ENGINEERING RESEAR~CH INSTITE UNIVERSITY OF MICHIGAN ANN ARBQR FINAL REPOR~T TEFATIGUE PROPERTIES OF FLAME-PLATED flEAT-TIREATED SAE 4130 STEEL By IvI J. SINNOTT Associate Professor of Chemnical and. Metallurgical Engineering Project 2.125 L INDE AIR PIRODUC TS C OMPANY SPEEDWAY LABORATORY Oc.tober., 1953

- l - ENGINEERING RESEARCH INSTITUTE ~ UNIVERSITY OF MICHIGAN THE FATIGUE PROPERTIES OF FLAME-PLATED HEAT-TREATED SAE 4130 STEEL INTRODUCTION The purpose of this investigation was to determine the effect of Flame-Plating on the fatigue properties of a heat-treated SAE 4130 steel. Replicate rotating-beam fatigue tests were made on Flame-Plated test specimens and the results compared with the results obtained on unplated specimens of the same type of steel. SUMMARY cycles limit, cycles Flame-Plating decreases the fatigue life, which is the number of of stress to cause failure at a given stress; and the endurance defined as the maximum stress which will not cause failure in 107 of stress. The decrease in fatigue life of the Flame-Plated steel as compared to the unplated steel varies from 15 percent at the shorter cycles to failure to 25 percent at the lower stress levels. The apparent decrease in endurance limit is approximately 19 percent of the endurance limit of the unplated steel. The variance of the results at any given stress level on the plated samples is less than the variance reported for unplated steels. The standard deviation (2-sigma limits) of the fatigue life of the Flame-Plated specimens has been determined, but the mean and standard deviation of the endurance limit have not been determined. 1

I - ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN EXPERIMENTAL PROCEDURE Test Material Bars of aircraft-quality 1/2-inch-round SAE 4130 steel, all from the same heat, were used in the experimental work of this investigation. The original stock was cut into 12-inch lengths and heat-treated by quenching from 1550~F into water and then tempering for 1 hour at 800~F. This treatment resulted in a hardness of RC 40-4.1, a proportional limit of 165,000 psi, an ultimate strength of 185,000 psi, an elongation in 2 inches of 10 percent, and a reduction in area of 60 percent. These results were obtained from two previous heats of this same material which were given this treatment. Each bar in the heat used in this investigation was checked for hardness after heat treatment; the results fell in the same hardness range, and the steel will have, therefore, the same general mechanical properties listed above. Test-Specimen Preparation The heat-treated 12-inch lengths of steel were cut into 4-inch lengths prior to machining to the final test-specimen specifications. Figure 1 is a sketch of the test specimen that was used in this work. It |16 37-TAPER 8/FT I" 0.480 0 -250 D — D RILL 2 S TAPi 20 16 RELIEVE 0.0 01 4 1 R FLAME- PLATED 64 Fig..1. Fatigue-Test Specimen Specification. generally follows the suggested standard test specimen to be used in R. R. Moore fatigue testing machines, but the design has been modified slightly -J I! 2

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN to facilitate preparation. Test results obtained with this type of specimen have been shown to be comparable to those obtained with the more usual type.1 The fatigue specimens were prepared by the N. A. Woodworth Company of Detroit to the specifications of Fig. 1 and then forwarded to the Linde Air Products Company for Flame-Plating. Prior to Flame-Plating the specimens were grit-blasted with 40 mesh alundum. After Flame-Plating (the area plated is shown in Fig. 1), the specimens were returned to the Woodworth Company for finish grinding and then they were shipped to the University for testing. Profilometer measurements made on several fractured specimens showed that the average surface finish was of the order of 8 microinches. The readings on the group tested ranged from a high of 11 to a low of 6. Metallographic measurements showed the finished, ground FlamePlate to be quite uniform in thickness and of the order of 0.0025 inch. In all phases of specimen preparation, care was taken to insure identical treatment in order to minimize any variations that might be introduced in the machining, Flame-Plating, or finish grinding. Fatigue Testing Standard R. R. Moore rotating-beam fatigue machines were used to obtain data on the fatigue properties. The speed of testing was maintained at 1800 rpm during all testing and two identical testing machines were used in obtaining the data. Replicate tests were made on both machines, and the results showed that there was no significant difference in performance between the two units. The maximum probable error in stress on any given specimen was of the order of + 200 psi. The load to be applied was calculated by the usual formula, using the 0. D. of the finished Flame-Plated specimen. The usual operating procedure, of selecting successively lower values of stress and determining the number of cycles of stress to failure until a stress level is reached at which the specimen does not fracture in 107 cycles of stress. was followed. Because of the known statistical nature of any fatigue failure, replicate tests were made at each of eight stress levels in order to determine the mean value and the standard deviation. If a material shcws a fatigue limit, i.e. some stress which can be applied for an infinite peria wi-,n no fal.uire, thie abtove-des-creJ procedure will picbk up such a limit, and by cueecssive raising adi.A —. iwng 52

TABLE I Fatigue Test Data on Flame-^Plated IHeat-Treated SAE 4.130 Steel Stress, psi 100,000 90,000,,, 80,000 75,000 72,500 70,000 1000 Cycles to Failure 52 28 24 24 29 28 45 79 43 52 4.1 42 92 6.1 62 78 5.1 102 119.167 188 83.168.101 138 90 109 66 145.10,264* 86 208 137 102 10,161* 10,484* 2,376 254.10,093* 10,000* 39,400 22,300 29,600 62,600 38,100 48,900 Mean, Cycles to Failure 94,100 55,000 72,100.-Sigma Limits.179,000 96,800.132,000 145,000 76,500.105,000.184,000 85,500 125,000 65, 000 60,000 *Did not fail. 4

ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN of the stress about this limit it can be determined with reasonable accuracy. In the present investigation there were insufficient specimens available to follow this technique; as a consequence, the fatigue limit has not been too accurately defined. Table I lists the experimental data, the mean values, and the standard deviations for the Flame-Plated specimens. Table II shows the data for the unplated specimens. The log-stress versus log-cycles-to-failure TABLE II Fatigue-Test Data on HeatTreated SAE 4130 Steel Heat A Heat B Stress, 1.000 Cycles Stress, 1000 Cycles psi to Failure psi to Failure 119, 000 108,200 95, 000 90,000 86, 200 84,400 84,200 83,200 82,700 80,400 20 32 174 222 402 1216 2753 10.126*. 103.11* 10178 134,000 131, 000 108,000 95,000 89, 800 88,500 87,200 84,600 82,000 80,050 79,400 77,450 74,200 22 37 56 131 272 184 132 569 1672 12706* 12500* 10500* 11450* *Did not fail. graph of these data, the S-N plot, is given in Figo 2. Also shown on this figure are the fatigue data taken on this same type of steel given the same treatment but not Flame-Platedo These data are taken from previous work of the author and were not part of the program being reported. The data on the unplated samples were not taken in such a manner that the means and deviations could be determined. As a consequence the lines drawn through these date are merely estimates of the mean values in order to draw comparisons between these samples and the Flame-Plated samples. 5

[ - ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN 140, 130 A MEAN LIFE 120 \ 0 (n (f)S - - 0 - ~ 8CA ' AsoK 70I I 7C0 ML 104 tOO CYCLES Fig. 2. S-N Plots of Unpiated and. Flame-Plated. Test Results. Curve is Flame-Plated. Results. Limits Shown are 1-Sigma. Lower In none of the failures was spailing of the Flame-Plating evid~ent prior to failure and. only slight evid~ence of spal~ling occurred. on actual failure. Table III is a comparison of the fatigue life of the Flame-Plated. and. unplated. specimens based. on the d~ata of Fig. 2. TABLE III Estimated. Fatigue Life of Flame-Plated. and. Unplated. SAE 4150 Stee~l I Cycles to Stress, psi Stress., psi Failure Flamne-Plated. Unplated._.10,000 50, 000.100,000 200,.000.106 107.125,000 98,000 75,500 67, 000 65,, oo 145,000o.125,000 97,9000 83, 000 8.1, 500 8.1, 500 I 6

- ENGINEERING RESEARCH INSTITUTE ~ UNIVERSITY OF MICHIGAN DISCUSSION The statistical treatment of fatigue data is a relatively new development and is not too well-known except to those following the work in this field closely. In order to assist in a more 6omplete understanding of the results of the present investigation, the use of statistical methods in fatigue investigations will be described briefly, although there are more comprehensive treatments to be found in recent literature.2'3'4 In any type of measurement there is some variance or scatter that occurs in making the measurement. In fatigue testing a scatter can be expected in the cycles-to-failure data when testing at a fixed stress level, or there can be a scatter in the stress required for fracture in a fixed number of cycles of stress. The only way to obtain a measure of the scatter is by making replicate tests. From such data a mean value can be obtained and the scatter described by determining the standard deviation. A more common term used to describe the standard deviation of a series of measurements is the term "sigma limits". The 1-sigma limits define, on the basis of the variance of the data, a range such that approximately 67 out of every 100 test points will fall within this range. The 2-sigma limits are such that 95 out of every 100 test points will fall within these limits. The greater the number of tests the more reliable is the mean and deviation. This means then that in place of the usual S-N plot of fatigue data an S-N-P plot should be constructed, where P is the probability of failure and can vary from 0 to 1, the mean value being P = 0.5. In order to estimate the mean value with reasonable accuracy, at least 6 and preferably 8 or 10 specimens should be tested at any given stress level. If an estimate requiring a P of 0.01 is required, the number of test specimens should be increased to 20.2 In the work being reported, some 6 specimens were testedat 6 stress levels and the mean values and standard deviations computed. As the fatigue or endurance limit, in this work 107 cycles of stress, is approached, some of the test specimens in a given group fracture, while others run out, There is no satisfactory method of treating statistically such a mixture of data other than to reject the run out data and base the mean and deviation on the specimens that failed. This results in a mean value which is lower than the true value and is probably as conservative a method as can be used to treat the data. When data of this type are encountered, the use of a technique known as "staircase" testing is indicated. This involves varying the strs? 7

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN in the vicinity of the fatigue or endurance limit in order to determine the mean and standard deviation on the stress for a given number of cycles of stress, usually 107. A complete fatigue curve obtained by utilizing this technique is shown in Fig. 3. Unfortunately, because these statistical studies 0 0 0 (I) (n w I 112 A ---1 -Q —0-a 110 I o MEAN LIFE 1o * Ia, \ i. 2r' 108 106 104 - -- ^- _ 102 Iu0 100 — a --- --- a --- —-- 98 -- - " Mean Endurance Limit - 96 _ - 105 ro6 107 CYCLES Fig. 3. S-N Plot, Statistically Determined, of SAE 4340 Steel, Quenched and Tempered (taken from ref. 3). are relatively new, specific information on SAE 4130 steel treated as in this investigation is not available. Figure 3, however, is such a study on a quenched and tempered SAE 4340 steel and shows quite graphically the large deviations that do exist. The present investigation has determined the fatigue life of the Flame-Plated steel with reasonable accuracy, but the endurance limit is not too well defined. The curve given for the unplated steel in Fig. 2 is at best a median value, since no statistical evaluation has been made. It can be seen, however, that if 2-sigma limits are placed on this median they would overlap the 2-sigma limits of the Flame-Plated steel, but it is also evident that a significant difference does exist between the means of the two groups. 8

- ENGINEERING RESEARCH INSTITUTE * UNIVERSITY OF MICHIGAN BIBLIOGRAPHY 1. Staff of Battelle Memorial Institute, Prevention of Fatigue of Metals, John Wiley and Sons, Inc., Net York, 194.1. 2. Freudenthal, A. M., "Planning and Interpretation of Fatigue Tests", ASTM, Special Technical Publication No. 121, 1951. 5. Epremiam, E., and Mehl, R. F., "The Statistical Behavior of Fatigue Properties and the Influence of Metallurgical Factors", ASTM, Special Technical Publication No. 137, 1952. 4. Fatigue and Fracture of Metals, Technical Press of Massachusetts Institute of Technology and John Wiley and.Sons, Inc., New York, 1952. I j 9

UNIVERSITY OF MICHIGAN IIIIll3 9015 03525 0482 3 9015 03525 0482