Elevated Temperature Fatigue of Pressure Vessel Steels

Transcription

1 Elevated Temperature Fatigue of Pressure Vessel Steels Total strain range to failure and crack growth rates change only moderately with temperature and cycle rate up to 800 F BY A. W. PENSE AND R. D. STOUT ABSTRACT. The Pressure Vessel Research Committee of the Welding Research Council has sponsored a number of investigations of the elevated temperature fatigue behavior of pressure vessel steels over a period of about ten years. These programs have spanned the period during which both full life behavior and crack growth rates have been of interest for engineering evaluation of service performance, and have included both aspects in the overall results. Two pressure vessel steels, A22 Grade B (equivalent to A56 Grade 70) and A57 Type F, were involved in both aspects of the study. These two materials were given full life tests at room temperature, 400, 600, 800 and 900 F, (204, 36, 427 and 482, and crack growth rate tests at room temperature, 500, 650, 800 and 950 F (260, 343, 427 and 50. The rate of cycling in the full life tests was 3.3, 0.3, 0.03 and Hz, while crack growth rate tests were performed at 0, 0., 0.0 and Hz. The results of the investigation showed that neither temperature nor cycle rate has a very significant influence on fatigue life or crack growth in the steels up to 600 F (36. At temperatures of 900 to 950 F (482 to 50, the 5000 to 00,000 cycle life A. W. PENSE is Professor, Dept. of Metallurgy and Materials Science, and ft D. STOUT is Dean of the Graduate School, Lehigh University, Bethlehem, PA 805. Paper was presented at the 56th AWS Annual Meeting held at Cleveland, Ohio, during April 225, 975. in the 3.3 to 0.03 Hz range decreased by about 25% from the ambient 0 Hz value, although considerable variation occurred depending on steel, temperature and cycle rate. Crack growth rates for approximately the same range of conditions were increased by a factor of about 4 over the ambient 0 Hz values. Cycle rates of Hz at 950 F produced increases in growth rates by factors of 0 or more over the ambient 0 Hz values, while under similar conditions, strain range for 5000 cycle life increased by 35%. The resolution to this anomaly appears to lie in the difference between crack initiation and crack propagation behavior in the steels. Introduction In the past ten to fifteen years, a gradual but significant change has taken place in engineering design philosophy with respect to fatigue. For many years, fatigue testing programs were restricted to full life types of tests, and although a full range of cycle life vs. stress range was obtained, the emphasis was most often on the endurance limit. For the construction of pressure vessels, however, it was recognized that endurance limit behavior was not appropriate, and a design philosophy was developed that encompassed the concept of short term cycle life as being appropriate to pressure vessel service. In assessing short term life, the concepts used in high cycle life can be applied, i.e., applied stress vs. cycle life curves can be developed and stress levels can be established on the basis of the allowable stresses for 5000 to 00,000 cycle life, depending on the service life anticipated. At this level of stress, however, the yield point of the steel is normally exceeded and, at least in the 5000 cycle life range, plastic strains rather than stresses calculated from the applied loads appear to be the important physical phenomena controlling fatigue life. For this reason, strains are used as trje controlling parameter in this type of fatigue testing. In these tests the data usually follow a relationship which may be written: N CAe () where N is the number of cycles to failure, J i is the total strain range and C and m are constants. The exponent m has been related to the strain hardening exponent of the steel and has a value between 2 and 4 with the average being about 3 (Ref. ). This equation appears to hold from E o < o b. A20 A302 A57F *** A302 ^ \ A20 APOl*^ i i I I I I I I i I 0000 looooo CYCLES TO FAILURE Fig. The fatigue strain reduction factor in full scale vessel tests " " W ELDING RESEARCH SUPPLEMENT 247s

2 the 5000 to 00,000 cycle life range for carbon and low alloy steels. The Pressure Vessel Research Committee of the Welding Research Council sponsored a number of studies of this type during these years, both on laboratory specimens and on model and full size vessels. From these tests the fatigue behavior of the vessels and the laboratory specimens could be compared. On making such a comparison, it was clearly established that the cyclic life of the actual pressure vessel was substantially shorter than the laboratory specimens; i.e., for the same total strain range in the laboratory specimens and in the membrane of the vessel, the number of cycles to 2.0 t> <0 o o: a io 0.8 5o.6 h Q2 / I I I I IN I I I I I III. S C»A20A a*a302b A57F OAD Calculated * Full Scale Tests $0: Cycles to Failure Fig. 2 Comparison of full scale and laboratory test data considering crack growth only 'Clip Gage Fig. 3 Lehigh cantilever bending specimen Loading A.0" {2"zx4mrr\) failure in the actual vessel was very much less than in the laboratory specimens. Figure illustrates this difference. As may be seen from this figure, the fatigue strain reduction factor varied between 8 and 4 for 000 to 00,000 cycle life. Since the failures in the actual vessels occurred at nozzles and other stress raisers, it might be argued that these stress concentrations served to elevate the strain range above the membrane strain level and thus to decrease the fatigue life. However, investigations of the influence of stress concentrations on fatigue life have indicated (Refs. 5, 6) that, even in the presence of rather high nominal stress concentrations, well above those found in pressure vessels, the actual fatigue strain reduction factor seldom exceeds 2.5, which leaves a substantial reduction in allowable fatigue strain unaccounted for. The resolution of this discrepancy apparently lies in the difference in the way in which fatigue cracks develop and grow in the laboratory specimens and in the full scale vessels. In the laboratory specimens, an appreciable portion of the fatigue life is consumed in crack initiation, and this portion contributes to specimen life. In the full scale vessel tests, natural flaws from fabrication are normally present in sufficient numbers to eliminate the need for crack initiation, and thus the vessel life is consumed in crack propagation only. An empirical fit of full scale tests and specimen tests on this basis was found. (Ref. 6) as is shown in Fig. 2. With the growing realization of the significance of crack propagation, rather than initiation, to the fatigue life of vessels, the emphasis on fatigue testing shifted to crack growth and laws governing crack growth behavior. The Pressure Vessel Research Committee, therefore, also sponsored work in this area to determine the applicability of these laws to the various stress and temperature Loading Fig. 4 (a) Crack growth specimen used for A57F; (b) crack growth specimen used for A22B. cycles that characterize pressure vessel life. The investigations (Ref. 7) showed that, as for many other materials, crack growth could be described by the relation: da/dn = C n (2) where da/dn is crack growth rate, C and n are constants, the is the fracture mechanics parameter, the stress intensity range. The stress intensity is calculated from the gross stress on the specimen, the size of the crack in the specimen, and a factor related to specimen geometry. The value of n for the pressure vessel steels varied between about 2.2 and 5.0, with the average being about 3.5. The tests showed that the growth rates were sensitive to temperature and, of course,, but not to the material tested. This was also in keeping with results on other materials, and served as an explanation for the behavior depicted in Fig.. The fatigue strain reduction factor is not material sensitive because crack growth is not. Moreover, high values accelerate crack growth according to about a fourth power law, thus crack growth for high strain ranges is much faster than at the lower strain ranges. Since these tests have established the relevance of the fracture mechanics approach to fatigue, the problem has become one of determining whether or not fatigue behavior can be described by this approach alone, especially in smaller components. In a large pressure vessel, preexisting flaws trigger crack growth. In a smaller component, however, growth may occur only after some "initiation" stage, perhaps the number of cycles necessary for a macroscopic flaw to grow from a microscopic (for example, an intrusion) or submicroscopic defect. In this stage, which may be an appreciable portion of the fatigue life, current fracture mechanics data suggest that crack growth should be either infinitesimally slow or not occur at all. Thus we must find some alternate approach to explain or define this behavior. For such components, the total strain range to failure approach still seems valid and may be employed. One area of fatigue in which both concepts may be useful is elevated temperature service. Many types of pressure vessel and other component fatigue service occur in the temperature range between ambient and about 900 F (482. This range generally falls below that for creep in the carbon and low alloy steels, (depending, to some extent, on the strain rate) and is one in which low cycle fatigue also frequently occurs. Examples of such applications might be 248s AUGUST 975

3 shells and components used in nuclear or chemical reactor service, gas and steam turbine components and parts used in piping systems. The Pressure Vessel Research Committee has undertaken two programs in this area, one using the total strain range approach (Ref. 8) and the other using the fracture mechanics approach (Ref. 7). Both test temperature and cycle rate were test variables in the studies. Some of these data have been published previously, some have not. It is the purpose of this paper to bring the elevated temperature fatigue data from these two types of programs together to place them in perspective. Materials and Procedures Two materials were given a full series of fatigue tests involving both full life and crack growth. They were a carbon steel, A22 Grade B (equivalent to A56 Grade 70) and a low alloy steel, A57 Type F. The A22 Grade B was tested in the mill normalized and stress relieved condition while the A57 Type F was tested in the mill quenched and tempered condition. The compositions and mechanical properties of the two steels are listed in Table. It should be noted that the yield and tensile strengths of the two steels are substantially different. The full life low cycle fatigue tests were carried out using the Lehigh cantilever bend specimen seen in Fig. 3. One end of the specimen was clamped in a fixed grip and the other end given a predetermined constant deflection. The strain was measured in the center of the reduced section using either an optical gage (room temperature) or a clip gage with a strain gage bridge. The clip gage measured strain over a in. (25 mm) gage length, while the optical gage length was shorter, about in. (6 mm). The geometry of the specimen was such that the centerpoint of the gage section had nearly complete transverse restraint and thus simulated wide plate bending. The crack growth measurements were made using the modified compact tension specimens seen in Fig. 4. The specimen seen in Fig. 4(a) was used for the A57F steel tests. As may be seen in this figure, crack tip displacements are measured by transmitting the relative motion of the crack faces to a clip gage above the specimen. The plastic range compact tension specimen seen in Fig. 4(b) was used for the A22 Grade B tests and was specifically designed to prevent plastic deformation in the specimen arms during elevated temperature testing. The specimen was specially calibrated in this work (Ref. 7). The specimens were heated in air to test temperatures between room temperature and 950 F (50 using resistance furnaces. The temperature of the specimens was controlled within ± 0 F (5.5 of the air temperature during the test. The loading cycle was sinusoidal except for very slow cycle rates in the crack Table Chemical Composition and Mechanical Properties of the Steels Chemical Composition Steel Mn P S Si Ni Cr Mo Cu Other Full Life Tests: A22 Grade B 2A22 Grade B A57 TypeF 2A57 TypeF Crack Growth Tests: 3A22 Grade B 3A57 TypeF B B Mechanical Properties Steel Yield str., ksi (MPa) Tensile str., ksi (MPa) Elong. in in., % Red. of area, % A22 Grade B 3A22 Grade B A57 TypeF 2A57 TypeF 3A57 TypeF 44.3 (305) 50.6 (349) 9. (820) 8. (83) 3. (779) 72.8 (502) 77.8 (536) 3. (903) 27. (875) 23. (847) ,.2 ID H CO Cyc 0000 Cyc. A COOOO Cyc. A22 Grade B 0.30 Hz IO" 3 F ~=BOKs^/ir\I.SeMPa^^) / 0"' * A22 Grade B A57 Type F ^ =50Ksf/n(55MPo/Ti) A "" " A^A A" >80Ksi,K(88MPa*fn) «5OKsi, i(55mpajlt0 A*""^ ; A s' " O0 6CO 800 looo O 200 4O Temperature F Temperature F icr : r A O OO Temperature F Temperature F Fig Hz The influence of temperature on fatigue strain range at Fig. 6 The influence of temperature on crack growth rate at 0 Hz WELDING RESEARCH SUPPLEMENT! 2498

4 growth test (0.002 Hz). These tests were roughly square wave; loading took 25 s, the maximum load was maintained 400 s, unloading took 25 s and the minimum load was maintained 30 s. The full life tests were conducted using a load ratio of R = (fully reversed bending) while the crack growth tests were conducted using a load ratio of R = 0. (tensiontension). Rate of cycling was a variable during the tests. The full life tests were performed at cycle rates of 3.33, 0.30, and Hz. The cycle rates used in the crack growth tests were 0, 0., 0.0 and Hz. Not all materials were tested in all temperatures and cycle rates. Because of the time required for the slowest cycle rate tests, these were limited to 500, 800 and 950 F (260, 427 and 50. The cycle rates and temperatures tested for each steel are V e_ e A22 Grade B ^8 CD A57 TypeF 0 ~~".0 0. Cycle RateHz 8_ O QOI e '»80 O <D900 OCX Fig. 7 The influence of temperature and cycle rate on fatigue strain range seen in Tables 2 and 3. Results and Discussion The results of the fatigue tests are shown in Tables 2 and 3 and Figs. 58. The full life test results are shown in Figs. 5 and 6, the crack growth test results in Figs. 7 and 8. Both the effect of temperature and the effect of cycle rate are illustrated in these figures. As temperature is increased both the strain range to failure and the crack growth rate experience moderate change. Between room temperature and 900 F (482, the strain range for 5000 cycle life decreases by about 35% for A22 Grade B and 20% for A57 Type F. The 00,000 cycle life strain range decreases only an average of about 5% for the two steels. The pattern of the higher cycle life total strains is complex. In some cases (0.30 Hz) the strain range holds about constant with temperature up to 800 F (427 and decreases again at 900 F (482. The overall trend is downward. Crack growth rates for both steels increase by a factor of about 2.4 between room temperature and 800 F (427. Such a change in crack growth rate is not considered particularly large as minor environmental changes can also produce such a variation. It is especially noteworthy that the fatigue behavior of these steels is not very sensitive to temperature in air environments in the most commonly experienced portion of this range, room temperature to about 600 F (36. The average decrease in strain range over this change in temperature is about 5%. Thus, much of the current fatigue data on low alloy steels can be applied at temperatures up to 600 F (36 with only minor adjustment. The change in crack growth rate in the room temperature to 500 F (260 range is also only about 50%, which is within the scatter band for the difference between several heats of the same steel. The effect of cycle rate on fatigue behavior is also only a moderate one between cycle rates of 0 to about 0.0 Hz. For a given temperature, the decrease in strain range for 5000 cycle life is normally about 5% or less. In only one case (A22 Grade B tested at 600 F) was this exceeded. The 00,000 cycle life strain range underwent somewhat more substantial decreases, as much as 40% for A22 Grade B and 35% for A57 Type F. The average decrease was about 20% for all of the temperatures tested. Over the same range of cycle rates, three orders of magnitude down to 0.0 Hz, the crack growth rate increased by a factor of only about.5. This is a surprisingly small change in growth rate in the range of temperatures listed. At the highest temperatures and slowest cycle rates tested, the two investigations produced apparently anomalous results. The 800 F (427, Hz 5000 cycle life strain range for A57 Type F, sharply increased by about 50% over the 0.03 Hz value, while the 950 F ( Hz crack growth rate also increased by a factor of as much as 6 over the 0.0 Hz value. The increased strain range suggests better fatigue resistance in the steel while the increased growth rates suggest poorer fatigue resistance. The resolution of this conflict may be in the nature of the two tests. Table 2 Full Life Elevated Temperature Fatigue Data Total strain ranges to fa ure for steel 74 F ( F ( F ( F ( F (482 A22 Grade B at cycle rates (Hz) of A57 Grade F at cycle rates (Hz) of s I AUGUST 975

5 Initiation of a fatigue crack does not play a part in the crack growth test, but is significant in the full life tests. Increasing the temperature of test to 950 F (50 and decreasing the cycle rate may produce more ductility in the steel. Since increased ductility generally produces better low cycle fatigue resistance, the fatigue life may be increased because crack initiation is retarded. Once initiation has taken place, however, large growth rates may be experienced and subsequent failure rapid. Some support for this explanation was found in the crack growth tests. Although data are limited, the tests at 950 F (50 seemed to show a threshold for crack_growth of about 20 kskin. (22 MPav'm). Below this level, crack growth rates dropped below 5 x 0~ 7 in./cycle (.3 x 0 5 mm/cycle). At 500 F (20, no threshold was observed down to values of 2 ksijh. (3 MPa\rh). There is also some evidence that this threshold has moved up to a value close to 50 ksklh. (55 MPa\m) at the Hz rate. Thus, the two investigations may be complementary in terms of predicting fatigue behavior. Summary This investigation may be summarized as follows:. Both the allowable strain ranges for 5000 cycle to 00,000 cycle life and crack growth rates for two pressure vessel steels, A22 Grade B (A56 Grade 70) and A57 Type F, were little influenced by temperature in the room temperature to 600 F (36 range. The strain range to failure decreased by about 5% and the crack growth rate increased by about 50% over this temperature range. 2. At 900 to 950 F (482 to 50 the 5000 cycle life strain decreased by c T3 o 2 u v3 & 0 4 o L. 0" J Fig. 8 KsivTn (88 MPavfn) 950 F 500 F about 25% of the ambient value, while the 00,000 cycle life strain was not strongly decreased, the average being about 5%. The crack growth rate increased by a factor of about 4 over the same temperature range (0 Hz data). 3. The effect of cycle rate over three orders of magnitude, between 3.3 and 0.0 Hz, had little effect on the total strain range to failure at room temperature, but did decrease the """ =50Ksiv^n(55MPavfri). 950 F 500 F i i i.0 0. i «A57 Type F / Cycle RateHz 77?e influence of temperature and cycle rate on crack growth rate ^*> f _ ~~ ~ i _o" 0.00 Table 3 Elevated Temperature Crack Growth Rate Data kskin. 75 F 500 F 650 F 800 F 950 F, (MPastn) 24 (260 (343 (427 (50, kskin. (MPa Nm; 500 F ( F (50 At 0 Hz, da/dn, 2.8 X 0".2 X 0" 5.8 X 0" 3.2 X 0 (7. X 0 4 ) (3.0 X 0 3 ) (.5 X 0 3 ) (8. X 0 3 ) The effect of temperature at 0Hz on steel A22 Grade B, da/dn, 4.2 X 0" 2.3 X 0" 6.0 X 0" 3.0 X 0" 6.0 x 0" 4.6 X X 0" 5.5 X 0" (.0 X 0 3 ) (5.8 X 0 3 ) (.5 X 0 3 ) (7.6 X 0 3 ) (.5 X 0 3 ) (.2 X 0~ 2 ) (.9 X 0" 3 ) (.4 X 0 2 ) The effect of cycle rate on steel A57 Type F At 0. Hz, da/dn, 5.8 X 0" 4.0 X 0" (.5 X 0 3 ) (.0 X 0 2 ) At 0.0 Hz, da/dn, 7.2 X X 0 (.8 X 0 (.4 X 0" A57 TypeF, da/dn.5 X 0" 8.0 X 0" 2.8 X 0".2 X 0" 4.5 X 0" 2.2 X 0" 5.8 X 0" 3.2 X 0" (3.8 X 0") (2.0 X 0 3 ) (7. X 0 *) (3.0 X 0 3 ) (. X 0 3 ) (5.6 X 0 3 ) (.5 X 0 3 ) (8. X 0 3 ) At Hz, da/dn, 3.5 X 0" 3.0 X 0".0 X X 0" (8.9 X 0 4 ) (7.6 X 0 3 ) (2.5 X 0 3 ) (8.9 X 0 2 ) WELDING RESEARCH SUPPLEMENT! 258

6 5000 cycle life strain by 5% and the 00,000 cycle life strain by 20% at higher temperatures. Over the same range of cycle rates, the crack growth rate, even at 950 F (50 was only increased by a factor of At the higher temperatures, 800 and 950 F (427 and 50 and at the slowest cycle rates employed, and Hz, 5000 cycle total strain range to failure increased by about 35% over the ambient 0 Hz value, while the crack growth rate also increased by a factor of 0 over the ambient 0 Hz value for high levels. This apparent anomaly is explained on the basis that, although crack growth is accelerated under these conditions, crack initiation is suppressed. Acknowledgment The authors gratefully acknowledge the technical and financial support of the Pressure Vessel Research Committee of the Welding Research Council during this work. They also acknowledge the assistance of Dr. R. A. DePaul and Dr. H. I. McHenry in performing the research involved. Bibliography. Stout, R. D. and Pense, A. W., "Effect of Composition and Microstructure on the Low Cycle Fatigue Strength of Structural Steels," Transactions of the ASME, Journal of Basic Engineering, June Gross, J. H. and Stout, R. D., "The Plastic Fatigue Properties of High Strength Pressure Vessel Steels," Welding Journal, Vol. 34 (4), April 955, Res. Suppl., pp 6s to 66s. 3. Dubuc, J. and Welter, G., "Investigation of Static and Fatigue Resistance of Model Pressure Vessels," Welding Journal, Vol. 35 (7), July 956, Res. Suppl., pp 329s to 337s. 4. Welter. G. and Dubuc. J., "Fatigue Resistance of Simulated Nozzles in Model Pressure Vessels," Welding Journal, Vol. 36 (6), June 957, Res. Suppl., pp 27s to 274s. 5. Kooistra, L. F. and Lemcoe, M. M., "Low Cycle Fatigue Research on Full Size Pressure Vessels," Welding Journal, Vol. 4 (7), July 962, Res. Suppl., pp 297s to 306s. 6. Welter, G. and Dubuc, J., "Fatigue Resistance of Simulated Nozzles in Model Pressure Vessels of T Steel," Welding Journal, Vol. 4 (8), Aug. 962, Res. Suppl., pp 368s to 374s. 7. Hickerson, J. P., Pense, A. W. and Stout, R. D., "The Influence of Notches on the Fatigue Resistance of Pressure Vessel Steels," Welding Journal, Vol. 47 (2), Feb. 968, Res. Suppl., pp 63s to 72s. 8. McHenry, H. I. and Pense, A. W., "Fatigue Crack Propagation in Steel Alloys of Elevated Temperatures," ASTM STP No. 520, Fatigue at Elevated Temperatures, DePaul, R. A., Pense, A. W. and Stout, R. D., "The Elevated Temperature Fatigue Properties of Pressure Vessel Steels," Welding Journal, Vol. 44 (9), Sept. 965, Res. Suppl., pp 409s to 46s. WRC Bulletin 88 October 973 "Behavior and Design of Steel BeamtoColumn Moment Connections" by J. S. Huang, W. F. Chen and L. S. Beedle This investigation is concerned with beamtocolumn moment connections that are proportioned to resist a combination of high shear force and plastic moment of the beam section. A theory based upon mathematical models and physical models is developed to predict the overall loaddeflection behavior of connections. Experiments were carried out on specimens made of ASTM A572 Gr. 55 steel, with fullywelded or with bolted web attachments having round holes and slotted holes. These specimens were designed incorporating all possible limiting cases in practical connection design, and were subjected to monotonic loading. Web attachments were fastened by A490 bolts utilizing a higher allowable shear stress of 40 ksi for bolts in bearingtype connections. A good correlation between the theoretical predictions and test results was obtained. It was concluded that flangewelded webbolted connections may be used under the assumption that full plastic moment of the beam section is developed as well as the full shear strength. "Test of a FullyWelded BeamtoColumn Connection" by J. E. Regec, J. S. Huang and W. F. Chen A test program has been developed which has the objective of investigating various symmetricallyloaded momentresisting beamtocolumn connections which are of extreme importance in design and construction of steel multistory frames. This report covers the testing of the first in a series of twelve specimens a fully welded beamtocolumn connection. In this report the design procedure is presented which forms the basis for this testing series. The test procedure is given along with a stepbystep description and analysis of the stress patterns in the section. It was found that this type of connection can be used in plastic design as adequate stiffness in the elastic range was developed along with sufficient strength and rotation capacity. The AISC Specification provided adequate rules in design of such a welded connection. Publication of these reports was sponsored by the Structural Steel Committee of the Welding Research Council. The price of WRC Bulletin 88 is $4.50 per copy. Orders should be sent to Welding Research Council, 345 East 47th Street, New York, New York S I AUGUST 975