THE EFFECTS OF A MACHINING-LIKE SCRATCH ON THE FATIGUE LIFE OF 4340 STEEL

Transcription

1 THE EFFECTS OF A MACHINING-LIKE SCRATCH ON THE FATIGUE LIFE OF 4340 STEEL R.A. Everett, Jr. Vehicle Technology Center U.S. Army Research Laboratory NASA Langley Research Center Hampton, VA J.C. Newman, Jr. and E.P. Phillips NASA Langley Research Center Hampton, VA ABSTRACT The main objective of this study was to assess the effects of a machine-like scratch on the fatigue life of a high strength steel. The steel chosen for this study was 4340 steel heat treated to an ultimate strength of 210 ksi. To accomplish this, constant amplitude fatigue tests were conducted on unnotched specimens with and without a inch deep machining scratch which was machined across the specimen s surface. Specimens with scratches that had been shot peened were also tested to see if the compressive residual stresses from the shot peening would provide any relief from the stress concentration caused by a scratch. The specimen was shot peened after the scratch was machined on the specimen s surface. Tests results showed about a 40 percent reduction in the material s endurance limit due to the machine-like scratch. However, the tests on the specimens that were shot peened after a scratch was machined into the specimen s surface showed that the shot peening almost eliminated the effects of the stress concentration caused by a scratch. In this study it was also shown that a small-crack fracture mechanics analysis can predict the total fatigue life of constant amplitude fatigue tests with and without machine-like scratches. INTRODUCTION The fatigue life of dynamic components of rotorcraft has mostly been determined by a safe-life analysis where the stress versus life curves are determined from constant amplitude tests on actual parts that contain no defects or flaws. In the late 1980 s the FAA added an additional requirement to the federal air regulations (FAR s) requiring a flaw tolerant safe-life evaluation for the larger transport category rotorcraft. Since the addition of the flaw tolerant safe-life requirement, some controversy has developed over this requirement to the point where certain influential groups want this removed from the FAR s. Therefore any data that would provide insight into how materials react to existing flaws, would aid governing agencies and rotorcraft manufacturers in their decisions on the flaw tolerance of rotorcraft parts. Since rotorcraft components are often complex configurations as opposed to flat parts, it would not seem unlikely that inadvertent flaws could be put into these components during the fabrication process. Flaws from a machining process, such as a machine-like scratch, are one type of flaw that could occur. In the life management of aircraft structures at least four different approaches to certifying the structure to be airworthy are currently being used by the civil and military authorities. Military agencies employ either damage tolerance, safe life, or fail-safe methodologies. These methodologies have been used ever since aircraft have been certified. In 1988 the civil authorities introduced a new methodology for rotorcraft which is a mixture of safe-life and a "type" of damage tolerance. This was introduced to address some of the in-service problems that had begun to appear in rotorcraft that were a result of either manufacturing defects or in-service defects. This new methodology is known by several names. Among them are the "flaw tolerant safe-life" and "enhanced safe-life". Sometimes it is even called damage tolerance, although classic damage tolerance as started by the U.S. Air Force in the early 1970's assumes that a pre-existing fatigue crack exists in the structure while the flaw tolerant method assumes a flaw (not a sharp fatigue crack) pre-exists in the structure. The flaw tolerant method still uses the Palmgren/Miner rule to establish retirement lives with the applied stress versus life cycle to crack initiation curves, S/N curves, coming from tests on structures that have pre-existing flaws such as nicks, dents, scratches, and corrosion (to name but a few of these flaw types). Presented at the American Helicopter Society 55 th annual Forum. May 25-27, 1999, Palais des Congres, Montreal, Quebec, Canada. Copyright by the American Helicopter Society, Inc. All rights reserved.

2 Because of the success that the rotorcraft community has shown with the safe-life method, the flaw tolerance method was thought to be an acceptable method to handle the previously mentioned manufacturing and in-service defect problems. However, much controversy currently exists with this method to the point where the Technical Oversight Group for Aging Aircraft, TOGAA, in its review of civil rotorcraft fatigue substantiation practices, is recommending that the flaw tolerance method be removed from FAR (Fatigue evaluation of structures for rotorcraft over 7,000 pounds). Currently, no rotorcraft as been fully certified using flaw tolerance with very little data in the literature to support the success of the flaw tolerant method. On the other hand, the U.S. Air Force has had almost thirty years of success using the classic crack-growth damage tolerance method. One of the principle attributes of the damage tolerance philosophy in the fatigue life management of aircraft structures is that a flaw is assumed to exist in the structure at the beginning of the service life of the aircraft. Inspection intervals are then calculated using fracture mechanics concepts assuming a flaw of a certain size exists in the pristine structure (ref. 1). Compared to this design methodology, the safe life design method calculates a retirement life normally using the Palmgren-Miner rule (ref. 2 and 3), which assumes the structure is free of any defects. In the course of a structural review of a helicopter which experienced a fatal crash, one of the fatigue design curves appeared to have one of the data points that defined the design curve as coming from a test article which had a preexisting flaw in the test article. During the design review this flaw was defined as a surface scratch which occurred during the machining process of the test article. Since the safe-life fatigue design methodology assumes no flaws are present in the structure when a retirement life is calculated, it was felt that this design test data point should not have been included in defining the design curve. As a result of this seemingly anomaly in the fatigue design curve a study was initiated to explore the effects of a machine-like scratch on fatigue life. The test program used for this study consisted of constant amplitude fatigue tests on 4340 steel. A set of control tests were run on specimens with no flaws. These fatigue lives were compared to tests on specimens that had a machine-like scratch machined onto the specimen surface. Since helicopter components that have a finite retirement life are usually shot peened, fatigue tests were also run on specimens that had been shot peened as well as specimens that had been shot peened after a machine-like scratch was machined on the specimens surface. Finally, a newly emerging fatigue analysis tool called FASTRAN, which uses only crack-growth data to predict fatigue life, is used to predict the fatigue life of the no-flaw test results as well as the simulated machine-like scratch test data. TEST PROGRAM To assess the affect of machine-like scratches on fatigue life, constant amplitude fatigue tests were conducted on the same material used in the helicopter where the machine-like scratch test data point appeared on the fatigue life design curve. This section describes the material, test specimen configuration, constant amplitude tests, shot peening, and the machine-like scratch put on the specimen surface. Material and Test Specimen Configuration The material used for this study was 4340 steel heat treated to an ultimate strength of 210 ksi. The fatigue endurance limit for this heat of material was determined to be about 68 ksi at a stress ratio, R = -1. This agreed with the value given in the Military Handbook 5B (ref. 4). Specimens were machined to have a surface finish of 32 rms which is the same finish used on the helicopter parts that were under investigation, The nominal thickness of the test specimens was 0.35 inches. Specimens were machined to an hour glass shape producing an elastic stress concentration factor, K T, of 1.03 as determined by the boundary force method (ref. 5). Figure 1 shows the test specimens configuration. Constant Amplitude Tests All constant amplitude tests were conducted in servohydraulic, electronically controlled test stands at a stress ratio, R, of minus one. The tests were run at a cyclic frequency of 10 hertz and loads were controlled to within 1 percent. All fatigue test lives reported herein were to specimen failure.

3 Shot Peening and Scratch Dimensions The shot peening procedure used on the specimens was done by the original manufacturer of the helicopter under investigation. This was considered to be the only way to have the specimen surface duplicate the condition of the actual helicopter part. X-ray diffraction measurements of the compressive residual stresses produced by the shot peening ranged from 60 to 90 ksi. The compressive residual stresses reached a zero stress level at about inches below the specimen surface. The machine-like scratch was machined into the specimen surface using an end mill. The scratch was machined across the entire width of the specimen, but only on one side of the specimen. The depth of the scratches was nominally inches with the width being about twice the depth. Each specimen scratch was measured to insure uniformity in geometry of the scratches for the test specimens tested in this study. Specimen Types Four different specimen types were used in the test program. For the baseline data, specimens were machined as pristinely as possible to provide fatigue life test data of specimens with no machining flaws. A second set of specimens were machined to the proper specimen geometry, then a machine-like scratch was machined on the specimen surface with the dimensions and procedures stated previously. A third set of specimens were shot peened after being machined. Finally, a fourth set was machined to the hour-glass geometry, then the machine-like scratch was machined on the specimen surface followed by shot peening. TEST RESULTS To assess the effect of machining scratches on the fatigue life of high strength steels, constant amplitude fatigue tests were conducted on unnotched specimens with and without the machine-like scratch. The scratch was machined onto one side of the specimen to a nominal depth of inches. Since the dynamic components of helicopters are often shot peened, a series of tests were also conducted on specimens that were shot peened after the scratch was machined onto the specimen surface. A few tests were also run on specimens that had been shot peened with no surface scratch. Figure 2 shows the results of the fatigue tests on the pristine specimens as well as the specimens that had been shot peened with no surface scratch (symbols with arrows indicate a runout, test stopped before failure.) The results of these data showed a definite increase in the fatigue life of the baseline material as a result of shot peening. Based on the limited amount of tests, the endurance limit of the baseline material appears to be about 10 percent higher with a shot- peened surface. It is also noted that the baseline material endurance limit of about 68 ksi agrees very well with that found in Military Handbook 5B (ref. 4). A general increase in fatigue life at all the stress levels tested is noted from the data shown in figure 2. The shape of the two S/N curves seems to be similar, but with the limited amount of tests run this similarity can only be assumed. As stated previously, the main objective of this work was to assess the effect of a machine-like scratch on the fatigue life of high strength 4340 steel. Figure 3 shows the results of the fatigue tests on the specimens with machine-like scratches compared to the pristine specimens. As seen in this figure the endurance limit was reduced from 68 ksi for the pristine specimen to about 40 ksi for the specimen with a inch deep scratch. This is a decrease of about 40 percent. A scratch of this size could easily be missed by an optical inspection of a helicopter component. With this magnitude of reduction in the endurance limit resulting from a fairly small machine-like scratch, it can be seen that some fabrication procedure is needed to protect against this possible loss in fatigue strength. As somewhat of a standard practice, most helicopter manufacturers shot peen their life limited dynamic components after machining. Figure 4 shows the effect on the fatigue life of specimens that have been shot peened after a scratch was machined onto the specimen surface. As shown by the test data, the fatigue life of the specimens containing a scratch with shot peening is almost the same as that of the pristine specimen. The small loss in the endurance limit shown here would be more than compensated for by the several ways the helicopter industry reduces the mean curve to the design working curve (ref. 6). Hence, shot peening is potentially a successful method in keeping the fatigue strength of a material at its pristine value when a fabrication flaw may be present in one of the helicopters components.

4 ANALYSIS Until just recently almost all rotorcraft dynamic components have been designed and life-managed using the safelife nominal stress method. As is stated in FAR "it must be shown that the structure is able to withstand repeated loads of variable magnitude without detectable cracks " over several differently defined time intervals. In fact these analyses are conducted assuming the structure to be free of fatigue cracks and other types of flaws. Only since the late 1980's has a criteria for rotorcraft certification been used which assumes flaws are present in the as-manufactured structure. As previously stated, in the introduction of this paper the flaw tolerant method for establishing retirement lives still uses the Palmgren/Miner linear cumulative damage rule. Without a large amount of conservative assumptions this design method is not capable of determining the fatigue life of a structure using only analytical means. Since about the mid-1980's a trend has been developing to predict total fatigue life (from the first load cycle to failure) using only fatigue crack growth considerations (Newman and his co-workers, ref. 7). In order for this to be accomplished a very small initial crack size ( to inches) has to be assumed to exist in the structure after the manufacturing process. Furthermore, as shown by numerous investigations (ref. 8-10) these very small cracks have crack growth characteristics that are considerably different than large cracks (cracks longer than 0.08 inches). In fact, these small cracks are considered to exhibit a "small-crack effect" as illustrated in Figure 5. The growth of these small cracks when described by linear-elastic fracture mechanics (LEFM), grow faster than long cracks at the same stress intensity factor range and grow at stress intensity factors below the long-crack threshold (ΔK th ). A review of the concepts involved in "small-crack" theory is given by Newman in reference (ref. 11). After conducting the experimental portion of this study, it was decided to investigate the ability of the small-crack crack theory to predict the fatigue life of the test specimens that contained the machine-like scratch. Since several studies by Newman and his colleagues (ref ) had shown a crack growth analysis using small-crack considerations could predict total fatigue life using only a crack growth analysis, it was logical to try to employ these concepts to predict the fatigue life of the test specimens with the machine-like scratches. In Newman's studies he used a crack-closure based (ref. 15) model along with small-crack growth characteristics to predict total fatigue life (ref. 11). While the work in reference 15 illustrates analysis only on aluminum alloys, this method has also been shown to work on high aircraft quality steels (ref. 12). As a first step in the analysis of the machine-like scratch, it was decided to use the small-crack analysis to predict the life of the as-manufactured test specimens in order to check the basic material data input into the crack-growth computer code. The computer code used for these studies known as FASTRAN (ref. 16) requires several parameters that are a function of the material one is analyzing. As stated previously, perhaps the most important parameter is the initial crack size. The long and small crack characteristics of the 4340 steel used in this study were thoroughly investigated as part of an AGARD study done during the 1980's (ref. 12). In the study conducted by Swain et al, examination of 35 crack initiation sites described the distribution of initiation sites as shown in Figure 6. The dominate initiation site was a spherical (calcium-aluminate) particle as shown in Figure 7. The mean defect size was determined to be a radius of about inches. From Figure 7 it is seen that defects found in this study ranged from a radius of about to inches. Most of the analysis done previously using FASTRAN has been done on specimen configurations that contained a hole with an elastic stress concentration, K T, of about three (ref ). However, as shown in figure 1 the test specimen used in this study has a nominal K T of one. In a K T = 1 stress distribution a much larger volume of material is available for the crack to initiate. In this larger volume of material, it is probable that a larger defect would exist that could initiate the crack. Newman showed this possibility in his analysis on the aluminum alloy 2024-T3 where his analysis required a inch radius crack to fit the constant amplitude test data for stress ratios, R = 0 and -1. In a K T = 3 stress field work done Newman showed an inch radius crack was needed to predict total fatigue life (ref. 11). However, for the analysis conducted on the K T =1 geometry used in this study a fatigue crack of inches was used to fit the test data as shown in Figure 8. While this prediction is not perfect, it must be remembered that one cannot predict total fatigue life from a Palmgren/Miner analysis perfectly. In fact, the safe-life analysis using the Palmgren/Miner linear cumulative damage rule as used by the rotorcraft industry, has been shown under at least two AHS round robins (ref. 17 and 18) to predict what the industry calls retirements lives that cannot predict the test data within a factor of ten. It has been said by different rotorcraft

5 industry analysts that a component with a mean life of 200,000 hours would have a design life of 2,000 hours. This paper is not written to debate the pros and cons of a Palmgren/Miner analysis compared to a crack-growth analysis, but to show that a newly emerging analysis tool maybe available in the near future. In order to predict the total fatigue life of the machine-like scratch test specimen two approaches were investigated. First, it was thought to model the scratch like a corner or surface crack and use the stress intensity solution for these crack geometries. The applied stress would be assumed to be a K T times the applied load. However, one of the weaknesses of this approach would be that the fatigue crack would be in the K T stress field from initiation until failure. The other problem in this approach was to determine the proper K T to be used. The K T that best fit the test results was K T = 2 (see figure 9). This was determined from a solution developed by Neuber for very sharp notches. Finally, the scratch was modeled as a single-edge crack under a tensile loading. With this option chosen in FASTRAN and using the average depth of the scratch, inches, from the several test specimens as shown in Table 1, and using the applied stress as the applied load divided by the cross-sectional area of the test specimen, the prediction is shown in Figure 10. Here, the agreement between the FASTRAN analysis and the test data is very good. CONCLUSIONS 1. Shot peening of high strength 4340 steel produced a higher endurance limit by about 10 percent. 2. A machine-like scratch in high strength 4340 steel reduced the endurance limit by about 40 percent. 3. Shot peening a material that contains a machine-like scratch restored the endurance limit of the material to within about 10 percent of its original value. 4. A crack-growth closure based analysis using small-crack theory predicted the total fatigue life of test specimens with a machine-like crack with very good accuracy. 5 REFERENCES 1. Gallagher, J.P., Giessler, F.J., and Berens, A.P., "USAF Damage Design Handbook, Guidelines for the Analysis and Design of Damage Tolerant Aircraft Structures," Air Force Wright Aeronautical Laboratory, AFWAL-TR , May Palmgren, A., Ball and Roller Bearing Engineering, translated by Palmgren and B. Ruley, SKF Industries, Inc. Philadelphia, 1945, pp Miner, M.A. "Cumulative Damage in Fatigue, Journal of Applied Mechanics, ASME, Vol. 12, Sept Military Standardization Handbook, MIL-HDBK-5B, August Tan, P.W., Raju, I.S., and Newman, J.C., "Boundary Force Method for Analyzing Two-Dimensional Cracked Plates," ASTM STP 945, Feb Thompson, G.H., "Boeing Vertol Fatigue Life Methodology," American Helicopter Society, National Specialists' Meeting on Helicopter Fatigue Methodology, St. Louis, Mo., March Newman,J.C., Jr., Swain, M.H. and Phillips, E.P.,"An assessment of the small-crack effect for 2024-T3 aluminum alloy", Small Fatigue Cracks, The Metallurgical Society, Inc., Warrendale, PA, , Pearson, S. "Initiation of fatigue cracks in commercial aluminum alloys and the subsequent propagation of very short cracks", Engng.Fract.Mechs., 7, , 1975.

6 9. Zoecher, H. (ed.) "Behavior of Short Cracks in Airframe Components" AGARD CP- 328, Miller, K.J. and de los Rios, E.R. (eds.) The Behavior of Short Fatigue Cracks, European Group on Fracture, Publication No. 1, Newman, J.C., Jr. "Fracture mechanics parameters for small fatigue cracks", Small Crack Test Methods, ASTM STP 1149, J. Allison and J. Larsen (eds.) 6-28, Swain, M.H., Everett, R.A., Jr., Newman,J.C., Jr., and Phillips, E.P.,"The growth of short cracks in 4340 steel and aluminum-lithium 2090", AGARD R-767, P.R. Edwards and J.c. Newman, Jr. (eds.) , Everett, R.A., Jr., "A comparison of fatigue life prediction methodologies for rotorcraft", 7Journal of the American Helicopter Society, Vol. 37, (2), April Newman, J.C., Jr., "Fatigue-life prediction methodology using crack-closure model", Journal of Engineering Materials and Technology, 117, , Newman, J.C., Jr., "A crack-closure model for predicting fatigue crack growth under aircraft spectrum loading", Methods and Models for Prediction Fatigue Crack Growth under Random Loading," ASTM STP 748, J.B. Chang and C.M. Hudson (eds.), Newman, J.C., Jr., "FASTRAN-II - A fatigue crack growth structural analysis program", NASA TM , Arden, R.W., "Hypothetical fatigue life problems", American Helicopter Society, National Specialists' Meeting on Helicopter Fatigue Methodology, St. Louis, Mo., March Everett, R.A., Jr., Bartlett, F.D., Jr., and Elber, W., "Probabilistic fatigue methodology for safe retirement lives," Journal of the American Helicopter Society, Vol. 37, (2), April Table 1. Crack Depth Measurements Specimen No. Left Edge Right Edge (inches) (inches)

7 R Figure 1. Test specimen configuration Alternating Stress, S a, ksi pristine specimen shot peened Runout Fatigue Life, N, cycles Figure 2. Fatigue Life of Pristine and Shot-Peened Specimens.

8 120 Alternating Stress, S a, ksi Runout 20 Pristine pristine specimen scratch specimen Fatigue Life, N, cycles Figure 3. Fatigue Life of Pristine and Scratch Specimens Alternating Stress, S a, ksi pristine specimen Scratched and Peened Runout Fatigue Life, N, cycles Figure 4. Fatigue Life of Pristine and Scratch/Peened Specimens.

9 R = constant S 1 < S 2 < S 3 da/dn or dc/dn Small cracks S 1 Small crack from hole S 2 S 3 Large crack from hole Large crack Large crack (ΔK-decreasing test) Steady state ΔK th ΔK Figure 5. Typical fatigue-crack growth rate data for small and large cracks. Cumulative distribution function Swain et.al. [13] 4340 Steel B = 0.20 in. w = 1.0 in. r = in. K T = Equivalent semi-circular defect radius, μm Figure 6. Cumulative distribution function for initiation sites in 4340 steel.

10 Figure 7. Crack initiation site at spherical-inclusion particle in 4340 steel

11 120 Alternating Stress, S a, ksi Runout 20 pristine specimen FASTRAN, K T = Fatigue Life, N, cycles Figure 8. FASTRAN prediction of pristine specimens fatigue life. 120 Alternating Stress, S a, ksi Runout 20 pristine specimen FASTRAN Predictions, KT = 2 Scratch Test Data Fatigue Life, N, cycles Figure 9. FASTRAN prediction of Scratch specimens fatigue life assuming K T = 2 for scratch.

12 120 Alternating Stress, S a, ksi Runout 20 pristine specimen Scratch Test Data FASTRAN, Edge Crack Fatigue Life, N, cycles Figure 10. FASTRAN prediction of Scratch specimens fatigue life modeling scratch as an edge crack.