temperature. However, for the corrosive environments that normally surround our par~ ticular structure, we have little evidence that it is a major problem in aircraft designing. Mechanism of Corrosion Fatigue The mechanism of the action of the corrosive environment on metals during the simultaneous action of repeated stressing is very complex and as yet not completely understood, A simple explanation might be pictured as follows: The cyclical stressing cracks the protective film or coating on the surface of a metal and allows the corrosive media to produce pits. These pits, not necessarily corrosion pits in the normal sense, may take the form of small intercrystalline cracks. As cracks, there is a considerable increase in local stresses and this condition strength- ens the attack of the corrosive agent and causes a more rapid des- truction of the protective film, This process continues at a rapid pace until the stress at the roots of the corroded fissures exceeds the fatigue limit of the metal and ultimately failure occurs as in the ordinary way but usually in a much shorter time, The principle of the problem of corrosion fatigue is therefore in the properties of the protective film which is formed instead of the resistance of the metal to corrosive attack. Corrosion Pitting This form of corrosion is usually seen in gear teeth, or in any form of rubbing and mating parts, as tiny pits on the abrading surfaces, This surface deterioration can develop into fatigue cracking in service. Explanations for this pitting are many, such as overloading producing high compression stresses, breakdown of the oil film which would allow localized overheating, incorrect alignment of the mating parts, inaccuracies in machining which could create dynamic stresses in high speed parts. These pits form only when there is relative movement between two parts and only then in the softer of the two. At the point of contact there is both a compression and an elongation of the surface layer of the metal, It is this alternating stress that eventually may fatigue crack the surface. These cracks are usually seen as arrowheads (sce sketch) and whey they have progressed suffi- ciently result in flaking of the material and are in plan view seen as oval shaped craters,43 —— DIR. ROLLING ~ INITIAL CRACK PLAN VIEW OF CRACK — CORROSION PIT 1.7 Corrosion by Cavitation ‘When metal is in contact with a liquid in motion (ship's propellor) the local eddies formed by cavitation rapidly deteriorate the metal, The process is purely mechanical and no chemical action is to be considered. The constant erosive action of the liquid in motion removes the protective film on the surface of the metal as rapidly as itis formed, Ina short time corrosion effects are observed, brought about by the high stresses that occurred, This progressive destruction of the surface (cracking) sets up notch effects and fatigue progresses once again in a more or less normal fashion. Stainless steels and other corrosion resis- tant metals are of no consequence here because cavitation damage occurs so rapidly that true corrosion does not have time to occur, 1,8 Fretting Corrosion or Chafing Fatigue ‘This type of corrosion is usually evidenced at the bearing surfaces of clamped or bolted together members. It is also found to occur in shafts with pressed on flanges or in shafts at their bearing collars. The rubbing or abrading action of the two members pos- sibly flake off small particles which become oxidized and act as an abrading agent between the two surfaces, Continuation of this process eventually forms notches or grooves in the parts which soon act as stress raisers and may ultimately fail the part in fatigue, again in the usual way. The red powder or dust (Fe,0,) that is usually seen around fatigue cracks in steel parts is the evidence of this chafing action, In the aluminum alloys this powder is usually black.1.9 44 Stress corrosion cracking believed to accompany only residual tension stresses ina part and not residual compression follows the grain boundaries as seen in the photograph below at 100 X. A 75ST aluminum alloy extruded angle fitting. For further study and information on the subject of corrosion fatigue, which we have stated is not at present a problem of major consequence with us in the aircraft industry, reference is made to the work of D, J, McAdam and H. J, Gough, J. Inst. Met, Vol. 49, 56, 60, 74and Proc. Amer. Soc. Test. Mat., Vol. 27, 28. 2, Elevated Temperature Effects In general, the fatigue strength. of metals is decreased with an increase in temperature. The results however are usually so erratic that no attempt should ever be made to extrapolate existing data. 2. Different materials, as well as the same materal ina different form, will yield a wide variety of properties, The most positive determination is a test under the prescribed conditions. At very high temperatures the fatigue properties of most of our structural materials is not a problem - at these high temperatures plastic deformation and yielding of our structures becomes the most cri- tical factor.44a ENDURANCE _ RATIO. @ fo) a ° ° 200. 400 600 800 1000 ° FAHRENHEIT a P PNESAE 2340 NizCR~-Mo.- STEEL . s ° = —= 12% CR- ste ENDURANCE _LIMIT- KS » ° ° 200 100045 The above diagram shows the characteristics of a few of the more common materials. The aluminum alloy appears to con- stantly drop off, while some lower strength steels actually rise in fatigue strength up to approximately 300°F. - then fall off, This phenomenon is also observed with the tensile strengths of these steels. 2.2 Somewhat analagous to corrosion fatigue, a fatigue limit at very high temperatures is not to be found. The SN curve continually descends. The metal is undoubtedly affected by oxidation pro- cesses at these high temperatures and also by the loss in strength. 2.3 The temperature effect may be further complicated by the existence of a superimposed steady state static stress. In this case the fatigueing of the metal is progressing simultaneously with the effects of creep. Creep however plays no part when the loading is purely alternating. A few SN curves for various materials are given in the Appendix at the end of this section of notes. 3, Surface Finish Shallow grooves or notches that do not purposely define the shape in the design of our member, but are rather in the form of tool marks from the machining operations, may have a considerable effect on the fatigue strength. These tool marks cannot always be completely removed; however, careful grinding and polishing operations may be used to reduce a coarser surface finishes and greatly benefit its resistance to fatigue. Whenever possible, machining or turning operations should be made so that the resulting tool marks are in the direction of the loading of the member rather than normal toit, The following tables may be used as a guide and a qualitative indica- tion of the effect of surface finishes on the fatigue strength. TABLE 1 Approx. Reduction in Surface Finish ae Finish Endurance Limit. oove a tee Design*! Percent Turning 001600 450 12 Coarse Filing 000827 225 19 Fine Filing 000394 110 7 Fine Emery 000158 45 3 Fine Grinding 000197 55 4 Polishing ° - ° Peak to Valley 3.5 %No//= Mean Deviation in Millionths of Inches =Surface Finish Polished with 0 and 00 emery cloth Polished with 1, 0, and 00 emery paper and rouge Circumferentially Ground Finely Turned Roughly Turned TABLE Ib UL T.S. Oz 50,200 51,100 & 45,800 + 43,600 42,000 Endurance Limit (Carbon Steels) = 98,000 psi % 100 103 92 87 84 UL TLS, Oz + 26,300 & 24,200 & 24,400 0 46 42,500 psi % 100 92 88 There is very little data on the effect of surface finishes on the fatigue strength of metals. data for his particular design. Ibis very seldom that the designer can find applicable There are too many variables and conditions to account for, such as the type of material, the tensile of the material, etc., to ever expect all cases to be covered and tabulated in the literature, If he cannot afford to conduct an experimental testing program for his set of cri- tical conditions, the best bet is naturally to call out polishing, grinding, and super-finishing.aT Appendix 9 The following assortment of laboratory test data presented in the form of S/N curves in one way or another relates to the surface condition of the metals and its subsequent effect on fatigue strength properties. Figure 34 Direct stress tests of riveted lap joints of heavy plate. These tests in- dicate a 35 ~ 40% improvement in fatigue strength at a temperature of -60°F. over tests at room temperature or up to 120°F, Contrary to some beliefs the colder material should react in a brittle manner and as such should have a greater notch sensitivity; however, tests indicate otherwise. At present, investigations are being conducted to determine the effect of heat cycling on the simultaneous fatigue cycling of materials. Although the data is not available, there is good reason to believe that heat cycling may have some adverse effect on the strength properties. This data would be use~ ful in the designing of turbine blades and jet exhaust rings. Figure 35 The first effect that may be noticed in this graph is that the bare material has a greater fatigue strength than the clad 75ST aluminum alloy. The ex- planation for this being that the softer and weaker cladding material would be the first to develop fatigue cracks which then directly influences the surface of the plate beneath the cladding and accounts for the decreased strength. How- ever, the effect of forming or bending either of the materials sets up high residual stresses which more than offsets any differences in plain clad or bare plate; with the net result that their is very little or no difference now in the formed members. A few test points, however, for the hot bent condition indicate that the forming at temperature has tended to reduce the residual surface layer stress. Figure 36 The effect of removing a thin surface layer of a prior stressed member has the beneficial effect of increasing the fatigue strength of parts. Itis as though we "turned back the clock" and started all over with a practically new part to be put into service. It is rather difficult to illustrate this effect by means of a graph - but a simple explanation along with the S/N curve should suffice, The S/N curve of this figure is for the small radius (0,15) specimens shown in the lower left hand corner. Five specimens were tested for 100,000 cycles ata maximum stress level of 25,000 psi. This life in cycles (from the curve) represented approximately one-half of their normal life. The five specimens were then Zyglo and Dychecked and no fine cracks were found. Specimens 1 and 2 were then re-machined - removing approximately . 0015" from the surface and enlarging the fillet radius to 1", These same specimens were then re-tested at a stress level of 42,500 psi and both failed in approxi- mately 18,000 cycles ~ which at this stress level represents approximately the other half of their life. What does this mean? Pictorially we show in the lower left hand drawing that the damage we imposed by the first loading cycles48 evidently extended beneath the thin surface layer that we removed. The metal has not forgotten this and even by changing the fillet radius the part was no better than if we had not removed any surface material at all, On the other hand, specimens 3, 4 and 5 had , 003 thickness of surface material removed which was evidently sufficient to cut below our hypothetical damage. See upper right hand drawing, These speciments when re-tested went far beyond the S/N curve for the small radius parts and, as a matter of fact, reached the S/N curve for new 1" radius parts (ref, figure 40), The di- mensions and amount of material removal in these tests should not be taken as absolute dimensions to use in design. They happened to work for this experiment, however, a much greater amount would undoubtedly be safer. This same procedure has also been found to be valid even after small cracks have formed, In this case, however, additional material beneath the most advanced crack should be removed. Figure 37 This graph again shows the difference between clad and bare plate (ref. figure 35), However, any benefits that appear to exist are lost when we do something with the materials such as fabricating joints. Figure 38 Similar to the experiment of figure 39. this graph shows the effect of material removal around the most highly stressed region of a part, The upper curve is for "par" specimens and the lower curve for unloaded hole specimens, In this experiment, hole specimens (no such thing as a hole speci- men?) were stressed in 50,000 cycle intervals - each interval followed by a drilling operation removing approximately .015 inches of material, It was possible to accumulate a far greater life for these specimens in this manner, as seen in the graph, The only mistake made in this test was that a few more operations should have been carried out so that the final fracture could have occurred in the basic material away from the hole. Figure 39 An extension of the direct stress experiments of figure 35 only for ex- truded material, From this S/N curve it can be seen that there is little dif- ference between the fatigue properties of 14ST, 24ST or 75ST extrusions, The curve also show the effect of cold bending 75S, hot bending 75S and the beneficial effect of cold stretching 75S. Figure 40 The S/N curves in this graph show the comparison between a homogeneous and porous surface in plate stock, It is believed that the greatest percentage of this difference is due to the porous surface condition of the specimens cut from the core of the large plate although some of the difference may be due to the strength properties of the weaker core material, Caution should be ex- ercised in manufacturing parts that are to be cut from core material and if possible use only the outermost material of a billet or thick plate or bar,49 OK DOUBTFUL 5 CT f 1 1 | 1 { ! USE 9 Figure 41, The effect of rolling threads in bolts is a recent process and has been found to considerably increase their resistance to fatigue. The lower S/N curve in the graph shows the fatigue characteristics of machine threaded bolts - the middle curve shows the improvement when the threads have been rolled - but this procedure should only be performed after heat treatment ofthe material, This procedure has now made the heads of bolts at the radius to the shank more critical than the threads. The latest step to gain addi- tional strength is to now roll the radius at the head of the bolt. Figure 42, The effect of decarburization in the surface layer of steel parts has long been observed to have a very noticeable effect on decreasing the fatigue strength of such parts. The S/N curves of plain and notched specimens in SAE 2340 steel are shown in this figure. For other materials see the refer- ence given on this graph.50 Appendix 10 THE CUMULATIVE DAMAGE THEORY This theory in its present form and the xperimental verification of it was conceived and carried out at the Douglas Aircraft Co. ; although some of the original ideas were taken from a theoretical discussion by B. F, Langer (Ref. Jounr. App. Mech., 1937). The concept of damage in fatigue is not new = and it has long been recognized that damage of some sort takes place. It might be well to discuss the older views even though they are not of much use today, There are some engineers however who still hold to the original concept of the "damage line", The determination of the damage line and a discussion of its faults follows: PRR “XN t+ & OVERSTRESS' NOS ojo a FAILURE "DAMAGE LINE” UNDERSTRESSING After determining the S/N curve for a member; additional specimens are run short of the cycles to failure at some overstress value. The speci- mens are then run at the endurance limit level and if they "ran out" (went 5 to 10 million cycles without failure) then an arrow was pointed towards the failure curve for the speciment (see specimen No, 1), Additional specimens are then tested for greater number of cycles until finally (specimen No. 3) a specimen is found that failed short of "running out!', The arrow on this test is directed towards a shorter life. By this procedure, sets of data will ul- timately define a limiting band, below which Hes a so-called safe region, Somehow this concept just doesn't add up. For surely if we applied stress cycles at many stress levels over and over again; and even though we appeared to stay below the damage line, the accumulative effect of this procedure would ultimately fail our part, It is believed that the apparent damage or no damage at the endurance limit stress level is obscured by the accuracy of stressing as well as the variables within the material itself at this peculiar stress level in nature that appears to approach infinity asymptotically. CUMULATIVE DAMAGE: PART I Part 1 of this Appendix is restricted to stress magnitudes, subjected to specimens, between the yield point stress of the material and the fatigue limit stress of the specimens established at 107 stress cycles,51 Part II will deal with understressing in fatigue and its effect, if any, to produce a lengthening of life, Overstressing, above the yield, will also be investigated to determine the effect of cold working on the fatigue life as well as any changes in the physical properties of the material "damaged" to some extent in fatigue, PartI - (General Theory) It seems probable that the process of fatigue failure in fatigue testing must follow some law in nature in which the "damage" is zero at the start of the fatigue test and 100% at fatigue failure of the specimens. It is assumed that the degree to which the "damage" progresses toward completion is represented by the quantity, A. Therefore, A = 0, or no damage, at the start of the test and A = 1 at 100% damage or failure. ‘These assumptions may also be subdivided into two intervals. (1) From the start of the fatigue test to the formation of a visible crack in the specimen and (2) during the growth of the crack, Due to the size of speciment usually tested, itis assumed A = 1 at complete fatigue fracture since the crack growth period is so short lived that usually it can not be accurately measured on small size laboratory samples. In establishing a means for calculating the "damage" effect of repeated loadig in fatigue the following nomenclature has been set forth, applicable to the conventional SN diagram, and henceforth will be used throughout this dis- cussion. _ ,,2 Known duties, or stress cycles subjected to the speci- 1 Bovveee mens at the corresponding stress level. N, 2,3. Stress levels of varying magnitude Stress cycles to failure, Values obtained from the SN curve at 2 c 2 yele ratio. N 2 Damage caused by 2, 2, 5... 1, 2,3. LL fa 28 = Damage ratio 4 Using the above notations with the aid of the figure below, the calcula- tions for the "damage" effect are presented in the following simplified form,52 E LIMIT_OF STRESS PART In fatigueing a specimen under repeated loads for 7 M' cycles we may assume the specimen has been "damaged" to the extent that one half of its useful life has been expended, Similarily, if a specimen is subjected to ™* SB cycles at 3 , one half of its life has been expended, It seems reasowable that (1), this relationship may be applied to any stress level within the limits outlined in Part I, and, (2) that the same degree of damage to the specimen has been attained in % cycles at 07 as ina greater number of cycles 7%, ata lower stress level Oz . The magnitude of 7, and 772 both being equivalent to 1/2 of their respective number of cycles at failure, N, and Ay. Expressed mathematically, the general relationship at stress level GF aH). Since so little is known as to how the "damage" progresses in fatigue the first assumption to be made is that ( /’) is a linear function, The equations derived from this assumption were then investigated ex- perimentally to prove their validity, Equation (1) may now be expressed as: a Ag pee Remembering that A is equal to 1,0 (one) at total "damage" or failure from fatigue, then: 2a LAG. 0)53 el » the total "damage" ratio = Re 2H A Equating D@; += A = 1,0 at fatigue failure and substituting in (4). =f Expanding equation (5) to stress levels of varying magnitude, the total "damage" or fatigue failure occurs when the following expression has been satisfied. 10 ce FAURE AT OF ......(5) e/g a Mn + ee Ihe NM SUMMARY OF RESULTS OF LABORATORY TESTS ‘The following statements are applicable only to the stress levels out- lined in Part I. The effect of "period of rest" has been neglected, since any differences in fatigue lifes were in-significant, Five to thirty minute "rest periods" were recorded, partially due to the "set-up" time in changing the loading cycle of the specimens from one stress level to another. n ee An average value of 1.03 = ZAy * at failure was obtained for 20 specimens, in calculating the "damage" effect of repeated loads of varying magnitude, (24ST and 75ST aluminum alloy) In plotting the "end points" of specimens subjected to loads of varying magnitude, with the aid of equation (6), failure occurred within the limits of "scatter" of the base S/N curve. ‘The order of load application, as to whether a high stress or low stress was applied first, was of no significance in obtaining an “end point" that fell within the Limits of "scatter" of the S/N curves. 4, Four series of different types of specimens were tested at various load levels to determine the validity of the cumulative damage concept for steel, ™ Values of (N) were taken at both the upper and lower limit of the S/N curve "scatter" band to reduce the probable error in selecting the proper (N) for calculation,54 Nine flat sheet steel specimens failed at an average value of ‘Sn/N =.94 under axial tension loads, Seven rotating beam flashwelded steel tubes gave an average value of 2n/N = 1.11 and four axial loaded flashwelded stecl tubes averaged En/N = 1.13, Thus the overall average was, n/N = 1.04, ‘Two loading methods and rates of load application were also employed to study the validity of the cumulative damage concept for steel specimens, Impact tests and axial load tests on NE4037 steel bolts were made and the damage phenomenon was shown to be comparable for the two types of loading, Some bolts were loaded under repeated impact for a portion of their total life then failed in the axial loading fatigue machine. The value of & n/N for this combined loading was 1.08 for the five bolts tested. ‘The validity of the cumulative damage concept in fatigue as presented here has been further established by later tests conduced at Battelle Memorial Institute, Representatives from Battelle, informed of the concept herein at the NACA Structural Sub~Committee Meeting at Washington, D, C. in November, 1944, conducted similar tests on an aluminum alloy of which one speciment in particular (by correspondence) was subjected to (4) stress levels of known duties, at a constant mean stress, and repeated 28 times before failure occur~ red, The end point or failure of this specimen employing the concept of cumu- lative damage in fatigue was within 1% of the basic SN diagram of the material established by virgin specimens, It is hoped that these tests will add stimulus to the investigation of the cumulative damage concept in fatigue for other materials, It is believed that the cumulative damage concept may become so complicated for some metals that it would be impractical while for others it may definitely not hold true. For example reference the "coaxing" of cast iron specimens by Kommers, The cumulative damage concept may bg fppyd pnlf to be validtfor un- stable metals such as Orowan has defined; %& have'no definite endurance limit in that the fatigue strength is still decreasing beyond 5x 108 cycles. This would indicate, for unstable metals, that there is no value of understress that would produce a lengthening of fatigue life at higher stress levels and would thus greatly simplify the problem, There is, however, a proposed method of evaluating the effect of under- stressing which for some steels, having definite endurance limits, is known to have a strengthening effect, This proposal was suggested by Langer and is outlined in Part IL, PART U The assumption is made that each stress level below the fatigue limit has a corresponding stress level above the fatigue limit and the effect of these two stress levels cancel each other out, If we wish to include in the equation ay ay >F 1.0 or (— + + = N Ny Np N,55 the effect of a stress 0% below the fatigue limit applied % times, we must find the percent increase in life at stress 02 the effect of which can be can- celled by the understressing Oy . Seefigure below. By subtracting the term 2% from the above equation, the lengthening of life produced by the undd*Stressing has been expressed, x a ™” Ny Ne aw woenae ubé (nl) mereared MN, cvCles 3 Ne ; BY My Cves 47 OG; 7 ONDERSTRENS N= No data are available as to which stresses bear reciprocal relationships to each other as for and . In order to apply this theory it would be neces~ sary to have a curve showing overstress as a function of the understress it will cancel, Itis possible to make a good guess as to what such a curve would look like having as little as (2) known test points to plot. KNOWN a UE i Euoyrance Livyr WERSTRESS - Op UNDER STRESS ~ Oo Assumed curve showing overstress as a function of the understress it will cancel. Understressing is not assumed to have much effect until it gets close to the endurance limit, This effect has been observed by other investigators and their results seem to be in close agreement,56 Since this theory was introduced (1944) there have been many investi- gators attempting to enlarge upon and refine the concept. They have notably shown that n/N does not equal exactly 1,0 but something less than 1,0 if low stress is applied first and greater than 1.0 if high stress is applied first, Their tests however have been based on only two different stress levels, (not a very realistic spectrum of loading). This discussion is not necessarily in defriese of the Cumulative Damage concept however, it has been shown experimentally that if all and many stress levels are considered that the end result does approach 1.0 with insignificant errors, Many invesi- gators appear to have lost sight of this aspect of the concept - that more than two stress levels should be considered,56.1 ee°Ols AN3WNOYIAN] _ 3ATSOugOD Ni ONISSAYLS G3LWad3ay dO 153443 SNO3NWLONWIS 35vsuns (aaaduuo5 4olud) div NI Qals3l Yyiv NI Q34S3.(+2) STRESS - 1000 Rs, EFFECT OF TEMPERATURE ON FATIGUE OF RIVETED JOINTS. —7———— ] 1 al 20 10 40° 60} ‘CYCLES TO FAILURE FIG. 34 CLAD AND gE Pet STRAIGHTENED AND UNBENT- AXIAL LOAD FATIGUE TESTS OF ——] \ 75ST PLATE ~ COLD BENT AND PLATE, | Z MAX.TENS. STRESS. - 1000 PSI TDs ‘and % PLATE. Benpng Ratio B+ 8. 40° CYCLES ~ FAILURE FIG.35 56.2(02) ne) MUOV STRESS = KEL 8 Lean Tet EFFECT OF REMOMING SURFACE LAYER AFTER PARTIAL FATIGUE DAMAGE OF 7557 PLATE. a iT {| feces rescore | x 18" 10 FIG. 36 s (000 PEL, y WAL LOAD TESTS ON BARE AND GAD Tish LATE TRB = Gein Hi TEIN Si wy 8 56.356.4 EFFECT OF INCREMENTAL HOLE ENLARGEMENT AT_50,000_CYC\. ITERVALS. res i Da. 40 --[nedtexsts —| 2 Neos ne ¢ Redes i os j | trove, - ‘SAMPLES? 2 30 4 a i Em a E-to 7 t wp ttevetes To rarure |i murions, | TSS.) oot Ot 0 20 FIG, 38 To 1 ereeer oF ronwins Ri] | GW FATIGUE PRapERTIES Reo - OF ISST EXTRUDED 3 ‘4 MATERIAL. 190 Onan Gato aENoNs 3 | Sher fracre, Roy Bee PRCTEY ue 4 i” : 2D alli ei 320 y = || R= oH | lease Bu au ||| 1h Ltrs 1 LL ios FIG. 3956.5 FATIGUE OF 7557 couwons cuT FROM 4 PLATE AND POROUS CORE OF 2h” PLATE. ye i TB) BH GRD, WOEL, WH To" eles To faltuhe FIG.40 ae FIG. 41Fed OOOT = plMAX. TENSION LOAD ON JOINT-«(LES.) 56.7 FIG .43 ite 000 ‘ ae iey\ ; 4 ‘SPEC. NO. 2 es \ aE | FAILED +3000 + | 7 | | 1 a w 10 5 CYCLES TO FAILURE57 Appendix 11 Suppose the following condition existed, A multi-row riveted lap joint in 75ST plate material is to be stressed from 3, 000 to 15,000 psi at a rate of 10 cycles per hour and the amibent temperature for this joint during ser- vice loading is -55°F, What is the service life for this joint? From figure 34 at O71... = 15,000 psi Ngge = 330,000 cycles Nooe = 33,000 hrs, and H, 55 If the original design only required a life of 20,000 hrs. for the joint the margin of safety may be calculated as follows. From figure 34 the 0” for 20,000 hrs. x 10 cycles = 200,000 cycles is read to be 17,000 psi ™2* (lower limit). max, joint _ 17,000 From £5. = max.Q design 15,000 =) ey) and M.S, = £5, -1 = 0,13, Now suppose a change in the original design; in that 1 cycle at a tem- perature of + 110° is interspersed with the 10 cycles at -55° every hour. From 5" = 1,0 we have N Boe Bagge 110" 235° = 10 at failure, By10e Nusse : + i = 0000451 85000 330000 Now mt = joint 52 N so Hine = 22,173 hes.The margin of sat since trail attempts would have to be made to satisty per hour for the 20, 000 hr. criterion in determi at N.ii0° 3 H, i and M.S, 58 ‘ety for this condition would only be an approximation 00005 "damage!" 8 Cmax = 15, 500 psi = 270,000 = 76,000 + oe = . 00008 a 000 * 270,000 1_ = 20,000 hrs, me N = 15500) 15000 0,0359 Chapter 4 EFFECT OF RANGE OF STRESS, COMBINED STRESS AND RESIDUAL STRESS IN FATIGUE Astonbt the end ond never stand to doubts ASG S So hard but seorch will find te out. = Rs Herrick If the first week's notes a brief discussion was made of the effect of range of stress and combined stresses in fatigue. This section of notes includes in more detail the stress range effects; that is, varying amounts of steady state stresses superimposed on cyclically varying stresses, Em- pirical equations are also given for computing endurance limits and fatigue strengths for any range of stress. 1. Range of Stres There are two methods of describing the range of stress. (1) The range may be designated as the magnitude in the change of stress from the maxi- mum to the minimum stress ina cycle but either the maximum or minimum stress should also be stated and (2) the range may be expressed as a steady stress with an alternating stress superimposed upon it, z 9 n zZ WwW a Ww of 94+ — ° Z|é f { | wn Ww oe a = oO oO COMPLETELY | RANGE WITH Pate aby REVERSED |STEADY TENSILE | Conpeecive RANGE STRESS _ STRESS60 Diagrams and test data may be presented in various forms, in that stress values may be expressed in absolute values or as a fraction of the ultimate or yield strength or in terms of endurance limit strength. Combi- nations of these systems will be used here for the purpose of discussion. Omax °* Tmax Omin ° Tin Om On * Tr Orn % Ty Oy 4 T On Te for 2, Range of Stress for Note! Nomenclature Maximum stress in range. Minimum stress in range. Steady stress or mean stress equals Omax * Omin 2 Alternating stress = Grmax —"Omin ,, 2 ‘maxx Tinin 2 Also used as endurance limit stress for ranges of stress other than OQ), equal to 0. Endurance limit for steady stress O,,, = 0 for unnotched members, As above for notched members. Yield strengths of material (tension and torsion), Ultimate strengths of material (tension and torsion). Ultimate compression strength considered to be same as compressive yield strength. ile Materials, 2,1 The effect of range of stress for torsional loading was discussed in the first week's notes, namely: that the endurance limit for any range of stress is equal to the endurance limit for complete reversals of shearing stress up to a maximum stress in the range of 0.8 of the torsional yield strength. Above this stress level the relationship appears to break down, see diagram H, page 12. This relationship however only holds true for notch-free speci- mens and does not hold for cases with stress concentrations. Equation: T, = TR61 2.2 Ranges of Axial Stress on Members, This case may be sub-divided into two parts, namely the steady stress may be either tension or compression. (a) When the steady state stress is tension, the relationship for range of stress, as determined by many tests, is as shown in the following diagram, Notice that the majority of test data lie above the straight line relationship, indi- cating conservative estimates will be made when using the straight line equation, Reference Table 1. La 00 uv ae wor aw Be a Zz Q 29 E2 ac Z ee > 52 az TENSILE STEADY STRESS Om IN TERMS OF UTS.Alloy Om. ksi 14ST6 -10,0 0 and + 5.0 +100 24ST4 +200 10.0 0 75ST6 cea +10,0 +20.0 TABLE I NOTCH FREE'DUCTILE MATERIALS ANC-5 Actual Test Values at 1,000,000 Cycles #26.0 £25.0 #25.0 £24,0 £22.0 £31.0 #29.0 £27.0 #26,0 22.0 Calculated Values from Equation_(a) 62 0, =Oy (i - os £28.99 425.0 223.1 421.2 217.1 432.7 #29.0 427.1 2 (@) The value of Op in equation is taken from ANC-5 test data at, = 0 R hn (b) When the steady stress is compression, the relationship appears very similar to the relationship seen in the torsional tests. The maximum stress diagram is used here since it gives a more striking interpretation of the data, 2,2 LSA= OR ° aos ° ° 120 MAX. COMPRESSIVE STRESS OF RANGE IN TERMS OF Oc= YIELD STRENGTH IN COMP Op “OI oO loc 100g ° ALTER. STRESS OF ANY RANGE IN TERMS OF Op.IN ALTERNATING STRESS 63 Ranges of Stress for Notched Ductile Materials, 3.1 The diagram for torsional ranges of stress in this case is pre- sented as follows é zi LIMIT— Tan TERMS OF ENDURANCE ST Lov Tensile Steady Stress T,,, of Range in Terms of Torsional Rupture Modulus T, The above equation may be written in oe form including the stress concentration factor K = T,/ T,, to 3,2 Ranges of axial stress on members may again be sub-divided into tensile or compression steady state stresses. The equation, when the steady stress is tension, is similar to the notch-free equation with the exception of the stress concen- tration factor: Om >) Ou 8 Orn tt - Om K . a ) Oo, 0 - Se There is no equation given for the axial compression steady stress since insufficient data is available. It would however be conservative, at least for ductile materials, to use the equation in section 2.2, (b) with the appropriate stress con- centration factor.64 4, Ranges of Stress for Brittle Materials. No sub-division for notch-free or notched members is made for brittle materials since it has been found that the relationships hold for both cases. However, an alternate equation is given since tests have indi- cated that the straight line relationship, used in the above-mentioned cases, is not as conservative when applied to brittle materials. Ow w 2b ude be BES yy na_ 2 25% oWO 2% a ES 20 0a YZ wi wee E 5g= 2 aes i OOF RANGE IN TERMS OF UTS. 10 Ge The above diagram applies to both steady state tensile stresses for axially loaded members and torsionally stressed members, It will probably be noticed in this diagram that no reference is made to torsional stresses. However, for the case in torsion at least for cylindrical specimens (Om = Tp) We use the normal stresses produced in torsion, since the normal str'@ss criterion seems to be best associated with fatigue failures of brittle materials whether it be for shear, tension or bending loads. In summary, the relationships may all be shown on a diagram similar to the one below and the regions defined in terms of the ratio Omax Trax65 UNNOTCHED DUCT'L Dex. = ° mr 40 MET'LS IN TORSION. \ \ S % On ‘ ‘ ° qe =2 (DUCTL MET'LS.) n MAX, = a ud B z Twax. FOR DUCTILE Z MATLS AND ALL 6 VALUES FOR | vi BRITTLE MAT'LS. 5 ° Ou Lo 5, As stated previously, we have two criterions for the design of our members, Failure by yielding if the loads are too high and failure by fatigue if'we are in the "unsafe range! and the part experiences a great number of repeated loadings. Depending upon the ratio of yield strength to ultimate strength for high and low strength materials, this criterion may be illus- trated as follows: 189,000, HI~STRENGTH MAT'LS.45 000K/ (Fry) LO -STRENGTH MAT'LS. Bry Err. Low. Fro 66 Ow" RSI. 6. Stress-Ratio System for Interpreting Torsional Ranges of Stress. The following diagram illustrates another form for which we may plot our test data in terms of "R" value, a more familiar term, Min, O Max.) RTuy, IN TERMS OF Te ENDURANCE LIMIT 67 a at PROPOSED EQUATION FOR ' / ANE SPECS. 2 Taax.* eo / ae (CONST. RANGE EQUATION) 2 Tans.” Re (MOORE EQUA.) 1 i —_e -10 ° +10 R= Teun. Tmax. ™ Combined Stresses in Fatigue: The equations for design, for members under combined loading, have been given in the first week's notes on page 13, These equations need no further explanation nor are they to be expanded upon. Possibly a more con- venient form than figure 9 for presenting the data would be as shown in figure 44, The table below gives experimental verification of the correla- tion between tor sional and bending stresses and is the basis upon which the equations for design are formulated.68 % THE CORRELATION OF TORSION AND BENDING FATIGUE RESULTS FOR A FEW STEELS rin safe torsional fatigue stress _ 9g. 57 distortion safe bending fatigue stress energy theory. Maier, A. F, Feversed shear (safe) 5 9.50 for tubes. reversed bending (safe) Moore and Kommers torsional endurance limit CBr, eee (C.steels of Sending endurance limit . “P “ widely dif- vending 0,48 min. .64 max. ferent H.T.) Gough torsional fatigue a ee bending fatigue 0.44min. .80 max, 49 polished specimens. Weibel torsional fatigue = 2/3 hypothetical. bending fatigue Weibel torsional fatigue 0,225" dia, wire. dending fatigue Weibel torsional fatigue = 0.67 0.221" dia. wire bending fatigue ground to size from above. Weibel torsional fatigue, = 0.72 0,225" dia. wire bending fatigue shot blasted. * ASME Trans., Vol. 57, p. 502. 8. Residual Stresses. 8.1 Residual stress is an item we discuss on a qualitative basis but very seldom consider on a quantitive basis, Measurements of residual stresses is a complicated procedure and the calcu- lations of these stresses may often prove erroneous. However, we do know that the presence of residual stresses in our members may act in two ways - to either increase or decrease the fatigue strength characteristics. Residual stress ina member may be introduced in some of the following ways.() 8 (a (e) (9 69 During the machining and forming of a member. During plating and coating of a manber. During installation of members - mismatched parts are sometimes forced to fit. During heat treatment of odd-shaped parts with abrupt changes in'section, During welding and cooling of a member. Ete, Depending upon the kind of stress and its magnitude the residual stress may produce either beneficial or harmful stresses. 8.2 Beneficial Effects (a) (b) (c) Ig the residual stress is compression it will tend to di- minish the magnitude of the applied tensile working stress and therefore increase the fatigue strength and fatigue life. (Shot-peening, rolling, large interference press-fit bushings, etc... If the residual stress is compression it is favorable towards the control of stress-corrosion or season cracking (season- cracking only occurring in the presence of residual tension stress. High residual tension stresses may however react ina favorable way, contrary to some beliefs, For instance, if the steady state residual tension stress is much higher than the cyclical tension stress, there will be little or.no change in stress in the working part. In this sense the member doesn't even know that cyclical loads are being imposed upon it. (This case is possible in interference fit if bushings where the amount of cold working is still in the elastic range, Also this high residual tension state might be considered as the beneficial effect on increasing the fatigue lifes of torqued bolts ~ although this process is usually called pre-stressing. ) 8.3 Harmful Effects (a) (v) (c) Overheating during improper machining operations may cause residual tension stresses. These stresses com- bined with tensile working stresses in our members will shorten their fatigue lifes. Residual tension stresses caused by forming, particularly cold forming, plating and coating of members have the same effect aS in (a) above. Residual tensile stresses will cause stress-corrosion. In this sense it is not even necessary to have combined working tensile stresses.70 SUMMARY By now we have briefly reviewed the mechanism of fatigue, stress concentrations in fatigue, the effect of different types of loading and dif- ferent ranges of alternating stresses in fatigue. We have briefly mentioned the two main strength theories* that appear to apply most readily to ductile and brittle materials, Namely** lL Brittle materials follow a combination of the Maximum Normal Stress criterion and Mohr's Theory. (T, = Oy) 2. Ductile materials follow the Maximum Shear Theory and R, von Mises criterion, (% 577 Oy) By the use of these two strength theories for static loads we may de- fine design margins for alternating loads (M.S. = f.S. - 1) and extend these margins to cases for limited numbers of cycles, ref. Appendix 13, The design equations given are not as accurate or absolute as an actual test wil dictate however, inthe absence of tests the equations are usable and are in a simple and convenient form, For a more detailed derivation of the design margin equations reference "Stress Concentration Design Factor", by R. E, Peterson, J, Wiley and Sons, 1953. * There is no derivation as to why ductile or brittle materials will follow these theories, Test data for a wide variety of materials has been found to match these theories more closely than others, The difference between brittle and ductile materials is sometimes taken at the dividing line of a 5% elongation. #% Reference Appendix 12.AXIAL OR TORSIONAL STRESS BENDING ENDURANCE LIMIT 8| AS rps & 5 f A “~~ 3 + FIG. 44 .2 wd 12 3 4 5 6 7 8 9 BENDING STRESS BENDING ENDURANCE LIMIT 70.1TENSION STRESS - 1000 Psi. 60 70.2 EFFECT OOF STRESS RATIO ON FATIGUE. STRENGTH TSSTALC (AKIALLY LOADED SHEET) | < | | 2 [\Re5, | _ N / | To 0 CYCLES TO FAILURE FIG. 45 rept oor7 Appendix 12 STRENGTH THEORIES FOR BRITTLE AND DUCTILE MATERIALS Brittle Materials Test data have indicated that the behavior and failure of brittle materials most closely follows a combination of the maximum normal stress criterion and Mohr's Theory. This theory states that failure occurs when either a principal tensile stress or compressive stress reaches the corresponding uniaxial strength. For brittle materials however, (. is usually considerably greater than Oy: Reference figure A. For tofsion(j, = -O1 teincot! ) which indicates that for equal magnitudes of principal stregs the normal stress condition for failure is T) a Briefly Mobr's Theory may be illustrated by the following diagram. O,, and ©, representing ultimate principal stresses, T shear stress, and and (J5 any other stress condition. Mohr's Theory states: when any clvéle, stich as the one enveloping (, and ()> circles, intersects the tangent lint - Qj, and (J, circles “~' that faiture will occur. The combination of the two criterion for brittle materials may now be shown as in the following diagram. The diagram is conservatiye for shear stresses however since 0 T,, is usually used instead of OT! .12 © rams, i Leis. NORMAL O° OR Ov “be Oz TENS, — FIG, A For design purposes we may now assume for static and fatigue stress~ ing the strength ratios are the same Gu, Oa Oc O-R Te. Oa Or-TR OR ee Ou ic LG3 Compare to equation in section 4, page 64, in 4th week's notes. This equation is in terms of ultimate strengths rather than steady state mean stress although the results obtained would be similar. The diagram above has been drawn for complete reversals of stressing although there is no reason why endurance limit stresses or limited fatigue life stresses for all ranges of stress could not be represented on such a diagram. Ductile Materials Test data have indicated that the behavior and failure of ductile materials most closely follows the Maximum Shear Stress Theory and R. von Mises Criterion, The Maximum Shear Stress states that failure occurs when the maximum shear stress reaches the value of the shear in auniaxial bar at failure. For principal stress (J), and (, the maxi- mum shear stresses are Oi “Oz 01 -O3 Oz -O3 a ee 2 2 2 For the case when 0 = 0 andO\>O2, both tension, failure will occur when14 it O, =-O) failure will occur when Or - Oy) OR 2 2 for torsion oO, = -O, =T wot Ty = BE ‘These values for ductile materials are just half the values for the brittle materials and the Maximum Normal Stress Theory. Mises criterion for failure by yielding is expressed as (0, -0,)% + (0, - 05% + (0, - 05)” = 1 2! 2 3x 1 3 v 2 for the case of O, = 0 Oy [Oi - Oz + 02" and if O> = -O; =T it can be shown that TORSION FAT. LIM. BENDING FAT. LIMIT 19,00075 Appendix 13 The following example problem will illustrate a few of the various methods we now have available for calculating the life of a part. Suppose an 1/8" thick, 1 1/4" wide, 12" diam, 75ST Cl. reinforcing ring with 1/4" holes is stressed to 21,900 psi (0 to max, pulsating stress). The material has an Fpy of 7,000 and an endurance limit 6 of £29,000 psi. ‘The equation on page 63 for ductile notched materials is now used to calculate the alternating endurance limit for Gj, = 0. On: ° Om For the stress condition of our problem (Onin = 0) the following relationship holds true O2 “Om So Oa 77,000 Oa = 29,000 (1 - Oa = 21,100 psi Ca * Om = 424200 psi °F Omax Expressed in terms of maximum stress the fatigue strength for 10° cycles is now 42, 200 psi; in terms of alternating stress the fatigue strength may also be expressed as #21, 100 psi, In both cases, however, the minimum stress is zero and the values are for notch free specimens. From the geometry and dimensions of the reinforcing ring the stress concentration factor (Ky) is calculated to be 2.51 from (D/W of hole = . 2).76 Maximum nominal stress 55,000 psi 0 Minimum nominal stress 0 Fy —& ‘ Oye _ LOG 10 G-6, lOGN, Snax. % 267 We 70600" 77000 - 42200 _ Logio®_ 6 77000-55000 LogN Log N Log N = 3.8 N = 6,300 cycles Using Ona K = 21,900 2.51 = 55,000 psi Omin = 0 R= ° and the S/N curve figure 25 N = 25,000 to 30,000 cycles ee ee) The actual life, from laboratory tests, on axial loaded plate with holes at a net stress of 21,900 psi (no need for calculating a stress concen- tration factor) is from 60,000 to 100,000 cycles. .,...,...,.,(3)7 It would appear that both methods are conservative although the method, (2), gives more reasonable values. ‘The method in (1) has been used by designers when only meager data exists. However, it is believed that this method is far too conservative and the designer should be aware of this fact whenever such a calculation is made.78 Chapter 5 METHODS TO IMPROVE THE FATIGUE RESISTANCE OF MATERIALS Fatigue strongth ike Beauty is only shin decpe- 1, This section of notes relates to methods of improving the fatigue re- sistance of designed parts as a function of their surface strength. As we have mentioned before, the majority of fatigue failures are surface phenom- enon and if our efforts and interests are directed towards the strengthening of surface material we should therefore be able to design better parts. It will not be the purpose of these notes to ever attempt to teach the "old pros” how to design their parts but rather to relate to them a few methods that may improve an already well designed member, So that good designers may not become too confident it might be well to review the record, There have been vary few catastrophic failures attributed to fatigue failures, how- ever, there have been many minor service failures that are both annoying and costly. Purportedly the C.A,A, have revealed that minor aircraft failures occur at a rate of 30 per month, A study of such failures reveals that at least 90% are traceable to design and production defects, The re- maining 10% may be the responsibility of the metallurgist as defects in material, material specifications or heat treatment, It should be remembered that very few designs have ever been successful because of a change of material. The causes of failure have usually been traced to either design or fabrication, 1.1 For the designer's review the following design guides are itemized according to Mr. E, Harpoothian*. The guides used are simple but much attention is required to get the desired results. (a) Proper choice of materials, (b) Use of large fillets. (c) Avoidance of sharp corners, (3) Gradual changes in section, Provision for large edge distance at holes. () Reduction in bearing stresses. (g) Eccentricities, joggles, and felexing of structure - reduced to a minimum, 2. It is doubtful for repeatedly loaded structures if we are actually getting more net work out of our materials today than we were 20 years ago, Basically, the useful strength of our materials is the same, with the exception of a very few higher strength alloys, however some new techniques have been developed that tend to increase the fatigue resistance of our structures. It is fortunate that it has been possible to develop such techniques since the design trend to higher working stresses has demanded some means of improved life. ™ CH, Structures Section,2.2 19 Surface Prestressing A few of the more common techniques that will improve the fatigue resistance of structural members without requiring changes in design or material are listed below: a) Shot Peening ) Cold Rolling (Rolled threads, fillets, etc.) (c) Nitriding Case Hardening (Carburizing) A few of these techniques are known to increase the physical properties of the surface material, however, they may all be classified as methods of obtaining surface-compression and it is in this sense that improved fatigue strength is obtained, A word of caution ~- it should be not inferred that large improve- ments in fatigue strength can be obtained at all stress levels and for all types of loading. The greatest amounts of improvement are always evidenced in the elastic stress ranges of the materials and the greatest benefits are seen for members that are pri- marily loaded in bending or flexing. Shot peening for instance has been found to have little or no effect on increasing the fatigue life of axially loaded members. (One explanation that has been offered is given on page 94 in the seventh week's notes "stress gradient effect") Shot Peening Sand-blasting, a relatively uncontrolled method of obtaining sur- face-compression has given way to the improved shot~peening method. By now, most of the principal variables to be considered in the study of shot-peening have been worked out and can be found in various tests and investigations on the subject. There will never be any need for the designer to specify the type and diameter of shot, the exposure time, intensity required, nozzle velocity, etc; therefore, there will be no need for us to make a study of these variables. We are mainly interested in the bene- ficial results to be gained and the few basic rules required to ob- tain the results. The process of peening a part sets-up compres- sive stresses in the surface much in the same manner as ob- tained in striking the surface of a flat bar with a ball-peen hammer. (Compression stress evidenced by the curling-up of the flat bar about the peened surface - peening of fittings and members how- ever not so severe that such a great change in shape is evidenced.) Graphically the effect of shot pening may be as shown in the dia~ gram below. Figure (a) shows the probable stress diagram for an unloaded beam with a prestressed surface. Figure (b) shows the stress diagram for a loaded (in bending) and prestressed beam and figure (c) shows the influence of the depth of prestressed layer.80 (A) ¢B) (c) PRESTRESS DEEP RESULTANT i [COMPRESSION Lf | psuatow t — AN, LOADING TENSION) Remembering that fatigue failures are primarily due to tension stresses (no fatigue failures due to compression) it can be seen that a compressive prestressed surface will tend to cancel or algebraically decrease the magnitude of the applied working ten- sile stresses. In this sense our part is working ata lower stress level and it follows that the part will last much longer. How- ever, as in most cases, you can carry a good thing too far. It must be remembered that for a state of equilibrium the pre- compressive stresses at the surface must be balanced by internal tnesile stresses and if the peening operation is carried too far the excessive internal tensile stresses (figure (c)) will damage our part, The effect of peening intensity on fatigue life as es- tablished by tests is as shown in the diagram below. ahs INCR. FATIGUE DECR. STRENGTH This same effect, that of increasing or deercasing the fatigue strength, is obtained by cold rolling and the results are comparable to shot peening. The precise amount of pening or rolling is only known for a few cases. There are methods available of fairly accurately measuring the amount of surface’ stressing which81 combined with experience and a proper sense of proportion is the only approach to the problem. On a comparative basis, testing should then be carried out to indicate the percentage improvement of the prestressed parts. 2.21 Advantages of Shot Peening (a) Can be applied to irregular shapes, in which heat treating might cause distortion, (b) Can be applied to finished parts either locally or all over. (c) Can be applied to fillets and grooves to reduce stress concentration, (a) Produces a surface resistant to pitting corrosion and stress corrosion, (ec) Can sometimes be used as a surface finish in place of polishing (reduces cost). (f) Used with apparent success to increase resistance to fretting corrosion, decarburization and surface damage such as light bruises. (g) Laboratory tests have indicated that as a salvage pro- cedure for uncracked parts that have been in service it may be used to "turn back the clock" and result in a greater fatigue resistant part than the original, Much more research and study is required before the limi~ tations and advantages of shot peening are fully realized, One thing certain however, is that its usefulness will be enlarged upon in the future, A recent modification of the shot peening process has been devised (U. S, Patent #2, 608, 752) in which the peening op- eration is carried out while the surface of parts are under tensile strains of approximately 80% of the material yield strength. Preliminary tests have indicated that the amount of residual compression stresses obtained in this manner arejgreate® than in the ordinary peening methods and with no greater internal tensile stresses. Nitriding, Honing, Carburizing. These methods similar to shot peening will also cause residual compression stresses to be present in the surface layers of members. Their control and limitations are again similar to those previously mentioned, however, these methods are not as widely used in the aircraft industry as shot peening and do not concern us as much, A rather interesting equation that has appeared to work with reasonable accuracy for nitrided parts;82 as a function of the depth of nitriding, is as follows, This equation however has its limits since too much surface com- pression can result in damaging internaltensile stresses. AG, C100) _ _100 Op DL 2d Where ( Og ) is the endurance limit of unnitrided parts and (0) is the depth of the surface layer affected, AG. When more shot peening data become available it would be in- teresting to apply it to the above equation, Rolled Thread Bolts It has long been established that rolled thread bolts are superior in fatigue strength to lathe, die cut or ground thread bolts, The reasons are similar to shot peening or rolling in that residual compressive surface stresses are set-up. Also similar to shot- peening; it has been found that heat treating after rolling tends to reduce the fatigue strength. The compressive stresses are re- Leved by the heating. Decarburization at the root of the threads and in the threads will also reduce the fatigue strength. Heat treatment should be carried out ina controlled atmosphere to minimize or prevent decarburization, but in no case should it be performed after the rolling of the threads of bolts. Bolt failures in fatigue usually occur between the first and second thread engagement with the nut, Photo-elastically it can be shown that under load this is the region of highest stress. With the ad- vent of rolled thread bolts a greater percentage of failures began to occur in the fillet between the head and shank of the bolt. In order to "squeeze" the last bit of usefulness from the bolts the process of rolling the head to shank fillet was devised. This step has just about "evened-up" the head or thread failure location and has resulted in a slight additional increase in fatigue life. Ref. figure 41 for rolled thread v.s. ground thread bolts.2.5 83 Prestressed Bolts - or how tight should a bolt be? Torqueing of bolts in connected members is an established method of preloading for increased fatigue life. Not so well established are the actual magnitudes of the preload in bolts due to specified torque values. Extensive testing to determine the relationship between the torque and the axial preload has resulted at best in a 6:1 scatter in the results obtained, It has long been recognized that torquemeters are not entirely satisfactory means of con- trolling the preload in bolts, However, when other methods are not available the torquemeter is to be used, Methods assuring greater accuracy are: 1, Measuring elongation (micrometer) of bolt from which the strain and therefore the load can be calculated. 2, Calibrated Wedge Washer, Patent #2, 188,356 (See Sketch), 3. Preload Indicating Washer (PLI) a DACo development, WEDGE RING INSIDE TAPER WASHER Euhbinahene 2 Seine ene, 90 that usher and wdge ring are flush. Often it is impractical to use or require the more accurate methods. In these cases specified torque values can be assumed to give the following average tension loads in bolts.84 TABLE I %Axial Preload in Pounds For a One~ Bolt Size (Fine THD. Series) Inch Pound Torque #8-32 30,0 #10-32 27.0 1/4-28 21.5 5/16-24 11 3/8-24 14.9 7/16-20 12.6 1/2-20 14 9/16-18, 10.3 5/8-18 9.24 3/4-16 7.64 7/8-14 6.78 1-14 5.99 11/4-12 4.80 11/2-12 4.16 % Values are for dry assemblies, lubricated assemblies give higher load values for same torque. Installation torque is applied to nut; if applied to bolt head a 20% increase in torque is sometimes specified. It has been shown that preloading or prestressing of bolts will definitely increase their resistance to fatigue. Unlike the sur- face-compression methods previously described the prestressing of bolts is a tension phenomenon, Imagine a bolt highly pre- loaded in tension and superimposed upon this preload a cyclical tension working load of lesser magnitude. It is conceivable that if the preload were high enough the bolt would nat be aware of any lesser alternating load that may be applied. Our greatest problemin determining the effect and amount of preload to apply, resolves itself into the determination of the resultant of combined steady and alternating loads. There are those that believe that an externally applied load will add directly to the preload in a bolt; there are those that believe that an externally applied load will not add to the preload in a bolt until it has reached a value equal to the preload, Both of these assumptions are wrong; The resultant load in a preloaded bolt is equal to the preload plus some fraction of the externally applied load. The fraction of theQQ) LOAD — W 85 applied load that is effective is a function of the elasticity of the bolt and of the members which are being connected, This effect can be expressed by the following equation and illustrated in the following diagram: By Ay w, + Wa iea * r internal applied E,A,tE,a,+ 5, Ay + Where W, is thd resultant load in the bolt and B,, A, = EA Value of bolt « E,A, = EA Value of clamped spacer material EA, = EA Value of = additional affected com- ponents and assuming length of bolt = length of material LOAD -STRAIN CURVE FOR BOLT if \é RESULTANT LOAD CURVE ' FOR BOLT WITH EXTER- |? NALLY APPLIED LOAD. f PRELOAD_ STRAIN STRAIN- &2.7 86 Bushed Holes "Bushing" the holes of pin-loaded moving parts has in the past been carried out for the purpose of reducing wear. Recently it has been shown that interference fit bushings have also resulted in greater fatigue resistance for pin loaded members. (See figure ) British and DACo tests however, indicated that the amount of interference was of prime importance in order to get the optimum results. Pressed in bushings with .001" to .003" interference gave no appreciable increase in fatigue lives how- ever, for interferences of .044" to .008" on 7/8" diameter holes the fatigue lives were increased nearly 10 fold. The explanation for this increase in analogous to the residual compressive sur- face layer phenomenom we have discussed in previous processes. It would be normal to assume that when we enlarge a hole by pressing in an oversized bushing that the resultant stress would be tension, This is not necessarily true. When a large inter- ference fit bushing is pressed into an undersized hole there is a thin layer of material around the edge of the hole in the plastic region, Tangentially this material is in’ tension but the large mass of material pushing back against this layer results in larger radial compressive stresses. The combined net stress in this plastic region is then compression, (Ref. "Plasticity" by A. Nadai. Chaper 28.) Expanded Holes A similar process to bushed holes is to draw a mandrel or com- bined broaching tool and oversized tapered mandrel through the holes. The resultant residual stresses around the edge of the hole, after springback, are compressive but not of the same order of magnitude as in the case of bushing a hole. Leaving the bushing in the hole seems to have a more pronounced effect on the fatigue life than the drawn taper pin method,87 Appendix 14 The following laboratory tests that have been made illustrate the effect residual stresses have on fatigue strength, Figure 46 ‘The stress concentration at the edge of an unloaded hole results in decreased fatigue strength as compared to "par" specimens. However, when a rivet is driven into the hole (but still unloaded) the "effect" of the hole has diminished to the point where these specimens have fatigue strengths equal to members containing no holes. The test points plotted on the curve are based on net stresses across the hole and appear to be even better than plain members, If gross-area stresses were calculated these same test points would plot at lower stresses and all fall within the scatter band for plain "par" members, In this case the driven expanding rivet has stressed the material around the hole; and remaining in the hole prevents distortion and high plastic straining at the hole's edge under external loading. This effect is not as pro- nounced when pin-loaded holes are considered. Figure 47 Supposedly, no fatigue failures will occur if the cyclical stressing is 5 all in the compression range. This is probably true, however in this particular case the range of stress was all compression and fatigue failures were obtained on axially loaded members containing holes, The stress levels required fo obtain such failures were extremely high; high enough so that the material at the edge of the hole was undoubtedly in the plastic region. It is believed that the hi-compression stress in plastically yielding the material has resulted in a residual tension stress of the order of 15,000 to 20,000 p.s.i., which was the cyclically damaging stress in failing the member. There were no residual stress measurements actually made, the comparison has been based on stress levels (compression curve to tension curve) at a given life. Figure 48 The S/N curves show the increase in life of prestressed holes over plain drilled holes. The holes in the extruded bars were plastically yielded (6% max. expansion of dia.) by drawing a tapered pin through the holes. Con- traction of the material around the hole after the pin had been pulled through has resulted in residual compression stress.811 FIG. 46 TSSTALC! SHEET WITH AND | | | WITHOUT RIVET FILLED HOLES | il 350 él a 40h 2) 2 230 z|| 6 : ee x fo SAMPLES.) ii 2 alll I I oo of 10 20 S, CYCLES TO FAILURE - MILLION: FIG.47 2asTaBAL SHEET. CENTRAL HOLE. elo i ho 7 Le 3 Elo ty 5 o-ASIMPLIFIED SPECTRUM FOR ONE HOUR HIGH PRESSURE PUMP LINES FIG. 69.20 LARGE SCALE STRESS CONCENTRATION Figure 7. Large Scale Streas Concentration. TIME ——— DETAILED PART LOAD SPECTRUM Figure 8. Detail Part Load Spectrum,FACTOR TENSION STRESS, 1000 p.s.i. | oF 2.6 A-A_| lL WRF tot 105) loe CYCLES to FAILURE STRESS CONCENTRATION IN FATIGUE (REFERENCES 2,3) CLAD 75ST6 Figure 9, Stress Concentrations in Fatigue, Clad 75-ST6 alueinue alloy. Axial tension, minim stress one-fifth ma urve (1.)"Average of experimental scatter band for plain semples. Curve (2.) Average of scatter band, saxples with central hole, K 2.6. Lover curve labelled 1/K calculated from upper curve (1.) by dividing each stress by K. (Data: Reference 2; Kt Reference 3) EXTRUDED SHEET BAR 75 STG JTH 4180 200,000 p. 2" 19" 60X10” w bea SIF es iN a SS STRESS CONCENTRATION vs FATIGUE Figure 10, Coxparison of Theoreticul Stress Con- centration, K and Apparent Stress Concentration Determined by Fatigue Tests. R is the ratio i fntress in axial tension fatigue xinisun to mex test. (References 2,3,5,6.) 90.2190,22 T . + ep AONGITUDIN 7 POLIS! q 50 4 2 40 =| 100" 4 g Re.050% Gf 0" 2 35--— — 8 Fen ay E =; —— ei ! ! % to 105 toe = NUMBER of CYCLES to FAILURE SURFACE FINISH vs FATIGUE Figure 11, Surface Finish va. Fatigue Strength, Extruded 75-ST6 Aluminum Alloy. ial tension, minimum stress one fifth the maximur. (Reference 7.) VALLEYS ZUuUFE | DESCRIP] © SKETCH R PER IN| D FORMED TION | AT S500 X [ES AT500x ERED | EZZZZZZA h . | ISMOOTH <4 | .02” | 100% tari |BETWEEN BETWEEN Ot eet ; ETWEEN | ey ROUGH ~2 ms any {Ee ou? 7 VERY oboe ROUGH ANY | >,O1t | <.01" | 38% waDeae: EFFECT or SEAMS and ROUGHNESS on FATIGUE LIFE, AMS5566 STAINLESS STEEL TUBING Figure 12. Effect of Senns and Roughness on Fatigne curvature of surface valleys. (Reference 1.)a} os 0 PRINCIPAL STRESS SPECTRUM CRITICAL ELEMENT Figure 13. Principal Stress Spectrum, Critical Element. | |24STS6AL SHEET- | [aaa EET) ? | ale zl * gyn i er + {Mo il UTED AND a IN Be bade $ le Sn NE SSon g i ° Bae ieaeg antares 1 t a or STRESS z Rete || | a t | z i 310 | z a 0 evcres “ro _eavure Figure 14 ‘lures in Compression, Clad 24-ST86 Aluminum Sheet ining « Central Hole, Ratio of tress was one fifth for hosh w ression and axial tension testa, a pe, 902 3902 4 f(2) —__o oe 1 200 20 “* 2000 —~ > 200 SIMPLIFYING THE DETAIL PART STRESS SPECTRUM Figure 15. Simplifying the Stress Spectrum. Upper curve, specti stress function. Lover curve, spectrum simplified by grouping sinilar axplitudes. ny Wty CYCLING AT VARYING MEAN STRESSES Figure 16. Cycles of Sane Auplitude but Differing in Mean Stress.MEAN STRESS, 1000 p.s.i, 9025 Figure 17. Effect of Mean Stress on Pernissible Auplitude in Axial Loading. Clad 75-ST6 sheet, +072" thick, one willion cycles. (Reference 2.) 1 7 7 T 072"75ST6 4 oor SHEET L 10 CYCLES 4 40k | = 20+ Co 4 ° L i ° 5 10 2 HALF_RANGE, STRESS, focal EFFECT of MEAN om pias Figure 18. Effect of Cyel idual Stress Pat in Changing « Re~ ox 15-ST6 h 028" diameter iron Curves before and after fatigue cycling inl tension. (Reference 10.) TO 60 50 40 30 AS - PEENED P28 EXTRUDED 75 ST6 65001 aS 3 000 ps.i. 2500, 000 CYCLES RESIDUAL STRESS, 1000 p.: 8 ° 10 2030) DISTANCE from SURFACE THOUSANDTHS of an INCH EFFECT of CYCLING on RESIDUAL STRESSES+2 & MAX. STRESS, 1000 ps.i, ATR 8 b é w 3 R 8 3 90.26 I PEENED, P28 sal (A PEENED, P46 © UNPEENED - . — ee (24 ! + SCATTER —_— an -— BAND a + he 10% 10> Toe CYCLES to FAILURE Figure 19. Effect of Shot Peening Extruded TEISTE on dts Fatigue Streusthe Hloundedvedge coupons, nd sO46" diencter shots Axl fennfon, sfaivam stress enecTifth the bextou {feference 10. SLE = ef | 4 + J & 40h 4 3 8 = 2 3 oo & £ 9 20} 2 Zz Ba -40 a @ a ~60, SURFACE MID-POINT SURFACE DISTANCE FROM SURFACE BENDING 75 S16, ¥=8 Figure 20. Residual Streaa Gradient, Extruded 75-ST6 ‘i ied to Hadine of Eight Tines |. (Reference90.27 70 ; — PLAIN 2 60} ¥ STRETCH g ROOM TEMP. a % \ ~ : nye AXWXx< 5+ aq — g T R/T= Be Boe xX < 2 1o* to oe, CYCLES TO FAILURE EFFECT of FORMING on FATIGUE PROPERTIES of 75 ST EXTR (AXIAL) Figure 21. Effect of Forming and Stretehing on the Fatigue Strength of Extruded 75-ST6. Axi Tension, minimum stress one fifth the maximum. (Reference 12. Figure 22. Typical Fracture Face for Simulated Bent Coupon. Hesidual tensile stress and fatigue nucleus are at upper surface, (Reference 12.)90.28 ioe - ESTIMATED for CHANCES ARE 4 NO re 50 SAMPLE! O65, ACTUAL BL CURVE BETWEEN REST g 8 UMITS > 24 2° 80 100 120 (40 rest 2'C 24h. 0 (2- CYCLES TO FAILURE @TO°F. ¢ pf @io000 Z8r | w = Y INCREASE IN ENDURANCE LIMITS of gt eo & TEST Qeaccuraren RESULTS from RESULTS EFFECT _of REST £50,000 p.s.1., 775T6, ROTATING BEAM Figure 23, Effect of Rest on Life of 75-ST6, Rotating Beam, 50,000 p.. (Experimental Data, Reference 13. TH O IRON € STEEL lef @goeeer — Up ——e TE iy iy 2 20 40 60 80 100 120 THOUSANDS of CYCLES PER MINUTE EFFECT of FREQUENCY on ENDURANCE LIMIT Figure 24. Effect of Frequency on Endurance Limit. “(References 21, 22, 23, 24, 25.)E.L. 5x108 ROTATING BEAM CKS.1.) 5S % CHANGE IN ENDURANCE LIMIT 90.29 23 T ‘A= HEAT TREATED *= COLD WORKED) x 20 5 10 5 0 20 40 0 80 [00 Feu CK S.L) ENDURANCE LIMIT vs ULTIMATE STRENGTH ALUMINUM ALLOYS Figure 25. Endurance Limit vs. Ultimate Strength, Aluminum Alloys. (Reference 26.) Figure 26. Effect of Artificial Aging on Endurance Limit, “Rotating beam testa, 5 x 10 cycles. (Reference 26.) NSTé 53ST6 — OIST6 STS 625T6 5 10 > 20 25 50 Y%e INCREASE IN ULTIMATE by ARTIFICIAL AGING EFFECT og AGING on ENDURANCE LIMIT ROTATING BEAM. 5X 108 ~~ENDURANCE LIMIT 30.36 120 gion / | / / ii /; a 7) GROUND yilt 8 eo ile . ee g TUES HOT ROLLED oa LES bop] 3 ALUMINUM LPL IVVOLEY LEE OP B {aus 11S Ned i eS cee ao l 3 15 40 80 120 160 20 240 w TENSILE STRENGTH, 1000 p.s.i. a e TENSILE STRENGTH vs, ENDURANCE LIMIT, STEEL Figure 27. Tensile Strength vs. Endurance Limit, Steels. (Reference 27.) FATIGUE OF METALS ea . ‘a ge in 08m el in ton ogni de oan a ‘overheating (x 150), (Caza) a Figure 28. Effect of Grain Size on Endurance Linit. (Reference 23.)ULTIMATE STRENGTH EL. ROTATING BEAM ot UAL OND O E.L. ROTATING BEAM 90.8\ EFFECT of STEEL MICROSTRUCTURE on ENDURANCE LIMIT Fru RATIO Figure 29. Effect of Steel Microstructure on Ratio of Endurance Limit to Tensile Strength. (References 28, 29, 30.) ‘47 ERRITE AFC 8 AUSTENITE 7 r a Bu w za E FZ ge |. >i, ee Sean so sz wa ot ut fe on az su Be be 2s 5= YIELD STRENGTH EFFECT OF STEEL MICROSTRUCTURE ON ENDURANCE LIMIT / Fey RATIO Figure 30, Effect of Steel Microstructure on Ratio of Endurance Limit to Tensile Yield Strength. (References 28, 29, 30.)a as] Ser *0 CS a 2 J I = ' 7 LONGITUDINA HEAVY SHAPE MAX, TENSION STRESS, 1000ps.1 a) THIN EXTRUSIONS °° + 1 40 o [bz 7 TRANSVERSE zope —Fa-]t Wenvy SHAPE CYCLES TO FAILURE 10¢ DIRECTIONAL AND HEAVY SECTION PROPERTI ES EXTRUDED 75ST6 Figure 31. Effect of Section Size and Stressing i N Curve, Extruded 75-ST6, Axial m stress one-fifth maximus. 2) WZ N 35 ° + & 30 a 2 s DEFECTS 3 2 20| 3 NO DEFECTS { & Ss wy DEFECTS 5 2 10 A= } i ao oe ge CYCLES TO FAILURE Le los 10% 10> oe EFFECT of SEVERE RING CONDITION exTRUDED 75STo Figure 32. Effect of Severe * Ring-Co S-N Curve, tension, mi (Reference90.33 10 = be [ : eB Fisher, ale } STAINLESS s cf: & cea = al 8 L = 2b a L z , i AIR- CORROSION FATIGUE DIRECT STRESS, COMPLETELY REVERSED PERCENT of FATIGUE STRENGTH 2200 CYCLES PER MINUTE Figure 33. Corrosion Fatigue in Air, Axial loading, completely reversed, 2200 cycles per minute. (Reference 34.) 129) * STRESS AT LIFE °F 10% te 10” CYCLES EFFECT of SURFACE TREATMENTS Figure 34. Effect of Surface Treatuenta, (References 38, 39.)RESIDUAL STRESS, 1000 ps.i. MAX. STRESS, 1,000 p.s.i. 9034 Figure 35. Typical Residual Stress Grad: in Nitrided and Carburized Steel. (References 40, 41.) T +80)~——\ ——— LARGE CARBURIZED GEART| +60 TENSION + & 7 | _NITRIDED & "THICK NITRALLOY PLATE COMPRESSION 0 30 ° pisTANCE oa ee -THOUSANDTHS oF Am, RESIDUAL STRESSE: NITRIDED AND CARBURIZED STEEL 100 NITRIDED 50} —+— 4 UN -NITRIDED ° | 1o# too oe 10? 108 09 CYCLES te FAILURE EFFECT of NITRIDING NITRALLOY on THE CORROSION FATIGUE STRENGTH Figure 36. Effect of Nitriding Nitralloy on the Corrosion Fatigue Strength. Rotating beam. (Reference 42.)91 Chapter 7 LABORATORY TESTING OF STRUCTURES AND COMPARATIVE TESTING There uss never in the world tuo opinions alike, hovnore than tuo bgirs or tuo grains; the 0st Mniversal quabity t5 diversity: He de Montaigne 1, There are undoubtedly as many types of fatigue machines as there are configurations of fatigue loads for aircraft structures, Basically, the test- ing machines may be classified into loading types such as direct stress (ten- sion), bending or flexure and torsion. Further, the machines may be sub- divided into types such as direct constant loading, direct constant straining (constant amplitude) and resonant loading types. Much time could be spent on the details and mechanisms involved in the construction of the various fatigue machines. For the time being, at least, it may be better to leave these details to the operators of the testing machines and to assume that any mechanical variations from one like machine to another is too insignificant to cause any variations in the fatigue lifes or strengths of our tested struc- tures, We may also assume that the latest of laboratory equipment is avail- able and in use and that the errors in setting up dynamic loads is no greater than 1-2%. (Dynamic loads used in this sense should not be confused with vibration, There are many who erroneously use the term vibration and fatigue interchangeably, Itis true that fatigue failures may be caused by vibrations from some resonant excitation; however the fatigue loading we will be mainly concerned with in these notes are from direct service loadings such as pres- surization, gust and maneuver loading, etc. .) 2, Two of the most difficult problems to determine in any fatigue test are (i) the proper laboratory test loads to apply to the structure and (2) the degree of completion of the simulated test specimen, (a) In determining the test loads to apply to laboratory specimens it is most often necessary to refer to the formal stress analysis of the structure in chosing the critical and most frequently applied load. In composite structures, having many components, it is usually rather difficult and often impossible to determine the most critical member of the structure. Stress analyses are often based on assumptions, e.g., load and pressure distributions, and in this sense there may be some doubt as to the validity of the chosen critical condition for test. There are no hard and fast rules to follow in order to insure a legitimate test, It will be left to the discretion of the design engineer as to the impor- tance of his test and how thorough the experimental analysis should be. (v) Unlike the "one-horse shay"' the members of a composite aircraft structure are not all going to fail at once. Some one member is92 more critical than another and if our engineering designer has that supreme capability of picking out the most critical member, -~ redesigning it for improved life, ~-then the next most critical member becomes critical, This process could be carried out almost indefinitely, but obviously it is not practical. How then does this effect our choice of a test specimen? Obviously, the most complete simulated structure would yield the most reliable results, However, this too is often impractical and like the choice of loads is left to the discretion of the engineer, This "fatigmatic! state of affairs now leads us to the following methods of attack: 3. Comparative Testing By far the comparative test is the easiest, most often conducted, and probably the most useful test. For example, this type of test may be used in determining such items as directional strength properties, attachment methods, material properties and heat treatment processes. As the name im~- plies, the comparative test can be used to compare the fatigue properties of similar or radically different designs and in this respect the ones with the best test re.ults are chosen for use. Actual values of service loads and com- pleteness of structure need not always be duplicated as the comparative test is used to indicate percentage improvements. On a quantitative basis the com- parative test can, however, sometimes be used to indicated approximate ser~ vice lifes. For instance, the comparative test of simulated structures for which service experience and its past performance is known can be used to determine the performance of a new design based on their laboratory test comparisons. 4, Test on Components in Analyzing Full Size Built-Up Structures As previously stated comparative testing is a very useful and necessary method, but for some reason it appears to be the lazy man's way. The cal- culations for service life of similar structures in predicated on past service experience (a-posteriori) and little or no knowledge of working stresses need be known, In component testing the fatigue analysis of new designs requires a thorough understanding of the behavior of the various materials, the working stresses most frequently applied and the (a-priori) statistical probability of frequency and magnitude of service loading. Further, the engineer must realize the limitations to which laboratory tests of simulated structures may be equated to the actual structure loaded in service. Component testing may be sub-divided into the following categories: 4.1) Full size testing of components 4.2) Sub-size testing of geometrically similar components 4.1 A full size test may either imply that the full scale or actual size member is to be tested, or the full size composite structure containing many members is to be tested, The disadvantage of the composite structure is that itis costly. However, if the stress analysis is accurate enough and possibly supplemented with strain gauged-static load tests then the designer may feel that a fatigue test of the single member, based on the stress analysis93 loads will give reliable answers, Reference Appendix 20, Comparisons of well instrumented tests have shown that equal fatigue lifes may be obtained for tests in which only the critical member is directly loaded and for tests in which the same cri- tical member is loaded by virtue of the loading of the composite structure, Sub-size specimen testing is often a practical method of testing when very large parts are to be considered. A word of caution, however, to the designer in this type of testing. - Experimentally it has been shown that the fatigue behavior of large specimens and geometrically similar sub-size specimens are different, Possibly you have heard of the term "size-effect" which has often been used very loosely to explain the difference between large and small specimens. As a matter of fact, it is often used in ignorance and is a handy term to apply to some unexplainable difference in the behavior of our structures. There is very little substantiating data on this subject, however, two of the most readily acceptable views are explained as follows: 4.21 It has been shown by test that the fatigue strengths for large bars in steels and light alloys will decrease with an increase in dia of bar, For one particular steel 2.5" in diameter the bending fatigue strength was only 60% that of a similar bar 0,6" diameter. The region of the greatest decrease in fatigue strength, however, is in the small range 0,2 to 0,6" in diameter. The graph of figure shows the test results on a light alloy for different types of loading and for notched and unnotched specimens, From this data the following con- clusions have been drawn: (a) There is a pronounced drop in the unnotched bending or flexual fatigue limit with an increase in size. How- ever, in this type of loading as in others it is con- cluded that the fatigue strengths approach some asymp- totic value and that no further reduction is significantly evidenced in members larger than about 1" in diameter. (b) This same drop is evidenced in bending tests on notched specimens. (c) Under torsional loading, both notched and unnotched, the decrease in fatigue strengths are also noted. (a) Unique amoung the different types of loading, direct tension loading of unnotched specimens does not exhibit this decrease in fatigue strength with increase in size, For notched direct tension stresses, however, there is a decrease and this decrease as well as for other types of loading is explained by the stress gradient effect. (see below)94 STABILIZING EFFECT (Stress-Gradient) The greater fatigue strengths of the smaller cross-section specimens is explained by the stabilizing effect of the lower stressed inner fibres on the higher stressed outer fibres, This stabilizing effect is dependent on the stress gradient and.its mechanism is explained by the following action, (1) The slip at the most highly stressed fibre would not be the same if the entire cross-section where stressed at a constant value and (2) there are different distributions of stresses from the higher to the lower ones. The theoretical treatment of this problem is due to H. A. Philipp, 1942. His assumptions are based on the stress distributions shown in the figures below: It can be seen that with decrease in test size the fatigue strength could not increase indefinitely but only to a value corresponding to the distribution shown in figure C, From this distribution an upper limit for flexural and bending fatigue strengths are obtained and are equal to 1.5 times direct stress fatigue limits for square bars and 1,7 times direct stress fatigue limits for circular bars. A further assumption was that the depth of the stabilized layer(s) was a constant dependent upon the shape of the material and type of loading. Philipp calculated the value for (s), for a few types of materials and loads, all of which proved to be in agreement with test results. S - inches Material and Loading 0.12 Steels - Bending and Flexure -06-. 08 Steels - Torsion +04 Light alloys (el ste! s \ ls Ss (8.) - = REVERSED FLEXURE (A) O,= " (DIRECT) TENSION95 4,22 One other explanation for "size-effect" is based on the probability of occurence of foreign particles in our other~ wise homogeneous materials that we use in fabricating structural members. The greater the number of foreign particles, acting as stress raisers in our parts, the greater the probability of failure, It is also assumed that the larger the part the greater will be the number of these small stress raisers. This explanation has been proposed along with test results on large billets and is explained diagram- matically as shown below: LARGE SPECIMENS QONTAIN ALL NEAR-SURFACE DEFECTS (FATIGUE STRENGTH FULL SIZE 1S LowesT ) SMALLER COUPONS: CUT FROM LARGE BULET. SOME SPECIMENS COWTHIN ONLY ONE OR TWO SUCH DEFECTS. (PROBABILITY OF HIGHER FATIGUE STRENGTH,) An empirical equation that has been found to have reasonable agreement with the maprity of test data for bending fatigue strengths is also given. Originally this equation was derived from the comparisons of large diameter specimens to the standard (RR Moore) rotating beam specimens 0.3 inches in diameter. However, it has been found to be valid within 10% for predicting endurance limits of large diameter specimens from the known endurance limits of small specimens in the range of 3/16" to 1/2" in diameter for the small specimens.96 Om = On (0-28 eee Where Oru = is the calculated endurance limit in bending for large diameter specimens Or = endurance limit for small standard, specimens (e.g. 0.3" d. R, R, Moore) D,d = Diameter in inches of large and small specimens respectively, Testing to Analyze Service Fatigue Failures The procedures and precautions to be taken in this type of testing will naturally follow those that have previously been discussed. It is mainly through this type of testing that we are able to put much faith in our designing for limited life, Ib is usually rather difficult to duplicate service fatigue frac- tures in the laboratory. However, when this condition has been accomplished by means of a careful analysis of frequency of loading then the accelerated laboratory test may be equated to the service life of the part, In this respect the statistical laboratory loading spectrum may be used for similar affected structures throughout the aircraft, When more data become available from analyses of this kind it will be possible for the designer to evaluate his design with much greater accuracy.FIG. 53 — SCALE EFFECT, LIGHT ALLOYS 96.1 a 230 7 é REVERSED | BENDING v, UNNOTCHED a 25 x ' E 2 20) BENDIN § AXIAL Z NOTCHED UNNOTCHED Z| 15 ec ‘0 2 a z wW 10 TORSION “No AXIAL. UNNOTCHED “| _f NOTCHED — 5 °c 2 3 4 5 G SPECIMEN TEST DIAM~IN.7 Appendix 18 Comparative testing has been shown to be the most practical type of testing to be used for determining the best attachment methods for struc- tural joints, Further, this method has also been very useful in evaluating the production tooling for hot forming 75ST rivet dimpling dies. The following example is used for illustrative purpose: Suppose a machine countersunk riveted wing joint was experiencing fatigue cracking at an average of 10,000 hours of service. Also suppose, that other riveting methods were feasible for this joint and that laboratory tests on riveted panel. joints indicated that dimpled sheets were superior to m'c'sk sheets, Figure 5: shows the fatigue strength characteristics of the various riveting methods. ‘The stress analysis for this particular wing joint indicates that the maximum working stresses vary from a maximum of 25,000 psi to a maximum of 10,000 psi, For the purposes of our calctlations* we will assume the average maxi- mum working stress that causes the greatest amount of fatigue damage is 1/2 (25,000 + 10,000) or 17,500 psi. From theS/N curve at this stress level the following calculation for dimpled riveted joints may now be made: Hp = (22) we eee A) uy 19,000 = 10,000 (191000 acy 23,700 Hrs, ¥ In subsequent notes we will go into the statistically determined loads ex- perienced by aircraft structures in service. In order to make calcula- tions such as the above it will be necessary to have a more realistic evalua- ton of the actual loads applied in the operational stressing of aircraft.98 Appendix 19 ‘The designer has been cautioned in the use of fatigue strengths as obtained from small sub-size laboratory specimens, The following ex- ample shows very good agreement between the actual fatigue strengths and the predicted fatigue strengths of large sections from small laboratory coupons; but it should not be inferred that such good results are always obtained, The S/N curves of figure55 show the fatigue strength character- istics of shot peened 14ST forged material as obtained on small 3/4" diameter rods tested in rotating bending. The dashed curve is the predicted S/N curve for the large 2 1/4" diameter fittings, geometrically similar in dimen- sions, as calculated from the equation. On =O, -(1 SG 2) 15 The few plotted points from actual tests on the large shot peened fittings are also shown on the graph and it can be seen that the agreement is good. The above equation, in this example, has been used to calculate fatigue Strengths for limited lifes - originally the equation was only meant to be used for endurance limit calculations. The similarity of the large and small specimens was controlled by the following ratios: 50 0.161 2.25 0,725 =) 0s22 and from Oy, OL 0,89 Os99 Appendix 20 A good example of a test relating the behavior of a composite struc- ture to that of its component parts was conducted by DACo in 1945, The frame to stringer intersection at the floor line was believed to be marginal due to repeated cabin pressure cycling. Various small laboratory speci~ mens were tested in order to develop a satisfactory design, Ultimately a full scale section of a fuselage was tested. Remarkable correlation was ob- tained between the small specimen tes ts and the full scale section, This study was conceived in the design stages of an airplane and was not the result of fatigue failures occurring after the component part had cracked in service. For the same average working loads both the small coupons and full scale section developed fatigue cracking in 29,000 cycles. See figure 56. This correlation confirmed the original views on the questionable joint and steps were immediately taken to improve the design, The design improve- ment was then determined with the smaller laboratory coupons and on a comparative basis the improved life for the full scale fuselage was established,99.1099.4% Appendix 21 Load-Spectrum Testing This type of testing has been gaining popularity in the past 12 years and requires the least amount of calculation, However, it is often based on the assumption that the working loads which our structures will experience are known, In some cases the service loading is fairly accurately known while for others, gust loading for example, more data are needed, Research by thN.A.C.A, and the Military are constantly being carried on. Soon possibly gust loads and their frequency distributions will be firmly estab- lished. The statistics that are available on gust loading will be given in Chapter 9. For the purpose of describing this type of testing, the following example is shown. Essentially it is a problem of initiating into the test program the mixture of high and low loads that will be encountered in service and to reproduce a certain number of load steps and their corresponding. frequencies. Figure (a) below shows the actual record of the load variations in operation and Figure (b) shows the reproduction of these loads for test. FIG. (A) LOAD FACTOR (B)99.2 Now since the actual operating loads are constantly varying and randomly displaced it is as accurate and as practical, for test purposes, to group equal magnitude loads according to their frequencies. These loads may then be applied to test specimens as shown by the diagram in Figure (c). In turn, this grouping of loads may be called a "part sequence!’ representing a certain number of operational hours and in testing may be repeated over and over until failure occurs. The life in hours is then determined from the number of part sequences applied and their corresponding time of life in hours. LOAD REVERSALS LOAD FACTOR PART SEQUENCE =H, HOURS FIG. (C)100 Chapter 8 DETAIL DESIGN FOR FATIGUE IN AIRCRAFT STRUCTURAL JOINTS ‘This section of notes has been taken from a lecture given by Mr. E Harpoothian (DACo) at the A. T, A¥ Conference held in Santa Monica, California on May 11, 1954. The notes will be given in the form of a des- cription ami discussion of the various diagrams and slides that were shown at that meeting. The design guides that were presented in this lecture have been previously itemized and are given on page 78, The guides (rules) that have been established were derived through the careful study of the follow~ ing collection of laboratory fatigue tests. Figure 57 The S/N curves compare the fatigue strengths of 14ST, 24ST and 75ST plate, rod extruded and forged products. Both notched and unnotched speci~ mens are shown as well as the large amount of scatter that should often be expected, An important conclusion that can be made here is that large gains in fatigue life can be expected by a reduction in stress concentrations that may result from design improvements. For instance a reduction from 17,000 psi to 11,000 psi can result in a 10~fold improvement in life. Further the data shows that there is no particular advantage in using one or the other of the materials from the standpoint of their resistance to fatigue loading. How- ever, since the scatter of test data is so large here, a more conclusive check of the fatigue properties as affected by directionality and size will be neces- sary for design purposes. This can be shown in the next figure. Figure 58 The S/N curves are for single-rivet lap joints (actually not very effi- cient joints but they may be used for comparative purposes), The curves are for three separate materials and tempers, 75ST, 24ST and 14ST, The results all fall within a fairly narrow sactter band and indicate that no one type of alloy is superior to the others. Figure 59 This figure shows the details of a box beam specimen for a flexural fatigue test. The beam specimens were loaded as shown in the diagram and are representative of aircraft type riveted structures, The test data for these box beams are shown in figure 60, Figure 60 ‘The materials used in the fabrication of the box beams are as noted on the graph. The results from this test are similar to the previous tests in that the fatigue strengths for all the alloys used fall within a fairly narrow scatter band,101 Figure 61 To this point, we have discussed the insignificant effect of material changes on the fatigue strengths of our designs. We should be now realize that a change of material, in our light alloys, is not necessarily going to solve our problem, Any small differences scen as a function of material change will always be over-ridden by a change in design. Figure 58 and 60 have revealed to us the small differences between 24ST and 75ST in fatigue, In this figure (61) however is the evidence of the effect of design of joint on the fatigue strength. (The lower (and we mean low) S/N curve shows good reason for not using single row riveted lap joints. It is true that the flexing due to the eccentricity of this type of joint may account for some of the de~ crease in fatigue strength. However it is still an inefficient joint, Figure 62 The S/N curves in this diagram show the effect of rivet type on the fatigue strengths of joints. It may be inferred from this graph that the sharp edges of m'c'sk. sheet have higher stress concentrations than dimpled or plain hole sheets; with the final result that m'c'sk, riveted joints are the most critical, Figure 63 A comparison is shown between the S/N curves for notched 24ST and 78ST axially loaded sheet in .091 thickness, The notch has aK, = 4.0, which is high, and never reached in good design. The S/N curves show an advantage with 75ST in the high stress Fange but no difference in the lower working stress ranges and beyond about 1000 cycles. Figure 64 The S/N curves of this diagram are for 24ST and 75ST bare aluminum alloy sheet (.091) and comparisons between the different alloys (unnotched) and different testing agencies can be made, Figure 65 The S/N curves in this diagram show the effect of sharp corners and core material, Note approximately a 10-fold increase in life from the .015 to 1,0 inch fillet radius, Figure 66 This diagram shows the block specimens and fatigue test results on extruded 24ST, 14ST and 75ST material. The mean stress for these tests is 13,000 psi; which is a typical average mean stress for most commercial transport wing analysis, ‘The results of all of the materials fall within a fairly narrow scatter band and it can be seen that all three materials will re- sist a epperimposed variable strese of #8100 psi for one million cycles (area = 1.23 in’102 Figure 67 This figure shows two types of high strength bolted joints that are representative of spar cap splices - a single~scarf and double-scarf joint, The materials used for the joints were extruded bar stock of 24ST and 75ST. The joints were tested at various variable stresses superimposed on a mean stress of 13,300 psi. (16,000 pounds), typical of 1G wing struc- ture stresses. The test results for the joints are shown in figure 68, Figure 68 The fatigue test results of the joints in figure 67 are shown in this graph, The variable test loads for all of the joints was superimposed ona mean load of 16,000 pounds (13, 300 psi), The double-scarf joint has the highest fatigue strength and shows advantages for use in design, The bolts loaded in double shear greatly contribute to the higher fatigue strengths of the double-scarf joint. Comparison of joint and material, as in figure 66, indicate that the advantages do not Iie in the choice of material in this ex- ample but is instead a choice in design,103 SUMMARY ‘The load-cycle curve of Figure 69 shows the test results for many types of aircraft structures and component parts. The interesting feature of this plot is that the test results all fall within a fairly narrow scatter band for the many types of structures; even though the test cycle stress ratios for the specimens are slightly different. It should be noted that the abscissa of this graph is - max. load as a percentage of ultimate strength, This type of plot is often used for material comparison purposes and for cases where calculations of stresses would be questionable. A more usable graph would be one using max, stress from which actual working stresses could be extracted in making an analysis, It should not be inferred from this graph that all structures have approximately the same lifes. It should be realized that the working stresses at some percentage of ultimate strength is not necessarily the same magnitude 08 stress for another structure at that same % of ultimate, No two aircraft structures from different models will have identical service lifes, Different manufacturers use slightly different working stresses and design guides (rules). It is the aim of all, however, to develop the most efficient structure and in so doing the highest working stresses feasible should be used for the materials available to all of us. Ideally, from the standpoint of economy in maintenance and reliability of airworthiness, the manufacturer with the best designers is going to get the business (and stay in business).60,000 on rewrne tea na 40.000] 30,000] Stress, ps 20,000] 10000) 1 ee 0 D1 Cycles FIG.57 Comparison of fatigue strengths of 14S, 245, and 75S. a Yar 245 Tal bratier head rls 1.400 sive indie ates = 1,200] 0.064" x 1" Alclad sheet 2 All tests made with compete 3.000 reversal of stress = co ‘Aeiad 758-T6-[ SR ots sts z +5 4 Aad 248-81 ee — + Neled 145-70 400 . et iad 285 Hild 265 -TO6~SRE_ APRS AS-TS 200 Sees ‘ . ort 1010" 0° 10" 10" to” ig” a0" 20 eles Fig.88 Shear fatigue tests of aluminum alloy riveted joints \O3Apai Computed extreme fiber stress, seeranen nil SRST 33 40,000) - A ests made wit complete 36,000} ——1___ erat ot ste (chess oes 1) Legend 22,000] © 246-13, x 2 2asctas \ B 248-181 (ges see 28,000] TT H 248-781 Ggee ater ‘aoreston, 24,000) aa [Computes buchngsrengi of cover 373816 Dates eansered ss ued nt rat tnes \o\, aoe Oe at a om 20,000 16,000 Di AI75-13 ree oe ) y 12,000 3 ; od epson eiceeeeteeesemeteesemeeseeeeecssieteeseseteeaises Number of eyes to failure 10 10 Fic.6O Results of flexural fatigue tests on aluminum alloy box beams, 103.2-1000_PsI. stress: TENSION 24STAL # 75STAL_RWETED SHEET so JONT COMPARISONS, % ; i | | [oousne | fe Sale aut conves Deawa Nee MEAN OF “SCAT 0 ta | > ht ‘ | ol ‘0 10° LHhroe ‘STRESS__CYCLES FIG.61 ’ Wok by Linch wide. Rivets, Btnchdlameta pi cette eee eee ste pt atin she oor rae 103.3-1o00_ RSI. STRESS 103.3 ‘2USTAL # TSSTAL_RWETED SHEET 30 JOINT. COMPARISONS, ase \GE | % HNL NP ae a Se aoe % ALL coves | DRAWN ARE MEAN OF “SCAT TENSION 10 (ane ie he > i f 408 ilo! ‘STRESS CYCLES FIG.6I 0. 4 62 arnee ree? peas ot snne tle ee ids, pete, eas OTE ate 94 ct103.4 20 : | ALUMINM-ALLOY SWEET SPECIMEND. 4.05 Rel, FATIGUE TESTS 75 ST MATERIAL 24 ST MATERIAL 75 ST BARE SHEET (REE Aca rn ewe) 081 3 bs BATTE | 24 ST BARE SHEET I! (O91 BAT TEULE NA MAX. STRESS PSI x 1/1000 10" 10) 10 10” N= CYCLES TO FAILURE (R=0) FIG.64103.5 EEREPB SH Fiere6 cry1036 Diretctres face curves for varie hlebstrongt sluninunvalioy ots oats FIG.68103.7 3uMiiva OL S313 Ol Ol gOl Ol yOl Ol 20! Se of 02 4 Ov wy f+ 4 'SO113 G940d 1SSL 8 ag" “ade | a “YLINWNDV Stl 6 de AN « SHVTIOD DUOL Yvad . no » SNIM " w HOVLLY os " $9114 uv3a9 ” ah ; anv 69 Od XQ) tN ONIM ANVIdIvL © go -odn< xox oor HISNaeLS aLWWILIN %=- avo “xXvW104 Appendix 22 For the designer's convenience, the following set of S/N curves for various materials are presented. With the proper stress concentration factors the designer may apply his particular design to the materials noted in making rough estimates of finite lifes for his part. It is not always practical to test each and every part and for these cases the following fatigue strength properties may be used as anaid. They may also be used as an aid in planning a test loading program and in this sense prevent the testing engineer and designer from wasting costly test specimens by the improper choice of an initial experimental load,104./ “FANUVA =OL — SFITIAI Ol 9°! Ol Ol VIMF * ++] | 4~— TTT H-S7. ; oz 1-99 ov (a00%a. 2 SIMMS 9-H (20006) ofp ws a Waon oo! ° 9 (0002) ovSh 106 ozi|: aL “8/7 Ovi NOISNFIL TIX “XVW I) TSM - SSFALOL 08 joo! Ho2! lovl O09! jO8l |00¢105 Chapter 9 FATIGUE LOADING AND ANALYSIS OF AIRCRAFT STRUCTURES Statistica? Icon prowe anyshing by statistics except fhe, truth. canning 9. The majority of loads that may cause fatigue cracking of our structural members can be classified in either one or the other of the following categories: 1, Flight loads such as atmospheric gusts, maneuvers and buffeting. These are irregular cycling loads varying in magnitudes and frequencies. 2. Ground loads as experienced in landings, take-offs and taxiing. 3. Vibrational loads as engine and propeller vibrations, These loads may be fairly constant and are considered to cycle constantly (infinite number of cycles). 4, Miscellaneous loads such as cyclic cabin pressurization, fuel surge loads in wing tanks, pressure cycling in hydraulic cylinders and lines, others...... 9.1 Atmospheric gusts and normal maneuvers are the principle loadings which cause the greatest amount of fatigue cracking in aircraft wing structures; and therefore may be classified as the most important to us. It is generally agreed that more accurate gust data, as well as a more suitable method of interpreting the atmospheric gust phenomenon from airplane response, are urgently needed, In the United States we are presently engaged in this study with V-g-h instruments of NACA origin. "The c.g. acceleration has been selected in preference to a strain-rating of the wing since it is less influenced by local conditions such as weight distribution, dynamic response and the particular characteristics of the aircraft on which the data are being obtained." (Quote from A, T.C, Fatigue Sub-Committee Recommendations, P,2B-Page 3, 1951.) Some aircraft companies have developed strain-range counters which serve their particular problem, However, these data are not applicable to other types of transport aircraft, The aircraft companies in this country have asked the commercial airlines, as well as the military, to cooperate in obtaining flight data with presently available and later improved recording instruments. In order to minimize the labor in the reduction of the vast amount of data that would necessarily be collected it has been recommended that IBM accounting machines be employed, Until such refinements have been made, however, there is no reason why we should not use the data presently available.105 Chapter 9 FATIGUE LOADING AND ANALYSIS OF AIRCRAFT STRUCTURES Statistics? I can prove anything ty statistics except the truth. © = Ga canning 9. The majority of loads that may cause fatigue cracking of our structural members can be Classified in either one or the other of the following categories: 1. Flight loads such as atmospheric gusts, maneuvers and buffeting. These are irregular cycling loads varying in magnitudes and frequencies. 2. Ground loads as experienced in landings, take-offs and taxiing. 3. Vibrational loads as engine and propeller vibrations, These loads may be fairly constant and are considered to cycle constantly (infinite number of cycles). 4, Miscellaneous loads such as cyclic cabin pressurization, fuel surge loads in wing tanks, pressure cycling in hydraulic cylinders and lines, others. 9.1 Atmospheric gusts and normal maneuvers are the principle loadings which cause the greatest amount of fatigue cracking in aircraft wing structures; and therefore may be classified as the most important to us. Itis generally agreed that more accurate gust data, as well as a more suitable method of interpreting the atmospheric gust phenomenon from airplane response, are urgently needed. In the United States we are presently engaged in this study with V-g-h instruments of NACA origin, "The c.g. acceleration has been selected in preference to a strain-rating of the wing since itis less influenced by local conditions such as weight distribution, dynamic response and the particular characteristics of the aircraft on which the data are being obtained." (Quote from A, T.C. Fatigue Sub-Committee Recommendations, P.2B-Page 3, 1951.) Some aircraft companies have developed strain-range counters which serve their particular problem. However, these data are not applicable to other types of transport aircraft. The aircraft companies in this country’have asked the commercial airlines, as well as the military, to cooperate in obtaining flight data with presently available and later improved recording instruments. In order to minimize the labor in the reduction of the vast amount of data that would necessarily be collected it has been recommended that IBM accounting machines be employed, Until such refinements have been made, however, there is no reason why we should not use the data presently available.106 The aircraft designer's task would be much easier if the atmosphere was composed of either a stationary mass of air or one traveling at a constant velocity. Instead the air moves in a very irregular manner.and in all directions. As far as our structural design is concerned this movement of air is only important to us when there is a large change in velocity over a relatively short distance. These movements of air are described as (U) "Gusts" and the simplest concept upon which the gust equation is based is the sharp-edged gust shown in Figure (a) below. However, since it is unlikely that such an abrupt discontinuity would ever occur in nature, the concept of a graded-gust as shown in (b) has been devised. The envelope of U.S, data has suggested the following equation for graded-gusts which may be used in approximating actual gust velocities from equivalent sharp-edged gusts. Notably, the British have used this equation in setting up Airworthiness Requirements. (a) (8) AIR MOVING AIR_MOVING AT CONSTANT AT CONSTANT, velocity. VELOCITY STATIONARY AIR SHARP EDGED Sauce Ruse GRADED GUST For simplicity the gust load factor equation used in design, in this country, is based on the sharp-edged gust and the structure incremental limit load factor is calculated froma 30 ft. per second gust as: — KUV_ An = ma 375 (F) where + KU = effective sharp-edged gust (Ug) in fps. indicated airspeed - mph = slope of lift curve = wing loading+G 107 The result of gust loads on aircraft structures may now be classified into two categorie (a) Single gust life of structure (b) Fatigue crack life of structure due to frequently applied gusts vary- ing in magnitude from 0 to 30 fps. (IG. to limit gust load factor) 9.11 The single gust life of a structure is largely controlled by the pilot and for this reason the placard maximum permissible indicated level flight, diving and gliding speeds are indicated, In selecting a speed for operation in severe turbulence a compromise must be made between two limitations. It is desirable to keep the speed low to permit the structure to withstand the greatest possible gust velocities. At the same time, itis desirable to maintain a speed high enough to prevent gusts associated with turbulence from causing the airplane to stall. (Reduce speed in severely turbulent air.) Reference to the figure below illustrates the fact that, flying at a given weight, the severity of acceleration produced by a gust of a given velocity is a function of the flying speed at which the gust is encountered, VA EXPECTED VELOCITY MPH, —=— (The following is a quotation from DAco plublication on Basic Information regarding the Strength and Operation of DC-3 Airplanes, 1946). "All of the precautions and operating practices and safety facilities that have grown up around normal commercial operations over the last eight years have been conditioned108 by these facts. Avoidance of severe weather conditions, warnings to fly slow in rough air, skill in maintaining control and recovering control without picking up excessive speed in recovering, weather reporting facilities, radio, ete., are all seen to be related to the design of the airplane itself, Adherence to what are known to be sound operating practices will continue to be the only way to maintain the excellent record of structural adequacy which has been established for American commercial aircraft, DC-35 included. 9.12 The gust loads which are of most interest, pertaining to the fatigue of structures, are those over which the pilot has little or no control. These gusts, randomly dispersed, are relatively small in magnitude but so frequently applied that they contribute to the fatigue loading of a structure and may ultimately cause fatigue cracking, About the earliest known source of V-g data was that obtained by H.W. Kaul in Germany. The earliest methods of obtaining acceleration data used the scratched smoked-glass method; however, the latest methods now use counting accelerometers, A refinement of the present method, is the V-g-h record which accounts for frequency and magnitude of gust at altitude. However, these data are still meager. The statistical reduction of all acceptable gust data may be shown between the upper and lower limits of the graph below. At present this distribution may be used as the probably gust loading (fatigue loading) for our wing structures, 5 US, BRITISH ¢ GERMAN DATA FREQUENCY oe fey 10 RELATIVE 10 20 30 £U,-FRS.109 For purposes of analysis the above distribution of frequency and magnitude of gust may be reduced to the following tabular form, (U.S, data). Effective Gust Velocity - f.p. Relative Frequency 42,25 24, 350,000 26,75 540,000 11.25 50,000 15.75 5,000 20,25 500 24,75 50 429,25 10 £5, 000, 000 ‘The above table represents ten million gust; 5,000,000 up and 5,000,000 down, About 1% of the gusts are not accounted for in this accumulative frequency. Load factors from normal maneuvers are included in the above distribution, therefore there is no need for handling normal maneuver counts separately. Commercial aircraft will usually make from one to two 90° turns on take-off and two 90° turns in landing. The maximum bank angle is approximately 30°. The load factor may be calculated as 1__. Onoccasion the aircraft may make a bank cos. 30° as great as 40 - 50°, possibly once or twice a year, but the occurrence is so seldom that it is usually neglected. There is no flight data available to sub~ stantiate this value; it has been arbitrarily assumed from talks with pilots. For commercial aircraft, at least, itis felt that the frequency of maneuvers can be approximated from the number of landings and take-offs. (See paragraph 9.2) and that their importance in a fatigue analysis does not always merit detailed isolation. A recent report (NACA TN3086 Page 4) indicates 15% maneuvers and 85% gusts for the total rough air time experienced by commercial transports. 9.13 The majority of data that have been gathered from experimental flights, and upon which this distribution is based, has been for aircraft flying between sea level and 12,000 ft. It has now been definitely established that the atmosphere consists of varying turbulent layers and that both the frequency and magnitude of gusts decrease with an increase in altitude. This effect is shown in the graphs below: 50 x 50 ¢ 40}2 cle + FREQUENCY 2 30h 4 30} INTENSITY 2074 20 = ofE lore inapaen TOF IND. Ue HI, ALT. [<__ PERCENT ROUGH” AIR IND. Ue LOW ALT. 8 1624 ° 4 ib110 For the analysis given in the Appendix of this chapter, however, the effect of altitude of operation will not be considered. This effect will be covered in a later chapter. Assuming that the distributions of gust loads as given in the table above are accurate, the next step is the conversion of frequency and magnitude of gusts to miles or hours of operation, On a mile-per-gust basis this may be shown as in the figure below. 10% - 104 _ Bhi0% i 7 VEL. Wi}10% HOO} n GUST_SHAPE Bhio 2 7 _ 10s 10 5 20 25 30 tu, FPS. For analysis, however, it is far more convenient to equate a given number of gusts into hours of operation and this conversion is accomplished by the use of the NACA equation: RL F = 5,280 22 nace cere nee(l) ue where F = total number of gusts (say 1,000,000) R = gustiness path ratio (interval dependent upon terrain: 0,24 for over mountainous regions, 0,006 for over water smooth flights ( Use 0. 10-for world-wide conditions), L = miles flown in experiencing (F) number of gusts = as a function of airplane size the NACA determined that commercial transports will experience a measurable gust in flying a distance equal to ll times its M.A.C, This value is given in feet in the equation,1 In the above equation the value of (L) miles flown is usually calculated, From this value of (L) and an average block speed for the aircraft, the hours of operation in experiencing (F) number of gusts may be determined from: Hy Vv wenn wenn nen(2) 9.2 Ground loads as experienced in landings, take-offs and taxiing are not as well established as are gust loads. The design load criteria due to landing impact and the like have been regulated but the cyclically varying loads due to the aircraft moving along the landing or take-off strip are relatively unknown. The A.T,C. Recommendations to the NACA and Military have requested the experimental determination of these loads and possibly they may soon be available. Some data on DC3s that are available are shown in Figures 74 to 76. Even though the cyclically varying landing loads are unknown, there may be some use in the knowledge of the number of landings and take-offs made per year per aircraft. This data may be obtained from the statements of the Airline Transport Association - Finance and Accounting Conference. Up to and including December 1953, the statements revealed the following figures from which the "number of landings" calculations were made, DC-6, DC-6B Avg. Use per Day Avg. Hrs. per Flight Avg.Hrs. Year Reg. Combination Coach, 9:27 hrs. 2:09 3,340 Tourist and Cargo ‘The number of landings and the number of take-offs per airplane can be calculated from, a ee Drs. per Year = 1,640 Hrs. per Flight 9.3 Vibrational loads as induced by the engine or propeller r.p.m, are rather difficult to determine, Such items as brackets and clips may be affected by this loading and it is usually best to conduct a static calibration test for the determination of the magnitude of load, The frequency of loading should be considered infinite since many billions of cycles may be accumulated over a period of only a few years. (Example: 3,000 rpm x 60 minutes x 3300 hrs. per year x 10 years = 5, 940,000,000 cycles.) The repetition of even a small load for the above length of time has been known to result in fatigue cracking. 9.4 Under the heading of miscellaneous cyclic loading may be considered the loads due to (1) cabin pressurization, (2) pressure loading in hydraulic systems, (3) operational and functional loading as in control systems, (4) heat cycling and loading of engine and turbine components, etc.112 No specific statistical loading spectrum can be given for these conditions as they will vary with the type and operation of the particular aircraft. However, for transport aircraft, the design limit operating cabin pressure is often assumed to cycle once for each hour of flight. There is little or no data for this particular loading distribution; however, it is believed that a cycle per hour would at present be conservative for analysis purposes. Since the statistical average flight time in hours per year is known (see paragraph 9.2), it follows that the number of pressurization cycles per year are known, The above frequencies may then be applied to all fuselage structures (doors, windows, floors, etc.) that are loaded during pressurization, On some models a maximum design working stress of 25,000 psi has been recommended for typical fittings and components that are directly loaded from cabin pressure loads; and maximum stresses of 28,000 psi for the supporting structures immediately affected and loaded through such fittings or members, These stress values have purposely been mentioned to remind the designer that sometimes relatively low work- ing stresses need to be used'even though the tensile strengths of the materials are in the order of 70,000 psi. Cyclic pressure loads in hydraulic cylinders and hydraulic lines may vary with the particular design or operation, In these cases, a mock-up model of the actual system may be set up and the actual magnitudes of peak pressures determined with the aid of calibrated "pressure pick-ups", 9.5 Ibis believed that loads from flutter should not be considered as a problem in fatigue. Flutter, if present, may cause failure which is not associated with fatigue but rather with excessive build-up of loading.U3 INVilddlv:U4 160 200 120 40 240 80Ws116 Appendix 23 Although the available methods of accurately predicting fatigue cracking of structural members have not been too successful they do, at least, give rough approximations and may possibly aid the designer in determining what relative degree of improvement he needs, The following example is an actual case and shows very good agreement between the predicted fatigue cracking by analysis and test and the fatigue cracking due to operational stressing; however, it should not be inferred that such good agreement always occurs. Although there are opinions to the contrary, it is believed that attempted analyses of this kind are necessary. Not until this step is made will the refinements and necessary corrections be determined; which, we hope, may ultimately lead to the optimum design. The example is for a particular wing joint of a given model. The fatigue analysis of the joint later proved to be marginal. It should be realized however that when this particular joint was originally designed there were no available fatigue analysis methods. Fatigue of the structure from atmospheric gusts had been considered and, in the light of present day knowledge, a very satisfactory design was accomplished, A laboratory fatigue test of the particular wing joint was conducted and the S/N curve showing the fatigue characteristics of this joint are shown in Figure 76, The envelope curves in this figure are the maximum and minimum failure stress levels for the tested joints. Superimposed within this envelope is the statistical gust loading spectrum for the particular joint. The graph of Figure 7%and the table below now show the analysis method of making a fatigue crack-life prediction. The gust statistics of the NACA are reduced to the corresponding working stresses of the joint in question. The accumulative gust cycles are then compared to the stress cycles for which the joint is tested and either graphically or mathematically the crack life is calculated, From the NACA equation: F = 5280 = and for a total of 10° gusts u it can be calculated that the particular aircraft will fly (L) miles in experiencing this number of gusts. 10° = 520 (0-2) (L) (11.) (13.6) ‘ L = 283,000 Miles From Hy == and assuming an average block speed of 250 mph v Ho = 1,130 hrs.11? Using the "part sequence" of 500,000 gusts which now corresponds to Hy = 1,130 hrs, the following calculation is made: u, » N a ECs max (gusts) sw (test) es — — N 42,25 13,975 £435, 000 int. 0 6.75 15, 920 54, 000 1,000, 000 + 00540 11,25 17, 870 5,000 47, 000 110600* 15.75 19, 820 500 19, 000 102630 20.25 21, 780 50 9, 600 100520 24.75 23, 700 5 5,400 £00093 229,25 25, 900 1 a : The ratio 3-7, from the cumulative damage concept (Ref. Chap. 3) indicates that the “particular joint is only 7% damaged or that 7% of its useful life, before cracking, has been expended, From Hy = 1,130 hrs, in experiencing +500, 000 gusts therefore, He = oS = 182 = 7,900 nes. — 0.143, N The calculation proves to be low (conservative) since actual cracking in the skin and doublers of this joint occurred between 10,000 and 12,000 hrs, The designer should realize that the "state of the art" in making a fatigue analysis of this sort has not been developed to any high degree of accuracy. However, redundancy of structure lessens the severity of fatigue cracking and in most cases the normal inspection and maintenance periods will uncover the situation long before cracking has progressed to the danger point, (Hindsight, is a wonderful thing but, if possible, should not be used as a crutch in structural fatigue design.) * Note that the greatest amount of damage is caused by gusts between 6 and 15 fps. This has been found to be typical in all analyses made of this kind.M8 ° VIS - SI8ILE WOSNIL o a lof119 Chapter 10 TRENDS IN COMMERCIAL AIRCRAFT OPERATION AND DESIGN *ginplicate and aad Lightness” " Hi Stout 10.1 This section of notes briefly discusses the major trends that are becoming increasingly important to us in designing for fatigue, Individually, the effect of some of the trends will be a lengthening of service life; while for others, a foreshortening of life can be expected as compared to present day aircraft structures. It is believed that the designer familiar with the following trends will be better equipped to design for trouble-free service. 10.2 The trends that are discussed are accompanied with graphs, reference Figures 7 to 80. In some cases the values have been estimated from the best information available and in all cases the values represent design conditions rather than averages. For example, the design cruise speeds for some models range between 310 and 340 mph; yet a more realistic value for year in and year out operation is an average speed of 250 mph. It is believed that a proportionate trend for average working and operating conditions may be assumed from the design values, In the past 20 years the trends in commercial aircraft design have been as follows. 1, A decrease in positive design load factors from a maximum in the order of 5 or 6 to the present average of approximately 3.8. Similarly the negative load factors have decreased from -2,0 to =1.8, 2, With this decrease in load factors, however, wing surface stresses have increased, Tension stresses at design load factors have increased from 22,500 psi to approximately 61,000 psi, The one-load factor level flight stresses have also increased but not in the same proportions as the limit working stresses and design load stresses, This has resulted in a steadily increasing range of cycling stress from the level flight to the maximum working stress levels. In general, for the same operation or service, this effect is in the direction of shorter fatigue lifes for structural members.DESIGN LOAD FACTOR WING SURFACE STRESS- KSI. 120 1960 cS T FIG. 77 ne G.W.> 25000 oo ! 5 oS B83 L SSp4 / oN. 4-1 / 4 Dé / | D5 D4 | ~Dg-I,407_ __| DSB — — péA—— — + 3 1934 “40 ‘so | DESIGN PERIOD - YEAR 70 60 2 BOT. SURF. DES. 2 ie 7 FIG. 78 40 oF AO INCREASING —> 20% s | [oo abo +GO pony . rlALTITUDE - FT. 40000 30000 121 The trend toward higher altitude of operation is at first a com- forting thought, Meteorologists have determined the most turbulent region to be the troposphere, next the tropopause and the least turbulent the stratosphere. Aircraft flying 30,000 feet or more will therefore experience less turbulence, Statistical data have indicated that not only the frequency of rough air is reduced but also the intensity (See page 109), However, we should not lose sight of the fact that in reaching high altitudes of enroute operation the troposphere, or most turbulent region, has to be flown through both in climbing and in descent, Furthe: been shown that a considerable amount of the total flight ti is spent in the climb and descent and therefore the only gain that is realized is in the enroute hi-altitude stage. it has Trends in airspeed and range of flight are shown in figure 80, ‘The effect of increased range of flight for hi-altitude operation is in the dirgction of increased life or hours of operation since more and moresWill be spent in the least turbulent area, This effect will be considered in the example and calculations of Appendix 24, The effect of increased airspeed is two-fold. In Chapter 9 we discussed the importance of reducing speed in turbulent regions. Increased airspeed in turbulent regions will not only magnify the effect of a gust but it will decrease the number of hours flown in expericncing a given number of gusts. The NACA equation of paragraph 9. 13 indicates that the number or frequency of gusts is a function of miles flown; the wing chord being-a geometric parameter of the gust interval (gust interval= distance flown between measurable gusts.) oe 8 COMMERCIAL OPERATION of at —— DESIGN VALUES VV. FIG.79 & “{ ea! a. 6. ON ? Ti Hy coasts / oth ° ° ° of e Dees, 1 of FIG: 79 10000$-+-= =<] — | aesy | ¢i YV 1935 40 “50 1960122 TRENDS IN AIRSPEED ND_MILES FLOWN 1600 (NOT AVERAGES) E 50 z § 400 a x i 1200 = { | a =4 Wigg9 1 00 fr ===> o MILES PER’ TURBO PROP a | z t 2 a FLIGHT 1 ¢ JET 3 800! : TRANSPORTS g B30 | t & 600 : AIRSPEED @ : = 400 = 200 2 | 200 +3 1935 39 4) 4345 47 ~~49~ SI 'S3« 55 57 FIG.80 DESIGN PERIOD - YEAR 5. The trend in hours flown per day, or more commonly referred to as the "utilization rate", is not shown, Pre-war utilization rates were in the order of 5-6 hours. The 1947-48 Frankfurt-Berlink Airlift operation described in "Fortune" magazine revealed the unprecedented figure of 8-9 hours a day. At present commercial transports are averaging over 9 hours with some airlines reaching 12 hours a day. Pre-war commercial transports were averaging around 2100 hours per year.-- today these same aircraft are flying between 3200 and 3500 hours per year. 6, Trends in design are not as easily outlined as trends in operation and performance. Twenty years ago it required 9 pounds of airplane lo lift one pound. Today, it requires less than 4 pounds of airplane to lift the pound. This is an admirable advance in weight-saving but it has been made at the expense of higher design working stresses - since this weight-saving ratio is far greater than the strength ratios of the older to the newer materials that have been developed, ‘The aircraft structures of today are far more complicated than they were a few years ago. Paradoxically, this has been brought about through public demand for greater comfort and safety. Com~- plicating structures only increases the probability of Tailire—which now, therefore, brings us toa dilemma, It is a great temptation and challenge to the young cadet engineer to employ his academic knowledge in his first designs - why do it the simple way when it's so easy to complicate it? One of the many tasks of a Chief Designer123 is the constant checking of such design conscientiousness. Period- ically he has to remind his staff of their wanderings from the direct, simple and most economical design. 7, In the past 10 years there has been a definite trend toward increased structural fatigue failures, There is no rule, however, that pre~ vents this trend from being altered, A constant consideration of fatigue loading on each and every part of our design will go far to reverse this trend, The design philosophy of diffuse structures, such as multi-spar wings, is a step in the right direction. Recently, the British have been considering a requirement that the fatigue cracking of any one member of a composite structure shall reduce its ultimate strength by no more than 12 1/2%, It is not the purpose of these notes to set up such requirements. When requirements are pro- posed they often become general rules which in turn may regulate structural design - even though they are not applicable. Require- ments used in the sense of "design guides" would be far more appropriate, With this interpretation we may list a few guides, (a) Provide for multiple load paths. (b) Give extra consideration to tension loaded fittings. (c) Apply fitting factors of safety to net stresses around holes and cutouts. (a) Test newly designed joints and compare with "time- tried” structures. (ce) Utilize longitudinal grain direction of materials whenever possible. (f) Provide generous fillets and radii, (g) Break all sharp edges. - Polish critical regions if feasible. Additional insurance for the adequacy of our designs are the inspection methods both during the fabrication of our parts and throughout its functioning in service. 10,3 In preparing the background for the fatigue studies of present and future designs it is often helpful to employ the information gained from the design, the construction and the operation of prior designed aircraft already in service. Certain basic changes from the past to the present or future designs enlarge the fatigue problem for the latter. The basic materials are often the same but higher cabin pressures, higher speeds and altitudes of operation, greater ranges of working stresses from IG to limit stress, etc., are the principle factors which require specialized study for the newer designs. Fatigue failures of our designs can be broadly classified into two types - those which affect the airworthiness of the airplane and those which catse fatigue cracking in the sense of maintenance repair costs. The former type is undoubtedly124 important since a few catastrophic events have occurred, However, too much emphasis placed upon the importance of this category is admitting that no consideration whatever is to be given our designs in fatigue. These notes are mainly concerned with the latter type - their nuisance value - and the potential high cost of maintenance due to fatigue cracking which may dictate extensive preventive measures. It is common to think of the onset of a fatigue failure as reducing the service life of an airplane, Actually this is a serious mis- direction of thought. In the case of a small part it is true that that part's life may be ended by fatigue. However, as far as the airplane is concerned it merely means replacement of the part. This is a perfectly normal operation in which the part is replaced because it is worn out just as many other parts such as bushings, bearings, and accessory equipment which may be replaced many times during the life of the airplane. In larger parts, fatigue cracks normally require repair rather than replacement; which is still a normal maintenance procedure, 10.4 Because stress levels vary widely throughout the structure and may be heavily affected locally by the stress raisers of various sorts, itis impossible to deal with the complete airplane in a fatigue life study but only with its com- ponent parts. It has probably been noticed, in these notes, that only typical components have been considered in the general effects of service loading. It is pointed out that any particular item, much more strongly affected than those chosen as typical, would probably be examples of poor design from the fatigue standpoint. Also such items could probably be improved by service changes to permit continued use of the aircraft, Such changes have already been made in the faithful old DC-3 - and are characteristic of all high service life airplanes. It may be of interest to note the service life accumulated on some commercial transports. Model Service Life - Hours (Max, on December 1951) Dc-3 52,110 De-4 18,613 De-6 11, 830125 Appendix 24 10.5 The following fatigue analysis is made for a typical structural joint of a proposed model. Using a DC-6 as a standard specimen, compari- ‘sons are made between the proposed model and the standard. No attempt will be made to calculate a "'safe life" in hours but rather a relative fatigue Life of the proposed model as a percentage of the standard. The standard specimen in assumed to fly between 10, 000-20, 000 ft., have a route of 400 miles, operate at a mean stress of 13,000 psi with a Hmit gust load factor stress of 26,000 psi. The fatigue life of the proposed model for the same operation except at higher working stresses and at various 1G stresses is then made and compared to the above standard, Fatigue life calculations for other operating conditions, higher altitudes, longer flight routes etc., are then made for the proposed model. PROBLEM To determine for equal design quality what 1G stress for a proposed model corresponds to the 1G stress for a typical structure (DC-6) for equal fatigue or service life. The limit gust load factor for the proposed model being 3.65. There are so many variables that influence the fatigue strength of struc- tures that the above problem cannot be answered in a word, In order to cri- tically compare the two models the following variables have been considered, 1. Corrections for the intensity of Ue and an with respect to alti. tude, 2. Corrections for the frequency of Ue and an with respect to alti- tude. 3. Corrections for the slight increase in time-to-climb for the different range routes. 4. Corrections for the longer time "enroute" for the various range routes. 5. Corrections for the same total number of miles flown for all routes over a long period of service. For example, more 400 miles flights will have been made than 2600 mile flights in a total of a million miles flown, 6, ‘The effect of design or 1G stress on the two models, Our standard specimen has a 1G design stress of 13,000 psi and a Limit gust load factof of 2.0, Our proposed model has a design load factor of 5.48 or 3,65 limit which would prohibit a 1G stress level as high as 13,000 psi. A 1G stress of 8, 000-10, 000 psi will be dic- tated by the material allowables for our proposed model. Reducing the 1G stresses and holding the maximum working stresses about equal will increase the range of fluctuating stressing and is one factor which will tend to reduce the fatigue life of our proposed model over the standard. 7. The approximate percent rough air experienced by season (spring, summer, fall) is known but no correction or provision has been made since the calculations are meant to cyclically include all seasons,126 8 A correction may be applied for "path ratio" if the airline route is known with respect to terrian, world wide conditions. 9, No corrections have been made to include the effect of time. It may be assumed that the various models will have approximately the same utilization rate and will have accumulated about the same number of hours per year. These calculations are average However, if the proposed model flys twice as fast as the standard it will have flown approximately twice as far, In this respect the proposed model will have ex- perienceétwice as much rough air during the year. 10.6 ‘The following diagram shows a few typical flight plans. For a realistic approach, typical segments of 400, 1800 and 2600 miles have been chosen for our calculations. No, of Times Cabin Total Elapsed Flight No. Segment Time W. toE. Pressurized 20-DC-6 Hon - LA 10.50 Hrs, 1 602-DC-6B SF - Chi chi - NY 9.67 Hrs. 2 336-Convair 340 SF - Reno Reno - Salt Lake Salt Lake - Den 9, 68 Hrs. 4 Den - Omaha Oma - Chi127 The first step in the calculation is the evaluation of the rough air experienced according to flight stage-climb-enroute-descent, The meager data that are available are given in Table 1 below. . Table 1 Frequency Distribution by Flight Stage of the Total Rough Air Experienced (Operation from S,L, to 10, 000') Item Climb Enroute Descent An 8.5% 43% 48.5 Up 9.0 49 42 AVG 9.0% 46% : 45% Ref, NACA TN-2663 An = load factor increment Uy = gustiness Table 1 indicates that approximately 50% of the total rough air exper- ienced in flight is in the "enroute" stage. The remaining 50% being exper- ienced in the climb and descent, These figures are for relatively low alti- tude and short range flights. In the following calculations the standard specimen and proposed model will first be assumed to operate under these conditions. The available data on frequency and magnitude of gusts at altitude will then be applied to the “enroute st age for various altitudes of operation, Tables 2 and 3 list the values ona percentage basis. The index at sea level being taken at 100%. Note - the values have been bracketed in Table 2 as 10-20, 000 ft., 20-30, 000 ft, etesesee128 Table 2 (Ref, page 109) Rel. Freq. of Rough Air Alb, ft, % Rough Air Rel. % 3,000 28 100 5,000 19 68 100% 10,000 10 35.7 15, 000 6.5 2312 1 fa 20,000 5 17 25,000 4.5 16.1 29 30,000 4 14.3 35,000 4 14.3 27 40,000 4 14,3 45,000 345 est 22 50,000 3 10.7 55,000 2 aS Table 3 (Ref, page 109) Rel, Intensity of Rough Air Ue IND. Gust Intensity Alt, ft, R - Ue SL, posal 5,000 +99 30 10, 000 295 28 15,000 291 27 20,000 286 26 25,000 281 24 30, 000 +76 23 35,000 17 21 40,000 +66 20 45,000 260 18 50, 000 254 16 55,000 247 14 The next step in the calculation is to determine the damage ratio, ( $.n/N) for the standard specimen. The methods and following computations are similar to those given in Chaper 9, page 117.129 Standard Specimen (DC-6) Grnean = 13-000 pst Altitude = S. L, --10,000 ft. U, = Gust velocity -- f.p.s. R = min O max n = gust frequency-cycles N = specimen normal life-cycles (Figure 81) 2 = damage ratio N Item au Cmax, OMin, oR an N 2 © N Mictstk, 30 26,0000 o 1 3,600 ,00028 wt 24 23,700 +2300 «09 5 200 00069 20 21,780 «= +4220 11950 13,800 100362 iS. 19820; yo1s0, 91. 600 30,000 01670 11 17,870 +8130.» 45 5000 82,000 06250 6 15,920 #10080 163 54000 350,000 15400 2 13,975 #12025 «186 © 435000 10, 000,40 . 04350 Zz 0.28129 The calculated damage ratio index En/N = 0.2813 will now be used as an index in comparing the standard to the proposed model. In addition it will be assumed that the ratio 0.2813 corresponds to the damage due to the enroute stage and also to the combined climb and descent stage. It will be necessary in the calculations to separate the two stages since the altitude effect will only apply to the enroute stage. The portion of the damage ratio applying to the climb and descent stage will then remain approximately constant regard- less of the final enroute altitude. (Ref. Figure 83. Time and miles flown in climb and descent is fairly constant irrespective of route range and altitude of operation.) Similar calculations for the proposed model are now given. It will only be necessary to show the calculations for a few conditions -- such as alti- tudes of 10-20, 000 ft. and 1G design stresses of 8, 9 and 10,000 psi. A summary of calculations for additional sets of conditions is given in Table 4. The values of Table 4 as plotted in Figure 82 are meant to include the majority of variables that should be considered in such fatigue life calculations. As previously stated the data of Figure 82 is for a constant number of miles flown. Our proposed model, for example, at 8,000 psi 1G stress and at 20,000 ft,130 altitude will have approximately 150% the endurance of the standard model However, if our proposed model flys half again as fast as our standard and since both models will accumulate approximately the same number of hours per year -- then in the same time the proposed model will have flown 1 1/2 times as far, experienced 1 1/2 times as much rough air and we may expect service troubles, if any, to occur in about the same number of hours on the two models. for the same number of miles flown. Item Same as std spec. £U, e 30 24 20 15 i Mean = 10,000 psi LIMIT GUST L.F, = 3. ALT. 10,000 FT, Cmax. Gming BR 36, 500 -16, 500 = 45 31,800 -11, 800 -.37 27,900 - 7,900 -.28 23,900 - 3,900 -.16 19,950 +50 o 15,970 + 4,030 +.25 11,985, + 8.015 4.67 PROPOSED MODEL 65 50 500 5000 54000 435000 600 1600 5000 16, 500 84,000 107 00834 03120 + 10000 = 30300 64300 04350 EZ 1.12904131 PROPOSED MODEL Mean = 10,000 psi ALT. = 20,000FT. R, = 0,86 Indicated Itensity (Table 111) Ry = .40 Relative Frequency (Table 11) a Item OMax, OMin. Ro oneak, N i 31,400 -41, 400 236 0.4 900 000445 27,400 - 7,400 27 2 1, 760 .001138 24,000 = 4,000 16 20 4,800 + 004170 20, 500 - 500 202 200 13, 800 + 014500 17,150 + 2,850 +16 2000 48,000 041700 13, 700 + 6,300 +246 21600 465,000 046500 10, 300 + 9,700 +694 174000 -- -- ZL = 0.108453 PROPOSED MODEL Mean = 8, 000psi ALT. 10,000 FT, R 1,00 Ry 1,00 a Item _OMax. Oxin, R n a 29,200 =13,200 0-445 1000 +0010 25,500 ~ 9,500 7.37 5 2500 0020 22, 300 = 6,300 228 50 5600 0089 19,130 - 3,130 -.16 500 15500 +0322 15,950 + 50 o 5000 49000 1020 12,775 + 3,225 #425 © 54000 280000 +1930 9,590 + 6,410 +67 435000 - Zs .3391Item Prtax. 25,100 21,900 19, 200 16,460 13,710 11,000 8, 250 132 PROPOSED MODEL Mean = 8,000 psi ALT, = 20,000 FT, Orin. -9, 100 -5, 900 =3, 200 - 460 #2, 290 +5, 000 +7, 750 Ry = 0.86 Ry 40 5 a 236 0.4 2,750 . 000145 =.27 2 6,800 000294 -17 20 15,000 001330 =.03 200 45,000 . 004450 +17 2000 145,000 + 013800 +45 21600 4,200,000 . 005150 +294 17400 — E = .025169133 sML spareinsqus are saSvjuoszed aayyetea z3uj0 94 YOTYA wLOAy oATwA XOPUT %OOT 947 SE UAL) Woy) st UsuTToads prepurIs 94 105 aFeLLEP soounjsip UoATS & UPYIA ono aB[noHed yey) 10) SYST Jo Taquinu ayy Aq pardrnut stg uumyoo uy aFeurep Jeq0 94 oy] Buvs aynox oj yadsax YIM STSeq UOUILUOD ¥ 0} SUOTIPUOD snorIeA ay} Jo} BTeULEp ay 2onpas oj z9pz0 UT (4) +aBejs aqnozua uy pazaprsuos Aquo uonezedo ,999'0z 12 aBemep paonpay *,000 ‘OL UF Se uoMerado ,009 ‘02 1 BBeIs qoosap pue quilT> ut aBeUep sues — (2) Lig mL*0 BSE 20°T 08 19h 68020] 1950°0 ae 000'6 000 ‘oz 082] Te'T s6l 98°T er 6h'8 I9soe “I | 8zs9°o 82zs9°0 | 000°6 000 ‘oT 000T 9£°0 969] €s°0 $ST Ler? ee9e"O| zsz0°O 1eee*o | 000'8 000 ‘oz ors 89°0 Le L6°0 €8 One z829°0| T6ee*O T6EE"O | 000'S 000 ‘OT £54 etl S22 £9°T os 9ek igzet*t}| 9¢00°O o6zt"t | 000‘OT 000 ‘OF bre) stl 222] so°t oF SPL LLbT'T) 2810°O 06ZzT‘T | 000‘0T 000 ‘oe 962 Peer 902 BL iT st 40° Iszez‘1| sBoT*o (epeztt 000 ‘OI 000 ‘0z 29 92°2 €1t sz°e sz LOFT lossz "Z| O6zT'T o6zt "Tt | 000‘Or 000‘0T pasodorg a I os 00T (a)99"e 9Z9S "O| ET8z°O €18z'0 | 000'0T Oo00'OT §=pxzepueig aqnoz a1yur 0092 23NOs eTTEX OST synoz strut QOF| Prepueys 07 yodsoy wWiUA 19poW pasodord jo oT andyeg anew TA @TavL oney adeureq134 10,7 CONCLUSIONS The service life of an aircraft in hours or distance flown will in- crease with an increase in altitude of operation, 2. ‘The service life of an aircraft in hours or miles flown will in- crease with an increase in route range. 3. Aircraft operation at low altitudes and short routes will experience the greatest frequency and intensity of atmospheric gusts. 4. A reduction in the design 1G stress will increase the service life of an aircraft. The following additional conclusions were not determined from the above analysis, however, they are of interest and also applicable to this discussion. The following conclusions and findings have been taken from a British European Airways investigation of high-altitude clear-air turbulence over Europe. "Structural/Passenger Comfort (a) Clear air turbulence occurs in isolated patches of widely varying thickness and horizontal extent and at widely varying altitudes; a representa- tive example is 50 to 100 miles long by about 3000 ft, thick, It is characterized by the lack of warning of its presence and by its short sharp hammering nature resulting from the quick succession of positive and negative gusts. (v) The probability of meeting a gust of a given magnitude is roughly the same at any height. (Their investigations between 15, and 37,000 ft.) (c) The probability of meeting a gust of moderate intensity in clear air at high altitude is roughly the same as in clear air or cloud at low or mod- erate altitudes. Severe gusts almost certainly occur much less frequently in clear air, (a) Severe gust are met more often in thunderstorms, but heavy tur- bulence is sometimes encountered in clear air (U.S, Data). Meteorological Aspects (a) Analysis of the possible causes pointed to the present of marked wind gradient in the vertical as the most likely cause, An indirect method of checking this theory has yielded promising results, confirming that wind gradient in the vertical is of prime importance, (v) Evidence exists which suggests that the worst turbulence is associated with jet streams, and that turbulence often occurs near the tropo- pause. Recommendations to Operations (a) Avoid flying in areas where jet streams or high horizontal thermal gradients are known to exist,S °° ALTITUDE — FT. « 10 a ° °° 135 b) Avoid flying with"2000 ft, of the tropopause. ying (c) _ If heavy turbulence is encountered assume one is entering a jet stream and (i) climb or descend until turbulence is reduced and/or (ii) fly at right-angles to the local wind directions. (a) Should it appear that wing oscillations are being excited by the turbulence, alter speed as much as possible." TYPICAL ROUTES FOR PROPOSED MODEL. ELAPSED TIME ENROUTE ONLY * DESCENT TIME- APPROX. THE SAME A SLIGHT INCREASE IN TIME 70 CLIMB WITH INCREASE IN FLIGHT DISTANCE, f16. 83. ENROUYTE sy 50 min. ~— | __ 383 min. 525 min. wa z 2 fees Ss S\ \2 |S 3 S| |S ig $ ce OF _ 5 30 2 102 #2576 FLIGHT _TIME-MINUTESOl SH~DAD NI SSBYULS TWIXv ‘XV (GRLON 4)138 Chapter 11 FATIGUE ANALYSIS - METHODS AND PREDICTIONS agvarg, he ong ,foeta ahele a feesteneelelle -? This section of notes will discuss the better known analysis and prediction methods currently being used, No attempt will be made to critically compare the different methods since they all have some merit. ‘The various methods are listed and will be discussed in the follow- ing order, 1. Method employing the Cumulative Damage Theory. 2. Tangent Method. 3, British Prediction of "Safe Life". 4, Stress concentration Factor Method, 5. Method proposed by P, B. Walker, l.1 Primarily, the reasons for ever considering any kind of a fa - tigue analysis for a structure will fall into one or the other of the following categories, (1) Limit or operating loads may be sufficiently high and fre- quent to give life-strength values of significant magnitude, (2) Service experience on parts of similar function may show the need for improved fatigue life. (3) Parts subject to manufacturing or maintenance inconsis~ tencies and difficult to inspect will require special attention to establish a satisfactory operating life. The above reasons cover the majority of cases and may be applied to all types of structures. Particularly, in the aircraft in- dustry, those components of an airplane the failure of which could endanger the airplane or its occupants. 11.2 Basic Approach ‘The steps and calculations made in a fatigue analysis are usually rather simple, The following procedure will be necessary regardless of the analysis method used, (1) A detailed stress analysis is required of the structure mem~ ber or component being studied,139 (2) The frequency of occurrence of working loads and their varying intensities should be established (For example gust loads on wings or cyclic cabin pressurization). (3) The effect of the cycling loads at varying levels should be determined, Laboratory tests are useful in making this ‘determination, In many of the methods the Cumulative Damage Theory is employed in accounting for the cumulative effect of the varying loads. a (4) With the information in (1) to (3) above, a calculation of the life expectancy of the member in question can bemde, The life expectancy thus obtained, however, is subject to rather wide variations since a number of factors used depart greatly from their mean value, One serious difficulty at present is the accurate determination of the dynamic stresses in a complex aircraft structure, Our knowledge of the frequency and magnitude of induced service loads such as gusts, landings and engine vibrations has been increased by statistical reports on the subject in the past few years, However, much more data are needed and there still remains to be gained a more complete understand~ ing of their effect on structures. Admittedly, no rigorous or wholly reliable life expectancy calculation for aircraft can be made at this time, The term "life expectancy" and "fatigue life" has been used often in the preceeding paragraphs - the designer is reminded that we are referenc- ing the fatigue cracking of our structural member when using this term - and not retiring the airplane. Sometimes a calculation can be made that comes very close to the results already obtained in service, Such a calculation could hard- ly be called a prediction since the actual results are already known, Many may wonder of what use these a-posteriori predictions are? It is believed that they are of great use, and will eventually be the major factor in checking the refinements to or developments of new analysis methods - which eventually may yield valid and reliable predictions (a-priori). It is possible that some of us, familiar with the subject, have been too close for too long a time to come up with a new idea - even when standing back and taking an objective view il is difficult to see where the analysis improvements are to be made (fatigued engineers), It is probable that a fresh (non corroded start with a virgin (specimen) mind will devise the needed methods, In this respect it is hoped that some of you, familiarizing yourself with the subject for the first time, will come up with the desired recommendations, A review of the cur~ rent analysis methods is now made. 11,3. Method Employing the Cumulative Damage Theory (DACo) This method has been discussed previously, Examples are shown in Appendix 23, page 116. The method of "safe life" predic- tion is based on the comparison between laboratory fatigue tests of the structural component and the stress analysis of the structure140 from the spectrum of the anticipated service loads, The Cumulative Damage Theory is employed in calculating the extent of fatigue damage resulting from the varying service loads as compared to the fatigue lifes of the structure as determined by test. 11.4 Tangent Method ‘The tangent method is based on the intersection of the S/N curve of the structural member as established by test and the frequency of occurrence curve of the anticipated service loads. This method is not quite as complete an integrating process in determining the extent of fatigue damage to the left of the S/N curve as the Cumulative Damage method and therefore gives more conservative answers, The method is shown graphically in Figure 84, The load frequency curves for l, 5 and 7 "part-sequences*" are shown with respect to the laboratory S/N curve for the structure, 5 British Prediction of "Safe Life" Step l, The British gust load spectrum is determined for average loads. ‘We do not have this data, however it is possible for these calculations to use U, S, data (U. S, data represents a far greater sampling). For the prediction, the British then multiply the number of the loads by 1,5, ‘The gust spectrum is then plotted as frequency vs, the average variable loads. The variable load about 1G is expressed as a othe U,T.S. (See Table 1 and 2 for the frequency and magnitude of average loads, The gust load spectrum is also plotted in figure 85.) Step 2. With the above gust load spectrum (they) then predict (calculate) the fatigue life of their standard specimen. This is done by employing the Cumulative Damage Theory. In making this calculation they also multiply the magnitude of the average loads by 1.25. The computations are shown in Table 3. = BRITISH STANDARD SPECIMEN Variable load as%ofU.T.S, 20 18 16 14 12 10 9 8 7 6 5 43 Frequency (10) 15 .19 .28 .36 .50 .80 1,05 165 260 4.5 10 30 200 The British claim this standard has been established by many com- ponent tests, (Laboratory) Step 3. Component structures are now tested at one load level only, This load level corresponds to the variable load for a 10 fps gust cyéTiag about the 1G load, They establish the endurance from a minimum of 6 specimens and use an endurance value for WHICH At Teast 90% of their tested specimens will exceed, Note: The British use the term endurance for laboratory tested specimens and the term fatigue life when FeTerring to the aircraft structure, —_ * Ref, page 117.141 Step 4, The "safe life" is then determined from the ratio of the endurance of the component test and the endurance of the standard specimen, at the same single load level as the component test, multiplied by the fatigue life of the standard specimen as determined in Step 1 and 2, COMPUTATIONS: Step 1 TABLE 1, (U.S. Statistics) + gust £, - relative freq. (f) 2.5) 2.25 + 435,000 + 652, 500 6.75 = “54,000 = “81,000 11,25 5,000 7, 500 15,75 500 750 20.25 50 15 24,75 5 8 29.25 L 2 TABLE 2, Not having British average loads-use MopeL-A sta, xx analysis as average typical loads. : as as % rr eee stuss, « eens. 21.25 2,25 13,000 13,975 975 ial 214 6.75 " 15,920 2,920 5210 6137 1.25 " 17/870 4,870 8155 10. 70 15.75 " 19,820 6,820 12500 1500 20,25 " 21, 780 15.40 191 20 24.75 " 23,700 10,700 18. 80 23, 50 29.25 " 25,900 12,900 22. 60 28.20 % UTS, = 57,000 psi, (test) COMPUTATIONS: Step 2 The loads times 1,25 and the frequency times 1.5 for the average gust spectrum are now compared to the British Standard specimen (fig. 85). Employing the Cumulative Damage theory the computations are made and shown in Table 3142 TABLE 3, Variable load or stress as n-Freq. N-British n Damage % of UTS Standard N Ratio 2.14 652, 000 0 6.37 81, 000 3,500, 000 0232 10.70 7,500 700, 000 +0107 15,00 750 300, 000 +0025 19.20 15 165,000 + 0004 23.50 8 100, 000% nil 28.20 2 - nil * extrapolated - fig. 85 LR -0368 Using the NACA equation, the average gust load statistics of Table 1 are converted to miles and hours flown in experiencing the the gusts. (Ref., page 110) F = s2so- RL 6 = 52g0+ (ODL (11) (13. 6) L = 283,000 miles 2 HS % . 283,000 | 1130 hours iV" 350 mph The number of these "part sequences'' of average gust loads that the standard British specimen can withstand before failure (assuming pa = 1.0 at failure ) is calculated as: eee number = 1 Ip 0368 | The fatigue life for the British Standard is therefore: 1130 0.0368 H 30,700 hours COMPUTATIONS: Step 3 Applying the British method to a typical mon Step 3, is as follows: component test, for A typical structural wing joint such as sta. XX is tested in the lab- oratory. The SN curve for this data is plotted in figure 86, The test on the Joint at the variable load corresponding to a 19 {P® gust may be determined143 from Table 2 as a variable load of 7, 66% of the UTS of the joint, Sim- ilarly this corresponds to a. maximum stress of 17,330 psi. and from figure 86 the endurance of the component test is 60,000 cycles. COMPUTATIONS: Step 4 The endurance of the British Standard Specimen (Fig. 85) at the com- ponent test load level of 7. 66% is 1, 700,000 cycles, The "safe life for this joint in service is then calculated from: 60, 000 te7ebs500 (30,700) = 1,080 hrs, 11.6 Stress Concentration Factor Method This method employs element tests of basis materials and the application of stress concentration factors or possibly even the testing of notched materials having known stress concentrations. The next step, and one of the most difficult for the designer, is to evaluate the critical regions of his design into terms of stress concentration factors, This is not a wholly reliable determination and may often lead to dangerous conclusions and a costly project, Further, analysts and designers will sometimes extrapolate existing data of the basic materials to the appro- priate stress ranges of their design by means of the Goodman diagram. Again, this procedure may result in even further inaccuracies and de- part from the characteristics of a built-up structure, The main advantage, if it can be called an advantage, of this method is that a minimum of testing is required. Fatigue testing of composite structures is a time-consuming and costly project and not all industries are in a position or have the foresightedness to have a well-equipped testing laboratory, The calculations for this method are given in Appen- dix 25, 11,7 The British Prediction of Safe Life Method (11.5) has undoubt- ediy stemmed from the method proposed by Walker, Walker proposed an equation for the estimation of "safe life to be applied to the labora- tory tests of component structures, His equation is as follows: M = kmN where M safe miles. k = safety factor (2.5), m = distance in miles between 8 fps gusts, N = 2/3 of the endurance cycles for the joint tested ata stress level corresponding to 1.25 x stress caused by an 8 fps gust. “Head of Structures, Royal Aircraft Establishment144 Walker's 8 fps gust level in accounting for the region of the great~ est amount of fatigue damage, is probably valid, (In most gther calculations that have been made the greatest percentage damage ratio » is between the 6 and 15 fps gust levels.) However, the overall factor Df 2.5 does not appear to have any significant meaning other than making the prediction more conservative, The factor of safety is meant to cover discrepancies in the testing, assumed service loading and the many other variables that exist, Waller's method applied to the sta. xx.Model:A structural joint is given on Figure 87. SUMMARY 11,8 It has been noted previously that no attempt will be made to critically compare or crivicize the various analysis methods. They all have their advantages and disadvantages and no one method will be rated as the best, The Cumulative Damage (DACo) Method however, for the particular example chosen, does come the closest in predicting fatigue cracking as compared to actual service experience, It is to be pointed out that this method also requires the greatest amount of time and the most extensive testing, The summary of the calculations by the various methods is given in the table below: PREDICTION FOR FATIGUE CRACKING IN SERVICE* Cumulative Tangent British _Stress P. B, Walker Damage Method Predjgtion Concentration Method Method pee = scacior (Paco) Method 7900 hrs. 5650 hrs, 1080 hrs, SEE APPEWA. 930 hrs. * The Summary of Service Difficulties by the Airlines reports fatigue cracking for the sta. xx joint, used in the examples, between 10,000 - 12,000 hrs. Cracks occurred in the lower aft skin between center and rear spar, Chordwise cracks in the access door. Two cases reported, The same type of cracking has been duplicated in the laboratory tests of e joint.145 PHOTO SHOWING THE LABORATORY TEST FATIGUE FAILURE OF THE MING JOINT. FAILURE occURRED THROUGH SECTION I=1 (SEE SKETCH) IN THE SHEET AND STRINGER AT THE FIRST RIVET ROW146148 4 OL YXS-Y OW VS - S8PLLS NOISNIL150 APPENDIX 25 Stress Concentration Factor Method: The particular joint to which this method is to be applied is shown in the photographs and sketch on pages 145; 50 This is the same structural Joint for which calculations by the other methods were made. In order to ‘save considerable time in these calculations let us assume, for the present, that we are now experienced enough to point out the most critical regions. Referring to the sketch of the joint, these regions are 1-1 and 2-2. The section of the joint through 1-1 is the minimum cross section. Note that the tongue doubler stops here and the stringer reinforcement doubler does not continue across the gap. Further, we may now assume that the crit- ical spot in this section is likely to be at the rivet holes cither in the skin or in the stringer, These rivet holes are not pin loaded holes however, since both skin and stringer are strained as a unit, Neither are they unloaded holes, Actually the condition may fall somewhere between the two. The section at 2-2 however, at the first row of rivets - through the skin, tee section, tongue doublers and stringers may definitely be considered "rivet loaded holes", However the extra doublers at this section re- duces the working stresses to a less critical value therthestressestom sesscritieat-valte than the stresses at section 1-1, (We know the answer ahead of time in this case - so it's not quite fair to infer that we will always determine the critical area this quickly - for some cases we may never determine it, Our first step is to determine the fatigue characteristics of the material in this region. The stress analysis of this 75ST~6 joint is identical to the values given on page S/N curves for the basic material are shown in fig. 18, These curves may now be converted to a constant mean stress curve of 13,000 psi with the aid of the Goodman Diagram in Figure 16, The constant mean stress curve is then plotted as in Figure 88, A stress concentration factor is now applied to the basic curve with he aid of fig. 21 for unloaded holes (even though we stated they were not unloaded holes in the strictest sense). This curve is labeled "(A)" in fig, 88. If the stress concentration factor (K,) data of fig. 21 were not available, we might arbitrarily select a S.C.'F. of K, = 1.5, This factor applied to the basic S/N curve now yields the predicted S/N curve for our joint as shown in fig, 88 labeled "(B)". Using these two S/N curves as limits, the following calculations can now be made. The same gust load statistics are used in this method as in the previous examples151 TABLE A - CURVE "A" for Const, (a) N a On = 13,000 (Gusts) Fig. 88 N 13,975 + 435,000 190, 000 2.29 15,920 = “54,000 110, 000 +49 17,870 5,000 78, 000 106 19, 820 500 56,000 -01 21, 780 50 44,000 = 23, 700 5 36, 000 . 25,900 1 26, 000 : ¥H, = 1,130 brs. = 2.85 H, = 4130 2 396 hrs. J 2,85 TABLE B - CURVE 13,975 + 435,000 107 +043 15,920 = “54000 320,000 2169 17,870 5,000 125,000 $040 19, 920 500 76,000 2006 21, 780 50 52,000 - 23, 700 5 36,000 - 25,900 1 26,000 Z= 0.258 ¥H, = 1,130 hrs, = 1,130 0, 258 H = 4,400 hrs. Within the limits specified it can be seen that there is a wide discrepancy in the calculated time to fatigue cracking, Ifa$.C.F. of K,. = 1.2 had been used the predicted life would have come out greater anid also closer to actual experience. However, who is to decide what factor to use? This is the main problem that confronts the designer and at present it would seem that the alternative methods, requiring the testing of the actual structure, would yield more accurate determinations, *Ref, page 116153 Chapter 12 DESIGN PRI IPLES design by rule not rote. This section of notes will review the preliminary design rules, or more appropriate - the design GUIDES, that apply to component struc- tures; guides which are aimed at or of major importance in providing for improved resistance to failure from fatigue. ws-erimtredection 5 ar : ri A eee ee introduction to this review, there is quoted below a portion of the 39th Wilbur Wright Memorial Lecture delivered in 1951 by Dr. A. E, Raymond, Douglas Vice-President Engineering, which represents a philosophy that, over the years, has been proven to result in good design, 12,1 “EXCELLENCE OF DETAIL DESIGN Two aircraft designed to the same specification may differ widely in size, cost, and weight simply because of relative quality of detail design. In fact, much of the progress being made in the art consists in learning how to design a small, efficient aircraft to do the job pre- viously done by a large complicated one. Not only does this hold true of the aircraft as a whole: it is also true of its components, down to the smallest parts. It is not always appreciated that a pound saving in weight empty may be compounded into ten or more pounds saving in gross weight on modern designs, Quality, weight saving, and simplicity are all closely related, The simplest design that fulfills all desired characteristics is likely to be the lightest, with best performance, and the most dependable and re- liable in its operation. In comparing a complicated design with a simple one, engineers often tend to compare them with both designs fully devel- oped and working perfectly, just as the designer intended. In sucha comparison the more complicated design often looks better. Unfortunate- ly, all the intangibles act against the complicated design and in favor of the simple one. Some of the intangibles are: Design time, Aircraft designs are not made in ivory towers; the design Sf eack component must be scheduled, often far in advance of the time that the degree of complexity is known, Thus the complex design rarely gets as much time, in proportion to its needs, as the simple one. Very Often the complex design can only be carried to the point where it barely works, and then must be frozen, under pressure from the schedule, while the simple design, with time to spare, can be refined. Thus those com- ponents which are simple tend to be more highly developed and free of service difficulties than those WhICH are complex, because of the amount araime required to get the complicated components to work at all leaves little or no time for improvement,154 Carry-over of learning from previous designs is a major aid in reduc- fg troubles im service. A design developed from a previous design, if the basic principles are still correct in the new application, is usually better than a completely new one, It takes engineers a few hard knocks to make them realize this; some never do, An engineer likes to take pride in a new and original design; he gets more satisfaction from this than from improving an existing one. It is a problem in psychology to get designers to take pride in the reliability and trouble-free operation of theix product, rather than in the originality of their concept, On the other hand, using an old design in new conditions, to which it is basic- ally unsuited, can be equally as'bad, Refinement of design, When refinement of design is carried on with the dea of making it more nearly perfect, making it meet requirements more exactly, etc., it often results in more complication, higher cost, and in general, a poorer overall design. This is the disease referred to as, "perfectionism". Where refinement takes the form of conscious simpli- fication, more often than not the result is a real improvement; unless carried too far, when it may delay the entire project, Even simplification, carried on too long, results in the basic concept falling behind the state of the art, Producibility. Aircraft production at best is not high volume production, ‘and the type of tooling required, except for a few items such as standard bolts and nuts which are used many times ina single article or are used universally, is not high production tooling. Such tooling is far too in- flexible to accommodate itself to the constant stream of changes which is an inevitable concomitant of aircraft development, This lesson was learned the hard way by the automobile industry during the last war when it was pressed into the manufacture of aircraft, We have also progressed considerable since aircraft were made by hand with little or no tooling. The correct middle course is to provide comparatively simple tools which can be augmented or duplicated if production is stepped up and which can be thrown away if the design is changed, This type of tooling often dictates the design, The true producibility of a design, as that term should be used, is the degree of its adaptability to the use of this type of simple tooling. Engineering, tooling, and shop should keep close to each other, mentally and physically, during construction of the prototype. It is a team effort and must be carried through in that spirit." In summary: 1, Keep your design simple. 2, Take full advantage of "Carry-over" from past airplanes,155 12,2 To be more specific, our review is now made of some general design details. The details are actually rather simple - so basic in fact that their applications are often overlooked even by an experienced designer, Probably the most important factor in good detail design is structural efficiency, Structural efficiency is the ability to carry maximum loads with minimum weight. It is obtained by the use of a good basic arrangement, efficient members and uniform structure throughout, Every part that is larger, stronger or less structurally efficient than it needs to be is too heavy. 12.3 Basic Arrangement ‘A good basic arrangement is one which minimizes the loads in the members or parts making up the structure, by a good geometrical arrangement, The best geometrical arrangement is the arrangement that spreads the loads into wide-spread trusses and deep beams as soon as possible. { 1 nor —— THIS THIS THIS NOT THIS ‘A. good geometrical arrangement also avoids eccentricities, and thus clifeinafes bending in axially loaded members. This is important in the design of efficient structures, as an eccentricity equal to 1/2 the depth of an axially loaded member will more than double its weight. KE * Taken from a Detail Design Memorandum by H, W. Adams, Chief Design Engineer, DACo.156 12,4 Efficient Members ‘The shape of a member has a great influence on its structural efficiency. The optimum shape varies with the type of loading. Axially loaded members should be straight, should have their material distributed far from the ¢, especially near the middle of the member, and should be supported to prevent buckling when prac- ticable, This goes back to the first and simplest order of Greek architecture ~ the Doric column, Ee THIS NOT THIS Bending members should be as deep as practicable and should have the material at the extreme fibre, with only material between the "caps" to carry the shear. i J THESE NOT THESE Torsion members should be closed sections with as much enclosed area as possible. THESE NOT THESE 12,5 Uniform Strength Throughout If a given load is to be carried with a minimum weight of structural material, it is obvious that the material should be distributed to match the load.’ Beams should have larger cap cross-sections in regions of larger bending moments, etc.157 —— J - [eeceeog | _ THIS NOT THIS As mentioned before these details are elementary. The problem of structural efficiency is not lack of knowledge of principles, it is lack of application of these principles. For example, bellcranks often have been designed with shallow trusses when they could have been designed with deep trusses, "A good method of design for high structural efficiency combined with manufacturability, is to design a good structural arrangement or "framework", then mentally wrap the material around this framework. Considering manufacturing methods first usually leads to a patched up, inefficient design. Perhaps asking yourself these questions may help in realizing the maximum structural efficienc: 1, What kind of structure is this; what is the "structural frame work"; which members are columns; which are beams, which are torque members? Can the basic arrangement (the geometry) be improved in the interests of better structural efficiency; can the trusses be made more spread out; can the beams be made deeper; can the torque boxes be made larger; can eccentricities be eliminated? 3. Can the member shapes be improved for better structural efficiency; is material far from the neutral axis; are the torque members closed sections? 4, Can material be removed from the lightly stressed regions to give more nearly uniform strength throughout?158 If you look at your design and ask yourself these questions, the chances are that the structural efficiency of your design will be good; poor structural efficiency is usually the result of lack of recognition of these problems; not lack of ability to solve them," 12.6 Specific Design Details To be even more specific in the general details just discussed let us now consider a few actual examples. It is true, in the case of the uniform strength beam, that additional material should be added to the most highly stressed region. However, this material if in the form of a doubler should be added in a gradual manner, For example, a leaf doubler or a tapered tongue doubler should be used to prevent an abrupt change in cross-section. Similarly, the bolts or rivets attaching the members should be in a tapered distribution both in size and pattern, ——_ SS, oO (Geecce eas] locoo0 0 THIS NOT THIS It is well known that doublers around cutouts are required to replace the strength of the material lost in the cutout, It may equal- ly be known that additional doubler material should be provided to account for the load concentration at the corner of the cutout, Not so well known however is the effect of the doubler attachment method at the corners of the cutout. Employing the effect of "superposition of notches" as diagrammatically shown below will give us our guide for the cutout doubler attachment method, Laboratory fatigue tests of the following examples will result in diminishing fatigue lifes in the order in which they are shown (1 to 8), | 2 3 4 5 6 7 8 o || D> ||]

) Stresses in members supporting such fittings, webs, doublers, etc., should not exceed 28,000 psi. 3. A design guide for estimating frequency of loading may be based on an overall average of 2 hrs. per flight. (Ref. Page 111.) Special safety factors have been used for windows. In addition the windows are to be dated and a replacement schedule devised to replace crazed and deteriorated windows. Windows should be double-strength, (double-paned), any one pane capable of carrying the full load, by a large margin.163 (d) Moveable Surfaces Elevator, Rudder, Aileron, Flap 1, No particular fatigue design safety factor policy has been applied to control surfaces. Flap skin cracking and the like that has occurred has been controlled by increasing the ma- terial skin gage. 2, No sharp or sheared edges should be allowed on formed brackets of plate stock. The edges are to be champfered and in critical regions may be hand-polished. 3. Attachment methods as Huck lock bolts and drive fit pins may be used to keep slip at a minimum wherever practical, 4. All eyebolt installations should be torqued to maximum recom- mended values, 5. No cantilever overhang of tabs should be allowed, Stiffer tabs are not as susceptible to buffeting or engine vibration frequen- cles. 6. If possible design the deflection of tabs and operating control systems stiff enough to have resonant frequencies above the propeller resonant frequency, This criterion should be used whether the surface is in the slip stream or not. (c) Nacelles and Engine Mounts 1, The present use of titanium for nacelle skin, frames and hat stiffeners has resulted in excellent service life, It is believed that this good service life has resulted from rather extensive specification testing which was carried out during the design stages. 2. On newly designed parts static tests should be conducted, Design guides are usually based on past designs since service experience backlog is very large. Occasionally there are parts where the surface finish has been desig- nated, Longitudinal grain direction should be specified and most of the time adhered to; there are isolated cases where this is not possible, 3. Welded parts should be thoroughly inspected and all inspection procedures unique to the part are to be set up, X-ray and Magnaflux methods may be used, Important flashwelded parts should be proof-tested 100%. (®) Landing Gears & Hydraulic Systems 1, In general the usual safety factors or fitting factors (1.15) are to be applied in design, For operating pressure systems a 2,5 factor in design is to be used, (1. 5)+(1.15)+ (1.45) = 2.5;164 the 1,45 factor accounts for surge pressures and repeated loads, Handbook’ values for stress concentration values may be used in the preliminary design stages, however in the final analysis it is sometimes not feasible to allow generous fillets; in these cases the design is often "eyeballed"' based upon past experience and testing. 2, In general landing gear parts have been primarily designed to static criteria only. However, with the newer materials (4340 HT, 256-280, 000) fatigue test criteria are in the stage of being formulated. 3, Shot peening of landing gear components has in the past been primarily a salvage procedure after service trouble has been experienced, It will probably be used more in the future on new designs and parts. 4, Serious consideration is being directed towards limited use of chrome plating. Chrome plating has often been one of the con- tributing causes of decreased fatigue life. The use of some other plating materials should possibly be considered, 5, There has been a gradual trend towards the discontinued use of dural for landing gear components. In all cases where dural has been replaced by steel there has been no further service trouble. 6, Surface finishing of parts is to be considered and called out on drawings according to the classification of the component: whether subject to impact, repeated or static steady load: However, surface finish as a function of fatigue strength is a subject about which very little test data are available and about which much less is known, The finer the machine tool marks does not necessarily mean the greater the fatigue resistance, Where and whenever possible, the machining marks should be in the direction of the loading of a part. A typical example of this condition occurred in the case of a main gear retract cylinder piston head, Finer machining in the fillet radius of this part still gave trouble, It was not until hand-polishing of the critical section was accomplished that the part was satisfactory, Ina sense "experience paid off", There were a few later designs, similar to this part, that were then designed in the same manner. (g) Control Systems No special margins of safety have been used in the past when designing for repeated loads. The usual fitting factors apply and have been found to be satisfactory, 2, Items such as belleranks, linkages, push rods, etc. should be designed stiff enough to have resonant frequencies above the highest exciting frequencies. Rolled threads after heat treatment are to be specified on push rod ends,165 4, Service experience on control systems has revealed very few fatigue failures; the main cause of trouble is wear. 5, Employing the design philosophy that single failure shall be catastrophic, the control system is unique in that the "stand by" system in a two system rigging is a safety factor far greater than any one empirically designed system even with large margins for repeated loads. In closing this chapter the following allegory seems particularly appropriate, The famous engineer-artist-sculptor Leonardo da Vinci was once asked to show some visitors his works, He guided them through his studio, pointing to the tiniest statuettes and taking particular pains to call the attention of his visitors to such details as eyes, ears, finger- nails, etc. At this point one of the visitors asked why he did not show them his big work instead of bothering about such mere "details" and "trifle: To which Leo replied, "My friend, you see it is just these details, these trifles, which make perfection; and perfection, you will concede, is neither a detail nor a trifle." (A little conceited maybe - but true.) s To extend the life of our designed members require’ much attention to the details and trifles and by no means not just the few discussed here. Details and trifles are the stress raisers in our present structural designs.9. 1a. 167 FATIGUE DESIGN GUIDE - CHECK LIST Gan fillets, fillet radii and spotface radii be made larger without interfering with mating or matching parts? Is edge distance adequate, 2D if possible, is there a possibility that this my fall short due to tooling set-ups in production - can an extra margin be provided to prevent this? Have bearing stresses in riveted and bolted members been reduced to the minimum? Is the menber critical or important enough that surface finish in machining coulé ‘be improved? (Economically feasible) Is the part subject to corrosion, clevated temperature, ete., and have necessary precautions been token to account for their effects on the design? Is plating or a protective coating part of the design and if so, has attendant reduction in fatigue strength been accounted for? Xs dt possible to design the member or structure vithout either hot or cold forming any of the components? Are the Joints, parts, fittings, ete., eccentric - if so, is it possible to design ‘them otherwise? Has longitudinal grain direction been specified? If the part 4s to be loaded in bending or flexing, has shot-peening been consider- ed? Is it practicol for the part in that its fatigue strength or life may thus be improved? Are doublers, reinforcements, etc., graduel rather than giving an abrupt change in cross-section at the minimum section? Has extra consideration been given tension loaded fittings - net stresses in pin- loaded and notched menbers for example? Have all sharp edges been broken? On mating parts, b rather than scarfed-off so that chaffing is minimized? Eos inepection in service been considered? If possible, can the structure be designed so that easy access to inspection procedures can be made? Hos inspection in the fabrication and manufacturing stage been set-up and are the inspection personnel versed enough to know what to look and check for? Is it possible to make the part or fittings from forged billet material rather than from extrusions or heavy plate stock? Is the part designed so that the mismatching on installetion will be at a minimum, thecefore resulting in the least possible residual tension stress? we edges been radiused Has the superposition of notches been prevented, i.e-, if more than one is present in the design?168 Chapter 13 SERVICE FATIGUE FAILURES He have 40,090,000 reason. forvfarire bat net a Single excuse. wR Kipling ‘The following notes will attempt to list the major fatigue failures that have occurred on transport aircraft in the past few years. The source of information has been from DACo Development Laboratory Reports and from correspondence between the Service Departments and the field representatives of the various airlines. In some instances of which there are no records, the fatigue failures have been recalled by the "old timers" of the Engineering Department. Improved inspection methods, coupled with the current interest the subject, have undoubtedly had a large affect upon uncovering the amount of fatigue failures that have been recorded, In the past, many fatigue cracks went unnoticed for a long time. An awareness of fatigue damage and the uncovering of fatigue cracks is a healthy state of affairs if its importance is kept on a common sense level, We are not taking the attitude that we are trying to hide something - fatigue is not a bad word - and certainly we do not want any insidious little cracks to grow in some dark corner until they eventually become a real threat, The operators of aircraft are gradually becoming educated to the effects of fatigue loading on their structures and have been informed by the manufacturers that it is through their mutual efforts that the prob- lems in fatigue can be best met. Periodic inspections, knowing approxi~ mately where and what to look for, are a few items that maintenance crewmen have to become acquainted with, Unfortunately, there is one thing they have not been briefed on - and that is the proper handling of the service fatigue cracked part they uncover. By the time the part reached the manufacturer, for its examination by the metallurgists, stress engineers and designers, the fractured surface is usually so ob- scured by saw marks, hammering marks and greasy fingerprints that it is often impossible to tell whether it was a fatigue failure or not, The parts do not have to be handled with tongs and wrapped in surgical band~ ages but the "hammer mechanics" should be informed that a relatively "clean" fracture means something to us. The study of fatigue fractured SUrtates is a promising field - and some day we may be able to determine with reasonable accuracy the service history of stressing, both quan- titatively and qualitatively, the nucleus of the fracture and why it occured in the location that it did, ‘When this is possible the percentage improve- ment in life for our part may be determined, Employing "carry over" from past designs in this respect will enable us to design even better members. ‘There is a growing trend, however, of increased numbers of fatigue failures and if we are to stem this trend in the present and future then serious consideration will have to be given to detail design,169 Summary. The majority of fatigue failures that have been compiled for a few transport aircraft are shown in figure A below, The failures are plotted at their earliest known occurrence, No attempt will be made to give des- criptions of each and every fatigue failure case. The majority of failures were due to one or the other of the following details ~ surface finish irregularities, too sharp a corner radius, eccentrically loaded members, among the many other general causes discussed in Chapter 12, Also, no attempt has been made to plot the actual number of failures but rather the number of different cases or types. In some cases, as many as 16 fatigue cracks of the same type have occurred on as many airplanes, while in other cases only one case of one type was ever recorded, The continued occurrence of any one type of failure in any one location is usually arrested by the service rework, pach OPERATIONAL HOURS « 10° 10 26 390 40 80 60 mopeLA [oe ee eo TF MODEL: B | ] MODEL: C O- SERVICE CRACKING The apparent trend in increasing fatigue failures is shown in figure B, This plot was determined by dividing the known types of service fatigue failures, for the various models, into the current operational hours accu- mulated by the models, In many respects this is not an absolutely true comparison between the different models and design periods, since there is a greater quantity of the later models which also have higher utilization rates, The probability of experiencing greater numbers of failures therefore will be increased, There has been a lot of discussion and concern about fatigue failures in recent years as well as in these notes. Before it gets out of hand we should remind ourselves by taking a look at the record - the airline service bulletins show between 200 and 300 service troubles per 1,000 hours per fleet, Of these service troubles less than 5% are directly caused by fat~ igue and the majority of these may be classified no greater than their nuisance value. It is often difficult to conceive why we should be concerned about such a relatively insignificant phenomenon, However, there are a few cases on record in 1947 and 1954 that not only involved millions of dollars but lost considerable prestige with operators and the public alike, Fatigue, you will concede, is an excrusiatingly contradictory subject !170 The following photographic examples are mainly presented for those not familiar with the appearance of fatigue fractures - either by labora~ tory test loading or service loading. 8006 ° al 6000; 4006; N 2001 SS 1934 ‘38 42 ‘46 ‘so FIC.B. MODEL DESIGN PERIOD - YR. OPERATIONAL HOURS PER STRUCTURAL SERVICE FAILURE (Wer. LER AIRPLANE)17 Fig. 90. Service failure of 1/2" wing attach bolt, Western , Airlines Plane (209, Fatigue failures in threads of these bolts have occurred between 4000 and 8000 hrs, The older type bolts are now replaced with rolled thread bolts, This bolt may have been improperly torqued since it undoubtedly has been experiencing the effects LEUS of loading for a long, long time, Note fatigue failure progression 90% through dia, of bolt, Fig. 91, Service failure of "RM stud in front spar fitting at 8,576 hrs, The nucleus of the crack occurred in the tool marks of a too sharp relief radius, Fig 92, Laboratory test failure of NAS bolt, Fatigue crack in head to shank fillet, Note tool marks. The head to shank fillet fo some bolts are now rolled,172 Fig. 93. Fatigue failure of wing stiffener and doubler, Note "arrest lines" in the" progression of the fatigue crack, The crack originated at the rivet hole - also note the insufficient rivet edge distance. This type of fatigue cracking has occurred in skin, doubler and stiffener in service between 12, 000 and 13,000 brs. Fig. 94, Fatigue failure of extruded hat section-wing stiffener in test of wing joint, This is a good example of the failure location in the minimum cross-section and at the first loaded bolt or rivet, The life of this member could have been considerably increased if the doubler reinforcements had overlapped.Fig. 97 Fig, 95. Fig, 96. Fig. 97. 75 ST ear of splice plate. Note fatigue crack "A" at base of sharp fillet radius. (.015") Enlarged view showing face of fatigue fractured surface and extent of growth of the fatigue crack, Service failure of this part occurred in approximately 7,300 hours, This is a good example of the importatif of providing for generous fillets and radii in design, Fatigue failure of Noise Dampena Retainer Nut in approximately 12,000 hrs. of service. Fatigue failure originating in sharp tool mark in machined radius,Fig. 98. Fatigue cracks in door jamb and door lock bolt cutout on laboratory fatigue test structure in 7500 cycles. Cracks caused by cyclic cabin pressure stressing. Similar cracks have occurred in service in 9,946 hrs. Test and service experience show good correlation for an assumed pressure cycle each hour of flight. Senet * 1902 Pig upper oper eoP Leo e Te hours ne -160¢2 Ragnt frgnt al me Fig. 99. (DC3) Did you ever consider drilling "stop holes" for fatigue cracks? - Well, don!Fig, 100, R etracting Nut on Main Landing Gear. Service Failure occurred in 11,040 hrs, This case is a good example of superposition of notches. The internal threads of the nut intersect the spanner holes, Fig. 101, Forged magnesium supercharger impellor. Service failures have originated in as little as 800 hours - at 12,000 rpm this would correspond to 580,000,000 cycles. The fatigue cracks in these parts originated in two locations. Either at the steel pin inserts or at the point of maximum combined radial and circumferential stress "A",116 Fig. 102, Nose Gear Yoke End Fitting. The average service life on eleven failed parts was 5,870 hours, The early fractures were believed to be caused by an under- cut fillet radius, The parts are now shot peened and closer inspection is given the machined fillet radius. i ‘ EXTENT OF GROWTH OF FATIGUE CRACK. STATIC FAILURE. NUCLEUSFig. 103, Stress corrosion cracking of front spar splice angles in 8, 620 hours service, The appearance of this type of fracture is often mistaken for a fatigue fracture in the usual sense. (Not corrosion fatigue. ) Fig. 104 Stress corrosion cracking of Nose Gear Fork.178 Chapter 14 FATIGUE MONITORS What we anticipate seldom oeeurst hat ue tease 2 expected” general ly happens. Ben Disraeli In the past ten years fatigue damage detector strips or fatigue crack monitors has been a subject frequently discussed by aircraft design specialists and members of various aircraft Fatigue Committees, Odd as it may seem for a subject so often discussed and considered, there has been no information available as to actual experience in the use of fatigue crack monitoxs, A hypothetical analysis using laboratory data has shown that a series of sensitive and bonded fatigue crack monitors, that will progressively fatigue crack due to the repeated working loads in an aircraft structure, may be used to determine the following information: 14,1 A means of assessing fatigue damage of structural members in an aircraft, 14,2 A method of comparing the difference in relative service loading of different aircraft and the same type aircraft operating by various airlines by the quantity of bonded monitors that become fatigue cracked, 14,3 A means of forewarning other aircraft after a series of cracked monitors and eventually a minor service failure has occurred on one par- ticular model, The lesser number of cracked monitors on identical aircraft will yield a relative crack expectancy of those aircraft with respect to the model that has already experienced a service failure. This use may only be justified if the cracked bonded monitors are close or ef adjacent to the location of a service failure. At present the information, as well as the accuracy of the data to be derived from the use of the monitors bonded to the aircraft, is in the laboratory stage. The proof of such a device lies ina service test, The benefits from this device may be many; for the operators a more efficient structure with less repair and maintenance; for the engineering designer increased knowledge to be able to design a better structure, The monitors however are not intended to substitute for but to supplement present inspect- ion efforts. 14,4 Monitors To be of any practical use the fatigue crack monitor, like a fuse, should naturally fail well in advance and give warning of the impending damage that may occur to the member to which it is bonded. The Douglas Company, as well as many others,have conducted laboratory tests in which a fine wire or series of wires are bonded to a sample coupon and encircling the stress raiser, As soon as the smallest of visible fatigue cracks originate they intersect and fail the fine wire and by electrical means this minute179 fracture can be simply detected, This same system can also be used to determine crack propagation data by subsequently failing a series of bonded wires spaced a known distance apart, Unfortunately, this system might require a considerable number of such bonded wires throughout the airframe and a fairly accurate determination of the location to bond the wires, Also the wires do not usually fail until the structure to which they are bonded has started to fail, Often a marginal structure is subjected to reversals of loading and for these cases it would be rather difficult to load the bonded wire. (You can't push on a rope.) It is believed that the simplest and most practical fatigue crack moni- tor may be in the form of a strip coupon containing a hole or notch to isolate its type and location of failure and of such a design to work at a higher stress than the structure to which it is bonded. A set of such coupons either in parallel or series and bonded to the airframe in regions of par- ticular interest would not be costly or time-consuming to cither install or visually monitor in obtaining the data. The method proposed in this chapter is based upon the relative distribution between the lifes of a series of progressively failing fatigue monitors on the airplane structure in service, This data correlated with the laboratory tests of specific Joints on which similar monitors are bonded is believed to be a means of estimating the fatigue crack life of an actual service-loaded structure. Only those members which are primarily axially loaded were considered in the preliminary laboratory investigations, Wing skin and fuselage Plating are not considered for this type of monitor duc to the frequent occurrence of elastic skin wrinkling and buckling, 14.5 Design One of the simplest forms of a stress raiser in a strip coupon is in the form of an unloaded hole. Since there seems to be more test data on this type of notch than any other itis the one that was recommended for the preliminary tests. To meet another requirement of having the coupon working to a higher stress than the member to which it is bonded, the coupon is "necked" down, The following analysis shows the design of a coupon "necked" down to yield a much higher strain than the member to which it is bonded, It should be noted that the coupon is only bonded at the larger shoulder ends of the coupon and not in its center Section, If the complete monitor were bonded it would be uniformly strained with the Targer structure to which it is cemented and no magnification of strain could be obtained in the monitor.180 ZL, = structural member gage length as well as unbonded coupon length A, = cross-sectional area of large shoulder end of coupon 1 8 P A = cross-sectional area of "necked" down section of coupon 2 A,, = cross section of structural member 4 , = total elongation struct, member dy 55 = total elongation of sections 1 & 2 of coupon Eo:€i: E, = strain in structural member, and sections 1 & 2 of coupon (P and E independent of geometry of coupon) Designing Ay = 3A, and ub £ ee so 1.34, 0, -60 wh iink Then 0.6656, = 6, or strain in coupon, = 1,505€, (strain in struct, member)181 By changing the ratios of "necked" down section and gage lengths almost any magnification of strain can be obtained, The magnification, however, should not be designed so high as to cause static failure of the monitors due to relatively low structure working stresses, The overall size of the crack monitors that have been investigated in the laboratory were ,032 x 3/4 x 5 1/2 24S-T3 alclad sheet weighing approximately 1/5 of an ounce. It is believed, however, that either 24ST or 75ST material would be satisfactory. Due to the corrosive environment experienced in service it would probably be best to use the clad materials, The preparation of the monitors is fairly simple. Blocks of a dozen or so coupons’ can be gang routed or a small punch die could be made and the coupons blanked out. Normal care in handling would be required so that the critical monitor test section be left reason- ably unscratched. In the laboratory tests the coupons were both riveted and bonded in their application and both methods were found to be satis- factory. The bonding method, however, seemed to be more practical for installation on actual structural members of the airplane since no holes need to be drilled in the structure, The bonding material "EPON".6, cured at room temperature, was found to be satisfactory in the tests. There is no assurance or experience in the use of bonding materials of this type over long periods of time. Accelerated laboratory tests indicate the bonding material to be satisfactory, however only the ex- posure under actual conditions of service will provide the proof, 14,6 Laboratory Testing of Fatigue Monitors Small coupons were designed as shown in section 14,5 and either bonded or riveted toa heavier piece of 0.102 ga. 75ST plate. The results of the laboratory fatigue tests are shown in figure A, The initial tests established the following: 1, The attachment method for the monitors was satisfactory, 2. By loading the monitor coupon indirectly the calculated loads and stresses in each member thay be acéurately determined, ‘The S/N graph of Fig. A shows the plotted data of the fatigued monitors and superimposed on this graph is the dashed S/N curve of previously established holed coupons that were directly loaded, The agreement between the curve and plotted data indicates that from a known load on a larger structure and the geometry of the bonded monitor the magnification of strain in the monitor was and can be accurately determined, 3, The fatigue strength characteristics of directly loaded or indirectly loaded monitors can be dupligated, and for the same Tatigue crack life. 4, The stress magnification of the monitors to the main members in the preliminary tests varied from 1,12 to 2,04,182 It is proposed that in the future as much data are collected as possible from the use of monitors bonded to structural panels and joints tested in the laboratory, The test data would be presented in the form as shown in Fig. QB. This plot has been made from actual data; however, the fatigue curves for the monitors have been translated to the stress levels of the "test joint" specimen. For example, the stress levels for monitor #3 are actually 2.0 times greater than the stress levels at which the data is plotted ({g7= 2.0). This type of plot readily shows the relative life distribution between the successively failing monitors and the structural Joint itself at the stress level to which the joint is being tested, 14.7 Location A preliminary study has been made as to the actual locations to bond the load indicating monitors (fatigue crack monitors). To be prac- tical, the locations should mect the following requirements: 1, Be covered or within the structure and not exposed to the weather, 2, Be readily accessible for examination and not in an area, for instance, that would require time to remove sealed doorgs - as the tank door areas near the center spar. 3, Be readily accessible and yet not in the open so that the monitor could be stepped on, bumped into, etc., by maintenance crewmen, 4, Be in fairly high stressed region: A few typical components that would be the most desirable to monitor would be the following: (a) Wing spar caps (root to outer joint), Principally those loaded in tension as front spar lower cap, (b) Horizontal and vertical stabilizer caps. (c) Flap hinge support brackets, (4) Main and nose landing gear components, An example of a monitor installation is shown in figure 14,8 Summary and Conclusions. As a first approximation for predicting the fatigue cracking in advance of a particular structural member, the following method is proposed, 1, The particular or marginal structural joint is to be fatigue-tested in the laboratory, Bonded to the test joint are to be a series of fatigue monitors of the same design as to be later bonded to the airplane structure, 2, The fatigue test data is to be presented in the form shown in Fig. B,183 3, The next step will be to periodically inspect the monitors that have been in service and log’ the hours of operation of the aircraft with respect to cracking of the monitors, 4, comparison of the life ratios of the monitors in the laboratory tests to the monitors in service will then be used to calculate the life or hours of operation before fatigue cracking of the par~ ticular joint on the airplane. yy 2g oint ‘joint = Fmonitor Nmnonitor See Table A for typical example calculations. Alternative uses for the monitor were mentioned in paragraphs 14,2 and 14,3, If close observance has been made of the fatigue monitors in service, the following may be deduced from the data. Figure shows the results that the monitors may provide. For a given airline "¥", let us assume the monitors crack as shown, The amount of damage is pro- portional to the distance along the abscissa and the number of hours for this damage is measured along the ordinate. Airline "Z" flies a gustier route and receives greater structural damage per operational hour. It has taken airline "Y" 6,000 hours to sustain the same amount of damage as airline "Z" sustained in only 3,000 hours. Airline "X", on the other hand flies the easiest route and may take about 12,000 hours for the same amount of structural damage. At some point in their histories, probably with airline "Z"', some fatigue damage may occur to the aircraft in the form of a cracked stringer in the vicinity of one of the monitors and it may occur at number 5. We can then say to airline "Y" that they should incorporate a service change to preclude this same trouble after the breakage of their number 4 monitor or at about 3,000 hours. This same warning applies to airline "X" at about 6,000 hours. There is good theoretical and experimental reasons for believing things will happen the way they have been described, It may be of in- terest to deliberately design some of the monitors to fail ata very low number of hours by over-amplifying the stress. This will in no way mean that the structure to which it is attached is damaged beyond a negligible amount.......... FATIGUED? (I mean you)STRESS a nT [RANGE _ OF MOST Lele. DAMAGING STRESS LEVELS - GUSTS FROM 6 To /5 FPS. MONITORS. CYCLES Nut Huot = HMONITOR ““KiMon.185 ot'g ost ‘ot 435198 09's sist oto Tos v SEM tote 0206 845% 42 96° s18t olor Tos v SOIAUGS NI LNd SNVIAMIY YALAV SNTLSWOS AATIVISNI SYOLINOW 005 "9 09% ‘OT sdyTp 42 $9 +z 096e 096€ 009 v ST O65 006 ‘8 spp we 52-2 096¢ 096€ 009 v see's ost ‘or sdy Sly 7 09's Stet stet 009 v sae a soz'h 0206 sd 793% 96°F stet sist 009 v ("saq) (*saq) uy (x01983) auerdzpy uo fy * “He Sa a Mir - zoxjuoW — sanofT quyor jo Buppoesy 1 MOM | poyoery Jo oWTL aw ‘oN a8ejasng ‘oN 1apoy e10pq aut], vied qe pue TeUOHIPPY Peseta = sUPLTTY Wort qe payetnumnooy SUE, TOITUOWY TRIO SNOILYTNI TY VLVd AYOLV¥ORVT| VLVG uXu SNITUIV "eS uoHEs07 10}UOHy INO GTANVXa Vv aTavadv) Y3MO1 avis LNOUS SNOLLVDOT BOLINOM | S3NMuIV SNoIYA 3HL JO SS3NISND Hive AHOMIS 40 ALIAS SHL 40 3uNSV3W vSiH, ONIGVOT SIANSS OL IAC 13s Vv NI SHOLINOW G3>4DVED 40 3N 2 39 S$ b6 ££ 2 1 4 KINO S1dNVX3 (DVaSaO SENOA —TWNOT.187 Acknowledgements Appreciation is expressed to H. W. Adams, who requested this course, and toW. W. Jenney, C. R. Strang and W. W. Bradley for their many helpful suggestions and continued interest throughout the compilation of these notes. Appreciation is also expressed to Miss Mary Anne Fuller and Mrs. Elsie Hillier for their typing and corrections of the notes.Pe. 12 23 20 ah 36 46 55 60 ae) 76 93 100 ut 133 ase 158 159 180 Middle of Page AD 1st Line line 23 line 15 line & Col. 4 of Table Lane 7 Idne 11 Line 6 Line 8 Line 10 near bot. Line 5 Line 9 4.2L. Line 7 *Footnote 9.2 Last Line Line 4 Col. 2 and Col. 3 Fig. at Bottom Line 6 Line 9 CORRECTIONS Description stress range” =o. Osa The influence of static tension on the - - - -. TeQ crosses the stress SW * SE zs PLAIN Ke" yorcmD Keg = (2.6) (30,100) = Last Line - change t2h,400 to = 23,400 wssto each other as foro, and o, + Spy OF 5h 2 ‘The equation on page 61 for ductile materials equation is o = 0, (2 - a) "and the $/N curve figure 15 N = 10,000 - 20,000 cycles" - - - ~~ - (2). figure 52 "s+ + dnterferences of .004 to .008 - - - - Figure 53 ATA - Air Transport Assoc. Change 1,640 to 1,550 Change 7°/, to 1/7th (2 places) interchange headings at top of colums Curve "A" is impossible.can't go below 13,000 psi ~ another fallacy of this particular method. dnterchange specimen numbers on 4 and 5 change - - - example in 5. to read --- example in b. FL from 6 = FE