This standard is for EDUCATIONAL USE ONLY.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions au

FftOFERTY

OF

university of Michigan
Liabraries
1017
AftTES 8C1ENTIA VBRITAS

The

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

A GUIDE FOR FATIGUE TESTING AND THE STATISTICAL ANALYSIS OF FATIGUE DATA

Prepar'ed by COMMITTEE -9 ON FATIGUE AMERICAN SOCIETY FOR TESTING AND MATERIALS

1963

Reg. U. S. Pat. Off.

ASTM Special Technical Publication No. 91-A* (Second Edition} Price: $5.00; to Members: $4.00

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by Published by the This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

AMERICAN SOCIETY FOR TESTING AND MATERIALS 1916 Race Street, Philadelphia 3, Pa.

BY AMERICAN SOCIETY FOR TESTING AND MATERIALS 1963 Library of Congress Catalog Card Number: 63-16331

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reprodu
Printed in Baltimore, Md. February, 1964

FOREWORD

The First Edition of this Guide was the composite work of many people who contributed a great deal of time to the discussion and writing of the text under the guidance of Task Group Leader, F. B. Stulen. A major portion of the statistical section was written by Miss Mary N. Torrey. George R. Gohn not only contributed to the discussion and planning, but also edited and arranged for the printing of the advance copies of the text. The coordination of contributions and discussions was done by H. N. Cummings. Appreciable contributions to the statistical parts of the Guide were also made by D. H. Shaffer. In addition to the above, R. E. Peterson, H. F. Dodge, D. P. Gaver, R. Hooke, W. T. Lankford, R. B. Murphy, W. C. Schulte, P. R. Toolin, and M. B. Wilk contributed to the discussions at various conferences. The original Task Group was organized under the leadership of J. T. Ransom, and a first rough draft was prepared in 1954 and revised in 1955. Other contributors to these drafts were E. W. Ellis, W. T. Lankford, F. A. McClintock, R. E. Peterson, E. H. Schuette, F. B. Stulen, and E. J. Ward. In 1956, F. B. Stulen became Leader of the Task Group and the Guide was completed under his direction. Upon the formation of Subcommittee VI on the Statistical Aspects of Fatigue, this subcommittee was asked to review the First Edition and to make any revisions necessary to bring the Guide up to date. As a result of this study, extensive revisions have been made in various sections as printed in this Second Edition. They include: (1) revisions in the definitions (Section II) and their separate publication as ASTM Tentative Definitions E 206,1 (2) an expansion of Section IV on the number of test specimens, (3) changes in Section V on tests of significance, and (4) the preparation of a new section, Appendix IV, on the use of the Weibull distribution function for fatigue Me. This work was carried out by four Task Groups headed by S. M. Marco, H. E. Frankel, Miss M. N. Torrey, and C. A. Moyer, respectively. Others who assisted in the preparation of the Second Edition were W. N. Findley, R. A. Heller, J. H. K. Kao, H. N. Cummings, W. S. Hyler, B. Ruley, and G. R. Gohn, Chairman of Subcommittee VI.
1 Definitions of Terms Relating to Fatigue Testing and the Statistical Analysis of Fatigue Data (E 206), 1962 Supplement to Book of ASTM Standards, Part 3.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduction
iii

NOTE.The Society is not responsible, as a body, for the statements and opinions advanced in this publication.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No furthe

CONTENTS

I. II. III. IV. V.

Purposes of Fatigue Testing Definitions, Symbols, and Abbreviations Test Procedures Minimum Number of Test Specimens and Their Selection Analysis of Fatigue Data Appendices Miscellaneous Reference Tables Additional Technique for Distribution Shape Not Assumed Analysis of Correlation Between Two Variables The Weibull Distribution Function for Fatigue Life References Index

PAGE

1 2 8 16 22 55 68 69 71 78 81

LIST OF T A B L E S

1.Allocation of Test Specimens for "Probit" Method of Test 11 2.Minimum Number of Specimens Needed for Determining 95 Per Cent Confidence Intervals of Stated Width for a Population Mean, p 19 3.Minimum Number of Specimens Needed for Determining 95 Per Cent Confidence Intervals of Stated Width for a Population Standard Deviation, a 19 4.Minimum Number of Specimens Needed to Detect if the Standard "Deviation of a Population Is a Stated Percentage of a Fixed Value 20 5.Minimum Number of Specimens Needed in Each Sample to Detect if a Standard Deviation of One Population Is a Stated Multiple of the Standard Deviation of Another Population 20 6.Minimum Number of Specimens Needed to Detect a Stated Difference Between a Mean and a Fixed Value 21 7.Minimum Number of Specimens Needed to Detect a Stated Difference Between the Means of Two Populations 21 8.Median Percentage of Survivors for the Population 24 9.Confidence Intervals for the Median 26 10.Approximate Confidence Intervals for the Mean 27 11.Confidence Intervals for Percentages 28 12.Fatigue Test Data 29 13.Fatigue Test Data 31 14.Percentages Surviving 108 Cycles 32 15.-"Probit" Test Data 33 16.Computations for Fitting a Response Curve by Method of Least Squares 35 17.Computation of Standard Deviation, s 36 18.Method of Computing Cent Confidence Limits for Per Cent Survival Copyright by ASTM Int'l95 (allPer rights reserved); Tue Jul 9 10:11:02 EDT 2013 Values 37 Downloaded/printed by 95 Per Cent Confidence Limits for Fatigue Strength 19.Method of Computing Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further Values 38 reproduct 20.Computations for Significance Tests 46 21.R. R. Moore Rotating Beam; Step Tests of 42 Specimens 50 This standard is for EDUCATIONAL USE ONLY. 22.Analysis of Data in Table 21 51 23.Prot Test Computations 52
v

TABLE

PAGE

vi
TABLE

CONTENTS
PAGE

24.Prot Test Computations 25.Minimum Per Cent of Population Exceeding Median of Low Ranking Points.. 26.Unpaired Rank Test 27.Percentiles of the x2 Distribution 28.Areas of the "Normal" Curve 29.Values of t 30.Percentiles of the x2/d-f- Distribution 31.Mo.25 and uo.yis for Runs Among Elements in Samples of Sizes Ni and Nz 32.F Distribution 33.k Factors for S-N Curves (Normal Distribution Assumed) 34.Working Significance Levels for Quadrant Sum 35.Ordinate Locations Corresponding to Per Cent Failed Values 36.Mean-Rank Estimates of the Per Cent Population Failed Corresponding to Failure Order in Sample 37.Typical Fatigue Test Data 38.Typical Fatigue Test Data, Without Runouts 39.Typical Fatigue Test Data, with Runouts

53 56 58 60 61 62 63 64 65 67 69 72 74
75

75 77

LIST OF F I G U R E S

FIGURE

1.Probability-Stress-Cycle (P-S-N) Curve for Phosphor-Bronze Strip 2.Response or Survival Tests 3.Illustration of Staircase Method 4.Representation of "Step" Testing of Single Specimen 5.Graphical Illustration of Prot Data 6."Normal" or Gaussian Distribution Curve 7.Response Curves for a Particular Type of Steel 8.Per Cent of Specimens Having at Least the Indicated Fatigue Strength at 107 Cycles 9.Prot Test: Stress as Linear Function of Stress Cycles 10.Log-Log Plot of Prot Data 11.Scatter Diagram 12.Typical Weibull Distribution Curves 13.Construction of Weibull Probability Paper from Log-Log Paper 14.Estimation of Weibull Distribution Function Parameters for Data in Table 38. 15.Per Cent Failed at Weibull Mean 16.Estimation of Weibull Distribution Function Parameters for Data in Table 39..

10 11 12 14 15 22 34 51 53 54 70 72 73 75 76 77

RELATED ASTM PUBLICATIONS

Abstracts of Articles on Fatigue (STP 9) Fatigue Manual (STP 91) (1949) Statistical Aspects of Fatigue (STP 121) (1951) Fatigue, with Emphasis on Statistical Approach (STP 137) (1952) Papers on Metals (STP 196) (1956) Fatigue of Aircraft Structures (STP 203) (1956) Large Fatigue Testing Machines and Their Results (STP 216) (1957) Basic Mechanisms of Fatigue (STP 237) (1958) Fatigue of Aircraft Structures (STP 274) (1959) Copyright by ASTM (all (1960) rights reserved); Tue Jul 9 10:11:02 EDT 2013 Acoustical Fatigue (STPInt'l 284) Fatigue of Aircraft Structures (STP 338) (1963) Downloaded/printed by

This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorize

STP91A-EB/Feb. 1964

GUIDE FOR FATIGUE TESTING AND STATISTICAL ANALYSIS

About 15 years ago, ASTM Committee E-9 on Fatigue prepared a Manual on Fatigue Testing.1 That Manual attempted to standardize the symbols and nomenclature used in fatigue testing, described the principal types of testing machines then in use, presented detailed instructions for the preparation of test specimens, outlined test procedures and techniques, and gave some suggestions for the presentation and interpretation of fatigue data. Since the Manual was first prepared, a number of new techniques have been developed for evaluating the fatigue properties of materials. Furthermore, the application of statistical methods to the analysis of the test results of samples offers a means for estimating the characteristics of the population from which the samples were taken. To take cognizance of these developments, this guide has been prepared. I. PURPOSES OF FATIGUE TESTING The purposes of fatigue testing are (1) to estimate the relationship between stress- (load-, strain-, deflection-) amplitude and cycle life-to-failure for a given material or component, and (2) to compare the fatigue properties of two or more materials or components. In order to specify the reliability of these estimates, they must be based on the results of testing a sample of fatigue specimens which have been drawn at random from a population of possible fatigue specimens and tested in accordance with acceptable testing procedures. The principal acceptable procedures discussed in this guide are:

A. "Standard" tests (constant amplitude or classical Wohler method).2 1. Single test specimen at each stress level. 2. A group of test specimens at each stress level. B. Response tests (constant amplitude). 1. "Probit" method. 2. Staircase method. 3. Modified staircase method. C. Increasing amplitude tests. 1. Step method. 2. Prot method. The primary purposes of the statistical analysis of fatigue data are: (1) to estimate certain fatigue properties of a material or a component (together with measures of the reliability) from a given set of fatigue data, obtained by testing a sample of fatigue specimens in accordance with one of the previCopyright by ASTM Int'l(2) (all to rights reserved); Tue Jul 9 procedures 10:11:02 EDTfor 2013 ous test procedures, and provide objective comparing Downloaded/printed by two or more sets of fatigue data to determine whether or not the data come Escuela Politcnica del Ejercito en Ecuador ESPEalso pursuant to License Agreement. on No further from similar populations. Statistical theory provides information (a) reproduc
1 2

Fatigue Manual, ASTM STP 91, Am. Soc. Testing Mats., 1949. This standard is for EDUCATIONAL USE ONLY. The term "standard" test, as used here, does not imply an ASTM standard.
1

Copyright 1964 by ASTM International www.astm.org

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

the most efficient use of a limited number of test specimens and (b) the number of test specimens required to give a specified degree of confidence in the test results. Even with some basic training, it is difficult to locate the techniques particularly useful in fatigue testing in the statistical literature. The purpose of this guide is to describe some statistical treatments that are suitable for the analysis of fatigue data obtained in any one of the foregoing test methods and to present these statistical treatments in a form useful to the test engineer. Definitions of certain statistical terms are included, but only enough of the basic concepts of statistics are included to make the methods understandable; theory is left to the references. Test procedures are discussed hi Section III while techniques for analyzing the data obtained in these tests are given in Section V and the Appendices. II. DEFINITIONS, SYMBOLS, AND ABBREVIATIONS Relating to Fatigue Tests and Test Methods: To encourage uniformity of terminology, the terms dealing primarily with fatigue testing and test methods are also published in ASTM Definitions E 6.3 The symbols used are, hi general, those recommended in the American Standard Letter Symbols for Mechanics of Solid Bodies.4 1. Fatigue (Note 1).The process of progressive localized permanent structural change occurring in a material subjected to conditions which produce fluctuating stresses and strains at some point or points and which may culminate in cracks or complete fracture after a sufficient number of fluctuations (Note 2).
NOTE 1.The term fatigue in the materials testing field, hasin at least one case glass technologybeen used for static tests of considerable duration, a type of test generally designated as stress-rupture. NOTE 2.Fluctuations may occur both in stress and with time (frequency), as in the case of "random vibration."

2. Fatigue Life, N.The number of cycles of stress or strain of a specified character that a given specimen sustains before failure of a specified nature occurs. Definitions 3 to 19, inclusive, apply to those cases where the conditions imposed upon a specimen result or are assumed to result in uniaxial principal stresses or strains which fluctuate in magnitude. Multiaxial stress, sequential loading, and random loading require more rigorous definitions which are, at present, beyond the scope of this section. 3. Nominal Stress, S.The stress at a point calculated on the net cross-section by simple elastic theory, without taking into account the effect on the stress produced by geometric discontinuities such as holes, grooves,'fillets, etc. 4. Stress Cycle.-^-The smallest segment of the stress-time function which is repeated periodically.
3 Definitions of Terms Relating to Methods of Mechanical Testing, 1962 Supplement Downloaded/printed by to 1961 Book of ASTM Standards (E 6), Part 3. This standard is for EDUCATIONAL USE ONLY. 4 Escuela Politcnica delAm. Ejercito en Ecuador ASA No. Z10.31948, Standards Assn.,ESPE 1948. pursuant to License Agreement. No further

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

DEFINITIONS, SYMBOLS, AND ABBREVIATIONS

5/ Maximum Stress, Smax- The stress having the highest algebraic value in the stress cycle, tensile stress being considered positive and compressive stress negative. In this definition, as well as in others that follow, the nominal stress is used most commonly. 6./ Minimum Stress, Sm,n. The stress having the lowest algebraic value in the cycle, tensile stress being considered positive and compressive stress negative. ?. Mean Stress (or Steady Component of Stress), Sm. The algebraic average of the maximum and minimum stresses in one cycle, that is,

8. Range of Stress, Sr.The algebraic difference between the maximum and minimum stresses in one cycle, that is 9. Stress Amplitude (or Variable Component of Stress), Sa-One half the range of stress, that is

10. Stress Ratio, A or R. The algebraic ratio of two specified stress values in a stress cycle. Two commonly used stress ratios are: The ratio of the stress amplitude to the mean stress, that is,

and the ratio of the minimum stress to the maximum stress, that is,

11. S-N Diagram. A plot of stress against the number of cycles to failure. The stress can be Smax, Smin, or Sa- The diagram indicates the S-N relationship for a specified value of Sm, A, or R and a specified probability of survival. For N a log scale is almost always used. For S a linear scale is used most often, but a log scale is sometimes used. 12. Stress Cycles Endured, N.The number of cycles of a specified character (that produce fluctuating stress and strain) which a specimen has endured at any time in its stress history. 13. Fatigue Strength at TV Cycles, SN. A hypothetical value of stress for failure at exactly N cycles as determined from an S-N diagram. The value of SN thus determined is subject to the same conditions as those which apply to the S-N diagram. NOTE. The value of SN which is commonly found in the literature is the hypothetical value of Smai , -Smt^ror Sa , at which 50 per cent of the specimens of a given sample could survive N stress cycles in which Sm = 0. This is also known as the median fatigue strength at N cycles (see definition 47). 14. Copyright Fatigue Limit, S/. The limiting value ofJul the9 median fatigue strength as N by ASTM Int'l (all rights reserved); Tue 10:11:02 EDT 2013 becomes very large. Downloaded/printed by
This standard is for EDUCATIONAL USE ONLY.

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant License Agreement. Nolimit. further reproduct NpTE.->-Certain materials and" environments preclude the to attainment of a fatigue

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

Values tabulated as fatigue limitsun the literature are frequently (but not always) values 6f SN for 50~per cent survival at N cycles of stress in which Sm = 0.

15. Cycle Ratio, CThe ratio of the number of stress cycles, n, of a specified character to the hypothetical fatigue life, N, obtained from the S-N diagram, for stress cycles of the same character, that is,

Theoretical Stress Concentration Factor (or Stress Concentration Factor), Kt-The ratio of the greatest stress in the region of a notch or other stress concentrator, as determined by the theory of elasticity (or by experimental procedures that give equivalent values), to the corresponding nominal stress.
NOTE. The theory of plasticity should not be used to determine Kt .

17. Fatigue Notch Factor, K/. The ratio of the fatigue strength of a specimen with no stress concentration to the fatigue strength at the same number of cycles with stress concentration for the same conditions.
NOTE. In specifying Kj it is necessary to specify the geometry and the values of Smax , Sm , and N for which it is computed.

18. Fatigue Notch Sensitivity, g.A measure of the degree of agreement between Kf and Kt for a particular specimen of a given size and material containing a stress concentrator of a given size and shape.
NOTE. A common definition of fatigue notch sensitivity is q = (Kf i)/(Kt 1\ in which q may vary between zero (where Kf = 1) and one (where Kf = t). K

19. Constant Life Fatigue Diagram.A plot (usually on rectangular coordinates) of a family of curves, each of which is for a single fatigue life, N, relating Sa, Smax and/or Smin to the mean stress Sm. The constant life fatigue diagram is generally derived from a family of S-N curves, each of which represents a different stress ratio, A or R, for a 50 per cent probability of survival.

Relating to Statistical Analysis:

20. Population (or Universe).The hypothetical collection of all possible test specimens that could be prepared in the specified way from the material under consideration. 21. Sample.The specimens selected from the population for test purposes.
NOTE. The method of selecting the sample determines the population about which statistical inference or generalization can be made.

22. Group.The specimens tested at one time, or consecutively, at one stress level. A group may comprise one or more specimens. 23. Frequency Distribution.The way in which the frequencies of occurrence of members of a population or sample are distributed according to the values of the variable under consideration. 24. Parameter.A constant (usually unknown) denning some property of the frequency distribution of a population, such as a population median or a population standard deviation. Copyright by Int'lvalue (all rights reserved); 9 10:11:02 EDT 25. Statistic. A ASTM summary calculated fromTue theJul observed values in a2013 Downloaded/printed by sample. This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito Ecuadora ESPE pursuant to License Agreement. No f 26. Estimation. A procedure forenmaking statistical inference about the

DEFINITIONS, SYMBOLS, AND ABBREVIATIONS

numerical values of one or more unknown population parameters from the observed values in a sample. 27. Estimate.The particular value, or values, of a parameter computed by an estimation procedure for a given sample. 28. Point Estimate.The estimate of a parameter given by a single statistic. 29. Sample Median.-The middle value when all observed values in a sample are arranged in order of magnitude if an odd number of specimens are tested. If the sample size is even, it is the average of the two middlemost values. It is a point estimate of the population median, or 50 per cent point. 30. Sample Average (Arithmetic Mean).The sum of all the observed values in a sample divided by the sample size. It is a point estimate of the population mean. 31. Sample Variance, s2.The sum of the squares of the differences between each observed value and the sample average divided by the sample size minus one. It is a point estimate of the population variance.
NOTE.This value of s* provides both an unbiased point estimate of the population variance and a statistic that is used in computing interval estimates and several test statistics (see definitions 34 and 42). Some texts define s2 as "the sum of the squares of the differences between each observed value and the sample average divided by the sample size," but this statistic is not as useful.

32. Sample Standard Deviation, s.The square root of the sample variance. It is a point estimate of the population standard deviation, a measure of the "spread" of the frequency distribution of a population.
NOTE.This value of 5 provides a statistic that is used in computing interval estimates and several test statistics (see definitions 34 and 42). For small sample sizes, j underestimates the population standard deviation. (See the ASTM Manual on Quality Control of Materials5 or texts on statistics for an unbiased estimate of the standard deviation of a Normal population.)

33. Sample Percentage.The percentage of observed values between two stated values of the variable under consideration. It is a point estimate of the percentage of the population between the same two stated values. (One stated value may be or + .) 34. Interval Estimate.The estimate of a parameter given by two statistics, denning the end points of an interval. 35. Confidence Interval.An interval estimate of a population parameter computed so that the statement, "the population parameter lies in this interval," will be true, on the average, in a stated proportion of the times such statements are made. 36. Confidence Limits.The two statistics that define a confidence interval. 37. Confidence Level (or Coefficient).The stated proportion of the times the confidence interval is expected to include the population parameter. 38. Tolerance Interval.An interval computed so that it will include at least a stated percentage of the population with a stated probability. 39. Tolerance Limits.The two statistics that define a tolerance interval. (One value may be or + .) 40. Tolerance Level.The stated probability that the tolerance interval includes at least the stated percentage of the population. It is not the same as a confidence level but the term confidence level is frequently associated with tolerance intervals. Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 41. Significant.Statistically significant. An effect or difference between popu6

Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Soc. Ejercito en Ecuador ASTM STP 15-C, Am. Testing Mats., ESPE 1957. pursuant to License Agreement. No further re

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

lations is said to be present if the value of a test-statistic is significant, that is, lies outside of predetermined limits.
NOTE.An effect which is statistically significant may or may not have engineering significance.

42. Test-Statistic.A function of the observed values in a sample that is used in a test of significance. 43. Test of Significance.A test which, by use of a test-statistic, purports to provide a test of the hypothesis that the effect is absent.
NOTE.The rejection of the hypothesis indicates that the effect is present.

44. Significance Level.The stated probability (risk) that a given test of significance will reject the hypothesis that a specified effect is absent when the hypothesis is true. Relating to Statistical Analysis of Fatigue Data: 45. Median Fatigue Life.The middlemost of the observed fatigue life values, arranged in order of magnitude, of the individual specimens in a group tested under identical conditions. In the case where an even number of specimens are tested it is the average of the two middlemost values.
NOTE 1.The use of the sample median, instead of the arithmetic mean (that is, the average), is usually preferred. NOTE 2.In the literature, the abbreviated term "fatigue life" usually has meant the median fatigue life df the group. However, when applied to a collection of data without further qualification the term "fatigue life" is ambiguous.

46. Fatigue Life for p Per Cent Survival.An estimate of the fatigue life that p per cent of the population would attain or exceed at a given stress level. The observed value of the median fatigue life estimates the fatigue life for 50 per cent survival. Fatigue life for p per cent survival values, where p is any number, such as 95, 90, etc., may also be estimated from the individual fatigue life values. 47. Median Fatigue Strength at N Cycles.An estimate of the stress level at which 50 per cent of the population would survive N cycles. NOTE 1.The estimate of the median fatigue strength is derived from a particular point of the fatigue life distribution, since there is no test procedure by which a frequency distribution of fatigue strengths at N cycles can be directly observed. NOTE 2.This is a special case of the more general definition 48. 48. Fatigue Strength for p Per Cent Survival at N Cycles.An estimate of the stress level at which p per cent of the population would survive N cycles; p may be any number, such as 95, 90, etc. NOTE.The estimates of the fatigue strengths for p per cent survival values are derived from particular points of the fatigue life distribution, since there is no test procedure by which a frequency distribution of fatigue strengths at N cycles can be directly observed. 49. Fatigue Limit for^ Per Cent Survival.The limiting value of fatigue strength for p per cent survival as N becomes very large; p may be any number, such as 95, 90, etc. (See Note, definition 14.)

50. S-N Curve for 50 Per Cent Survival.A curve fitted to the median values of Copyright ASTM Int'l (all rights reserved); 9 estimate 10:11:02 EDT fatigue life at by each of several stress levels.Tue It Jul is an of 2013 the relationship between Downloaded/printed applied stress by and the number of cycles-to-failure that 50 per cent of the This standard is for EDUCATIONAL USE ONLY. Escuelawould Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorize population survive.

DEFINITIONS, SYMBOLS, AND ABBREVIATIONS

NOTE 1.This is a special case of the more general definition 51. NOTE 2.In the literature, the abbreviated term "S-N Curve" usually has meant either the S-N curve drawn through the means (averages) or the medians (50 per cent values) for the fatigue life values. Since the term "S-N Curve" is ambiguous, it should be used in technical papers only when adequately described.

51. S-N Curve for p Per Cent Survival.A curve fitted to the fatigue life for p per cent survival values at each of several stress levels. It is an estimate of the relationship between applied stress and the number of cycles-to-failure that p per cent of the population would survive; p may be any number, such as 95, 90, etc.
NOTE.Caution should be used in drawing conclusions from extrapolated portions of the S-N curves. In general, the S-N curves should not be extrapolated beyond observed life values.

52. Response Curve for N Cycles.A curve fitted to observed values of percentage survival at N cycles for several stress levels, where N is a preassigned number such as 106, 107, etc. It is an estimate of the relationship between applied stress and the percentage of the population that would survive N cycles.
NOTE 1.Values of the median fatigue strength at N cycles and the fatigue strength for p per cent survival at N cycles may be derived from the response curve for N cycles, if p falls within the range of the per cent survival values actually observed. NOTE 2.Caution should be used in drawing conclusions from extrapolated portions of the response curves. In general, the curves should not be extrapolated to other values of p.

LIST OF SYMBOLS AND ABBREVIATIONS The following terms are frequently used in lieu of or along with the terms covered by the preceding definitions. In general the symbols are those recommended in the American Standard Letter Symbols for Mechanics of Solid Bodies (see footnote 4). For stress, the use of S with appropriate lower case subscripts is preferred for general purposes; for mathematical analysis the use of Greek symbols is generally preferred. Symbol Term

A Area of cross-section, Stress ratio C Cycle ratio c Distance from centroid to outermost fiber D or d Diameter E Modulus of elasticity in tension or compression e (epsilon) Strain f t-lb Unit of work G Modulus of elasticity in shear 7 (gamma) Shear strain / Moment of inertia i Subscript denoting ith term in.-lb Unit of work / Polar moment of inertia Int'l (all rights of reserved); Tue Jul 9 10:11:02 EDT 2013 ksi Copyright by ASTM Thousands pounds per square inch Downloaded/printedor by kips per square inch This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized. Kf Fatigue notch factor

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA Term Theoretical stress concentration factor Unit of torque Unit of torque Bending moment Poisson's ratio

Fatigue life; number of cycles JNumber of stress cycles endured; Sample size Load Pounds per square inch Probability of failure; Per cent failure; Probability of survival; Per cent survival Fatigue notch sensitivity 9 Stress ratio R s Sample standard deviation s2 Sample variance S Nominal stress S or <r (sigma) Normal stress Stress amplitude Sa Compressive stress Sc Fatigue limit Sff Compressive yield strength Jcj/ Mean stress SJ> Maximum stress S^ max C . Minimum stress LJrmn SN Fatigue strength at N cycles Sr Range of stress Ss or T (tau) Shear stress Tensile stress St Stv Tensile yield strength Tensile strength Su (7 (sigma) Standard deviation; Stress .A (7 Estimate of standard deviation (T Variance Standard deviation of x <rx Variance of x <rx2 T Torque; Temperature t Time T (tau) or Ss Shear stress
2

Symbol Kt Ib-ft Ib-in. M n (mu) preferred, or v (nu) (generally used in applied mechanics) N n P psi P

III. TEST PROCEDURES Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Until recently, there was only oneThis accepted method of conducting laboraDownloaded/printed by standard is for EDUCATIONAL USE ONLY. tory fatigue tests on aEjercito material or component. This "standard" test, using Escuela Politcnica del en Ecuador ESPE pursuant to License Agreement. No further reproduct single specimens at several stress levels, is described in Section V of the

x TEST PROCEDURES

Manual on Fatigue Testing (STP91). Experience showed, however, that this test method did not give adequate information for many of the purposes for which fatigue data are needed. Therefore, within the last ten years, a number of new methods for performing more meaningful fatigue tests have been introduced, each method having certain advantages. The choice of test method depends upon the objective of the test and the number of available test specimens. When the objective is to determine an S-N curve, the "standard" tests (Sections I Al and A2) are generally the most suitable. To determine the long-life fatigue strength or the fatigue limit, response tests (Sections I Bl, B2, and B3) or increasing amplitude tests (Sections I Cl and C2) are recommended. The latter methods also are used for comparing the long-life fatigue properties of different materials or different methods of processing. All seven of these experimental fatigue testing techniques are described in the following paragraphs. For analysis of the data, see Section V. A. "STANDARD" TESTS (CONSTANT AMPLITUDE) 1. Single Test Specimen at Each Stress Level: In the "standard" test method described in STP 91, each fatigue specimen is cycled at a different constant stress (or strain) amplitude until fracture occurs. The stress levels are usually selected to cover a series of stresses ranging from high values, at which failure will occur within a limited number of cycles, to low values at which no failure will occur (runouts) or at which failure will occur only after an extremely large number of cycles. If the primary interest is in the longlife end of the S-N relationship (often called the fatigue limit), the investigator usually has some preconception of this value for the material or component to be tested. In this case, the first stress level is selected somewhat above the estimated fatigue limit. Depending upon the results of the first test, succeeding specimens are then tested at stress levels either above or below this value, until a stress level is reached at which the specimen does not fail within the prescribed number of cycles. Near the fatigue limit, some specimens must be run at stress levels high enough to produce failures in order to have data from which the fatigue limit may be estimated. This method of test is used when the investigator has available only a relatively small number of specimens for test. Such is generally the situation when (1) the fatigue specimens are expensive, (2) the supply of material is limited, or (3) machine parts, full size sections, or assemblies are being tested.

2. Group of Specimens Tested at Each Stress Level: Since the "standard" test, using only one specimen at each stress level, gives very little information concerning the variability of the material or component and test procedure, it is more satisfactory to test several specimens at each of a number of different stress levels. In this procedure, each group should consist of at least four specimens in order to estimate the variCopyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 ability of the data. Ten or more specimens are preferable to obtain some Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. indication asPolitcnica to the shape of the fatigue life values. Three orfurther repr Escuela del Ejercito en distribution Ecuador ESPE of pursuant to License Agreement. No more different stress levels must be investigated for the determination of the

10

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

S-N curves for p per cent survival (see Fig. I).6 Generally, at least four or five stress levels are used in a test of this nature. To determine the fatigue limit of the material, a number of groups also should be tested at stress levels in the vicinity of the fatigue limit. This would increase the total number required to at least seven groups. Furthermore, to obtain approximately an equal degree of precision throughout the range of the S-N curve, more specimens should be tested in the long-life than in the short-life range.

FIG. 1.Probability-Stress-Cycle (P-S-N) Curve for Phosphor-Bronze Strip.

B. RESPONSE TESTS (CONSTANT AMPLITUDE) 1. "Probit" Method: In the "Probit" method, one or more groups of specimens are tested for a fixed number of cycles at four or five different stress levels distributed about the stress of interest. This test has been used primarily for estimating thefatigue limit of a material, that is, the stress at which 50 per cent of the test specimens will fail prior to, and 50 per cent will survive, the preassigned cycle life, N. The test is not limited to this application; it is just as valuable for estimating the fatigue strength or the fatigue limit at any other percentages of survival, provided that the specimens are properly allocated to the various stress levels. When used to estimate the fatigue limit at 50 per cent survival, at least two stress levels should be selected so that the percentage of specimens surviving N cycles will be less than 50 and two more stress levels selected at which the percentage of survivors will be more than 50. A fifth stress level producing approximately 50 per cent survivors is Copyright by essential. ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 desirable but not
6

Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. Such curves are sometimes referred to as probability-stress-cycle (P-S-N) curves.

TEST PROCEDURES
TABLE 1.ALLOCATION OF TEST SPECIMENS FOR "PROBIT" METHOD OF TEST.
Expected Per Cent Survival Relative Group Size"

11

25 to 75 15 to 20, 80 to 85 10,90 5,95 2, 98

1 1.5 2 3 5

a The group size is the number of specimens included in a test at one stress level. Thus, whatever group size is chosen for testing at stress levels for which the expected per cent survival is between 25 and 75, the sizes of other groups must be increased by the factor in the second column to obtain the number of test specimens required for testing at stress levels for which the per cent survival is expected to be larger, or smaller, if similar precision is to be obtained in the test results. If the stress levels are chosen successively, starting with levels requiring the smallest group size, the group size required for the other levels will be determined more easily. Previous data for the same material or similar materials should be used as a guide for choosing the stress levels, whenever they are available; otherwise a preliminary test such as that described under Sections IIIA1 or B2 may be required. A properly designed "Probit" test will give more useful fatigue data than any of the other response or increasing amplitude tests.

FIG. 2.Response or Survival Tests.

In "Probit" tests a group should consist of not less than five specimens and the total tested at all stress levels should be at least 50. The distribution of the total number of available test specimens will depend upon the purpose of the test. The relative group sizes for different stress levels are shown in Table 1. This allocation is suggested so that the observed percentage survival values will have approximately equal weight, a condition necessary for fitting the response curve by the usual method of least squares. This allocation also by facilitates the (all computation of confidence onEDT the 2013 reCopyright ASTM Int'l rights reserved); Tue Jul 9 limits 10:11:02 by sponseDownloaded/printed curves. As an alternative to the use of the relative group sizes (TaThis standard is for EDUCATIONAL USE ONLY. Escuela Ejercito en used Ecuador ESPE pursuant to License Agreement. ble 1), groupsPolitcnica of other del sizes can be at each stress level, provided that

No fu

12

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

weighting factors are employed and the analysis conducted as indicated in references (1-3) .7 Figure 2 presents data that might be obtained in a "Probit" test of the type described if the preassigned number of cycles were 107. Although not required for the "Probit" analysis, the actual number of cycles-to-failure should be recorded for each specimen that fails before 107 cycles, so that the data may be available for other types of analysis, such as the plotting of P-S-N Curves. 2. The Staircase Method: The staircase (or "up-and-down") method of testing is.a variation of the "Probit" method. It may require fewer specimens than the latter but is likely to be useful only when the primary interest is in the mean fatigue strength corresponding to a preassigned cycle life, N* The advantage gained

FIG. 3.Illustration of Staircase Method. NOTESpecimens numbered in chronological order. Number of cycles for each test is constant unless failure occurs beforehand.

in reducing the number of specimens tested may be offset by an increase in the time required to conduct the test. In the staircase method the specimens are tested sequentially, one at a time. The first specimen is tested at a stress level equal to the estimated value of mean fatigue strength for the prescribed number of cycles or until it fails, if it fails before that number of cycles. If the specimen fails, the next specimen is tested at a stress level that is one increment below the first stress level. If the first specimen does not fail, the second specimen is tested at a stress level that is one increment above the first stress level, and so forth. The data are recorded as shown in Fig. 3. The specimens that did not fail are designated by the o's and those that failed as #'s. The chart shows at a glance the stress level that should be used for the next test. The selection of the proper increment of stress level is very important.

7 The boldface numbers in parentheses refer to the list of9 references appended Copyright by ASTM Int'l (all rights reserved); Tue Jul 10:11:02 EDT 2013 to this guide. 8 Downloaded/printed by The staircase method as described in the literature estimates the mean fatigue strength, This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito enis Ecuador ESPE pursuant to License Agreement. No further re not the median fatigue strength which used elsewhere in this guide.

TEST PROCEDURES

13

Ideally, most of the tests should be made at three stress levels, so chosen that about 50 per cent of the test specimens survive at the middle stress level, about 70 per cent survive at the lower stress level, and about 30 per cent survive at the higher stress level. Previous data for the same or similar materials are needed in order to choose the stress levels efficiently. If none are available, some preliminary testing may be required. Such data are discarded up to the first pair of data giving opposite results; for example, in Fig. 3, data for tests 1, 2, and 3 should be discarded. Since the testing is concentrated at stress levels near the mean fatigue strength value, the number of specimens tested may be less than for the "Probit" method, which gives results for a wider range of stress values. In general, at least 30 specimens should be tested because, at most, only half of the test results are actually used in the computation of the mean fatigue strength. If data obtained by the staircase method are analyzed by response curve methods, the results may be statistically biased because of the sequential nature of the staircase method. Further, if the main interest lies in estimating the response curverather than the mean strengthat N cycles, the staircase method is not an efficient experimental procedure. 3. Modified Staircase Method: The time required to complete a test by the staircase method can be reduced by dividing the one long staircase program into several shorter, independent staircases and conducting these several tests simultaneously. This treatment is known as the modified staircase method. In the modified procedure, the total number of specimens, T, is divided into r groups of n each, so that rn = T. Each group is tested as a separate staircase program, with a separate chart for each group. Thus several machines may be used simultaneously. In the modified staircase method, as in any other test in which specimens in a group are tested on more than one machine, a check should be made to determine whether the machines give significantly different results. If the results are not significantly different, the data may be combined for statistical analysis. C. INCREASING AMPLITUDE TESTS 1. Step Method:

In many cases the "Probit" or staircase methods of test require more specimens than are available. When only a few parts are available for determining the fatigue limit, a natural desire is to test each part until it actually fails instead of just counting the number of runouts. When testing a limited number of specimens sometimes the practice is to run each specimen at several stress levels for a large number of cycles, say 107. If typical response curves for the material are available, the test may be started at a stress level corresponding to a percentage survival of approximately 90 per cent. For each successfully completed run, the applied stress level is inCopyright by ASTM Int'l (all rights Tue 10:11:02 EDT 2013 creased by an amount corresponding toreserved); a decrease inJul the 9 probability of surDownloaded/printed byand the test is repeated until failure of the specivival of about 5 per cent This standard is for EDUCATIONAL ONLY. EscuelaWhen Politcnica del Ejercito Ecuador ESPE pursuant to may License No fu men occurs. response curvesen are not available, the USE test be Agreement. started

14

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

at a stress level equal to about 70 per cent of the estimated fatigue limit, and the stress increments should be approximately 5 per cent of the estimated fatigue limit. In the past, this method has not been considered an acceptable technique because the fatigue strengths of some materials will be increased or "coaxed" by stressing them at stress levels below their fatigue limits, whereas the fatigue strengths of other materials may be decreased by damage due to "under-stressing." However, in recent years it has been observed that understressing does not greatly affect the true fatigue limit of some alloys, such as many of the alloy steels and a few of the nonferrous materials (4). For those materials with which neither appreciable coaxing nor damage occurs, it is possible to estimate the fatigue strength of each specimen or part by stressing it at consecutively higher levels until the specimen fails.

FIG. 4.Representation of "Step" Testing of Single Specimen.

This method is illustrated graphically in Fig. 4. In this manner, the fatigue strength corresponding to a preassigned value of N for each specimen or part may be estimated. The main disadvantage of the procedure is that thespecimens are run initially at a sufficiently low stress level so that failure will not occur. As a result, a number of stress levels of runouts are usually necessary before failure of the specimen occurs. Before the step technique of fatigue testing can be safely used, the effect of coaxing or under-stressing the material must be known. Certain steels, sensitive to strain-aging, will have their fatigue limits artificially raised by coaxing or under-stressing at low stress levels. In other cases it is thought that coaxing or under-stressing may damage the material artificially and cause premature failures. Although step tests have been made with a single specimen, four or more are needed to estimate the median fatigue A larger sample gives Copyrightby ASTM Int'l(allrights reserved); TueJul 910:11:02 EDTstrength. 2013 greater precision by in the estimates of the median and the variability of the Downloaded/printed This standard is for EDUCATIONAL USE ONLY. Escuela PolitcnicadelEjercitoenEcuadorESPEpursuant toLicenseAgreement.Nofurtherreproductionsauthorized. fatigue strength.

TEST PROCEDURES

15

2. The Prot Method (5-8): In 1945 Marcel Prot, in France, devised a rapid method for estimating the fatigue limit of a material. By using the Prot method, a good estimate of the fatigue limit may be obtained in a fraction of the time required by other methods but at the expense of more uncertainty than is present in most of the other test methods. The use of this technique is restricted not only to those materials which are not sensitive to coaxing effects, as discussed in Section III Cl, but also to materials that apparently have a fatigue limit. In contrast to the step method, it is suggested that at least 20 test specimens be used to obtain the data needed for the Prot analysis because of the wide scatter in fracture stress usually found in Prot fatigue data. To date, it has been found that, by the use of the Prot procedure, the fatigue limits of many alloy steels may be obtained within a few per cent

FIG. 5.Graphical Illustration of Prot Data. !, 2 , s, indicate different loading rates in psi per cycle.

of the estimate found from constant amplitude methods. It is not certain, however, that the long-life fatigue strengths of nonferrous alloys can always be evaluated by this method. In the Prot method, the test on a specimen is first started at an alternating stress of about 60 to 70 per cent of its estimated fatigue limit and the stress is raised at a constant rate. A number of specimens is tested at the same rate of loading until each specimen fails. At least three rates of loading are used to establish and check the linear relationship between stress and the power of the loading rate, which is required in the Prot analysis. The lowest rate should be as small as practicable and the highest rate should be low enough so that the specimen does not fail by yielding before fracture. The type of data observed is shown in Fig. 5. One of the simplest methods for obtaining a constant rate of loading in a fatigue test is to use a stream of water flowing at a constant rate into the loading container. Another is toreserved); arrange Tue for small weights, such as shot, Copyright by ASTM Int'l way (all rights Jul 9 10:11:02 EDT 2013 to be poured into a container at a constant rate. Fairly good results have Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. been obtained by adding small weights byESPE hand, one attoaLicense time, at small and Escuela Politcnica del Ejercito en Ecuador pursuant Agreement. No further rep

16

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

equal cycle increments. Any device that increases the stress at practically a continuous, constant rate can be used.
NOTE.For constant rates of loading, all the points obtained at a given rate should fall on the same straight line. Small variations in the rate of loading or variations in the testing speed may cause scatter such as that shown in Fig. 5.

IV. MINIMUM NUMBER OF TEST SPECIMENS AND THEIR SELECTION The practicability of fatigue tests is based upon the assumption that establishing the fatigue characteristics of a given material by studying the performance of a random sample selected from a larger body of possible specimens (the population or universe) is possible. Implicit in these tests is the assumption that the sample tested is "representative" of the population. By random selection and allocation of the test specimens, using a table of random numbers (9, pp. 366-370) the influence of all variability inherent in the material and testing procedures is given a fair chance of being reflected in the test data. There are innumerable stages in the testing program in which any one specimen or any one group of specimens may be affected differently from others from the same population. For example, if one bar of a batch of bar stock is tested, it is often tacitly assumed that the remaining bars are the same as the one tested. Usually they are not because, for example, such blanks are heat treated in batches. For each batch the furnace settings are slightly different. Within each heat-treatment batch, those specimens near the walls of the furnace are under slightly different conditions from those in the center. Specimens prepared at the start of the day are machined with sharper tools than those that succeed them. Specimens tested at the beginning of a program may have the advantage of being tested on newer, more perfect testing machines than those that are tested later when wear of the machines has modified their characteristics. These are but a few examples of* the many factors that may produce significant biases in the results unless controlled by appropriate randomization. The following are some of the factors for which randomization might be considered: Position of specimen within the whole batch of material Heat-treatment batch Position of specimen in heat-treating furnace Order of quenching Order of polishing Assignment to testing conditions (stress level and so forth) Order of testing Assignment to testing machine Machine operator This list will suggest other variables of importance in particular programs. Onp job of thp and is to decide the EDT randomization Copyright bypnginppr ASTM Int'l (allstatistician rights reserved); Tue Jul 9 how 10:11:02 2013 sb_n^1H he rqm'eH r^it by A common misconception is that randomization can_ Downloaded/printed This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica Ejercito en Ecuador ESPE pursuant to License Agreement. No further be accomplished by del rough-and-ready procedures such as reaching blindly

repr

NUMBER or SPECIMENS AND THEIR SELECTION

17

into the box of specimensjQr^the^iextjto^b^tested, but it is summing how chance procedures. The sample can be "biased" by unconscious and unrecognized trends of human behavior as well as by unknown patterns of arrangement. The best procedureJs_tQ_ae-iiD the program on the basis of random numbers as previously suggested (9"). To obtain fatigue data that can be used most efficiently, atrained^tatisti^ cian shoiild_b_OJiau]lejJ_whenever possible, in jjlanning the experiments and specimen selection. In most cases the statistician will be able to plan the experiments to measure not only the effects of the main variables under study but also the effects of the more important secondary variables as well, and do this without requiring many, if any, additional specimens. Tests conducted in accordance with such a plan can be analyzed to give an estimate of the importance of each of the known variables that contribute to the scatter in the test results. The techniques of experimental design are too involveoL however, to be included in this Guide. ' "Some indication of the minimum number of specimens needed for a given degree of confidence in the results obtained when using the different test procedures has been given in Section III. For supplementary references on this subject, see references (10-12). 9 The following sections discuss the minimum number of specimens needed for each type of analysis given in Section V when the sample size is fixed before testing. All samples are assumed to be randomly selected samples from the population under consideration. A. LITE DISTRIBUTION SHAPE NOT ASSUMED 1. S-N Curves: The minimum number of fatigue test specimens needed at each stress level depends on: (1) which per cent survival curve is desired and (2) what confidence level is desired. For a 50 per cent confidence level and one group tested at each stress level, Table 8 in Section VA1 shows the number of specimens needed for several values of per cent survival. For example, a 95 per cent survival curve requires at least 13 specimens at each stress level. Table 2510 provides similar information for one or more groups tested at each stress level and seven values of confidence level, including 50 per cent. For example, from Table 25, one group of five specimens at each stress level is needed for an 87 per cent survival curve corresponding to a 50 per cent confidence level. For a 90 per cent confidence level, at least five groups of 10 specimens at each stress level are needed for an approximately equivalent S-N curve. When several S-N curves are to be drawn from the same data, Table 25 should be studied carefully to find the best combination of number of groups and group size.
Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

9 See also Recommended Practice for Choice of Sample Size to Estimate the Average Downloaded/printed by Quality of a Lot or Process (E 122), 1961 Book of ASTM Standards, Part 3. This standard is for EDUCATIONAL USE ONLY. 10 Escuela del Ejercito I, en Ecuador pursuant to License Agreement. No f For TablesPolitcnica 25 to 33, see Appendix pp. 55-67. ESPE

18

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

2. Estimates of Parameters, Single Stress Level: The minimum number of specimens needed depends upon the desired width of the confidence interval for each parameter. In general, as the sample size increases, the confidence interval for any given confidence level becomes narrower and the difference between the observed value and the universe value becomes smaller. For the median at a confidence level of 0.95, confidence limits are equal to the observed minimum and maximum values up to a sample size of nine, when the width of the confidence interval becomes less than the observed range. (See Table 9 on page 26.) If ranges for prior samples from the same population are known, a sample size can be chosen so that the interval will have approximately the desired width. If the number of specimens are only 3, 4, or 5, Table 10 (see Section V A2) gives procedures for computing confidence intervals for the mean. For per cent survival values, Table 11 (Section V A2) gives values of 95 per cent confidence limits for four sample sizes. Comparing the widths of the confidence intervals gives some idea of the size of sample needed. If a good estimate of p = per cent survival/100 is available, the minimum sample size is approximately:

where E = one half the desired width of a 95 per cent confidence interval (see ASTM Recommended Practice E 122)9. It is more difficult to determine the minimum number of specimens needed for a confidence interval of a given width for fatigue life corresponding to a stated value of per cent survival other than 50 per cent. See footnote 15, page 29, for equations for setting up tables similar to Table 9 (in Section V A2) for other percentage points. In general, the sample sizes would be larger than for medians. 3. Tests of Significance: The minimum number of specimens needed depends upon the desired magnitude of the difference that should be detected and the size of the risks that can be tolerated. When the rank test is used to test the differences of group medians, it is difficult to relate the desired values and the criteria for the significance test given in Table 26 for two groups and in Table 27 for more than two groups. At least five specimens should be included in each group. For differences of two or more percentages (other than 50 per cent) no precise estimate of the minimum number of specimens needed is possible unless prior estimates of the percentages are available. At least 15 specimens should be included in each group. 4. Response Curves:
Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by A discussion of the minimum number of specimens and their allocation This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica Ejercito III en Ecuador ESPE pursuant to License Agreement. to stress levels is given del in Section Bl.

NUMBER or SPECIMENS AND THEIR SELECTION B. LITE DISTRIBUTION SHAPE ASSUMED

19

1. Normal Distribution: It is assumed here, as well as in Section V B that the fatigue data can be transformed so that they will be approximately Normally distributed. A Normal distribution is assumed in all cases. Each sample is assumed to be drawn at random from its population. 2. S-N Curves: Table 33 gives k factors for computing points on 75, 90, 95, 99, and 99.9 per cent survival curves for four values of confidence level, including 50 per cent, and for n = 3 to 25. The minimum number of specimens should increase as the per cent survival increases, but there is no definite criterion for choosing a particular group size except for the relative magnitudes of the k values. (Note that the rate of decrease is less as, increases.) The number of specimens tested at each stress level can be smaller than the group sizes needed when the life distribution is not assumed (Section V A).
TABLE 2.MINIMUM NUMBER OF SPECIMENS0 NEEDED FOR DETERMINING 95 PER CENT CONFIDENCE INTERVALS OF STATED WIDTH FOR A POPULATION MEAN, /*. Standard Deviation, a, Assumed Known.
Width of Interval
0 2<r 0.4 06
...

TABLE 3.MINIMUM NUMBER OF SPECIMENS" NEEDED FOR DETERMINING 95 PER CENT CONFIDENCE INTERVALS OF STATED WIDTH FOR A POPULATION STANDARD DEVIATION, a. Some Estimate of <r Available.
Width of Interval Number of Specimens,

Confidence Limits
X
0.1<7

Number of Specimens, n

08

2.0

10 1.2 1.4 1.6 1.8

0.2 =t 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

384 96 43 24 15 11 8 6 5 4

0.14<7 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

385 190 84 47 30 21 16 13 10 8

. _ n = Where:

E =

width of interval
2
0 Based on Fig. 1 of Greenwood and Sandomire (13). * The values of re given in Table 30.

x = sample mean. "ASTM Designation E 122 (see footnote 9).

3. Confidence Intervals: For the Mean.If a good estimate of the population standard deviation, 0, is available, Table 2reserved); gives the minimum Copyright by ASTM Int'l (all rights Tue Jul 9 10:11:02 EDT 2013 number of specimens needed for confidence intervals of stated width for the mean, p, of the population. Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Politcnica delused Ejercito ento Ecuador ESPE pursuant License Agreement. authorized. If the Escuela sample is estimate ato as well as No /*,further thereproductions sample sizes should be

20

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

larger, since ^0.975 values from Table 29 should be used instead of the 1.96 in the equation for n (Table 2). For the Standard Demotion.In order to find the minimum number of specimens needed for determining confidence intervals of stated width for the standard deviation, <r, of a population, some estimate of a must be available, since the width of the interval is measured in units of a. However, Table 3 can be used as a guide even if no good estimate of a is available. For example, if n = 8, the sample-standard deviation, used to estimate the population standard deviation, may be above or below a by 0.5 <r, whereas an estimate based on n = 30 will not be expected to deviate from the true value by more than 0.25<r.
TABLE 4.MINIMUM NUMBER OF SPECIMENS" NEEDED TO DETECT IF THE STANDARD DEVIATION OF A POPULATION IS A STATED PERCENTAGE OF A FIXED VALUE. Number of Specimens, n Percentage of Fixed Value Chance of Detection
7 8 9 12 15 20 28 42 80%

TABLE 5.MINIMUM NUMBER OF SPECIMENS0 NEEDED IN EACH SAMPLE TO DETECT IF A STANDARD DEVIATION OF ONE POPULATION IS A STATED MULTIPLE OF THE STANDARD DEVIATION OF ANOTHER POPULATION. Number of Specimens,
Multiple

Chance
Detection

90% of
8 9 12 14 19 26 38 55

40 45 50 55 60 65 70 75
a

Chance of Detection
1.5. 2.0 2.5
3.5 4.0
0

80%

Chance of Detection
52 20 13 9 8

90%

3.0

39 15 9 7 6 5

Based on Fig. 2 from Ferris et al (14).

Based on Fig. 3 from Ferris et al (14).

4. Tests of Significance: Difference Between Two Standard Deviations.The sample sizes for testing the difference between two means are given in Tables 6 and 7. In some cases, the principal interest is in the difference between standard deviations. 1. One Standard Deviation a Fixed Value.If one standard deviation is a fixed valuefor example, the long-time standard deviation of data based upon an old procedureand if the other standard deviation is to be computed from data based upon a new procedure that may reduce the variability, Table 4 gives the minimum number of specimens needed to detect a reduction of a stated amount. These sample sizes apply when the observed standard deviation, s, for the new procedure is indeed smaller than the fixed value, and the ratio s2/(fixed value)2 is compared with 1/F0.95, corresponding to o and n 1 degrees of freedom for numerator and denominator respectively. (See Section V B4(a) and Table 32.) 2. Two Sample Standard Deviations. If the problem is to test whether Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by the variability of procedure 1, say, is greater than the variability of proThis standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercitobeen en Ecuador ESPE pursuant to License Agreement. cedure 2 (the numbers having assigned prior to taking the data), Ta- No

further

NUMBER or SPECIMENS AND THEIR SELECTION

21

ble 5 gives the minimum number of specimens needed in each sample to detect that si is a stated multiple of s%. If the observed value of si is indeed larger than the observed value of sz, compare s?/s<? with ^0.95 corresponding to (HI 1) degrees of freedom for numerator and denominator (since n\ = HZ). (See Section V B4(a) and Table 32.) In this case it is not correct to make the test if s22 is greater than Si2. Difference Between Two Means: 1. One Mean a Fixed Valise.If one mean is a fixed valuefor example, the long-time mean of data based on an old procedure or a commonly used materialand the other mean is to be computed from data based upon a new
TABLE 6.MINIMUM NUMBER OF SPECIMENS0 NEEDED TO DETECT A STATED DIFFERENCE BETWEEN A MEAN AND A FIXED VALUE. a = Unknown Standard Deviation of the Population Being Estimated.
Number of Specimens, Difference Difference
80% Chance of Detection 90% Chance of Detection

TABLE 7.MINIMUM NUMBER OF SPECIMENS0 NEEDED TO DETECT A STATED DIFFERENCE BETWEEN THE MEANS OF TWO POPULATIONS. <7 = Unknown Standard Deviation of Each Population; <n = az
Number of Specimens, n 80% Chance of Detection
64 29

90% Chance of Detection

86 39 23 15 11 9 7 5

0 50<r 0 75 1 00 1 25 1 50 1 75 2 00 2 50
0

34 16 10 7 6 5 4 3

44 21 13 9 7 6 5 4

0 50<r 0 75 1 00 1 25 1 50 1 75 2 00 2 50

17 12 9 7 6 4

Taken from Table E of reference (11).

0 Taken from Table E.I of reference (11)

procedure that may shift the mean, Table 6 gives the minimum number of specimens needed to detect a shift in either direction, measured in terms of the population standard deviation of the new procedure. These sample sizes apply when the computed value of

is compared with 0.975 in Table 29. (See Section V B4(6).) No F-ratio test is needed. 2. Two Sample Means.The minimum number of specimens needed in each sample to detect a difference in two population means, stated as a multiple of their equal universe standard deviations, is given in Table 7. TheseCopyright sample by sizes apply when the two standard deviations are ASTM Int'l (all rights(1) reserved); Tue sample Jul 9 10:11:02 EDT 2013 not significantly different and (2) the computed value of / (see Section Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. V B4(6)) is Politcnica compared /0.9?5 in Table 29. Escuela del with Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductio

22

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

V. ANALYSIS OF FATIGUE DATA A bask concept of statistics is that a group of one or more specimens is a sample taken from a larger body or population. Such a sample is considered to be just one of a "number," often very large, of samples that could have been taken. The sampling procedure used delimits the-population being estimated. The results obtained from tests on a random sample from the population can be used to estimate the characteristics of the whole population and to measure the precision of the estimates. In the case of fatigue tests the data observed are usually the lives of specimens tested at a constant applied stress (strain or deflection) amplitude. Since the cycle life varies from specimen to specimen, this measurable char-

FIG. 6."Normal" or Gaussian Distribution Curve.

acteristic is not a fixed value and is best described by a frequency distribution. The graphical presentation of the distribution of cycle lives for the population of specimens that have lives between certain limits is known as a frequency distribution curve. Such a distribution curve may be estimated from the raw test data or from transformed test data, that is, either from values of N or from values of log N, log log N, N1'2, and so forth. When the frequency distribution curve has a particular kind of bell shape, as shown in Fig. 6, the data are said to have a "Normal" or Gaussian distribution. This Normal probability distribution curve, f(x), is represented by the equation:

The

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USEand ONLY. <r, the populaconstants in the formula are population mean,to /*, Escuela Politcnica del Ejercito enthe Ecuador ESPE pursuant License Agreement. No further re

ANALYSIS or FATIGUE DATA

23

tion standard deviation (a measure of the dispersion).11 It should be emphasized that values of the parameters of the population can only be estimated from tests on the specimens in the sample; to obtain exact values would require that the total population be tested. While some fatigue tests, particularly those made in the finite life range of an S-N curve, may yield approximately Normal distributions of cycle life, generally a transformation to log cycle life is required. Others do not yield Normal distributions, even after various transformations are performed on the data. This is particularly true in the case of tests made at applied stresses near the fatigue limit where runouts are observed. Hence, other distributions, such as the Weibull distribution,12 the "extreme value" distribution with and without lower limits, as used by Freudenthal and Gumbel (IS), and other distributions, that are just as normal in the usual sense, as the Normal or Gaussian distribution, have been applied to the analysis of fatigue data. While references to some ,of these distributions are included in this Guide, analysis of the fatigue data has been confined mostly to methods that require no assumptions of distribution shape or to the methods based upon the assumption that the raw data or the transformed data have a Normal distribution. As stated previously, however, any set of observations to which these statistical methods are applied is assumed to come from a random sample from the population of interest. If a series of samples is drawn, procedures for testing for statistical control are given in the ASTM Manual on Quality Control of Materials (see footnote 5). Lack of statistical control in data indicates that the series of samples does not come from the same population. A. LIFE DISTRIBUTION SHAPE NOT ASSUMED 1. S-N Curves: These techniques should be used when the actual shape of the distribution of fatigue life values for a given material is unknown or sketchy and the number of specimens tested at each applied stress level is too small, say less than 50, to estimate the shape of the distribution. In such cases, these techniques give conservative results. (a) One Group at Each Stress Level.Usually the first step in the analysis of fatigue data is to draw the S-N curve for 50 per cent survival; it is the curve fitted to the medians of the groups at the several applied stress levels. The median, an "order statistic," is the middlemost value when the observed values are arranged in order of magnitude, or the average of the two middlemost values if the group size is even. Other S-N curves, those for p per cent survival (where p is not 50), may be fitted to other order statistics if the group size is greater than 1. If the group values are arranged in order of magnitude, NI is the minimum cycle life value, or the first order statistic, A7 2 is the second observed value, or the second order statistic, and so forth. The estimated percentage of survivors for the population at cycle life
11

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 by InDownloaded/printed the Normal distribution, the median and the mean are equal. This standard is for EDUCATIONAL ONLY. Agreement. No further rep 12 Escuela Politcnica ESPE pursuant toUSE License See Appendix IV, p.del 71.Ejercito en Ecuador

24

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

values of Ni, or 7V 2 , depends upon the group size. Table 8 gives the median percentages at Ni and Nz for several group sizes.13 Some of these percentages also are given in Table 25 for one group, m = 1, and the 50 per cent confidence level. The median percentage of survivors at the maximum value of the sample, Nn , is 100 (per cent for N\), etc. Examples of how Table 8 may be used follow: 1. The 50 per cent survival curve may be estimated from the median of any sample size. 2. If three specimens are tested at each applied stress level, the 79, 50, and the 21 per cent survival curves may be estimated from the entries in Table 8 and their complements. The value 79 per cent is found opposite sample size 3 in the second column, the value 50 per cent is taken from the median
TABLE 8.MEDIAN PERCENTAGE OF SURVIVORS FOR THE POPULATION.
c ~i, Size, c; n Sample

M-

Vajue

^ ^ At the Lowest At the Next Lowest

ValuCj

2 3 4 5 6 7
8.".

:' 9. 10 " : 11.-. 12 13 14 15.' 16,

. 50 70 79 84 87 89 90 91 92 93 94 94 95 95 95 96

30 50 61 69 73 77 80 82 84 85 86 87 88 89 90

S-N curve, and the value 21 per cent is obtained by subtracting the value in the second column from 100 per cent. 3. If 7 specimens are tested at each applied stress level, the 90, 77, 23, and 10 per cent survival curves may be estimated from the entries in Table 8 and their complements. The 50 per cent survival curve may be estimated from the median. At least 13 specimens must be tested at each applied stress level to estimate the 95 per cent survival curve. In practice, values of per cent survival less than? 50 usually are not wanted. Hence, if all of the specimens in a sample are tested simultaneously, the tests may be stopped as soon as the specimen having the median value of fatigue life for the sample has failed, unless the data are required for other purposes.

13 These are called "median percentages" because, half of the time, the true*percentage will be Copyright larger, andby for the other the time, smaller. They close to, but usually ASTM Int'lhalf (allof rights reserved); Tue Jul are 9 10:11:02 EDT 2013 not equal to, the "expected" by percentage of survivors, which is equal to 1 i/(n + 1), where Downloaded/printed standard is for EDUCATIONAL ONLY. i is the number of the order statistic and nThis is the sample size. TheUSE confidence level associEscuela Politcnica del Ejercito Ecuador ESPE size, pursuant to License Agreement. ated with expected percentages varies en with the sample whereas it is constant for No furt median percentages.

ANALYSIS or FATIGUE DATA

25

As mentioned previously, the percentage survival values given in Table 8 are median values; they are based on a "confidence level" of 50 per cent.14 Percentage survival values corresponding to higher confidence levels, such as 95 or 99 per cent, are given in Table 25 for a single sample when m = 1. For example, if three specimens are tested at an applied stress level, 79 per cent of the population are expected to survive N\ cycles (50 per cent confidence level), but the statement that at least 37 per cent of the population will survive N\ cycles may be made with greater confidence (confidence level = 95 per cent). If estimates of the population percentage are made from a series of samples tested at one applied stress level and the statement is made that at least 79 per cent of the population will survive N\ cycles, 50 per cent of such statements are expected to be incorrect. If the statement is made each time that at least 37 per cent will survive N\ cycles, only 5 per cent of such statements are expected to be incorrect. However, S-N curves corresponding to a 50 per cent confidence level are usually shown. The effect of fitting a curve to the same order statistics at several stress levels probably increases the confidence level; how much is not known. If S-N curves are based on other confidence levels, the fact should be plainly indicated on the chart. (Z>) Several Samples, or Groups, at Each Stress Level.If it is not possible to test all the specimens in a sample simultaneously and if stopping the tests before all the specimens have failed is desirable to save time, the required sample may be divided, at random, into two or more groups (see references 17 and 18). Then the median of the particular order statistics (the first, second, and so forth) for the several groups may be used for constructing the S-N curve. Table 25 gives values of percentage survival for several numbers of groups and several confidence levels. EXAMPLE.With five testing machines available, 15 specimens were tested at a constant applied stress level in three groups of 5 each. For each group, all machines were assumed to be stopped after the second failure. (Actually, all machines were allowed to run until fracture occurred or until 10 million cycles of fatigue stressing had been applied, so that the time saved could be estimated for this particular set of tests.) The test data are:
Group Life, kilocycles

2 3

162, 229, (261, 668, 2 281) 105, 131, (140, 245, 10 000+) 275, 373, (5 503, 8 695, 10 000+)
Lowest Ranking Points Median

From these data we have:

2Vi.
#2

162, 105, 275 229, 131, 373

162 229

Entering Table 25, under "Lowest Ranking Points," in the column for m = 3 groups, opposite n = 5 in each group, and at a confidence level of 50 per cent,

14 Technically speaking, the S-N curves on order are "nonparametric Copyright by ASTM Int'l (all rightsbased reserved); Tue statistics Jul 9 10:11:02 EDT 2013 tolerance limits," which are described by Murphy (16). The probability that at least p Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. per cent of the population lies above Ni cycles, where Ni is the ith order statistic of the Politcnica del Ejercito en Ecuador pursuant to License Agreement. sample, Escuela is properly called a "tolerance level"; but ESPE the term confidence level appears to No furth have been used more frequently.

26

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

read 87.05 per cent. This value is an estimate of the percentage of the population from which the original 15 specimens were selected that will survive 162 kilocycles. Similarly, at a confidence level of 95 per cent, 67.03 per cent or more of the population are estimated to survive the 162 kilocycles. Again, for the "Second Ranking Points," at a confidence level of 50 per cent, 68.61 per cent of the population are estimated to survive 229 kilocycles and, at a 95 per cent confidence level, 45.40 per cent or more of the population are estimated to survive 229 kilocycles. Additional information can be obtained from the preceding test data by considering all 15 specimens as one "group" and determining the percentage of the population expected to survive 105 kilocycles, which is the lowest ranking point for m = 1 and n = 15 in Table 25. For a 95 per cent confidence level, straight-line interpolation between 74.11 per cent for n = 10 and 86.09 per cent for n 20 TABLE 9.CONFIDENCE INTERVALS FOR THE MEDIAN." Confidence Level = 0.95.
Confidence Limits Sample Size, n Lower Upper

6 7
8 9 10

# #saa #
#2

#8 N8

#6 #7 #9

30

12 13 14 15 20 25
35

11

#2 N2
#3 #3

#2

N4
# N6 # #1 #11 #1 . #1

N10 #W #11 #12 N12 N15 #W

#24 #27

40. 45 N13 50 " Based on a table in Nair (19).

#30 #3

gives about 80 per cent. From this, it is estimated that at a 95 per cent confidence level about 80 per cent of the population will survive 105 kilocycles. 2. Estimates of ParametersSingle Stress Level: (a) Median Fatigue Life: 1. Point Estimate.A point estimate of the population median is the sample median, described above in Section V Al(a). 2. Confidence Interval Estimate.A confidence interval for the median that does not assume a particular frequency distribution for the population may be computed if the sample size is larger than five. The n observed values of fatigue life, N, are arranged in order of magnitude as follows: Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

Downloaded/printed by This standard is for USE ONLY. NiN*gNf EDUCATIONAL N. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductio

ANALYSIS OF FATIGUE DATA

27

The confidence limits corresponding to a confidence level of at least 0.95 are given by the order statistics designated in Table 9, p. 26. EXAMPLE.Assume that ten specimens are tested at a particular stress level and the observed values of fatigue life in kilocycles are 201, 224, 226, 230, 232, 238, 24 244, 245, and 248. The point estimate of median fatigue life is the average of the two middlemost values, namely 235 kilocycles. The interval estimate is defined by A/2 and Ng (see Table 9), which are 224 and 245 kilocycles, respectively. The population median may be above or below the sample median235 kilocyclesbut the chances are at least 95 in 100 that the statement, "the median lies between 224 and 245 kilocycles," is correct if the sample came from one population. (b} Mean Fatigue Life: 1. Point Estimate.A point estimate of the population mean is the sample average. TABLE 10APPROXIMATE CONFIDENCE INTERVALS FOR THE MEAN.
Confidence Level = 0.95. Sample Size, n Procedure". 6 Length of Interval

3 ...... add the range of the observed values to the largest value and subtract it from the smallest value: that is, Ni - (N3 - JVi) and N3 + (N3 - Ni). 4 ...... add (range) /4 to the largest value and subtract it from the smallest value:

3 X range lj X range

5 ...... use the range: N\ and Ns

0 6

I X range

See Youden (20) for n = 3. Private correspondence from W. J. Youden, for values of n greater than 3.

2. Approximate Confidence Interval Estimate.An approximate confidence interval estimate for the mean that does not assume a particular frequency distribution for the population may be computed as shown in Table 10, if the sample size is 3, 4, or 5. (c) Per Cent Survival for a Stated Value of Fatigue Life: 1. Point Estimate.A point estimate of the percentage of the population that has fatigue life values equal to or above a stated value is the sample percentage of observed values equal to or above the same stated value. 2. Confidence Interval Estimate. Confidence limits corresponding to possible values of sample percentage, p, for four sample sizes are given in Table 11. Values for other sample sizes may be read from a chart from Dixon and Massey (9), p. 415, from which many values in Table 11 were taken.
Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

Downloaded/printed EXAMPLE. Using the by data given in This thestandard above example of this Section and 230 is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. kilocycles as the stated value of fatigue life, the following estimates of the popu- No furth

28

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

lation value of per cent survival are obtained: (1) point estimate: 70 per cent and (2) interval estimate: 34 to 94 per cent.

A.larger sample size will give a shorter interval estimate (see Table 11). (d) Fatigue Life for a Staled Value of Per Cent Survival: 1. Point Estimate.A point estimate of the population value of fatigue life for a stated value of per cent survival is based on order statistics as
TABLE 11.CONFIDENCE INTERVALS FOR PERCENTAGES." Confidence Level = 0.95.
Sample Size

=s
#,per cent
100

n = 10 P, per cent

= 20

= 40

Limits Lower Upper

48 100

Limits Lower Upper

P, per
cent
100 95

Limits Lower Upper

P, per cent 100 97.5 95 92.5 90 87.5 85 82.5 80 77.5 75 72.5 70 67.5 65 62.5 60 57.5 55 52.5 50

Limits Lower Upper

100

68

100

82 75 68 . 62 56 51 45 40 36 32 27

100 100 98 97 94 92 88 85 81 77 73

90

54

100

90 85

80

29

99

80

43

98

80 75

70

34

94

70 65

60

15

95

60

25

89

60 55

50

18

82

50

91 87 83 79 77 73 70 67 64 61 58 56 53 51 48 46 43 41 38 36 34

100 100 99 98 97 96 94 93 91 89 87 85 83 82 79 77 *75 73 71 68 66

Where: p = sample percentage (for example, percentage surviving). Confidence limits corresponding to (100 p) per cent are: lower: 100 (tabular value for upper limit corresponding to p, per cent); upper: 100 (tabular value for lower limit cor-' responding to p, per cent). 0 Based on chart from Dixon and Massey (9), p. 415, and, for n = 40, on chart from Pearson and Hartley (2), p. 204.

outlined in the Section on S-N curves: "One Group at Each Stress Level" (Section V A). A particular value is the median, corresponding to 50 per cent survival. Another point estimate may be derived from the. cumulative frequency distribution of the observed values. In general, the two point estimates would not be exactly equal. 2. Confidence Interval Estimate. Interval for 2013 medians '(50 per Copyright by ASTM Int'l (all rights reserved); Tue Julestimates 9 10:11:02 EDT cent survival) are described in Section V A2(a). Interval estimates for fatigue Downloaded/printed by
This standard is for EDUCATIONAL USE ONLY.

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions a

ANALYSIS or FATIGUE DATA

29

life values corresponding to other percentage points may be computed by using reference (21) ,15 3. Tests of Significance: (a) Differences of Group MediansSingle Stress Level.-If two or more groups of specimens are tested, the question of whether the observed differences in the values are due to chance or to some differences in the populations from which the groups were drawn often arises. The observed differences, for example, could arise because of differences in material lots or differences in the characteristics of the testing machines. The rank tests given in this section assume that the several groups are independently and randomly drawn from populations that are of the same shape but may differ with respect to their medians. All the observed values in one group are assumed to come from one population. Since the populations are assumed to be of the same (though unknown) shape, only those groups that are tested at the same stress level should be compared, since the form of the distribution tends to change with change in stress level. 1. Rank Test for Two Groups.In the rank test for two groups the rank of each observation in the two groups combined is determined. The lowest value is given the rank of 1, the next higher observed value is given the rank of 2, and so forth. If one value appears several times, that is, there is a tie, the average of the ranks for those numbers is assigned to each one. For example, if the llth, 12th, 13th, and 14th values are all equal, they are each given the rank of (11 + 12 + 13 + 14)/4 = 12.5. The ranks for the two groups are totaled separately and the total for one of the groups (the one with the smaller number of observations if the group sizes are unequal) is compared with the critical values given in Table 26 for sample sizes equal to the group sizes. If the observed value falls within the range of values given in Table 26 for the chosen significance level (5 or 1 per cent), the groups may be considered to have come from one population. If the observed value falls outside the range of values given in the table, the two groups are said to be significantly different, that is, to have come from two populations with different medians. The use of the 1 per cent significance level gives a smaller risk of calling the
15 The interval Nk to Nm may be computed as follows: (1) k is chosen so that

(2) m is chosen so that

where by ASTM = 1 Copyright (confidence level), Int'l and (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 p = (stated value of per cent survival)/100. Downloaded/printed by

This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further r

30

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

groups significantly different when they are actually drawn from one population and the observed difference is due to chance. EXAMPLE.To compare two machines, the rank test was applied to the data from 27 specimens randomly assigned to two testing machines. (See Table 12.) According to Table 26, the rank total for Machine A in Table 12, which has "the smaller number of measurements," should be between 101 and 179 (Ni = 10, NZ = 17) for the 5 per cent level of significance, and between 89 and 191 for the 1 per cent level of significance. This means that the actual total, 87, would not be expected to occur as often as once in a hundred samples due to chance alone, if the two machines were completely interchangeable. Thus, on the average, the machines give significantly different fatigue life values.
TABLE 12.FATIGUE TEST DATA.
Machine A Rank Kilocycles
624 662 681

Machine B Rank Kilocycles

1 4 5 6 8
2

9 10 24
18

732

688 99

774 781
865

7 11 12 13 14 15.5 15.5 19 20 21 22 23 25
26 17

715 811 822 833 841 842 842 869 892 903 944 946 1 , 032 1,067 1,092
849

667

948

87 Total

27

291 Total 2. Rank Test for More than Two Groups.The method of assigning ranks is the same as for the two-group test, ranking the observations for all the groups combined. The ranks are totaled separately for each group and the following test-statistic, H, is computed from the rank totals (22):

where: k = number of groups, Hi = number of observations in the ith group, N = y^,ni, the number of observations in all groups combined, and Ri = sum of the ranks in the ith group.

Copyright by ASTM (all rights approximately reserved); Tue Jul 92 10:11:02 EDT The test-statistic H isInt'l distributed as x with k 1 2013 de2 Downloaded/printed by grees of freedom if each Ui is at least five. a discussion of x , see referThis standard(For is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No fur ence (9)). Thus, the value of H calculated from the observed data may be

ANALYSIS or FATIGUE DATA

31

compared with the values of x2 given in Table 27 to determine whether there may be a significant difference among the populations from which the groups were drawn or not. If H is greater than the x2 value for k 1 degrees of freedom and the chosen significance level, the populations are said to be different; that is, the groups may be said to have been drawn from two or more populations. Inspection of the rank totals will usually show which groups are different from the others if the difference is significant. EXAMPLE.To compare five machines, the rank test was applied to the data from 25 specimens, randomly assigned to the five machines (see Table 13).
TABLE 13.FATIGUE TEST DATA.
Machines

c
(3)539 (12)651 (14)662 (15)675 (19)744

Total

(5)596 (10)640 (11)646 (18)733 (24)807 Sum of ranks, Ri. ..

Bf

(6)599 (13)661 (21)760 (22)774 (23)781

(2)530 (8)624 (9)638 (16)684 (25)889

(1)477 (4)568 (7)607 (17)719 (20)757

68
4624 924.8

85
7225 1445.0

63
3969 793.8

60
3600 720.0

49

R*

2401 480.2

4363.8

H = 80.56 - 78 = 2.56 Entering Table 27 with 4 degrees of freedom, one less than the number of groups, gives x2 = .49, corresponding to a 5 per cent significance level or a percentile of 95. Since the computed value of H, 2.56, is very much smaller than 9.49, the observed values of fatigue life may be considered to be from one population; the machines may be considered to be interchangeable. (ft) Differences of Two or More Percentages (for example, per cent survival values).The test-statistic used to test the significance of the differences among percentage values computed from observed data is x2- The formula for x2 may be written in two ways; the second one is usually better for computation purposes. 1. When the sample sizes are unequal:

where: k =Copyright numberby of ASTM samples; Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

Downloaded/printed by standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del en This Ecuador ESPE pursuant to License Agreement. No fu sum over k Ejercito samples;

32

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

, = size of ith sample (i = 1, 2, k); Xi = observed number of "events" in the ith sample; an event may be a failure, a survival, etc.; pi = observed fraction for the ith sample: pi = #*/;; and P = Z_,Xi/_,ni = average fraction for all samples combined. 2. When the sample sizes are equal the formula reduce to

where n = sample size and x = ^ Xi/k (23, pp. 175-178). The other terms were defined previously. The computed value of x2 may be compared with the tabular values given in Table 27 for k 1 degrees of freedom (d.f.). If the computed value of X2 is larger than the tabular value corresponding to: percentile = 100 (chosen significance level), the percentages are said to be significantly different; that is, the samples were drawn from different populations. If the computed value of x2 is smaller than the tabular value, the samples may be considered to have come from one population. Another use of the x2 test is to test whether or not the observed percentage values are significantly different from an arbitrary value, such as 50 per cent. The method of computation is the same as that given previously, except that: (1) the first way of writing the formula for x2 is used for the computations, (2) the arbitrary value, which may be called p', replaces p, and (3) d.f. = k. EXAMPLE.To compare six lots of phosphor-bronze strip, the x2 test was applied to the data given in Table 14, using a significance level16 of 10 per cent. TABLE 14.PERCENTAGES SURVIVING 108 CYCLES. Stress = 25,000 psi.
Lot
Sample Size, ntPer Cent Surviving, V)0pi Number Surviving, Xi
Xi*

1 2 3 4 5 6
Total

15 20 17 25 19 14

60.0 40.0 58.8 48.0 57.9 50.0

9 8 10 12 11 7
57 =

5.40 3.20 5.88 5.76 6.37 3.50

30.11 =

110 =

Copyright by ASTM Int'l (all rights reserved); Tue Jul Downloaded/printed by 16 for EDUCATIONAL USE ONLY. Significance levels commonly areThis 10standard and is 5 per cent. ESPE pursuant Escuela Politcnica del used Ejercito en Ecuador

9 to

10:11:02 License

ANALYSIS OF FATIGUE DATA

33

The tabular value of x2 corresponding to degrees of freedom = 5 and percentile = 90 per cent is 9.24. Since the computed value, 2.28, is much smaller than 9.24, no significant difference among samples is indicated by these data and the samples may be considered to have come from one population. If the computed value of x2 were larger than 9.24, the lots from which the samples were drawn would be considered to be significantly different.

4. Response Curves: Instead of drawing S-N curves when specimens are tested at several levels of applied stress, response curves may be constructed from the data if the applied stress levels have been properly chosen. A response curve is constructed for a stated value of the number of cycles-to-failure, such as 107, 108, and so forth. It is based on the values of per cent survival at N cycles that are observed for several (at least 3) values of applied stress. This method of analyzing fatigue data is especially useful when some of the specimens "run out," that is, survive the duration of the test. While a response curve may be drawn on any type of graph paper, experience has shown that the per cent survival values tend to lie along a straight line when the data are plotted on Normal probability paper.17 (a) Fitting a Response Curve. If the observed per cent survival values lie along a straight line when plotted on Normal probability paper, a straight line may be fitted to the points by eye or by the method of least squares. The latter method is, of course, more precise and not subject to the biases that may be introduced by a person fitting a line by eye. The equations for the slope and intercept of a line fitted by the method of least squares to per cent survival values having equal weight are:18

where : X = applied stress value (usually coded to reduce the size of the number) , p = per cent survival in each group of specimens tested, Y = z, a value of the Normal deviate obtained from Table 28. The value of 0 corresponding to p is obtained by entering the column headed "Area" with 1 (/>/100) and reading the corresponding value of z. Interpolation between tabular values may be required. Y is called the transformed value of p, X ^L,X/k = average of .X values, Y = ^^ Y/k = average of Y values, k = total number of groups tested, or total number of observed p values, and 2^ = sum from 1 to k.
Intercept, a = Y
17

No. 3127, Codex Book Co., Inc., Norwood, Mass; or No. 358-23, Keuffel & Esser Co.,18 New York, N. Copyright by Y. ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 If the specimens have been allocated as indicated in Table 1, the per cent survival by equal weight. If this has not been done, references (1), values Downloaded/printed will have approximately This standard is for EDUCATIONAL USE ONLY. Politcnica del Ejercito en Ecuador ESPE the pursuant (2), or Escuela (3) should be consulted for a method of analyzing data. to License Agreement. No furthe

34

FATIGUE TESTING AND STATISTICAL ANALYSIS OP DATA

fY = a + b (X X}JThe subscript,/, The equation for the fitted line is is used to denote the fitted line values.
TABLE 15. "PROBIT" TEST DATA.
AppliedStress, Ksi

Applied Stress, ksi Number^ Specimens Numberof Specimens

15

Per Cent

Survival, ,

40.0. 41.5. 43.0. 44.5. 46.0.

8 5 8 15

14 6 3 2 1

93.33 75.00 60.00 25.00 6.67

TABLE 16.COMPUTATIONS FOR FITTING A RESPONSE CURVE BY METHOD OF LEAST SQUARES.

Applied Stress, ksi Coded0 Value, X
1 2

Per Cent Survival, p

Transformed Values, Y

1 4

XY

Fitted Values
F/

Pf
93.4 78.2 52.0 24.8 7.9

40.0 41.5 43.0 44.5

46.0

3 4 5 15

93.33 75.00 60.00 25.00 6.67

-1.50 -0.67 -0.27 + 0.67 + 1.50 -0.25

F - -0.05

9 -0.75 16 > +2.68 25 +7.50

-1.50 -1.34

-1.51 -0.78 +0.05 +0.68 + 1.41

Sum

55

+ 6.59

X = 3

" The stress values are coded by assigning numbers 1 to 5 to the successive applied stress values.

EXAMPLE.Fatigue tests of specimens of a certain steel gave the data in Table 15 for five applied stress levels. The number of specimens tested at each stress level was chosen in accordance with the procedure outlined in Section III Al, so that the per cent survival values would have approximately equal weight. In order to fit a straight line to these data, the p values must be transformed to Y values, as outlined above and demonstrated in Table 16; also it is convenient to code the applied stress values when the increments between successive values of applied stress are equal.
Slope, Intercept, a = 0.05

Yf = -0.05 + 0.730T - X) pf = 100 (1 - "Area" value) where an Area value corresponds to a value of z = F/ in Table 28. Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Figure 7 shows the observed values of per cent survival at 107 cycles

Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions autho

ANALYSIS OF FATIGUE DATA

35

(a) Confidence band for per cent survival values. (b) Confidence band for fatigue strength values. FIG.Int'l 7.Response Curve Tue for a Particular of Steel. Copyright by ASTM (all rights reserved); Jul 9 10:11:02Type EDT 2013
This standard is for EDUCATIONAL USE ONLY.

Downloaded/printed by Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authori

36

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

(open circles) plotted on Normal probability paper. To draw in the line fitted to the observed values by the method of least squares, select at least two values of applied stress and the corresponding values of pf from Table 16; connect these plotted values by a straight line. (The confidence limits are discussed in the next section.) In fatigue,tudies the derived estimate, ^(F), corresponding to a stated value of p (or F) is often required. The equation for the derived estimate, .called X, ajfunction of F, is:

where F = the transformed value of the stated per cent survival value. This derived estimate is called the fatigue strength for p per cent survival at N cycles.
TABLE 17.COMPUTATION OF STANDARD DEVIATION, s.
Applied Stress, ksi 4 0 . 0 41.5 Y Yf 1 51 (F - F/)*

1 50
0 67 0 . 2 5 + 0.67

43.0 44.5 46 . 0 Sum

+ 1 . 50

0 78 0.05 + 0 68 + 1 41

0 0001 0 0121 0 . 0 4 0 0 0.0001 0.0081

0.0604

(V) Confidence Limits for a Response Curve.19The method of computing symmetrical20 confidence limits fora response curve plotted as a straight line on Normal probability paper when the test specimens have been allocated according to Table 1 is presented here. While no assumption about the distribution of fatigue life values is required, the computation of confidence limits by the method outlined here assumes that the Normal transformation of the per cent survival values is suitable, that is, that the transformed percentages are Normally distributed about the fitted line. Two types of confidence limits may be computed for a response curve that has been fitted by the method of least squares: 1. Confidence limits for per cent survival at N cycles for a stated value of applied stress (Fig. 7(a)). (This method is an alternate for the one described in Section V A2(c) for a single stress level.) 2. Confidence limits for fatigue strength at N cycles for a stated value of per cent survival (Fig. 7(6)). The two types of confidence limits are computed differently, but the resulting confidence bands are numerically equivalent (Fig. 7). For either type of confidence limits, the standard deviation of the ob19 20

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Much of the material in this section is based on Chapter 18 of Hald (24). Downloaded/printed by The risk, equal to 1 (confidence level), is evenly divided between the two limits. This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further re

ANALYSIS or FATIGUE DATA

37

served Y values about the fitted line is needed. Provided that all Y values have equal weight (see Section III Al), the following equation may be used:

The symbols have been defined previously; the computation procedure is given in the following example. EXAMPLE.Table 17 shows the method of computing the standard deviation of F values about the fitted line, using the values of F and F/ given in Table 16.

1. Confidence Limits for Per Cent Survival Values at N Cycles.In computing such confidence limits for a stated value of applied stress, the transTABLE 18.METHOD OF COMPUTING 95 PER CENT CONFIDENCE LIMITS FOR PER CENT SURVIVAL VALUES.

Confidence Limits,
Applied _ Stress, ksi

Coded Value, X

i
4 1 0 1 4 10
0.40 0.10 0.0 0.10 0.40 0.77 0.55 0.45 0.55 0.77 0.47 0.34 0.27 0.34 0.47

Y Values Lower Upper

p Values Lower Upper

40.0 41.5 43.0 44 5 46.0

Sum . . .

1 2 3 4 5

-1.51 -1.98 -1.04 -0.78 -1.12 -0.44 -0.05 -0.32 + 0.22 + 0.68 + 0.34 + 1.02 + 1.41 + 0.94 + 1.88

97.6 86.9 62.6 36.7 17.4

85.1 67.0 41.3 15.4 3.0

formed values of per cent survival are used. The formula for symmetrical confidence limits is:

where: X' X ^(X X)2 FZ ,k -2

= the stated value of applied stress (coded), = average of X values for the data used in fitting the line, is computed for the X values for the data, and = a factor obtained from Table 32 for 2 and k 2 degrees of freedom for numerator and for denominator, respectively.21

The other symbols have been defined previously. The confidence limits should be computed for all the applied stress levels
21 Hald (24) uses t = (Fi, k - 2)1/2 instead of (2F2, k - 2)1/2. The wider limits given here ensure Copyright the desired level confidence made using byconfidence ASTM Int'l (all when rights several reserved); Tue Jul statements 9 10:11:02are EDT 2013 the same fitted line. Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further

38

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

used in the experiment and may be computed for any other value within the range of the applied stress values. The method of computation is given in the following example. EXAMPLE.Table 18 shows the method of computing 95 per cent confidence limits for per cent survival, at 107 cycles, using the value of 5 computed in Table 17 and the Yf values from Table 16, where: n = 5, the number of observed p values, ^2,3 = 9.55, the value of F0.95 in Table 32 for 2 d.f. for numerator and 3 d.f. for denominator, (2F2,3)1/2 = (19.10)1/2 = 4.37. s(2F2,s)1/2 = 0.14 (4.37) = 0.61, and p = 100 (1 Area value); an Area value corresponds to a z = Y value in Table 28.

The confidence limits computed in Table 18 are plotted in Fig. 7(a) and connected by dotted lines to form a confidence band. The band applies to
TABLE 19.METHOD OF COMPUTING 95 PER CENT CONFIDENCE LIMITS FOR FATIGUE STRENGTH VALUES.
Es(2F* i)"t Confidence Limit! CJ

Coded X(Y) Values

Coded Values Lower Upper

1.81 2.55 3.46 4.54

Fatigue Strength, ksi Lower Upper

90 75 50 25 10

-1.28

-0.67
0

+ 0.67 3.99 + 1.28 4.82

1.32 2.15 3.07

2.82 0.72 0.00 0.98 3.31

0.282 0.072 0.000 0.098

0.68 0.51 0.43 0.53 0.331 0.72

0.62 0.46 0.39 0.48

0. 66

3. 07 4. 06 4. 96

1. 19 2. 09

3.58 4.30

0.57 1.63 2.68

5.62

43.87 45.31 44.95 46.93

39.36 41.22 40.95 42.32 42.52 43.69

" 38.50 + (1.5 X coded value); 38.50 corresponds to zero on the coded scale, see Table 16. For example: lower confidence limit for fatigue strength corresponding to p = 90 is 38.50 + (1.5 X 0.57) = 38.50 + 0.86 = 39.36 ksi.

any value of applied stress within the range used in the experiment.22 For example, for an applied stress of 41 ksi, the point estimate of the population of per cent survival at 107 cycles is 85 per cent. The confidence interval estimate, corresponding to a confidence level of 0.95, is 74 to 92 per cent. 2. Confidence Limits for Fatigue Strength at N Cycles.These are confidence limits on the derived estimate, X(Y}. The exact formula for the confidence limits is:

where: F = the transformed value of the stated per cent survival value and
Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

The other symbols have Downloaded/printed by been defined previously.

22

This standard is for EDUCATIONAL USE ONLY. Escuelashould Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further r Caution be used in drawing conclusions from extrapolated portions of the response curve or the confidence band.

ANALYSIS OF FATIGUE DATA

39

In many cases C deviates only slightly from b and the approximate confidence limits will be close to the exact ones:

EXAMPLE.Table 19 illustrates the computation of 95 per cent confidence limits for fatigue strength values at 107 cycles, using the exact formula. In table 19: p = stated value of per cent survival; Y = transformed value of p (See Section V A3 (a);

The confidence limits computed in Table 19 are plotted in Fig. 7(6) and connected by dotted lines to form a confidence band. The band applies to any value of per cent survival within the range of observed values, that is, the extent of the fitted line. It will be noted that this confidence band differs from the one in Fig. 7(a) only at the ends and that, numerically, the bands are equivalent. Thus it is recommended that confidence bands be computed according to the simpler method illustrated in Table 18 and the confidence limits for fatigue strength be read from the intersections of the confidence band and the stated value of per cent survival. For example, for 50 per cent survival, the point estimate of the population value of fatigue strength is 43.1 ksi. The confidence interval estimate, corresponding to a confidence level of 0.95, is 42.5 to 43.7 ksi. Likewise, for 80 per cent survival, the point estimate is 41.4 ksi and the confidence interval estimate is 40.5 to 42.0 ksi. The method illustrated in Table 19 should be used when a confidence band is not presented or when the estimates are needed to more decimal places than can be read from a chart. B. LITE DISTRIBUTION SHAPE ASSUMED 1. Normal Distribution of Fatigue Life: The preceding analyses, which made no specific assumptions about the This standard is for EDUCATIONAL USE ONLY. shape of the life distributions, are Jul always applicable but, because of Copyright by fatigue ASTM Int'l (all rights reserved); Tue 9 10:11:02 EDT 2013 their generality, may Downloaded/printed by not give results that are precise enough for all pur-

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduction

+40

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

poses. In order to sharpen the inferences which may be made from the data, particular functional forms can be assumed for these distributions and the statistical analyses carried out relative to these forms. It should be pointed out, however, that the precision so gained is real only when the particular distribution chosen honestly reflects the random process being studied. (a) Choice of Distribution Shape.In the literature are several specific fatigue life distributions which have some basis in theory. However, empirical evidence gathered to date has not allowed a clear-cut choice to be made among them. In this position, it is sensible to choose a form for the fatigue life distribution according to the following two criteria: 1. There should be satisfactory agreement between the data gathered and the results predicted by the "theoretical" distribution. 2. The distribution form should be mathematically tractable or, at worst, one on which extensive studies have already been made so that laborious computations can be kept at a minimum. These two conditions can be fulfilled by finding a transformation which, when applied to a set of fatigue life data, results in new data which closely follow the Normal or Gaussian distribution. Once this has been done, all the results of Normal distribution theory are available to analyze the transformed data. The particular transformation which will do the job mentioned above is not necessarily the same for all cases. It has been found that replacing the observed fatigue life, N, by log N will often give a distribution which is approximately Normal. Some situations may arise, however, where this transformation is inappropriate. The joint efforts of the test engineer and experimental statistician will be needed to make a specific choice which is suitable for the problem being studied. If usable inferences are to be made, the data transformation should not be so variable as to be a function of tjje individual sets of data, but the problem of what transformation to use is not a cut-and-dried statistical one. In Section B, all random variables, unless otherwise noted, are assumed to have been transformed so that they are Normally distributed. (b) Estimates of Parameters.On the assumption that the variable being studied is Normally distributed, the information in the sample can be used to obtain estimates of the parameters for the particular population. In many situations it is sufficient to obtain point estimates of n and a. More meaningful estimates, however, may be gotten by the use of confidence intervals, since these utilize not only the point estimates, but also depend on the amount of variation expected in these estimates from experiment to experiment. A 95 per cent confidence interval on /*, for example, can be described in the following way: Once the data from a particular experiment have been collected, an interval which presumed include the Tue trueJul parameter p, can be calcuCopyright by is ASTM Int'l (allto rights reserved); 9 10:11:02value, EDT 2013 lated from the sample values in a prespecified way. Since this interval is a Downloaded/printed by function of thedel observations, it is itself a random quantity, both as posiEscuela Politcnica Ejercito en Ecuador ESPE pursuant to License Agreement. Noto further reproduction tion and length, so that it will not be certain to include /x in any single experiment. The process for constructing the interval from the sample can This standard is for EDUCATIONAL USE ONLY. be determined, however, in such a way that the chance of \L being covered

ANALYSIS OF FATIGUE DATA

41

by the interval in a given experiment is exactly 0.95. It is then said,, "With a confidence level of 0.95, the true value, >, is included in the interval so computed." The mean, /*, of the Normal population, which is equal to the median of the population, may be estimated either by the sample mean or by the sample median. The sample median is the less efficient of the two, in the sense that it experiences a larger variation from experiment to experiment but has the advantage of being simpler to compute. The standard deviation, a, is estimated by the sample standard deviation. The sample quantitiesthe mean, x, and the standard deviation, sare summarized below, where Xi is the tth sample value from a sample of size n:

where ^ is the sum from I ton. These quantities as they stand serve as point estimates of the population parameters. If these estimates are in terms of transformed variables, they should be converted to cycle life values for presentation. The construction of appropriate confidence intervals is described in Section V B3. 2. S-N Curves: Suppose the cycle life (or some transformation of cycle life) of specimens from a certain population is a Normally distributed random variable, the applied stress being fixed. If the parameters of this distribution, /x and a, were known, then for any preassigned percentage, p, a number, K, could be determined so that p per cent of the population would have cycle lives exceeding /j, Ka. In general, however, n and a are unknown and can only be estimated by information obtained from a sample drawn from this population. What is done in this case is to determine a number k such that the probability of the random variable x ks not exceeding /x Ka is exactly 7, where 7 is a confidence level chosen in advance. It is then said, with a confidence level of 7, that at least p per rcent of the population is greater than x ks, where x and s have been computed from a sample of size n as described in Section V Bl. The numbers, k, which are called "one-sided tolerance factors," are functions of p, 7, and n. Table 33 lists some values for k. Consider the problem of constructing a specific S-N curve, say, for 90 per cent survival. Any point (Si, NI) on the curve should give the following information: If the applied stress is Si, then 90 per cent of the specimens to be tested will survive Ni cycles. Since the parameters of the fatigue life distribution are not known, the above defined S-N curve cannot be constructed. One Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 ^2) 10:11:02 2013 can, however, construct a curve whereon any point (2, has EDT the folDownloaded/printed lowing meaning: if the by applied stress is Sz , then with a confidence level of standard is for EDUCATIONAL USE ONLY. Escuela en This Ecuador ESPE pursuant to License 7, at least 90Politcnica per cent del of Ejercito the specimens to be tested will survive Nz Agreement. cycles.

No furt

42

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

The confidence level, 7, may be chosen in advance in accordance with the specific purpose of the S-N curve. An S-N curve is determined by fitting a curve of appropriate shape to points computed from data at several stress levels as described in the next paragraph. The confidence level associated with the fitted curve cannot be specified precisely but is probably larger than the confidence level used at each value of applied stress. The mechanics of constructing a point for fitting an S-N curve may be described as follows: Given a sample of n cycle lives for a fixed stress level S, compute the mean, x, and standard deviation, s, of the transformed cycle lives. From Table 33, read the value of k corresponding to the per cent survival, p, the confidence level, 7, and the sample size, n, that are being considered. The value x ks is then the appropriate abscissa for the ordinate, S, on the S-N curve. As in other instances, for presentation this value may be transformed back to read cycle life directly. Other sources discuss tolerance limits, which are different from those given here. These are basically of two types: (a) Two-sided Tolerance Intervals.These are discussed in reference (9). Tables are given there for two-sided tolerance factors. These factors are not applicable directly to fatigue data analysis, nor is it possible to transform them in an elementary way to obtain one-sided factors. (b) One-sided Tolerance Intervals of the Form (-co, x + ks).These are completely symmetric with the intervals of the form (x ks, <) as given here. The tolerance factors, k, are the same for both types of onesided intervals. 3. Confidence Intervals: (a) Confidence Interval for the Mean*.The statistics, x and 5, in addition to serving as point estimates of n and a, are also used in constructing confidence intervals for these parameters. The construction of a confidence interval for n is done as follows: 1. Select a confidence level, say 7 (where 7 is a proportion, 0 < 7 < 1, not a percentage), which is appropriate. Keep in mind that there is a risk of 1 7 that the interval being constructed will not contain p. Also remember that the greater the confidence, the wider the interval will be. 2. Having chosen 7, compute fti = (1 7)/2 and j82 = (1 + 7)/2. From Table 29, where t$ is tabulated as a function of degrees of freedom (d.f.) for several values of ft, read fa and fa , taking d.f. = n 1. (Note that fa is negative.) 3. The desired confidence interval has the limits:

(b) Confidence Interval for the Standard Deviation.A confidence interval for a Copyright has analogous meaning that for /x and isJul similarly constructed: by ASTM Int'l (allto rights reserved); Tue 9 10:11:02 EDT 2013 bylevel, 7. The 1. Downloaded/printed Select a confidence remarks about 7 made before are still This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further re pertinent.

ANALYSIS

OF FA

43

2. Compute ft\ and (82 as before. From Table 30, read:

using percentiles 100 ft and 100 ft respectively, with d.f. = n 1. 3. The desired confidence interval has the limits

EXAMPLE.This example will illustrate the construction of confidence intervals. The entries in the following table are logarithms of fatigue life values as observed in a test on eight specimens:
i
x{ = log Ni

1 2 3 4 5 6 7 8

4.8388 4.9243 4.9445 4.9542 4.9731 4.9777 5.0334 5.0828

The estimates are readily determined to be: x = 4.9661 and s = 0.0726. These are obtained using the denning formulas from Section V Bl. It is now desired to have 90 per cent confidence intervals on both n and <r. The following quantities are relevant (see above):

The desired confidence interval for /* has the limits

or
4.9175, 5.0147

Similarly, the desired confidence interval for <r has the limits:

or
0.0512, 0.1304

by ASTM (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Note Copyright that the pointInt'l estimates and confidence interval estimates are logDownloaded/printed by arithms and may be converted to cycle life values for presentation. This standard is for EDUCATIONAL USE ONLY. Escuelamedian Politcnica del en X Ecuador ESPE pursuant to cycles. License Agreement. No further reproductions authori Sample = Ejercito antilog = 92 thousand (This value may be

44

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

found from the original, imtransformed observations.) Confidence limits for the population median are derived directly from the confidence limits for JLI: antilog 4.9175 = 82,700 cycles and antilog 5.0147 = 103,400 cycles. Sample average, N = antilog (X + 1.1513<r2).23 Since o-2 is unknown, an approximate value of the sample average may be found by replacing er2 by s2. Then:
sample average, N = antilog (X 1.1513 s2) = 93,792 cycles

where = means "approximately equal to". Sample standard deviation:

where N' is the population mean in terms of cycle life. Replacing N' by N gives:
Sample standard deviation

Approximate confidence limits for the population standard ^deviation are derived from the confidence limits for v.

4. Tests of Significance: If cycle life data are collected at several stress levels for the purpose of estimating the life distributions of these several levels, the techniques of analysis just described are applicable. If, however, the principal interest is a comparison of distributions in the sense of deciding whether their paranreters are the same, then a different set of analyses is needed in order to use the information in the data more effectively. A summary of these techniques, classified according to the questions they are designed to investigate, is given in this section. Except for the part on "Differences Among k Means," this section is concerned with a comparison between two sets of data. (a) Difference Between Two Standard Deviations.Before testing whether or not the means of two samples are significantly different, it is important to investigate whether the standard deviations are significantly different or not. Let $i be the standard deviation for a sample of size n\ from the first Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 population and let s2 be similarly defined for the second. Compute the ratio Downloaded/printed by
23
This standard is for EDUCATIONAL USE ONLY.

Escuela del Ejercito en Ecuador ESPE to License Agreement. No further rep See Hald Politcnica (24), p. 161, and Soyero and Olds (25), p!pursuant 671.

ANALYSIS OF FATIGUE DATA

45

of the squares of the standard deviations, V, putting the larger standard deviation in the numerator; if then

In order to decide if the difference is statistically significant, a significance level must have been chosen in advance. If the chosen significance level (as a proportion) is denoted by a, compute 0 = 1 (a/2) (see Note) and, if /3 = 0.95 or 0.975, read from Table 32 the value of Fp corresponding to n\ 1 ''Degrees of Freedom for Numerator" and n% 1 "Degrees of Freedom for Denominator." (If s-? > Si2, the degrees of freedom are reversed.) If V > Fp , then the standard deviations are considered to be significantly different. If V ^ F$ , the standard deviations are considered to be not significantly different.
NOTE. -In "two-sided" tests, one-half of the critical region of the distribution curve of the statistic is in the upper tail, while the other half is in the lower tail; consequently, when using tables to obtain Fp and //j values, these values are obtained for significance levels of one-half of the level stated.

(Z>) Difference Between Two Means: 1. Standard Deviations Not Significantly Different. -If V ^ Fp (see (a) above) , the hypothesis that the population means are equal may be tested by the following procedure: compute s2, an estimate of the common variance, by

Next compute the test-statistic

where x\ and 2 are the sample means for the first and second samples, respectively. With a preassigned significance level, a, compute 0 = 1 (a/2) and find tp , from Table 29, corresponding to d.f. = n\, + nz 2. If \t \ > t$, conclude that the populations are different in mean or in variance or both. (On the average, identical populations will be erroneously judged different about 100 a per cent of the time.) If the samples are large enough so that the test for the standard deviations would have Copyright by difference ASTM Int'l between (all rights two reserved); Tue Jul 9 10:11:02 EDT 2013 probably detected any by important difference in the variances, a value Downloaded/printed standard is for EDUCATIONAL USE ONLY. \ 11 >Escuela fy canPolitcnica be attributed to different population means. With the same No fur del Ejercito en This Ecuador ESPE pursuant to License Agreement.

46FATIGUE TESTING AND STATISTICAL ANALYSIS or

DATA

reservation about sample size, if I / 1 < tp, one may conclude that the population means are not significantly different. 2. Sample Standard Deviations Significantly Different.Ifthe sample standard deviations are significantly different, the hypothesis that the population means are equal may be tested by the following procedures.24 Compute the numbers:

and

From Table 29 read a value /j, corresponding to

If this value is not an integer, use the nearest smaller integer. If I /' I < tp, the population means are judged equal and if I /' I > tp, they are judged unequal; a = 2(1 /3) is the approximate proportion of the time when the means are actually equal but will be incorrectly judged unequal. TABLE 20.COMPUTATIONS FOR SIGNIFICANCE TESTS.
First Surface Finish
Xi = log Ni

Second Surface Finish

*,- - x

(*,- - 5)* 162.1 X 10~ 17.5

4.7 1.4 0.5 1.3
4

*< = log Ni

Xi

(*< -i)

4.8388 4.9243 4.9445 4.9542 4.9731 4.9777 5.0334 5.0828 39.7288

-0.1273 -0.0418 -0.0216 -0.0119 +0.0070 +0.0116 +0.0673 +0.1167

45.3 136.2

4.5315 4.6232 4.6232 4.6435 4.6435 4.6532 4.6721 4.6902 4.6902 4.7243 46.4949

-0.1180 -0.0263 -0.0263 -0.0060 -0.0060 +0.0037 + 0.0226 + 0.0407 +0.0407 +0.0748

139.2

16.6 16.6 56.0

6.9 6.9 0.4 0.4 0.1 5.1

x io-4

369.0

248.2

EXAMPLE.Assume that, after testing specimens with one surface finish, another lot is fabricated with a different finish and tested at the same stress level. The tests are to be analyzed to determine whether the Tue change surfaceEDT finish significantly Copyright by ASTM Int'l (all rights reserved); Jul 9in 10:11:02 2013 affects the fatigue life at Downloaded/printed by the stress level used for the tests. The distribution of
This standard is for EDUCATIONAL USE ONLY.

24

This is an approximate test; see Hald (24), pp. 397-398.

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reprod

ANALYSIS OF FATIGUE DATA

47

fatigue lives at this stress level is roughly log-normal; therefore, the computations are in terms of log N rather than N. (See Table 20.)

We now have the mean, x, and the variance, s2, of each sample. To test them for significant differences, use the F-ratio test and then the /-test, choosing a significance level of 0.05.

From Table 32, F0.m (d.f. numerator = 7) (d.f. denominator = 9) = 4.20. The variances of the two samples are considered to be not significantly different. Thus the test given in Section V B4(6)l may be used:

From Table 29, t0.976 (d.f. = 16) = 2.12. Since t = 10.8 is larger than t0.y,5 = 2.12, the mean of the first sample is judged to be significantly larger than the mean of the second sample. In the matter of fatigue life at the test stress level, the second surface finish appears to be inferior to the first. (c) Difference Among k Means.: If k (k > 1) sets of data have been obtained, each of which is a random sample from a Normal population, these populations being taken as having a common standard deviation, then one can test the hypothesis that these populations have a common mean. Let:% be the/th observation from the iih group, Xi be the mean of the ith group, and Ui be the size of the tth group. Define sw2, average variance within samples, and sb*, variance among samples, by

and

where Downloaded/printed x is the mean of the numbers xt . Compute the ratio by

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en This Ecuador ESPE pursuant to License Agreement. No fur

48

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA From Table 32 read the value of F$ (/3 = 1 a), corresponding to k 1

"Degrees of Freedom in the Numerator" and T^ w,- jfe "Degrees of Precis dom in the Denominator." The hypothesis that the means of the k populations are equal is accepted only if V < Fp . It is recommended that a statistician be consulted if a more complete exploration of this situation is considered necessary. 5. The Staircase Method: As pointed out in Section III B2, the staircase method has been found useful for estimating the mean fatigue strength at a preassigned cycle life, N. The procedure for experimentation and some discussion of the advantages and limitations are given in Section III. (a) OtUline of A nalytical Method: 1. Determine whether failures or runouts are the less frequent event. Only the less frequent event is used in the analysis. 2. Number the stress levels and make a table as follows:
Stress i Ni iNi

Sz Si So

2 1 0

N2 Ni No

2N2 N! 0

Sum:

where: i = 0 is assigned to the lowest stress, S0, on which the less frequent event occurs; i = 1 is, assigned to the stress level for i0 + d, etc.; Ni = number of the less frequent events at the corresponding levels; iNj = product of i and Ni at each level; and A = "iNi. 3. Estimate the mean fatigue strength by

where: N = total number of the less frequent events, d = preselected stress increment, and S0 = first stress level. And use +1 if the less frequent event is a run out, or ^, if the less frequent event is a failure. 6. The Modified Staircase Method: Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 by Sizes.For sample sizes less than 15, use the fol(a) Downloaded/printed For Small Sample This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further lowing simplified procedure:

ANALYS

1. Start the staircase procedure as before, but use a step size, d, equal to twice the preselected stress increment until the first pair of opposite results is obtained. Subsequently, use a step size, d, equal to the preselected stress increment. 2. In the analysis, disregard the specimens used up to the first pair which give opposite results. Let n = total number of specimens tested, starting with the first pair of opposite results. Let C = sum of the stresses used on the last n 1 tests plus the stress which would have been used on the next test had it been run. Then,
C m n

or merely the average stress used on the last n 1 tests plus that which would have been used on the (n + l)th test. EXAMPLE.The following data were obtained in a modified staircase fatigue test:
Stress, ksi

2
h

4
h

6
h h

8
h

10

11

12

13

14
h

15

16
h

17

62 61 60 59

h h

h h

h h

58

Where x = specimen has failed and o = specimen has not failed (as in Fig. 3). The n is taken as 1 with the first specimen of the pair which gave opposite results. The (n + l) th , or 17th, test would have been run at 61.0 ksi. The last n 1 tests started at n = 2. Therefore, the total C is composed of the stresses from n = 2 (62.0), to n = 17 (61.0).

C 980 m = - = 61.25 ksi n 16

(V) For Reducing Testing TimeFor staircase runs involving more than 30 specimens, the time required to complete the staircase method is long because of the sequential nature of the procedure. This long time can be reduced by dividing one long staircase program into several shorter, independent staircases and conducting these simultaneously. 1. Proceed by dividing the total number of specimens, T, into r groups of n each, so that rn = T. Test each group as a separate small-sample staircase program, as described previously. Test as many groups simultaneously as desired. 2. The best estimate of m for the whole set of T specimens is then:
Copyright by ASTM Int'l (all srights iMii reserved); Tue Jul 9 10:11:02 EDT 2013 mby= = arithmetic average of r values of m. Downloaded/printed h This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.

50

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

EXAMPLE.The step size is 5 ksi; and the data arose from testing five groups of 15 specimens each.
No. 2 No. 3 No. 4 No. 5
Group No. 1 m 101

102 97 99 99

TABLE 21.R. R. MOORE ROTATING BEAM; STEP0 TESTS OF 42 SPECIMENS. SAE 4330 Steel: Tensile Strength, 140 ksi (Nominal). Failure Stress,
ksi

Kilocycles at Failure Stress

928 585 2 082 665 969 < 1 567 1 522 490 1 085 1 285 10 390 289 595 384 479 1 039 < 973 329 1 489 1 265 2 076

Failure Stress,
ksi

Kilocycles at Failure Stress

950 692 1 077 1 177 5 184 7 661 997 1 524 566 630 108 815 3 850 1 166 681 967 606 680 886 735 1 479

65

60

70

65

(.Continued, nex ! column)

0

Starting stress = 55 ksi. Steps = 5 ksi. Run = 107 cycles or to failure.

The estimate of the mean fatigue strength of the population is given by:

7. The Step Method: The data from the step method described in Section III can be analyzed by the use of arithmetic or logarithmic probability paper. This is illustrated by analysis of the data presented in Table 21. In this test, all 42 specimens were first tested at an applied stress of 55 ksi, at which stress level all survived 107 cycles. The applied stress was then raised to 60 ksi and the test repeated; the 31 surviving specimens were testedCopyright at 65 ksi; Int'l and so reserved); forth.Tue In each the surviving specimens are by ASTM (all rights Jul 9 10:11:02case, EDT 2013 subjected to an applied stress that is 5 ksi higher than the preceding value. Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. This is repeated until specimens have failed. The mean between the last Escuela Politcnica del Ejercitoall en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.

ANALYSIS or FATIGUE DATA TABLE 22ANALYSIS OF DATA IN TABLE 21.

Estimated Fatigue Strength, ksi

51

Number of Specimens Per Cent of 42 with This Per Cent of 42 Having at with This Fatigue Least This Fatigue Fatigue Strength Strength hhh
11 16 11 4 0

57.5 62.5 67.5 72.5 (77.5)

26.2 38.1 26.2

9.5 0

100 73.8 35.7 9.5 0

stress survived and the failure stress is the estimated fatigue strength for each specimen. The data are then tabulated as indicated in Table 22. The data in the last column can be plotted on arithmetic probability paper, as shown in Fig. 8.

FIG. 8.Per Cent of Specimens Having At Least the Indicated Fatigue Strength at 107 Cycles.

The stress levels of 100 per cent and 0 per cent survivals are indicated on Fig. 8 by short arrows. It Tue is Jul interesting to note that the dash-line exCopyright by ASTM Int'l (all rights reserved); 9 10:11:02 EDT 2013 trapolations of the Downloaded/printed by "reasonably straight line" suggest that, if a very much This standard is for EDUCATIONAL USE ONLY. Politcnica delspecimens Ejercito en Ecuador ESPE pursuant totested, License Agreement. No further reproductions authorized. larger Escuela number of had been some of them would probably

52

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

have a fatigue strength higher than 72.5 ksi and some might not have survived a stress of 55.0 ksi. Extrapolation on probability paper of the type used for Fig. 8, however, depends critically on the assumption of a Normal population; this assumption is almost certainly not safe. G. W. Snedecor (26) says that extrapolation is mostly a guessing game. If a straight line does not reasonably represent the plotted points, the population distribution is probably not nearly Normal enough to warrant analysis by logarithmic or arithmetic probability paper. In such cases, a statistician should be consulted to determine whether some transformation will Normalize the data. If the straight line is acceptably close to the plotted points, the mean fatigue strength is approximately the value of the abscissa at P = 0.50. Otherwise, that value is the median, that is, the value above which half of the specimens survived, but it is not necessarily the mean. If the straight line is acceptably close, the approximate fatigue strength corresponding to any given mortality ratio (per cent failure) can be estimated by reading directly from the line, but only between the first and last plotted points. Extrapolation to very low or very high probabilities is entirely unjustified and may be very misleading.
TABLE 23.PROT TEST COMPUTATIONS.
Group Fracture Stress, S (Mean), psi (Mean), psi per cycle 0.208

Log a - 3 1 ,7723 - 3 0 ,8482 - 3 9, ,3181 - 3

9, .3181

Line'i

Change in Log a (A Log a)

A C
0

C B

92 700

80 600 71 800 92 700 See Fig. 11.

0.0592 0.00705 0.208

C - B B A C - A

0.5458 0.9241 1.4699

8. The Prot Method: The Prot method is based upon the assumption that the curve relating the failure stresses and some power of the rate of increase of stress per cycle is a straight line, provided that (1) the power is properly selected and (2) this straight line will intersect the axis of failure stresses at the fatigue limit. This may be expressed by the equation: where: ,5 = fracture stress (final breaking stress), psi, E = fatigue limit, psi, K = a constant, a = rate of increase of stress, psi per cycle, and n exponent of a that makes S a linear function of an. A mathematical solution for rights the value of Tue the Jul fatigue strength Copyright by ASTM Int'l (all reserved); 9 10:11:02 EDTand 2013the optimum value of n, the Downloaded/printed by power of a that will place the plotted test values This standard is for EDUCATIONAL USE ONLY. on a straight line, anddel for the standard deviation of thetofatigue is No Escuela Politcnica Ejercito en Ecuador ESPE pursuant License strength Agreement.

further

ANALYSIS or FATIGUE DATA

53

proposed by Corten et al (5). The following "cut-and-try" method provides an alternative to the mathematical solution. The arithmetic mean of the individual values of a in each of the three groups is determined, as well as the arithmetic mean of the individual values of Sthe fracture stressin each group. (Strictly, the geometric means should be used instead of arithmetic means, but the difference would usually be unimportant.) This calculation gives three pairs of values, such as those included for illustration in the second and third columns of Table 23. They represent the entire set of test data fairly well. The test data may also be plotted as shown in Fig. 9.

Stress Cycles FIG. 9.Prot Test: Stress as Linear Function of Stress Cycles. TABLE 24.PROT TEST COMPUTATIONS.
Assumed E, psi Group

S -E

Log (S-E)

Line

Change in Log (S E) [A log (S - E)]

Slope, n = A log (S - E) -5- A log a

71 000. ..

ic
69 000 . . .

21 700 9 600 800 21 700 23 11 2 23 25 13 4 25 700 600 800 700 700 600 800 700

4.3365 3.9823 2.9031 4.3365 4.3747 4.0645 3.4472 4.3747 4.4099 4.1335 3.6812 4.4099

C -B B -A C -A
C -B B -A C -A C -B B -A C -A

0.3542 1.0792 1.4334 0.3102 0.6173 0.9275 0.2764 0.4523 0.7287

0.3542/0.5458 = 0.65 1.0792/0.9241 = 1.17 1.4334/1.4699 = 0.98 0.3102/0.5458 = 0.57 0.6173/0.9241 = 0.67 0.9275/1.4699 = 0.63 0.2764/0.5458 = 0.506 0.4523/0.9241 = 0.49 0.7287/1.4699 = 0.496

f B h 1C f B h

67 000. . .

The cut-and-try method depends upon the fact that Prot's equation can be written thus:
log (S-E) = log K + n log a

This plots as through the Copy r iag straight h t b y line AST M I n t points ' l ( a l C, l rB, i g and h t s A, r efor ser v e correct d); Tue Jul value D ofoE, with slope n. The procedure is to assume values of E, as shown wnloaded/printed by EDUCATIONAL USE ONLY. in Table slopes C-B, B-A, and E s c24. u e The la P o l i t cof n ilines ca d e This l standard E j e is r for c i t oC-A e n are E ccomputed. u a d o r E Since SPE pursu

54

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

C, B, and A, for the correct value of E in the logarithmic equation, all lie on a straight line, the slopes of the three lines must come out the same when the correct value of E has been determined.25 Table 24 is a suggested form for systematizing the trial computations. A value for E close to but a little less than the value of the mean failure stress for A must first be assumed. The quantities indicated in Table 24 are then computed, using values of S and A log a given in Table 23. If the

(a) Incorrect assumed value of fatigue limit. (b) Correct assumed value of fatigue limit. FIG. 10.Log-Log Plot of Prot Data.

slopes are not within 1 or 2 per cent of one another (see Fig. 10), another value of E must be assumed and the calculation repeated. The value of E that makes the slopes sufficiently close to one another so that the three points lie on the same straight line is the estimate of the fatigue limit as* given by the Prot method.
25 Although theoretically only three groups of specimens need be used, in some cases more groups with fewer specimens per group may be advisable. The "cut-and-try" procedure described for three groups can easily be adapted to four or more groups. Some materials have been foundfor example, some case-hardened steelsfor which the Prot method gave no solution. Whether or not this means that they have no true fatigue limit is not known.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further r

STP91A-EB/Feb. 1964

APPENDIX I MISCELLANEOUS REFERENCE TABLES

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

Copyright 1964 by ASTM International www.astm.org

h h h h h
Minimum Per Cent of PopuNumber Confiof Points dence in Each Level, Group, n per cent
1 3

Lowest Ranking Points Number of Groups, m

Second Ranking Points Number of Groups, m

15

25

15

25

50.00 50.00 50.00 50.00 50.00 50.00 50.00 50 75 25.00 32.64 35.94 37.88 39.20 41.50 43.35 10.00 19.58 24.66 27.86 30.10 34.15 37.51 90 95 5.00 13.53 18.94 22.53 25.14 30.00 34.14 98 2.00 8.40 13.52 17.27 20.10 25.61 30.48 1.00 5.89 10.56 14.23 17.10 23.14 28.14 99 99.9 0.10 1.84 4.76 7.67 10.25 16.12 22.06
50 75 90 95 98 99 99.9 50 75 90 95 98 99 99.9 50 75 90 95 98 99 99.9 50 75 90 95 98 99 99.9 50 75 90 95 98 99 99.9 50 75 90 95 98 99 99.9
0

79.37 79.37 79.37 79.37 79.37 79.37 79.37 50.00 50.00 50.00 50.00 50.00 50.00 50.00 63.00 68.84 71.10 72.35 73.18 74.58 75.68 32.64 38.20 40.51 41.84 42.74 44.30 45.55 46.42 58.06 62.71 65.31 67.01 69.89 72.12 19.58 28.37 32.38 34.76 36.39 39.26 41.59 36.84 51.34 57.41 60.85 63.11 66.94 69.89 13.53 23.09 27.82 30.73 32.73 36.32 39.26 27.14 43.80 51.33 55.68 58.57 63.50 67.29 8.40 17.83 23.08 26.43 28.79 33.09 36.67 21.54 38.90 47.27 52.20 55.50 61.15 65.53 5.89 14.75 20.16 23.73 26.28 31.00 34.97 10.00 26.38 36.22 42.47 46.80 54.41 60.42 1.84 8.04 13.18 16.97 19.84 25.43 30.36 84.09 84.09 84.09 84.09 84.09 84.09 84.09 61.42 61.42 61.42 61.42 61.42 61.42 61.42 70.71 75.58 77.42 78.45 79.12 80.26 81.14 45.63 50.91 53.04 54.25 55.06 56.45 57.55 56.23 66.52 70.47 72.65 74.06 76.44 78.26 32.05 41.40 45.38 47.68 49.22 51.90 54.02 47.29 60.65 65.95 68.89 70.80 74.00 76.43 24.86 35.91 40.85 43.77 45.73 49.16 51.88 37.61 53.84 60.64 64.46 66.95 71.13 74.30 17.94 30.04 35.90 39.43 41.83 46.07 49.48 31.62 49.26 57.01 61.41 64.30 69.15 72.83 14.09 26.38 32.71 36.59 39.27 44.03 47.87 17.78 36.81 46.69 52.61 56.58 63.36 68.53 6.40 17.41 24.41 29.04 32.34 38.38 43.40 87.05 87.05 87.05 87.05 87.05 87.05 87.05 68.61 68.61 68.61 68.61 68.61 68.61 68.61 75.79 79.93 81.49 82.35 82.91 83.87 84.60 54.58 59.38 61.28 62.35 63.07 64.29 65.26 63.10 72.17 75.58 77.44 78.65 80.66 82.19 41.61 50.65 54.35 56.46 57.85 60.26 62.15 54.93 67.03 71.68 74.22 75.86 78.59 80.65 34.26 45.40 50.13 52.85 54.67 57,79 60.25 45.74 60.93 67.02 70.38 72.55 76.15 78.85 26.71 39.60 45.40 48.78 51.05 54.98 58.09 39.81 56.75 63.79 67.70 70.23 74.44 77.60 22.21 35.84 42.26 46.07 48.64 53.10 56.63 25.12 44.95 54.37 59.82 63.41 69.41 73.91 12.20 26.10 33.78 38.58 41.90 47.79 52.52 89.09 89.09 89.09 89.09 89.09 89.09 89.09 73.55 73.55 73.55 73.55 73.55 73.55 73.55 79.37 82.97 84.32 85.06 85.54 86.36 86.99 61.05 65.38 67.08 68.03 68.67 69.76 70.61 68.13 76.20 79.19 80.81 81.86 83.60 84.92 48.97 57.45 60.84 62.75 64.01 66.17 67.86 60.70 71.65 75.77 78.00 79.44 81.81 83.60 41.82 52.56 56.97 59.48 61.13 63.96 66.16 52.10 66.18 71.64 74.62 76.53 79.68 82.03 34.17 47.04 52.56 55.72 57.82 61.42 64.23 46.42 62.37 68.75 72.25 74.49 78.20 80.95 29.43 43.38 49.59 53.19 55.58 59.70 62.91 31.62 51.36 60.19 65.17 68.41 73.76 77.73 18.14 33.54 41.35 46.06 49.25 54.80 59.17 93.30 93.30 93.30 93.30 93.30 93.30 93.30 83.77 83.77 83.77 83.77 83.77 83.77 83.77 87.06 89.40 90.27 90.74 91.05 91.58 91.98 75.26 78.28 79.45 80.09 80.52 81.25 81.82 79.43 84.95 86.93 88.00 88.68 89.81 90.66 66.31 72.68 75.11 76.46 77.34 78.82 79.97 74.11 81.87 84.66 86.15 87.10 88.65 89.81 60.59 69.07 72.33 74.14 75.32 77.30 78.82 67.62 78.06 81.86 83.89 85.17 87.26 88.79 53.98 64.81 69.06 71.42 72.95 75.52 77.48 63.10 75.33 79.86 82.28 83.80 86.28 88.09 49.64 61.87 66.79 69.54 71.31 74.30 76.57 50.12 67.05 73.74 77.34 79.63 83.31 85.97 37.63 53.40 60.19 64.02 66.53 70.73 73.92 96.59 96.59 96.59 96.59 96.59 96.59 96.59 91.75 91.75 91.75 91.75 91.75 91.75 91 75 93.30 94.55 95.01 95.26 95.43 95.69 95.90 87.10 88.77 89.41 89.77 90.00 90.40 90.70 89.12 92.17 93.24 93.81 94.17 94.76 95.21 81.90 85.63 87.01 87.76 88.25 89.07 89.70 86.09 90.48 92.01 92.81 93.32 94.15 94.76 78.39 83.53 85.43 86.46 87.13 88.23 89.07 82.23 88.35 90.48 91.59 92.29 93.41 94.23 74.12 80.99 83.53 84.90 85.78 87.24 88.33 79.43 86.79 89.36 90.71 91.54 92.88 93.85 71.12 79.19 82.19 83.81 84.84 86.55 87.83 70.79 81.88 85.87 87.94 89.23 91.27 92.72 62.24 73.74 78.14 80.52 82.03 84.51 86.33

10

20

This standard is for EDUCATIONAL USE ONLY. From Schuette (27). Downloaded/printed by curves faired through test data, subtract three from the number of NOTEIn evaluating fatigue Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions author groups before entering table.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

56

Thhhhhhhhh lation Exceeding Median of: Third Ranking Points Number of Groups, m
1 3 5 7 9 15 25 1 3

Fourth Ranking Points Number of Groups, m

5 7 9

15

25

20.63 9.14 3.45 1.70 0.67 0.33 0.03 38.57 24.30 14.22 9.76 6.01 4.18 1.31 50.00 35.94 24.66 18.93 13.53 10.56 4.76 57.85 44.68 33.32 27.13 20.93 17.31 9.40

20.63 20.63 20.63 20.63 20.63 20.63 12.33 13.79 14.67 15.28 16.36 17.25 7.00 9.00 10.31 11.25 13.00 14.50 4.73 6.75 8.15 9.20 11.20 12.99 2.88 4.72 6.12 7.20 9.39 11.41 3.65 4.98 6.05 8.29 10.42 2.00 0.61 1.60 2.62 3.54 5.69 7.97

38 . 57 28. 7& 20.96 16.90 12.94 10.65 5.75

50.00 40.51 32.36 27.81 23.04 20.12 13.11 57.85 49.06 41.17 36.58 31.61 28.45 20.44

38.57 30.63 24" . 09 20.54 16.90 14.70 9.50 50.00 42.38 35.73 31.91 27.80 25.19 18.57 57.85 50.82 44.47 40.71 36.57 33.88 26.74

38.57 31.72 25. 99" 22.80 19.46 17.39 12.30 50.00 43.46 37.70 34.35 30.71 28.37 22.25 57.85 51.84 46.38 43.13 39.52 37.15 30.75

38.57 32.46 27.29 24.38 21.29 19.35 14.45 50.00 44.18 39.04 36.03 32.73 30.58 24.90 57.85 52.51 47.65 44.76 41.53 39.39 33.57

38.57 33.76 29.61 27.23 24.66 23.01 18.69 50.00 45.44 41.37 38.98 36.32 34.58 29.85 57.85 53.67 49.88 47.60 45.04 43.35 38.65

38.57 15.91 15.91 15.91 15.91 15.91 15.91 15.91 34.80 6.94 9.40 10.53 11.22 11.69 12.54 13.24 31.51 2.60 5.30 6.83 7.83 8.56 9.92 11.09 5.11 6.18 6.97 8.52 9.91 29.60 1.30 3.56 0.51 2.16 3.56 4.62 5.45 7.12 8.68 27.51 26.15 0.25 1.50 2.74 3.76 4.58 6.28 7.93 0.03 0.46 1.21 1.97 2.67 4.29 6.03 22.51 50.00 31.38 46.44 19.38 43.25 11.21 7.64 41.37 39.26 4.69 37.87 3.27 34.05 1.02 57.85 54.60 51.65 49.87 47.87 46.54 42.83

31.38 31.38 31.38 31.38 23.06 24.63 25.55 26.17 16.64 19.21 20.77 21.84 13.35 16.30 18.14 19.44 10.17 13.35 15.42 16.90 8.35 11.58 13.75 15.33 4.48 7.43 9.66 11.38

31.38 31.38 27.26 28.15

23.78 25.37 21.80 23.77 19.67 22.03 18.31 20.90

14.80 17.91

42.14 42.14 42.14 29.71 33.67 35.32 20.09 26.61 29.50 15.32 22.73 26.22 10.88 18.74 22.73 8.47 16.31 20.54 3.79 10.54 15.02
64.49 54.23 44.83 39.33 33.43

42. 14 42.14 42.14 42.14 36.27 36.91 38.03 38.93 31.22 32.38 34.43 36.09
28.32 29.76 32.33 34.42 25.20 26.92 30.02 32.57 23.21 25.09 28.51 31.36 18.07 20.29 24.47 28.06 64.49 60.38 56.59 54.29 64.49 64.49

74.14 64.46 55.04 49.31 42.95 38.82 28.15

86.84 81.33 75.52 71.74 67.26 64.17 55.26

74.14 67.80 61.67 57.87 53.51 50.58 42.41

86.84 83.27 79.66 77.31 74.53 72.59 66.86

74.14 69.11 64.29 61.30 57.86 55.53 48.92

86.84 84.03 81.22 79.43 77.31 75.83 71.48

74.14 69.85 65.76 63.24 60.33 58.35 52.73

86.84 84.45 82.10 80.60 78.84 77.62 74.02

74.14 70.33 66.74 64.52 61.95 60.21 55.26

86.84 84.73 82.66 81.36 79.83 78.77 75.66

74.14 71.18 68.41 66.70 64.74 63.41 59.61

86.84 85.20 83.62 82.64 81.49 80.70 78.40

74.14 71.84 69.71 68.40 66.90 65.89 63.00

86.84 85.58 84.37 83.62 82.76 82.17 80.45

64.49 64.49 64.49 57.69 59.07 59.86 51.40 54.06 55.58 47.60 51.02 52.98 43.34 47.59 50.05 29.71 40.54 45.30 48.08 20.46 32.94 38.98 42.59 75.79 77.93 78.76 79.23 69.58 73.98 75.68 76.63 65.63 71.48 73.74 75.00 61.04 68.54 71.47 73.10 57.93 66.52 69.91 71.80 49.13 60.64 65.36 68.00

61.28 61.99 58.34 59.71

45.04 49.33 52.74

51.68 54.52 56.76 49.94 53.16 55.71

56.55 58.33

81.94 81.94 81.94 81.94 81.94 81.94 81.94

79.55 80.08 80.50 77.26 78.32 79.14 75.83 77.23 78.31 74.17 75.97 77.36 73.03 75.11 76.71 69.73 72.63 74.85

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 USE ONLY. This standard is for EDUCATIONAL Downloaded/printed by Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.

57

58

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

TABLE 26.UNPAIRED Critical lower and upper rank totals for the 5 per cent and 1 per cent levels of significance. Values
h
4 5

10

11

12

13

Ni

4 5 6
7. 8. 9. 10

11. 12. 13. 14.

15 16

17.
18 19 20

21. 22. 23. 24.

25 26 27

28. 29. 30.

0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01 0.05 0.01

11 25 12 28 18 37 13 10 13 11 14 11 15 12 16 12 16 13 17 13 18 14 19 14 31 34 35 37 38 41 41 44 44 48 48 51 51 55 54 58 57 62

15 19 16 20 17 21 18 22 19 24 19 25 20 26 21 27 22 28 23 30 23

40 41 26 44 23 45 28 48 24 49 29 52 25 53 31 56 27 56 33 61 28 60 34 65 29 64 36 69 30 68 37 73 31 72 39 77 32 75 41 82 33 42 35

52 55 56 37 60 33 61 39 65 34 65 41 69 36 69 43 74 37 74 44 79 38 78 46 84 40 83 48 89 41 87 50 94 43 91 52 99 44 96 54 103 46 56 47

68 72 73 78 78 83 83 89 89 95 94 100 99 106 104 111 109 117 114 122 119 128

49 44 51 46 54 47 56 49 58 51 61 53 63 55 65 56 68 58 70 60 73 62

87 92 93 98 98 105 104 111 110 117 115 123 121 129 127 136 132 142 138 148 143 154

63 57 66 59 68 61 71 63 74 65 77 67 79 70 82 72 85 74 88 76 91 79

108 114 114 79 131 121 71 189 121 82 138 96 157 128 74 146 88 165 127 85 145 100 164 116 184 135 76 154 90 174 106 194 133 88 152 104 171 120 192 142 79 161 93 182 109 203 139 91 159 107 179 124 200 149 81 169 96 190 112 212 146 95 165 111 186 128 208 155 84 176 99 198 116 220 152 98 172 114 194 132 216 162 87 183 102 206 119 229 158 101 179 118 201 136 224 169 89 191 105 214 122 238 164 104 186 121 209 140 232 179 92 198 108 222 126 246 170 107 193 125 216 144 240 182 94 206 111 230 129 255 111 199 129 223 148 248 97 213 114 238 133 263 132 231 152 256 117 246 136 272 156 264 139 281

137 126 142 129 146 133 150 137 155 141 159 144 163 147 168 151 172 155 177 158 181 162

214 226 222 235 231 244 240 253 248 262 257 272 266 282 274 291 283 300 291 310 300 319

"Adapted from of Dr. Wilcoxon (28). Tue Jul Copyright by work ASTM Int'l Frank (all rights reserved);

9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductio

MISCELLANEOUS REFERENCE TABLES

RANK TEST.0 in the body of the table refer to the group with the smaller number of measurements, NI.
14 15

59

16

17

18

19

20

21

22

23

24

25

160 147 165 151 170 156 175 159 179 163 184 167 189 171 194 175 199 179 203 184 208 188

246 259 255 185 269 171 264 190 278 175 273 196 289 179 283 201 299 184 292 206 309 188 301 211 319 193 310 216 329 197 319 222 339 202 329 227 348 2 0 6 338 232 358 211 237 215

280 294 290 212 305 196 299 218 316 200 309 223 326 205 319 229 337 210 329 234 347 215 339 240 358 220 348 246 368 225 358 251 3 7 9 230 368 257 389 235 378 263 400 239 268 244

316 332 326 241 344 223 337 247 355 228 347 253 366 233 358 259 377 238 368 265 388 244 378 271 399 249 389 277 410 254 399 283 421 260 409 289 433 265 420 295 444 270 301 275

354 372 365 271 384 251 376 278 396 257 387 284 408 263 398 290 419 268 409 297 431 274 420 303 443 2 8 0 431 310 455 286 442 316 466 291 453 323 478 297 464 329 490 303 336 308

395 415 407 303 427 282 418 310 439 288 430 317 452 294 441 324 464 300 453 331 4 7 6 306 464 338 4 8 9 313 476 345 501 319 487 352 513 325 499 359 525 331 510 366 538 337 373 343

438 459 450 338 472 315 462 345 485 321 474 352 498 328 486 360 511 334 498 367 524 341 510 374 536 347 522 382 549 354 534 389 562 360 546 396 575 367 558 404 588 373 411 380

482 505 495 374 519 349 508 381 532 356 521 389 546 363 533 397 559 370 546 405 573 377 558 412 586 383 571 420 600 390 584 428 613 397 596 436 627 404 4 4 609 4 640 411

529 554 543 412 568 385 556 420 582 392 569 428 0 0 596 4 582 436 611 407 4 4 596 4 625 414 609 452 639 422 622 461 653 429 635 469 667 436 6 4 8 477 681 444

579 605 592 451 620 423 606 460 6 3 4 431 620 469 649 438 6 3 4 477 664 446 648 486 678 454 661 494 693 462 675 503 708 469 689 512 722 477

630 658 644 493 673 463 658 502 689 471 673 511 704 479 687 520 719 487 702 529 734 495 716 538 750 504 730 547 765 512

683 713 698 536 729 505 713 5 4 6 745 513 728 555 761 522 743 565 777 530 758 575 792 539 773 584 808 547

739 771 7 5 4 787 770 803 785 820 800 836 816 852

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduction

60

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

TABLE 27.PERCENTILES OF THE

Degrees 6 of freedom
1 2

2 x

DISTRIBUTION."

Percentiles
0.5

2.5

10

90

95

97.5

99

99.5

3 4 5 7
g 6

0.000039 0.0100 0.0717 0.207 0.412 0.676 0.989 1.34 1.73 2.16

0.00016 0.0201 0.115 0.297 0.554 0.872 1.24 1.65 2.09 2.56

0.00098 0.0506 0.216 0.484 0.831

0.0039 0.1026 0.352 0.711 1.15

0.0158 0.2107 0.584 1.064 1.61

2.71 4.61 6.25 7.78 9.24 10.64 12.02 13.36 14.68 15.99 17.28 18.55 19.81 21.06 22.31
23.54 25.99 28.41 33.20 40.26

3.84 5.99 7.81 9.49 11.07 12.59 14.07 15.51 16.92 18.31 19.68 21.03 22.36 23.68 25.00
26.30 28.87 31.41 36.42 43.77

5.02 7.38 9.35 11.14 12.83 14.45 16.01 17.53 19.02 20.48 21.92 23.34 24.74 26.12 27.49
28.85 31.53 34.17 39.36 46.98

6.63 9.21 11.34 13.28 15.09 16.81 18.48 20.09 21.67 23.21
24.73 26.22 27.69 29.14 30.58

7.88 10.60 12.84 14.86 16.75 18.55 20.28 21.96 23.59 25.19
26.76 28.30 29.82 31.32 32.80

9 10 11 12 13

1.24 1.69 2.18 2.70 3.25 3.82 4.40 5.01 5.63 6.26 6.91 8.23 9.59 12.40 16.79
24.43 40.48 91.58

1.64 2.17 2.73 3.33 3.94 4.57 5.23 5.89 6.57 7.26 7.96 9.39 10.85 13.85 18.49 26.51 43.19 95.70

2.20 2.83 3.49 4.17 4.87 5.58 6.30 7.04 7.79 8.55 9.31 10.86 12.44 15.66 20.60
29.05 46.46 100.62

14 15

2.60 3.07 3.57 4.07 4.60 5.14 6.26 7.43 9.89 13.79 20.71 35.53 83.85

3.05 3.57 4.11 4.66 5.23 5.81 7.01 8.26 10.86 14.95 22.16 37.48 86.92

20 24 30

16 18

32.00 34.27 34.81 37.16 37.57 40.00 42.98 45.56 50.89 53.67

40 60 120
a 6

51.81 55.76 59.34 63.69 66.77 74.40 79.08 83.30 88.38 91.95 140.23 146.57 152.21 158.95 163.64

From Dixon and Massey (9) by permission of the publishers. For large values of degrees of freedom the approximate formula may be used:
X2 = Yz\.Za + (2n - l)i/p

where Za is the normal deviate and n is the number of degrees of freedom. For example, hhhh Yl (2.326 + 10.909)2 = 87.6 for the 99th percentile, for 60 d.f.

CopyrightbyASTMInt'l(allrightsreserved);TueJul 910:11:02EDT2013 Downloaded/printedby This standard is for EDUCATIONAL USE ONLY. EscuelaPolitcnicadelEjercitoenEcuadorESPEpursuanttoLicenseAgreement.Nofurtherreproductionsauthorized.

hhh
TABLE 28.AREAS OF THE "NORMAL" CURVE."
z

61

Area 0.0013 0.0019

Area

-3.0 -2.9 2.8 -2.7. 2.6 -2.5 -2.4 -2.3 -2.2 2.1
2.0 1.9 1.8 1.7 1.6

3.0 ff 2 9 ff

2 7 ff 2 6 ff 2.5 2.4 ff 2 3 ff 2.2 ff 2 1 <r 2 1 1 1 1 0 r 9 <r 8 ff 7 ff 6 ff

2 8 <r

0.0026 0.0035 0.0047 0.0062 0.0082 0.0107 0.0139 0.0179 0.0228 0.0287 0 0359 0 0446 0 0548 0.0668 0.0808 0.0968 0.1151 0.1357 0.1587 0.1841 0.2119 0.2420 0.2743 0.3085 0.3446 0.3821 0.4207 0.4602 0.5000
0.001 0.005 0.010 0.025 0.050 0.100 0.150 0.200 0.250 0.300

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 15 1.6 1.7 1.8 1.9 2.0 2.1 2.2 2.3 2.4 2.5 2.6
2.7

At At At At At At At At At At At At At H At At At /t At M At At At At At At At At At At

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

0.1 ff 0.2 ff 0.3 ff 0.4 ff 0.5 ff 0.6 ff 0.7 ff 0.8 ff 0.9 ff 1.0 ff 1.1 ff 1.2 ff 1.3 ff 1 4 ff 1 5 ff 1.6 ff 1.7 ff 1.8 f 1.9 ff 2.0 ff 2.1 ff 2.2 ff 2.3 ff 2.4 ff 2.5 ff 2.6 ff 2.7 ff 2.8 ff 2.9 ff 3.0 ff

0.5398 0.5793 0.6179 0.6554 0.6915 0.7257 0.7580 0.7881 0.8159 0.8413 0.8643 0.8849 0.9032 0.9192 0.9332 0.9452 6.9554 0.9641 0.9713 0.9772 0.9821 0.9861 0.9893 0.9918 0.9938 0.9953 0.9965 0.9974 0.9981 0.9987 0.999 0.995 0.990 0.975 0.950 0.900 0.850 0.800 0.750 0.700 0.650 0.600 0.550

n
h h h h h h h h h

-1.5 1.4 1.3 1.2 1.1

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1. 0

u - 1.5 u 1.4 ff h 1.3 ff h 1 2 ff

M A* 1.0 ff

1.1
ff ff ff ff

h h h h h

0.9 0.8 0 7 0 6

h 0 5 ff h 0 4 ff /* 0.3 ff h 0.2 ff h 0.1 ff h h

A* M M A* M M A* At M A*

2.8 2.9 3.0

-3.090 -2.576 -2.326 -1.960 -1.645 -1.282 -1.036 -0.842 -0.674 -0.524 -0.385 -0.253 -0.126 0

3.090 2.576 2.326 1.960 1.645

+ 3.090 + 2.576 + 2.326 + 1.960 + 1.645 + 1.282 + 1.03 + 0.842 + 0.674. +0.524 + 0.385. + 0.253. + 0.126.

At At At At At
At At At At At

+ + + + +
+ + + + +

3.090 2.576 2.326 1.960 1.645

1.282 1.036 0.842 0.674 0.524

1.282 1.036 0.842 0.674 0.524

0.385 0.253 0.126

At At M -

0.350 0.400 0.450 0.500

At + 0.385 At + 0.253 At + 0.126

From Dixon and Massey (9) by permission of the publishers.

Copyrig Downlo This standard is for EDUCATIONAL USE ONLY. Escuela

62

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

TABLE 29.VALUES0-5 OF t.
Degrees of Freedom
to.9i

to .976

/0.9875

to.nt

^0.9975

1 2 3 4 5

6.31 2.92 2.35 2.13 2.01 1.94 1.89 1.86 1.83 1.81 1.80 1.78 1.77 1.76 1.75 1.75 1.74 1.73 1.73 1.72 1.72 1.72 1.71 1.71 1.71 1.71 1.70 1.70 1.70 1.70 1.68 1.67 1.66 1.64
'O.OB

12.7 4.30 3.18 2.78 2.57 2.45 2.36 2.31 2.26 2.23 2.20 2.18 2.16 2.14 2.13 2.12 2.11 2.10 2.09 2.09 2.08 2.07 2.07 2.06 2.06 2.06 2.05 2.05 2.05 2.04 2.02 2.00 1.98 1.96
<0.026

25.5 6.21 4.18 3.50 3.16 2.97 2.84 2.75 2.69 2.63 2.59 2.56 2.53 2.51 2.49 2.47 2.46 2.45 2.43 2.42 2.41 2.41 2.40 2.39 2.38 2.38 2.37 2.37 2.36 2.36 2.33 2.30 2.27 2.24
to.oit*

63.7 9.92 5.84 4.60 4.03 3.71' 3.50 3.36 3.25 3.17 3.11 3.05 3.01 2.98 2.95 2.92 2.90 2.88 2.86 2.85 2.83 2.82 2.81 2.80 2.79 2.78 2.77 2.76 2.76 2.75 2.70 2.66 2.62 2.58
/0.006

127

14.1 7.45 5.60 4.77 4.32 4.03 3.83 3.69 3.58 3.50 3.43 3.37 3.33 3.29 3.25 3.22 3.20 3.17 3.15 3.14 3.12 3.10 3.09 3.08 3.07 3.06 3.05 3.04 3.03 2.97 2.91 2.86 2.81
/0.0025

6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 40 60 120
00

Degrees of Freedom

When the table is read from the foot, the tabled values are to be prefixed with a negative sign. Interpolation should be performed using the reciprocals of the degrees of freedom. 0 From Dixon and Massey (9) by permission of the publisher. 6 The values in this table were computed from percentiles of the F distribution. See Pearson and Hartley (2).

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No furthe

MISCELLANEOUS REFERENCE TABLES

63

TABLE 30.PERCENTILES OF THE xVd.f. DISTRIBUTION.0-6

Degrees of Freedom
1

h
0.5

2.5

10

90

95

97.5

99

99.5

4 5

2 3

0.000039 0.00501 0.0239 0.0517 0.0823

0.00016 0.0101 0.0383 0.0743 0.111

0.00098 0.0253 0.0719 0.121 0.166 0.206 0.241 0.272 0.300 0.325 0.347 0.367 0.385 0.402 0.417 0.432 0.457 0.480 0.517 0.560

0.0039 0.0513 0.117 0.178 0.229 0.273 0.310 0.342 0.369 0.394

0.0158 0.1054 0.195 0.266 0.322 0.367 0.405 0.436 0.463 0.487 0.507 0.525 0.542 0.556 0.570 0.582 0.604 0.622 0.652 0.687 0.726 0.774 0.839 1.00

2.71 2.30 2.08 1.94 1.85 1.77 1.72 1.67 1.63 1.60 1.57 1.55 1.52 1.50 1.49 1.47 1.44 1.42 1.38 1.34 1.30 1.24 1.17 1.00

3.84 3.00 2.60 2.37 2.21 2.10 2.01 1.94 1.88 1.83 1.79 1.75 1.72 1.69 1.67 1.64 1.60 1.57 1.52 1.46 1.39 1.32 1.22 1.00

5.02 3.69 3-12 2.79 2.57 2.41 2.29 2.19 2.11 2.05 1.99 1.94 1.90 1.87 1.83 1.80 1.75 1.71 1.64 1.57 1.48 1.39 1.27 1.00

6.63 4.61 3.78 3.32 3.02 2.80 2.64 2.51 2.41 2.32 2.25 2.18 2.13 2.08 2.04 2.00 1.93 1.88 1.79 1.70 1.59 1.47 1.32 1.00

7.88 5.30 4.28 3.72 3.35 3.09 2.90 2.74 2.62 2.52 2.43 2.36 2.29 2.24 2.19 2.14 2.06 2.00 1.90 1.79 1.67 1.53 1.36 1.00

7 8 9 10 11 12
13

0.113 0.141 0.168 0.193 0.216

0.237 0.256 0.274 0.291 0.307

0.145 0.177 0.206 0.232 0.256

0.278 0.298 0.316 0.333 0.349 0.363 0.390 0.413 0.452 0.498 0.554 0.625 0.724 1.00

14 15

0.416 0.436 0.453 0.469 0.484

0.498 0.522 0.543 0.577 0.616 0.663 0.720 0.798 1.00

20 24 30 40

16 18

0.321 0.348 0.372 0.412 0.460 0.518 0.592 0.699 1.00

60 120
00

0.611 0.675 0.763 1.00

Interpolation should be performed using reciprocals of .the degrees of freedom. From Dixon and Massey (9) by permission of the publisher. The values in the table were computed from percentiles of the F distribution. See Pearson and Hartley (2).
0 6

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en This Ecuador ESPE pursuant to License Agreement. No fu

TABLE 31.-o.os6 AND o.975 FOR RUNS AMONG ELEMENTS IN SAMPLES OF SIZES Ni AND
K0.02B
MO, 975

Nl
tf
3 4 5 7 g 9 10
11 12
15
13

10

11

12

13

14

IS

16

17

18

19

20

2 4 5 5 5

3 6 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

10

11

12

13

14

15

16

17

18

19

20

2 2 2 2 2 2 2 2 2 2
2 2

2 3 3 3 3 3 4 4 4 4 4
4 4

8 g g 9 9 9 9 9 9 9 9 9 9 9 9 9 9

9 9 10 10 11 10 11 11 12 11 12
11 11 11 11 11 11 11 11 11 11 12 12 H 13 13 13 13 13 13 H 1? 1? 13 13 13 13 14 14 15 IS 15 15 16 16 16 16 16

2 2 3 3 3 3 3 3 3 3
4 4

3 3 3 4 4 4 4 5

3 4 4

4 5 5 5 6 6 6 6
6 7

s
5 5 6 6 6 7 7 g g 8
7 7

6 6 7 7 7 7 g g 9
8 g

5 5 7 7 7 g g 8 9 9 9 9 7 g g g g 9 9 5 S S 5 S
11 11 11 12 12

14 15 15 15 16 16 17 17 17 17 17 17

15 16 16 17 17 17 18 18 18 19 19 16 17 18 18 18
19 19

14
16 17 19 20

2 2 2 2 3
3 3

5 5 6
6 6

9 9

10 10 11

H 1S 14 14 14
15 IS 15 15 IS

18 18 19 19
20 20

19 19 20
20 21

20 21
21 22

21
22 22

18

2 2 2

3 3 3

4 4 4

5 5

s s

5 6 6

6 6 6

7 7 7

9 9 10 9 10 10 9 10 10 10 10 11 10 10 11

11

ii

S
12 13

11

12

12 12

n n n

s
14

5 S

19 20 20

20 21 71

21 22 22

22 22 23

23 23 24

22 23

24 24 ?4

74 24 25 7S

25 25 26

76 76

77

Nl = #2

0 .025

MO .976

Ni = Nz 40 42

Mo .025

0. 975

20 21 22 23 24 25 26 27 28 29 30 32 34 36 38.. Copyright by
0

14 15 16 16 17 18 19 20 21 22

28 29 31 32 33 34 35 36 37

27

44 48
50

46

31 33 35 37 38 40 45 49 54 58

50 52 54 56 59 61 66 72 77 83

The values listed are such that a number less than or equal to the Wo.o26 value will occur not more than 2.5 per cent of the time and a number greater than Mo.975 will occur not more than 2.5 per cent of the time. For values of Ni and -/V2 larger than 20, a Normal approximation may be used- The mean is

55 60 65 70

and the variance is

. For example, for Ni = N2 - 20, the mean is 21 and the variance is 9.74. 9 10:11:02 EDT 2013 The 2.5 and 97.5 percentiles are 21 + 1.96 (9.74)1/2 = 27.1 and 21 Downloaded/printed by 1 2 With permission from Eisenhart and Swed 1.96 (9.74) / = 14.9. Escuela Politcnica del Ejercito en (29). Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.
This standard is for EDUCATIONAL USE ONLY.

75 63 22 39 88 24 41 80 68 93 26 43 72 85 99 28 45 77 90 104 30 47 (all95rights reserved); 82 Tue Jul 109 ASTM Int'l 100 86 115

TABLE 32. F DISTRIBUTION.0 Degrees of Freedom for Numerator

1 2 3 4 5

10

12

IS

20

24

30 .

40

60

120

00

Upper 5 per cent Points (^0.95)

1 2 3 4 5 6 7 8 9 10 11 13 14 15 16 17 18 19 20 21 22 23 24 25 30 40. 60 120
12

252 242 254 249 248 244 246 239 241 234 237 216 225 253 250 251 230 161 200 19.5 19.5 19.5 19.5 19.5 18.5 19.0 19.2 19.2 19.3 19.3 19.4 19.4 19.4 19.4 19.4 19.4 19.4 19.5 8.62 8.55 8.57 8.59 8.53 9.55 9.28 9.12 9.01 8.94 8.89 8.85 8.81 8.79 8.74 8.70 8.66 8.64 10.1 5.75 5.66 5.69 5.72 5.63 7.71 6.94 6.59 6.39 6.26 6.16 6.09 6.04 6.00 5.96 5.91 5.86 5.80 5.77 4.36 4.50 4.46 6.61 5.79 5.41 5.19 5.05 4.95 4.88 4.82 4.77 4.74 4.68 4.62 4.56 4.53 4.40 4.43

5.99 5.59 5.32 5.12 4.96 4.84 4.75 4.67 4.60 4.54 4.49 4.45 4.41 4.38 4.35 4.32 4.30 4.28 4.26 4.24

5.14 4.74 4.46 4.26 4.10 3.98 3.89 3.81 3.74 3.68 3.63 3.59 3.55 3.52 3.49 3.47 3.44 3.42 3.40 3.39 3.32 3.23 3.15 3.07 Int'l (all 3.00

4.76 4.35 4.07 3.86 3.71 3.59 3.49 3.41 3.34 3.29 3.24 3.20 3.16 3.13 3.10 3.07 3.05 3.03 3.01 2.99 2.92 2.84 2.76 2.68 rights 2.60

4.53 4.12 3.84 3.63 3.48 3.36 3.26 3.18 3.11 3.06 3.01 2.96 2.93 2.90 2.87 2.84 2.82 2.80 2.78 2.76

4.39 3.97 3.69 3.48 3.33 3.20 3.11 3.03 2.96 2.90 2.85 2.81 2.77 2.74 2.71 2.68 2.66 2.64 2.62 2.60

4.28 3.87 3.58 3.37 3.22 3.09 3.00 2.92 2.85 2.79 2.74 2.70 2.66 2.63 2.60 2.57 2.55 2.53 2.51 2.49

4.21 3.79 3.50 3.29 3.14 3.01 2.91 2.83 2.76 2.71 2.66 2.61 2.58 2.54 2.51 2.49 2.46 2.44 2.42 2.40

4.15 3.73 3.44 3.23 3.07 2.95 2.85 2.77 2.70 2.64 2.59 2.55 2.51 2.48 2.45 2.42 2.40 2.37 2.36 2.34

4.10 3.68 3.39 3.18 3.02 2.90 2.80 2.71 2.65 2.59 2.54 2.49 2.46 2.42 2.39 2.37 2.34 2.32 2.30 2.28

4.06 3.64 3.35 3.14 2.98 2.85 2.75 2.67 2.60 2.54 2.49 2.45 2.41 2.38 2.35 2.32 2.30 2.27 2.25 2.24 2.16 2.08 1.99 1.91 1.83

4.00 3.57 3.28 3.07 2.91 2.79 2.69 2.60 2.53 2.48 2.42 2.38 2.34 2.31 2.28 2.25 2.23 2.20 2.18 2.16 2.09 2.00 1.92 1.83 1.75

3.94 3.51 3.22 3.01 2.85 2.72 2.62 2.53 2.46 2.40 2.35 2.31 2.27 2.23 2.20 2.18 2.15 2.13 2.11 2.09 2.01 1.92 1.84 1.75 1.67

3.87 3.44 3.15 2.94 2.77 2.65 2.54 2.46 2.39 2.33 2.28 2.23 2.19 2.16 2.12 2.10 2.07 2.05 2.03 2.01 1.93 1.84 1.75 1.66 1.57

3.84 3.41 3.12 2.90 2.74 2.61 2.51 2.42 2.35 2.29 2.24 2.19 2.15 2.11 2.08 2.05 2.03 2.01 1.98 1.96 1.89 1.79 1.70 1.61 1.52

3.81 3.38 3.08 2.86 2.70 2.57 2.47 2.38 2.31 2.25 2.19 2.15 2.11 2.07 2.04 2.01 1.98 1.96 1.94 1.92 1.84 1.74 1.65 1.55 1.46

3.77 3.34 3.04 2.83 2.66 2.53 2.43 2.34 2.27 2.20 2.15 2.10 2.06 2.03 1.99 1.96 1.94 1.91 1.89 1.87 1.79 1.69 1.59 1.50 1.39

3.74 3.30 3.01 2.79 2.62 2.49 2.38 2.30 2.22 2.16 2.11 2.06 2.02 1.98 1.95 1.92 1.89 1.86 1.84 1.82 1.74 1.64 1.53 1.43 1.32

3.70 3.27 2.97 2.75 2.58 2.45 2.34 2.25 2.18 2.11 2.06 2.01 1.97 1.93 1.90 1.87 1.84 1.81 1.79 1.77 1.68 1.58 1.47 1.35 1.22

3.67 3.23 2.93 2.71 2.54 2.40 2.30 2.21 2.13 2.07 2.01 1.96 1.92 1.88 1.84 1.81 1.78 1.76 1.73 1.71 1.62 1.51 1.39 1.25 1.00

4.17 4.08 4.00 3.92 Copyright by ASTM 00 3.84

2.69 2.53 2.61 2.45 2.53 2.37 2.45 2.29 reserved); Tue 2.37 2.21

2.42 2.33 2.34 2.25 2.25 2.17 2.17 2.09 Jul 9 10:11:02 2.10 2.01

2.27 2.21 2.18 2.12 2.10 2.04 2.02 1.96 EDT 1.88 1.94 2013

Downloaded/printed by NOTE.Interpolation T hhhhhhhhhhhhhhhhhhhhhhhhhhh 0 Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized. This table is taken, with permission of the Biometrika Trustees, from Table 18 of Pearson & Hartley (2). This standard is for EDUCATIONAL USE ONLY.

(Continued next page)

TABLE 32 F DISTRIBUTION Continued

Degrees of Freedom for Numerator
1 2

10

12

IS

20

24

30

40

60

120

00

Upper 2.5 per cent Points (^0.975)

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

969 957 977 948 864 997 985 993 1001 1006 800 922 937 1014 648 900 963 1010 1018 39.5 38.5 39.0 39.2 39.2 39.3 39.3 39.4 39.4 39.4 39.4 39.4 39.4 39.4 39.5 39.5 39.5 39.5 39.5 14.1 17.4 16.0 15.4 15:1 14.9 14.7 14.6 14.5 14.5 14.4 14.3 14.3 14.2 14.1 14.0 14.0 13.9 13.9 8.41 8.36 8.46 9.98 9.60 9.36 9.20 9.07 8.98 8.90 8.84 8.75 8.66 8.56 8.51 12.2 10.6 8.31 8.26 6.18 6.12 6'. 23 8.43 7.76 7.39 7.15 6.98 6.85 6.76 6.68 6.62 6.52 6.43 6.33 6.28 6.07 10.0 6.02

8.81 8.07 7.57 7.21 6.94 6.72 6.55 6.41 6.30 6.20 6.12 6.04 5.98 5.92 5.87 5.83 5.79 5.75 5.72 5.69

7.26 6.54 6.06 5.71 5.46 5.26 5.10 4.97 4.86 4.77 4.69 4.62 4.56 4.51 4.46 4.42 4.38 4.35 4.32 4.29

6.60 5.89 5.42 5.08 4.83 4.63 4.47 4.35 4.24 4.15 4.08 4.01 3.95 3.90 3.86 3.82 3.78 3.75 3.72 3.69

6.23 5.52 5.05 4.72 4.47 4.28 4.12 4.00 3.89 3.80 3.73 3.66 3.61 3.56 3.51 3.48 3.44 3.41 3.38 3.35

5.99 5.29 4.82 4.48 4.24 4.04 3.89 3.77 3.66 3.58 3.50 3.44 3.38 3.33 3.29 3.25 3.22 3.18 3.15 3.13

5.82 5.12 4.65 4.32 4.07 3.88 3.73 3.60 3.50 3.41 3.34 3.28 3.22 3.17 3.13 3.09 3.05 3.02 2.99 2.97

5.70 4.99 4.53 4.20 3.95 3.76 3.61 3.48 3.38 3.29 3.22 3.16 3.10 3.05 3.01 2.97 2.93 2.90 2.87 2.85

5.60 4.90 4.43 4.10 3.85 3.66 3.51 3.39 3.29 3.20 3.12 3.06 3.01 2.96 2.91 2.87 2.84 2.81 2.78 2.75

5.52 4.82 4.36 4.03 3.78 3.59 3.44 3.31 3.21 3.12 3.05 2.98 2.93 2.88 2.84 2.80 2.76 2.73 2.70 2.68

5.46 4.76 4.30 3.96 3.72 3.53 3.37 3.25 3.15 3.06 2.99 2.92 2.87 2.82 2.77 2.73 2.70 2.67 2.64 2.61

5.37 4.67 4.20 3.87 3.62 3.43 3.28 3.15 3.05 2.96 2.89 2.2 2.77 2.72 2.68 2.64 2.60 2.57 2.54 2.51

5.27 4.57 4.10 3.77 3.52 3.33 3.18 3.05 2.95 2.86 2.79 2.72 2.67 2.62 2.57 2.53 2.50 2.47 2.44 2.41

5.17 4.47 4.00 3.67 3.42 3.23 3.07 2.95 2.84 2.76 2.68 2.62 2.56 2.51 2.46 2.42 2.39 2.36 2.33 2.30 2.20 2.07 1.94 1.82 1.71

5.12 4.42 3.95 3.61 3.37 3.17 3.02 2.89 2.79 2.70 2.63 2.56 2.50 2.45 2.41 2.37 2.33 2.30 2.27 2.24 2.14 2.01 1.88 1.76 1.64

5.07 4.36 3.89 3.56 3.31 3.12 2.96 2.84 2.73 2.64 2.57 2.50 2.44 2.39 2.35 2.31 2.27 2.24 2.21 2.18 2.07 1.94 1.82 1.69 1^57

5.01 4.31 3.84 3.51 3.26 3.06 2.91 2.78 2.67 2.59 2.51 2.44 2.38 2.33 2.29 2.25 2.21 2.18 2.15 2.12

4.96 4.25 3.78 3.45 3.20 3.00 2.85 2.72 2.61 2.52 2.45 2.38 2.32 2.27 2.22 2.18 2.14 2.11 2.08 2.05

4.90 4.20 3.73 3.39 3.14 2.94 2.7,9 2.66 2.55 2.46 2.38 2.32 2.26 2.20 2.16 2.11 2.08 2.04 2.01 1.98 1.87 1.72 1.58 ,L-43 1.27

4.85 4.14 3.67 3.33 3.08 2.8,8 2.72 2.60 2.49 2.40 2.32 2.25 2.19 2.13 2.09 2.04 2.00 1.97 1.94 1.91 1.79 1.64 1.48 1.31 1.00

5.57 4.18 3.59 3.25 3.03 2.87 2.75 2.65 30 5.42 4.05 3.46 3.13 2.90 2.74 2.62 2.53 40 5.29 3.93 3.34 3.01 2.79 2.63 2.51 2.41 60 2.52 EDT Copyright Int'l (all rights reserved); Tue Jul 9 10:11:02 20132.30 2.67 2.39 3.23 2.89 3.80 120 by ASTM5.15

00 5.02 3.69 3.12 2.79 2.57 2.41 2.29 ^.19 Downloaded/printed by S further reproductions authorized. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No

2.57 2.51 2.41 2.31 2.45 2.39 2.29 ' 2 . 1 8 2.33 2.27 2.17 ^"2.06 2.22 2.16 2.05 1.94 2.11 ' 2.05 1.94 1.83

2.01 1.94 1.88 1.80 1.74 1.67 1.61 1.53 1.48 1.39

NOTE. Interpolation should be performed using reciprocals of the degrees This ofstandard freedom. is for EDUCATIONAL USE ONLY.

TABLE 33.k FACTORS0 FOR S-N CURVES (NORMAL DISTRIBUTION ASSUMED).

P
75 90 95 T = 0.50 99

99.9

75

90

95 y = 0.75

99

99.9

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

0.773 1.498 0.739 1.419 0.722 1.382 0.712 1.360 0.705 1.346 1.337 0.701 0.698 1.329 0.694 1.324 0.693 1.320 0.691 1.316 0.690 1.313 0.689 1.311 0.688 1.308 0.686 1.307 0.686 1.305 0.685 1.303 0.684 1.302 0.684 1.301 0.683 1.300 0.683 1.299 0.683 1.299 0.682 1.298 0.682 1.297
y = 0.90

1.939 1.830 1.780 1.750 1.732 1.719 1.709 1.702 1.696 1.691 1.687 1.684 1.680 1.678 1.676 1.674 1.672 1.671 1.670 1.668 1.668 1.667 1.666

2.765 2.601 2.526 2.483 2.455 2.436 2.421 2.411 2.402 2.395 2.388 2.384 2.379 2.376 2.373 2.370 2.367 2.366 2.364 2.361 2.360 2.358 2.357

3.688 3.464 3.362 3.304 3.265 3.239 3.220 3.205 3.193 3.183 3.175 3.168 3.163 3.157 3.153 3.150 3.146 3.143 3.140 3.138 3.136 3.134 3.132

1.464 1.256 1.152 1.087 1.043 1.010 0.984 0.964 0.947 0.933 0.919 0.909 0.899 0.891 0.883 0.876 0.870 0.865 0.859 0.854 0.849 0.845 0.842

2.501 2.134 1.961 1.860 1.791 1.740 1.702 1.671 1.646 1.624 1.606 1.591 1.577 1.566 1.554 1.544 1.536 1.528 1.520 1.514 1.508 1.502 1.496

3 152 2 680 2 463 2 336 2 250 2 190 2 141 2 103 2 073 2 048 2 026 2 007 1 991 1 977 1 964 1 951 1 942 1 933 1 923 1 916 1 907 1 901 1 895
y

4.396 3.726 3.421 3.243 3.126 3.042 2.977 2.927 2.885 2.851 2.822 2.796 2.776 2.756 2.739 2.723 2.710 2.697 2.686 2.675 2.665 2.656 2.647

5.805 4.910 4.507 4.273 4.118 4.008 3.924 3.858 3.804 3.760 3.722 3.690 3.661 3.637 3.615 3.595 3.577 3.561 3.545 3.532 3.520 3.509 3.497

= 0.95

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

2.602 1.972 1.698 1.540 1.435 1.360 1.302, 1.257 1.219 1.188 1.162 1.139 1.119 1.101 1.085 1.071 1.058 1.046 1.035 1.025 1.016 1.007 0.999

4.258 3.187 2.742 2.494 2.333 2.219 2.133 2.065 2.012 1.966 1.928 1.895 1.866 1.842 1.820 1.800 1.781 1.765 1.750 1.736 1.724 1.712 1.702

5.310 3.957 3.400 3.091 2.894 2.755 2.649 2.568 2.503 2.448 2.403 2.363 2.329 2.299 2.272 2.249 2.228 2.208 2.190 2.174 2.159 2.145 2.132

7.340 5.437 4.666 4.242 3.972 3.783 3.641 3.532 3.444 3.371 3.310 3.257 3.212 3.172 3.136 3.106 3.078 3.052 3.028 3.007 2.987 2.969 2.952

9.651 7.128 6.112 5.556 5.201 4.955 4.772 4.629 4.515 4.420 4.341 4.274 4.215 4.164 4.118 4.078 4.041 4.009 3.979 3.952 3.927 3.904 3.882

3.804 2.619 2.149 1.895 1.732 1.617 1.532 1.465 1.411 1.366 1.329 1.296 1.268 1.242 1.220 1.200 1.183 1.167 1.152 1.138 1.126 1.114 1.103

6.158 4.163 3.407 3.006 2.755 2.582 2.454 2.355 2.275 2.210 2.155 2.108 2.068 2.032 2.001 1.974 1.949 1.926 1.905 1.887 1.869 1.853 1.838

7 5 4 3 3 3 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2

655 145 202 707 399 188 031 911 815 736 670 614 566 523 486 453 423 396 371 350 329 309 292

10.552 7.042 5.741 5.062 4.641 4.353 4.143 3.981 3.852 3.747 3.659 3.585 3.520 3.463 3.415 3.370 3.331 3.295 3.262 3.233 3.206 3.181 3.158

13.857 9.215 7.501 6.612 6.061 5.686 5.414 5.203 5.036 4.900 4.787 4.690 4.607 4.534 4.471 4.415 4.364 4.319 4.276 4.238 4.204 4.171 4.143

0 In which: n = sample size, p = per cent survival, and 7 = confidence level.

Table 33 was originally prepared by D. H. Shaffer of the Westinghouse Research Laboratories. After this table was prepared, Technical Report No. 34 of Nov. 1 1957, on "Tables from One-Sided Statistical Tolerence Limits" by G. S. Lieberman was received from the Office of Naval Research. This report, prepared at the Applied Mathematics and Statistics Laboratory of Stanford University under Contract N6 ONRThis standard is for EDUCATIONAL USE ONLY. 25126 (NR-042-002), contains similar tabular values for sample sizes ranging from 3 to Copyright by ASTM Int'l (all r 50, with an explanation on the construction of the tables. Downloaded/printed by 67 Escuela Politcnica del Ejercito

STP91A-EB/Feb. 1964

APPENDIX II ADDITIONAL TECHNIQUE FOR DISTRIBUTION SHAPE NOT ASSUMED RUN TESTS In addition to the rank test for two groups given in Section V A3, there is another simple test called the run test that may be used, although it is probably not as sensitive as the rank test. First arrange all individuals of both samples in ascending or descending order. Then count the number of "runs" from each sample. For example, suppose the observations from the two samples are identified by A and B, and the ordered series gives ABBAAABABB Here the number of runs is six: A - BB - AAA - B - A - BB. From Table 31, can be established whether this number of runs is too small for the observations of both samples to have been drawn from one population. If the number of runs is too small, it may be concluded that the samples are probably different.
EXAMPLEInspection of the data in the rank test example, Table 12, shows a total of ten "runs." Ranks 1 and 2 count as a run, rank 3 counts as a run, and, similarly, 4 to 6, 7, 8 to 10, and so on. According to Table 31, eight or less runs should not occur, fof* NI = 10 and Nz = 17, more than once in 40 times (2.5 per cent), on the average. In the example used, the run test does not indicate clearly that the machines are not interchangeable. However, the fact that the ten runs found are close to the critical number of eight runs implies the desirability of another check, such as already has been provided by the rank test, for example.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

68
Copyright 1964 by ASTM International www.astm.org

STP91A-EB/Feb. 1964

APPENDIX in ANALYSIS OF CORRELATION BETWEEN TWO VARIABLES "QUADRANT SUM" CORRELATION TEST , Although statistical analysis of correlation between two variables is not discussed in this Guide, the following test is very useful in determining whether or not a correlation between two variables probably exists. This simple, quickly applied test is described by Wilcoxon (28) and credited by him to Olmstead and Tukey, referring to their paper, "A Corner Test for Association," (30). After the scatter diagram has been plotted, the following procedure is to be followed:1 "Two median lines are drawn so as to divide the points into two groups of equal numbers, horizontally and vertically. The plotted points now lie in four quadrants, the lower left and upper right being taken as plus (or minus) quadrants while the upper left and lower right are taken as minus (or plus) quadrants. Commencing at the right side of the diagram, lay a ruler or pencil parallel to the vertical axis and move it to the left, counting the plotted points passed over until the next point lies on the other side of the y-median. Next lay the ruler at the bottom of the diagram parallel to the horizontal axis and move it upwards, counting the plotted points passed over before a point is reached on the other side of the z-median. In a similar manner, move in from the left, and down from the top. Four values are thus obtained to which are attached the signs of the quadrants in which they lie. The algebraic sum of these four values is called the quadrant sum and its expected value is zero if there is no association of the two plotted quantities." Table 34 gives critical values of the quadrant sum, indicating significant association at various probability levels. These critical values are almost independent of the number of points. TABLE 34 WORKING SIGNIFICANCE LEVELS FOR QUADRANT SUM.
Significance Levels, per cent
10 5
a

Quadrant Sums
9 11

Significant Levels, per cent

2 1

Quandrant Sums
13 14 to 15

Use 14 for 14 or more points, 15 for fewer points.

When it is necessary to deal with an odd number of points, one of the coordinates of a point will lie on the ^-median and one of the coordinates of another point will lie on the j-median. In this case, substitute for these points a new point, with those coordinates taken from the original points which do not involve the medians. Then proceed in the usual manner. Another difficulty arises when tied values are encountered. Some of the tied points Copyright may be on the side of the median favorable being included in the quadby ASTM Int'l (all rights reserved); Tue Julto 9 10:11:02 EDT 2013 rant sum, and one or more Downloaded/printed by of the other members of the tied group may be on the
1
This standard is for EDUCATIONAL USE ONLY.

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc Quoted with permission of Frank Wilcoxon (28).

69
Copyright 1964 by ASTM International www.astm.org

70

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

other side of the median. In this case treat the tied group as if the number of its points before crossing the median were: number favorable for inclusion in quadrant sum 1 + number unfavorable

FIG. 11.Scatter Diagram. EXAMPLE.The scatter diagram of Fig. 11 shows inclusion size plotted against specimen life to failure and suggests a possible correlation. The above test was applied and shows a quadrant sum of 16, as indicated at the right of the figure. The conclusion is that a definite correlation probably exists between the two variables.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further

STP91A-EB/Feb. 1964

APPENDIX IV THE WEIBULL DISTRIBUTION FUNCTION FOR FATIGUE LIFE* On the assumption that fatigue failures are initiated at the "weakest link," the fatigue lives of a group of specimens tested under a given set of conditions may be represented by one of a family of frequency distribution functions:

where: N = specimen life, No ^ 0 = minimum life parameter, Na = characteristic life parameter occurring at the 63.2 per cent failure point for the population [63.2 = 100(e - \/e), e = 2.718], and b > 0 = Weibull shape (or "slope") parameter. This function is a simple exponential distribution function when 6 = 1; the Rayleigh distribution function when b = 2; and a good approximation of the Normal distribution function when b = 3.57, that is, when the mean and the median values are equal. The curve representing this function (Fig. 12) is usually skewed to the right, going on to infinity, and, for b > 1, reaches zero frequency (touches the life axis) to the left of the mode, which is the life value where the highest number of failures occur. The distribution is said to have a nonzero minimum life if the curve touches the life axis at a value of life greater than 0. In other words, any specimen from the population represented by such a distribution will have zero probability of

* This description of the Weibull distribution function, as an addition to ASTM STP 91-A, was originally prepared by a Task Group in Subcommittee VI on Statistical Aspects of Fatigue of ASTM Committee E-9 on Fatigue, composed of: C. A. Moyer, chairman, Physical Laboratories, Timken Roller Bearing Co.; J. J. Bush, General Motors Research Laboratories; and B. T. Ruley, New Departure Div., General Motors Corp. It has been revised, prior to publication in its present form, by another Task Group in Subcommittee VI, composed of: John K. H. Kao, chairman, New York University; Robert A. Heller, Columbia University; B. T. Ruley, New Departures Div., General Motors Corp.; J. M. Holt, Applied Research Laboratories, U. S. Steel Corp.; M. P. Semenek, International Harvester Co.; and G. R. Gohn and Miss M. N. Torrey, Bell Telephone Laboratories, Inc. There has been a demand from the roller bearing industry for the inclusion of an additional section covering the use of the extreme-value distribution originally proposed for the analysis of fatigue data by W. Weibull (31,32). Since Fisher and Tippett (33) are often credited with first showing that this distribution was one of three limiting types of the extreme-value distribution, it is sometimes referred to as "Fisher-Tippett Type III for smallest values." As pointed byreserved); Freudenthal and9Gumbel Copyright by ASTM Int'l (allout rights Tue Jul 10:11:02 (34), EDT this 2013 distribution has some theoretical basis,by assuming that fatigue failures are examples of extreme values, Downloaded/printed that is, they are smallest-strength or weakest-link It has also been used by others This standardvalues. is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc in the analysis of life test data.
71

Copyright 1964 by ASTM International www.astm.org

72

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

FIG. 12.Typical Weibull Distribution Curves, from Kao (35). TABLE 35.ORE)INATE LOCATIO NS CORRESPON DING TO PER CENT FAIL!]D VALUES.
F(N) X 100
log

1 - PUt)

F(N) X 100
52

Iog

1
1 - F(N)

2 4 5 6 8 10 12 14 15 16 18 20 22 24 25 26 28
30

32 34 35 36 38 40 42 44 45 46. 48 50

0.0088 0.0177 0 0223 0 0269 0 0362 0.0458 0.0555 0.0655 0.0706 0.0757 0.0862 0.0969 0.1079 0.1192 0.1249 0.1308 0 1427 0 1549 0.1675 0.1805 0.1871 0.1938 0.2076 0.2218 0.2366 0.2518 0.2596 0.2676 0.2840 0.3010

54 55 56 58

60
62

63 2
64 65

66 68 70 72
74 75 76 78

84 85 86 88

80 82

90 92 94 95 96
98

0 3188 0 3372 0 3468 0 3565 0 3768 0 . 3979 0 4202 0 4341 0 4437 0 4559 0.4685 0.4949 0.5229 0.5528 0 5850 0 6021 0 6198 0 6576 0.6990 0.7447 0.7959 0.8239 0.8539 0 9208 . 1 . 000 1.097 1.222 1.301 1.398 1.699

NOTE.All logs are to the base 10. failure prior to N0 life. Later it will be shown how to test for N0 values greater than zero, but if it is reasonable to assume N0 = 0, the frequency distribution function is simplified. Since the data are usually obtained in an ordered manner in fatigue testing, it is easy to fit a cumulative distribution function to fatigue life. The cumulative function for the of population failed to life N is 2013 Copyright by fraction ASTM Int'l (all rights reserved); Tue prior Jul 9 10:11:02 EDT

Downloaded/printed by standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en EcuadorThis ESPE pursuant to License Agreement. No further reproductions au

WEIBULL DISTRIBUTION FUNCTION FOR FATIGUE LIFE

This function can be transformed into the straight-line relationship

73

which allows a simple graphical method for fitting the Weibull distribution to the data and the subsequent graphical estimation of the parameters (b, N0, and Na) in the formula.
95

FIG. 13.Construction of Weibull Probability Paper from Log-Log Paper. Construction of Probability Paper Although Weibull probability paper can be purchased from a source such as Cornell University, Ithaca, N. Y., Columbia University, New York, N. Y., or Technical and Engineering Aids for Management, 104 Belrose Ave,, Lowell, Mass., it can be constructed rather simply from square log-log paper, that is, log-log paper in which the cycles are the same size in both directions. The paper is prepared by the marking off on the vertical logarithmic scale of the probability percentages F(N) corresponding to the values of

given in Table 35. For example, in Fig. 13, the ordinate of the 90 per cent failure Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. value is 1.000 on the vertical logarithmic scale. Similarly, the ordinate for the 20

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further r

TABLE 36. MEAN -RANI C ESTI MATE 3 OF THE PER CE NT PO PULArDION i^AILEI) COR]RESPONDINCJ TO FAILUEJE ORIDER IJf SAM!PLE. Samp' e Size, n. Order No., q
No. No No No No 1 2 3 4 5

10

11

12

13

14

15

16

17

18

19 5 10 15 20 25 30 35 40 45 50

20

50.00 33.33 25.00 20.00 66.67 50.00 40.00 75.00 60.00 80.00

16.67 33.33 50.00 66.67 83.33

14.29 28.57 42.86 57.14 71.43

12.50 25.00 37.50 50.00 62.50

22.22 33.33 44.44 55.56

11.11

10.00 20.00 30.00 40.00 50.00

9.09 18.18 27.27 36.36 45.45

8.33 16.67 25.00 33.33 41.67

7.69 15.38 23.08 30.77 38.46

46.15 53.85 61.54 69 . 23 76.92

7.14 14.29 21.43 28.57 35.71

6.67 13.33 20.00 26.67 33.33

40.00 46.67 53.33 60.00 66.67

6.25 12.50 18.75 25.00 31.25

5.88 11 76 17 65 23 53 29.41
35 29 41.18 47.06 52.94 58 82

5.56 11 11 16 67 22 22 27.78
33 33 38.89 44.44 50 00 55 56

5.26 10 53 15 79 21 05 26.32
31 58 36.84 42.11 47.37 52 63

00 4 76 00 9 52 00 14 29 00 19 05 00 23 81 00 00 00 00 00 28 33 38 42 47 52 57 61 66 71 57 33 10 86 62 38 14 90 67 43

No 6

No. 7 No 8 No 9 No 10 No No No No No No No No No No
0

85.72 75.00 66.67 60.00 87.50 77.78 70.00 88.89 80.00 90.00

54.55 63.64 72.73 81.82 90.91

50.00 58.33 66.67 75.00 83.33

42.86 50.00 57.14 64.29 71.43

37.50 43.75 50.00 56.25 62.50 68.75 75.00 81.25 87.50 93.75

1 12 1 14 15 16 17 18 19 20

91.67 84.62 78.57 73.33 92.31 85.71 80.00 92.86 86.67 93.33

64.71 70.59 76.47 82.35 88.24

61.11 66.67 72.22 77.78 83.33

57.89 63.16 68.42 73.68 78.95

55 00 60 00 65.00 70 00 75 00

94.12 88.89 84.21 80.00 94.44 89.47 85.00 94.74 90.00 95.00

76 19 80 95 85.71 90.48 95.24

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed Mean-rank estimatesby = 100 Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.

This standard is for EDUCATIONAL USE ONLY.

WEIBULL DISTRIBUTION FUNCTION FOR FATIGUE LIFE

75

per cent failure line is 0.0969 on the logarithmic scale. On such paper, the tangent of the angle 6 is an estimate of the Weibull "slope," b, for the population line. The angle 0 may be measured with a protractor, or the slope of the line may be computed. Plotting Positions on Probability Paper: The fatigue data for any one sample are first ordered from shortest to longest life, each specimen being given an order number, q, from 1 through n. The horizontal plotting position is its individual life value. All runouts are assumed to have longer lives than the last ordered specimen that failed, but such data are treated separately below under "Estimates of the Distribution Function Parameters." The vertical plotting position of the per cent failed (Fig. 13) is the estimate of the per cent of the population failed, F(N), based upon the specimen order number. Mean-rank estimates of the percentages of the population failed at successive TABLE 37.TYPICAL FATIGUE TEST DATA.
Number of Revolutions to Failure Order, q
Lot 1 Lot 2 2 3 5 8 11 13

TABL,E 38.TYPI CAL FATIGUE 1PEST DATA, WITHOUT RUNG UTS. Plot of N Versus 1?(N) Nonlinear
Order, q
No. 1. . .

Specimen

Number of Revolutions to Failure

No. 1 . . 1.1 X 106 No. 2 . . 2.3 No. 3 . . 4.0 No. 4. . 6.5 No. 5. . 8.6 No. 6 No. 7 No 8

0 X 106 7 0 0 5 0 20 0 23 5

No. 2 . . . No. 3 . . . No. 4 . . . No. 5. . . No. 6 . . . No. 7. .. No. 8. . .

No. 2 No. 5 No. 8 No. 1 No. 7 No. 6 No. 3

No. 4

4.0 X 105
5.0 6.0 7.3

10.6 13.0

8.0 9.0

order numbers are given in Table 36 for sample sizes ranging from 1 through 20. Mean rank, q/(n + 1), is an unbiased estimate of F(N); such estimates are recommended by Gumbel (36) and Weibull (37). Blom(38) suggests modified mean-rank estimates. For the data given in Table 37 for the sample taken from lot 1, the abscissa for the first specimen is plotted at its life value of N = 1.1 X 106 revolutions and the ordinate at F(N) X 100 = 16.67, the plotting position for the first of a sample of five based upon mean ranks given in Table 36. Estimates of the Distribution Function Parameters: 1. An estimate of the population cumulative distribution that corresponds to the data plotted in Fig. 13 can be fqund quickly by drawing a line by eye through the failed points. More refined techniques for calculating this line can be found by referring to Gumbel (36), Lieblein (39), or Kao (35). It is possible to calculate this line by the method of least squares, as illustrated in Section V A4 of this guide. For example:

and

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduct

76

FATIGUE TESTING AND STATISTICAL ANALYSIS or DATA

Comparisons using these methods as against the graphic method sho.w,,however, that the latter is usually adequate for small samples. 2. An estimate of the characteristic life, Na , is obtained from Fig. 13 by reading off the life value corresponding to the intersection of the fitted line and a horizontal line corresponding to F(N) X 100 = 63.2 per cent. 3. An estimate of the median life is obtained by reading off the life value corresponding to the intersection of the straight line of Fig. 13 and a horizontal line corresponding to F(N) X 100 = 50 per cent.

FIG. 14.Estimation of Weibull Distribution Function Parameters for Data in Table 38. 4. In Fig. 13, the minimum life, N0 , is assumed to equal zero, since the plot of the fatigue data is approximately linear. The plotted data from Table 38 result in a line which curves downward (Fig. 14(a)); thus the existence of a finite minimum life value greater than 0 would be suspected. To find an estimate of minimum life, N0 : (1) note the life value which the curve approaches asymptotically, (2) obtain the quantity N N0 for each point by subtracting the N0 value from each individual specimen life, and (3) plot this life difference on Weibull paper versus the same per cent failed values as before. Thus, by trial and error, the best estimate of N0 will be found so that the data shown in Fig. 14(o) will, when transformed, plot as a straight line, as shown in Fig. 14(6). 5. The slope parameter, b, is equal to the tangent of the angle 6 shown in Fig. 13. Another estimate of b can be made by computing the tangent of 6 from the logarithms of the ordinates and abscissas of two widely separate points, NI and Nz, on the fitted line. Thus

estimate of b
Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013

6. The skewness of theby Weibull distribution varies with the shape parameter, Downloaded/printed This standard is for EDUCATIONAL USE ONLY. b; andEscuela the Weibull mean, general, occur at various per-cent-failed values; No furth Politcnica delin Ejercito enmay Ecuador ESPE pursuant to License Agreement.

WEIBULL DISTRIBUTION FUNCTION FOR FATIGUE LITE

TABLE 39.TYPICAL FATIGUE TEST DATA, WITH RUNOUTS.
Order, q

77

Specimen
No. 2

Number of Revolutions to Failure

No. 3 No. 4 No. 5 No. 6 No. 7 No. 8

No.'l No. 2

No. 5 No. 4 No 1 No 6 No 3 No. 7 No. 8

1.30 X 106 1.60 1.75 2.10 2.35 2.70 runout runout

FIG. 15.Per Cent Failed at Weibull Mean.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. FIG. 16. ^Estimation of del Weibull Distribution Function Parameters for Agreement. Data in Table 39. reproduc Escuela Politcnica Ejercito en Ecuador ESPE pursuant to License No further

78

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

that is, the mean does not coincide with the median. Using the estimated Weibull slope, b, it is possible to read from Fig. 15 an estimate of the per cent failed at the Weibull mean and then refer back to the estimated population line on Weibull probability paper, as in Fig. 13, to read off the estimated mean life from the curve. Gumbel (36) and Kao (35) give methods for calculating the Weibull mean1 when the characteristic life Na and the slope b are known. 7. For data containing run-out specimens (Table 39), the n' broken specimens (6 in the example, Fig. 16), out of a total of n specimens tested, are plotted on probability paper at the mean-rank plotting positions, corresponding to a sample size n (8 in the example, Fig. 16(a)). The line drawn through these points will approach a horizontal asymptote, F/racture , which is equal to the ratio of the first plotting positions corresponding to sample sizes n and n', respectively (Fig. 16(a)). The parameters of this distribution may be obtained graphically by plotting only the n' broken specimens at mean-rank plotting positions, corresponding to a sample size n' versus N N0 , where N0 is again the estimate of the vertical asymptote approached by the curve. The slope of the resulting straight line (Fig16(6)), tan 6 = b, can be obtained as described in this Section. Na , at the probability level of 63.2 per cent, is taken directly from the plotted line. The estimated equation of the probability function for the complete sample of size n will then become

where F/ = rf rac ture . The curve of Fig. 16(a) may now be replotted by using, as ordinates, Fracture times the ordinates of the straight line and, as abscissas, N0 plus the abscissas of the straight line. Note that Na is, in this case, no longer the estimate of the? characteristic life parameter of the complete distribution, F(N). The value of N at the 63.2 per cent probability of failure level may be obtained from the plot in Fig. 16(o).
1

Weibull mean:

where T the gamma function; and for Weibull variance:

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further repro

STP91A-EB/Feb. 1964
FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA REFERENCES
(1) D. J. Finney, Probit Analysis, Cambridge University Press, 1952. (2) E. S. Pearson and H. O. Hartley, Biometrika Tables for Statisticians, Cambridge Uni versity Press, 1954. (3) R. A. Fisher and F. Yates, Statistical Tables for Biological, Agricultural and Medical Research, Fourth Edition, Oliver & Boyd, London, 1953. (4) T. S. Dolan, "Certain Mechanical Strength Properties of Aluminum Alloys 25S-T and X76S-T," NACA TN914, October, 1943. (5) H. T. Corten, Todor Dirnoff, T. J. Dolan, and Masaki Sugi, "An Appraisal of the Prot Method of Fatigue Testing, Part II," Technical Report No. 35 on the Behavior of Metals under Repeated Stress, ONR Contract N6-ori 071(04), University of Illinois. June, 1953. (6) E. Prot, "Fatigue Testing Under Progressive Loading, A New Technique for Testing Materials," translated by Edward J. Ward, Captain, USAF, WADC TR 52-148, September, 1952. (7) E. J. Ward and D. C. Schwartz, "Investigation of Prot Accelerated Fatigue Test," WADC TR 52-234, November, 1952. (8) A. P. Boresi and T. J. Dolan, "An Appraisal of the Prot Method of Fatigue Testing," Technical Report No. 34 on the Behavior of Metals under Repeated Stress, ONR Contract N6-ori-71, T.O. IV, University of Illinois, January, 1953. (9) W. J. Dixon and F. J. Massey, Jr., Introduction to Statistical Analysis, McGraw Hill Book Co., 1957. (10) E. L. Crow, F. A. Davis, and M. W. Maxfield, Statistical Manual, Dover Publications, Inc., New York, N.Y., 1960. (11) The Design and Analysis of Industrial Experiments, edited by O. L. Da vies, Hafner Publishing Co., New York, N.Y., 1956. (12) D. B. Owen, Handbook of Statistical Tables, Addison Wesley Publishing Co., Inc., Reading, Mass., 1962. (13) J. A. Greenwood and M. M. Sandomire, "Sample Size Required for Estimating the Standard Deviation as a Per Cent of Its True Value," Journal, Am. Statistical Assn., Vol. 45, 1950, p. 258. (14) C. P. Ferris, F. E. Grubbs and C. L. Weaver, "Operating Characteristics for the Common Statistical Tests of Significance," Annals of Mathematical Statistics, Vol. 17, 1946, p. 178. (15) A. M. Freudenthal and E. J. Gumbel, "Minimum Life in Fatigue," Journal, Am. Statistical Assn., September, 1954. (16) R. B. Murphy, "Non-Parametric Tolerance Limits," Annals of Mathematical Statistics. Vol. XIX, 1948, pp. 581-589. (17) E. H. Schuette, "A Simplified Procedure for Obtaining Design-Level Fatigue Curves," Proceedings, Am. Soc. for Testing Mats., Vol. 54, 1954. (18) E. H. Schuette, "The Prediction of Exceedances in Limit-Value Testing," Statistical Methods in Materials Research, proceedings for a short course conducted by the Pennsylvania State University, June, 1956. (19) I. R. $air, "Table of Confidence Intervals for the Median in Samples from Any Continuous Population," Sankhya, Vol. 4, 1940, pp. 551-558. (20) W. J. Youden, "Systematic Error in Physical Constants," Physics Today, Vol. 14, September, 1961, p. 32. (21) "Tables of the Binomial Probability Distributions," Applied Mathematics Series 6, Nat. Bureau Standards, U. S. Government Printing Office, Washington, D. C., 1949. (22) W. H. Kruskal and W. A. Wallis, "Use of Ranks in One Criterion Variance Analysis," Journal, Am. Statistical Assn., Vol. 47, 1952, pp. 583-621. (23) P. G. Hoel, Introduction to Mathematical Statistics, Second Edition, John Wiley & Sons, Inc., New York, N. Y., 1954. (24) A. Hald, Statistical Theory with Engineering Applications, John Wiley & Sons, Inc. Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 New York, 1952, pp. 550-551. Downloaded/printed by (25) N. C. Severe and E. G. Olds, "A Comparison Tests on the USE Mean of a LogarithicoThis standardof is for EDUCATIONAL ONLY.

79

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

Copyright 1964 by ASTM International www.astm.org

80

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

Normal Distribution with Known Variance," Annals of Mathematical Statistics, Vol. 27, No. 3, September, 1956, p. 670. G. W. Snedecor, Statistical Methods, Fifth Edition, The Iowa State College Press, 1946. E. H. Schuette, "The Significance of Test Results from Small Groups of Specimens," Proceedings, Am. Soc. Testing Mats., Vol. 57, 1957. F. Wilcoxon, Some Rapid Approximate Statistical Procedures, American Cyanamid Co., New York, N. Y., 1949. C. Eisenhart and F. Swed, "Tables for Testing Randomness of Grouping in a Sequence of Alternatives," Annals of Mathematical Statistics, Vol. 14, 1943, p. 66. P. S. Olmstead and J. W. Tukey, "A Corner Test for Association," Annals of Mathematical Statistics, Vol. 18, 1947, pp. 495-513. W. Weibull, "A Statistical Distribution Function of Wide Applicability," Transactions, Am. Soc. Mechanical Engrs.; and Journal of Applied Mechanics. Vol. 73, September, 1951, pp. 293-297. W. Weibull, Fatigue Testing and the Analysis of Results, Pergamon Press, New York N. Y., 1961. R. A. Fisher and L. H. C. Tippett, "Limiting Forms of the Frequency Distribution of the Largest or Smallest Member of a Sample," Proceedings, Cambridge Philosophical Soc., Vol. 24, Part 2, 1928, p. 180. Reprinted in Fisher's Contributions to Mathe matical Statistics, John Wiley and Sons, Inc., New York, N. Y., 1950. A. M. Freudenthal and E. J. Gumbel, "Physical and Statistical Aspects of Fatigue," Advances in Applied Mechanics, Vol. 4, 1956, pp. 117-158. J. H. K. Kao, "A Summary of Techniques on Reliability Studies of Components Using Weibull Distribution," Proceedings, Sixth Symposium on Reliability and Quality Control, Cornell University, Ithaca, N. Y., January, 1960. E. J. Gumbel, "Statistical Theory of Extreme Values and Some Practical Applications," Applied Mathematics Series 33, Nat. Bureau Standards, Feb. 12, 1954. W. Weibull, "A Statistical Representation of Fatigue Failures in Solids," Acta Polytechnica, Mechanical Engineering Series, Vol. 1, No. 9, 1949. G. Blom, Statistical Estimates and Transformed Beta Variables, John Wiley and Sons, Inc., New York, N. Y., 1958. J. Lieblein, "A New Method of Analyzing Extreme-Value Data," Technical Note 3053, National Advisory Committee for Aeronautics, Washington, D. C., January, 1954.

(26) (27) (28) (29) (30) (31) (32) (33)

(34) (35) (36) (37) (38) (39)

SUPPLEMENTARY READING FOR APPENDIX IV

(40) J. H. K. Kao, "The Design and Analysis of Life-Testing Experiments," Transac' tions, 1958 Middle Atlantic Conference, Am. Soc. Quality Control; and Reliability Training, Inst. Radio Engrs., 2nd Edition, Chapter II, March, 1960. (41) E. J. Gumbel, Statistics of Extremes, Columbia University Press,. New York, N. Y., 1958. (42) E. J. Gumbel, "Probability Tables for the Analysis of Extreme-Value Data," Applied Mathematics Series 22, Nat. Bureau Standards, July 6, 1953. (43) W. Weibull, "New Methods for Computing Parameters of Complete or Truncated Distributions," FFA Report 58, Aeronautical Research Inst. of Sweden, February, 1955.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authori

STP91A-EB/Feb. 1964

INDEX
Estimates of parameters, single stress level, pp. 18, 40, 76 Estimates, mean rankWeibull distribution function, Table 36 (p. 74) Estimation, p. 4, Fig. 14 (p. 76) Estimation, Weibull distribution function parameters, Fig. 14 (p. 76) Fatigue, p. 2 Fatigue data (see analysis of), p. 22 Fatigue life, pp. 2, 6, 27, 39 Fatigue life for a stated value of per cent sur63) vival, p. 28 X2 distribution, percentiles of, Table 27 (p. 60) 2 Fatigue life for p per cent survival, p. 6 X ,P-32 Fatigue limit, p. 3 Choice of distribution shape, p. 40 Computations for fitting a response curve by Fatigue limit for p per cent survival, p. 6 method of least squares, p. 34, Table 16 (p. 34) Fatigue notch factor, p. 4 Computation of significance tests, Table 20 (p. Fatigue notch sensitivity, p. 4 Fatigue strength, p. 6, Fig. 8 (p. 51) 46) Computation of standard deviation of values Fatigue strength for p per cent survival at N cycles, p. 6 about fitted line, Table 17 (p. 36) Fatigue test data, Table 12 (p. 30), Table 13 (p. Confidence coefficient, p. 5 31), Table 37 (p. 75), Table 38 (p. 75), Table Confidence interval, pp. 5, 26, 27, 28, 42, Table 39 (p. 77) 9 (p. 26), Table 10 (p. 27), Table 11 (p. 28) Fatigue tests, p. 1 Confidence level, p. 5 F-distribution, Table 32 (p. 65) Confidence limits (see confidence interval) F-ratio test, pp. 45, 47 Constant amplitude tests, pp. 1, 9-13 Frequency distribution, p. 4 Constant life fatigue diagram, p. 14 Construction of Weibull probability paper from Gaussian distribution curve, Fig. 6 (p. 22) Group, p. 4 log-log paper, Fig. 13 (p. 73) Correlation between two variables, Appendix III Increasing amplitude tests, pp. 1, 13, Fig. 4 (p. 14), Fig. 5 (p. 15) (p. 69) Cycle ratio, p. 4 Interval, pp. 5, 9 Definitions, p. 2 Interval estimate, p. 5 Definitions relating to fatigue tests and test ^-factors for S-N curves, Table 33 (p. 67) Least squares, method for fitting a response methods, p. 2 curve, Table 16 (p. 34) Definitions relating to statistical analysis, p. 4 Definitions relating to statistical analysis of Level, confidence, p. 5 fatigue data, p. 6 Level, significance, p. 6 Difference among k means, p. 47 Level, tolerance, p. 5 Limits, confidence, p. 5 Difference between two means, pp. 21, 45 Difference between two standard deviations, Limits, fatigue strength at N cycles, p. 38 pp. 20, 44 Limits, method of computing, Table 19 (p. 38) Limits, tolerance, p. 5 Distribution, p.by 4,ASTM TableInt'l 27 (all (p. rights 60), Table 30 Tue Copyright reserved); Jul 9 10:11:02 EDT 2013 Maximum stress, p. 3 (p. 63), Table 32 (p. 65), Downloaded/printed by Fig. 6 (p. 22) Distribution curves, Fig.del 12 (p. 72),en Fig. 6 (p. 22) confidence for,No p. 19 Escuela Politcnica Ejercito Ecuador ESPE Mean, pursuant to Licenseinterval Agreement. further reproductions Mean, definition, p. 5 Distribution shape, choice of, pp. 22, 40 Mean, confidence limits for, Table 10 (p. 27) Estimate, p. 5 Mean fatigue life, p. 27 Estimate, interval (see confidence interval) This standard is for EDUCATIONAL USE ONLY. Estimate, point, p. 5 Mean, sample, p. 21 Abbreviations, p. 7 Analysis of correlation between two variables, Appendix III (p. 69) Allocation of test specimensprobit method, Table 1 (p. 11) Analysis of fatigue data, p. 22 Areas of the Normal curve, Table 28 (p. 61) Arithmetic mean, p. 5 Average, sample, p. 5 X2/d.f. distribution, percentiles of, Table 30 (p.
81
Copyright 1964 by ASTM International www.astm.org

82

FATIGUE TESTING AND STATISTICAL ANALYSIS OF DATA

Mean rank estimates: per cent of population "Quadrant sum" correlation test, Appendix III failed corresponding to failure order in sample, (p- 69) Table 36 (p. 74) "Quadrant sum," working significance level, Table 34 (p. 69) Means, confidence interval for, p. 42, Table 2 Range of stress, p. 3 (p-19) Means, differences between two, p. 45 Rank test, pp. 25, 30, 31, Table 25 (p. 56), Table Means, differences among k, p. 47 26 (p. 58) Mean stress, p. 3 References, p. 79 Mean, Weibull, Table 36 (p. 74), Fig. 15 (p. 77) Response curves, pp. 7, 18, 33, 34, 36, 38, Fig. 7 Median, confidence limits for, Table 9 (p. 26) (p. 35), Table 16 (p. 34) Response or survival tests, pp. 1, 10, 12, 13, Median fatigue life, pp. 6, 26 Median fatigue strength at N cycles, p. 6 Fig. 2 (p. 11) Median percentage of survivors for the popula- Rotating beam tests, R. R. Moore, Table 21 tion, Table 8 (p. 24) (p. 50) Median, sample, p. 43 Run test, Appendix II (p. 68) Runs along elements in samples of sizes NI and Medians, differences of group, p. 29 Method of least squares, pp. 34, 35, Table 16 N2, Table 31 (p. 64) Sample, p. 4 (p. 34) Minimum per cent of population exceeding me- Sample average, p. 5 Sample means, p. 21 dian of low ranking points, Table 25 (p. 56) Sample median, p. 5 Minimum stress, p. 3 Modified staircase test method, pp. 1, 13, 48, 49 Sample standard deviation, p. 5 Moore rotating beam step test, Table 21 (p. 50), Sample percentage, p. 5 Table 22 (p. 51), Fig. 8 (p. 51) Sample variance, p. 5 Mo.025 and MO.QTS for runs among elements in sam- Scatter diagram, Fig. 11 (p. 70) Selection of test specimens, pp. 16, 17, 19 ples of sizes NI and N%, Table 31 (p. 64) Nominal stress, p. 2 Significance level, p. 6, Table 34 (p. 69) Significance level, for "quadrant sum," Table 34 Normal curve, areas of, Table 28 (p. 61) Normal distribution curve, Fig. 6 (p. 22) (p. 69) Significant, p. 5 Normal distribution of fatigue hie, p. 39 Normal distribution, ^-factors for S-N curves, S-N curve for 50 per cent survival, p. 6 S-N curve for p per cent survival, p. 7 Table 33 (p. 67) Number of test specimens, minimum, pp. 16-21, S-N curves, pp. 6, 7, 17, 19, 23, 25, 40, 41, Table 8 (p. 24), Table 25 (p. 56), Table 33 (p. 67) Table 2 (p. 19), Table 3 (p. 19), Table 4 (p. 20), Table 5 (p. 20), Table 6 (p. 21), Table 7 S-N diagrams, p. 3 Staircase test method, pp. 1, 48, Fig. 3 (p. 12) (p. 21) Parameter, pp. 4, 18, 26, 40, Fig. 14 (p. 76), Fig. Standard deviation, pp. 19, 44-46, Table 2 (p. 16 (p. 77) 19) Per cent of specimens having at least the indi- Standard deviation, confidence interval for, p. 42 cated fatigue strength at 108 cycles, Fig. 8 Standard tests, pp. 1, 9, 23, 25 Statistic, p. 4 (p. 51) Per cent survival for a stated value of fatigue Steady component of stress, p. 3 Step test method, pp. 1, 13, 50, Fig. 4 (p. 14) life, p. 27 Per cent survival values at N cycles, confidence Step tests, R. R. Moore rotating beam specimens, limits for, p. 37 Table 21 (p. 50), Table 22 (p. 51) Percentiles of the x2 distribution, Table 27 (p. Stress, p. 3 60) Stress amplitude, p. 3 Percentiles of the x2/d.f. distribution, Table 30 Stress concentration factor, p. 4 Stress cycle, p. 2 (p. 63) Stress cycles endured, p. 3 Point estim_ate, pp. 5, 26-29 Population, p. 4 Stress ratio, p. 3 Probability-stress-cycle curve, Fig. 1 (p. 10) Survival tests, Fig. 2 (p. 11) Probability paper, pp. 33, 73 Symbols, p. 7 Probit testallocation of test specimens, Table Test of significance, pp. 6, 18, 20, 29, 30, 31, 45, 46, 47, Table 20 (p. 46) 1 (p- 11) Probit test data, Table 15 (p. 34) Test procedures, pp. 8, 9, 10, 12, 13, 15, Fig. 3 Probit test method, p. 10 (p. 12), Fig. 4 (p. 14T, Fig. 5 (p. 15) Copyright Int'l23 (all rights reserved); Tue specimens, Jul 9 10:11:02 EDT number, 2013 Test minimum p. 16 Prot test, pp. 1, by 15, ASTM 52, Table (p. 52), Fig. 9 This standard is for EDUCATIONAL USE ONLY. Downloaded/printed by Test specimens, selection of, p. 16 (p. 53)

Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

INDEX
Test-statistic, p. 6 Theoretical stress concentration factor, p. 4 Tolerance interval, pp. 5, 41, 42 Tolerance level, p. 5 Tolerance limits, p. 5 2-test, p. 45 /-values, Table 29 (p. 62)

83

Universe, p. 4 Unpaired rank test, p. 30, Table 26 (p. 58) Values of t, Table 29 (p. 62) Variable component of stress, p. 3 Weibull distribution, p. 71 Weibull mean, Fig. 15 (p. 77), Footnote p. 78 Wohler test method, p. 1

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions au

This page intentionally left blank

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by This standard is for EDUCATIONAL USE ONLY. Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproduc

THIS PUBLICATION is one of many issued by the American Society for Testing and Materials in connection with its work of promoting knowledge of the properties of materials and developing standard specifications and tests for materials. Much of the data result from the voluntary contributions of many of the country's leading technical authorities from industry, scientific agencies, and government. Over the years the Society has published many technical symposiums, reports, and special books. These may consist of a series of technical papers, reports by the ASTM technical committees, or compilations of data developed in special Society groups with many organizations cooperating. A list of ASTM publications and information on the work of the Society will be furnished on request.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 This standard is for EDUCATIONAL USE ONLY. Downloaded/printed by Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions authorized.

This standard is for EDUCATIONAL USE ONLY.

Copyright by ASTM Int'l (all rights reserved); Tue Jul 9 10:11:02 EDT 2013 Downloaded/printed by Escuela Politcnica del Ejercito en Ecuador ESPE pursuant to License Agreement. No further reproductions au