Material Aspects of Impact Fatigue of Valve Steels R

Purdue University Purdue e-Pubs

International Compressor Engineering Conference School of Mechanical Engineering

1978 Material Aspects of Impact Fatigue of Valve Steels R. Dusil

B. Johansson

Follow this and additional works at: https://docs.lib.purdue.edu/icec

Dusil, R. and Johansson, B., "Material Aspects of Impact Fatigue of Valve Steels" (1978). International Compressor Engineering Conference. Paper 254. https://docs.lib.purdue.edu/icec/254

This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact [email protected] for additional information. Complete proceedings may be acquired in print and on CD-ROM directly from the Ray W. Herrick Laboratories at https://engineering.purdue.edu/ Herrick/Events/orderlit.html MATERIAL ASPECTS OF IMPACT FATIGUE OF VALVE STEELS

Robert Dusil and :Sorje Johansson Steel Research Center, Uddeholms Aktiebolag S-683 01 Hagfors, Sweden

UTRODUCTION Tensile fatigue testing was done on a high frequency Amsler pulsator, Fraotographic TyPe 2HFP, testing studies of service failures of com frequency 70 - 90 Hz, maximum pressor valves showed load 20 000 N. The that the mechanical limita pulsator works on the resonru1ce principle tions of the material are and is primarily associated equipped with an electromagnetic system vhich with the conditions which prevail above the valve provides the d:r:iving force, Specimen shape is seat (1-3).Impact stresses induced when the valve shown in Fig. 3, component hits the seat or valve stop were ana lysed and several investigators indicated that despice the fracture occurence, the calculated or measured stresses were lower than fatigue strength of the valve material (3 -G). Presented study was conceived to determine the applicability and the limitations of the valve material fatigue strength at impact loading. Beside impact fatigue, a tensile fatigue test was performed. Two standard high strength valve steels were tested, UHB 20 and U}lli Stainless 716, the first one was hardened ru1d tempered to tensile strengths of I 2400 - 300 J11N/m2 (200- 330 ksi). The fractographic features that allow distinction between various loading modes are diGcussed. Using the Weibull concept of criti cally stressed volume an attempt is done to esti Jnate the magnitude of impact stresses in the tested specimens.

TESTING PROCEDURE

Fig. I In order to simulate the dynamic performance of - Impact fatigue testing equipment, compressor valves the testing was made in a special testing equi.~Jment, Fig. I. The specimen, dimension 100 x 20 x 0.38 mm, see Fig. 2, was operated by short-duration compressed air pulses generated by a fluidic device. The reed was repeatedly lifted The specimens were blanked. from the strip parallel from and striked against the seat. The operating with the roL:.ing direction, the edges were groli..'l:J. frequency was 250 Hz which is equivalent to the and polished. No surface prepaxation was made, specimen natural frequency. When the reed hits the seat, the impact intensity was detected by a piezo The .impact and tensile fatigue tests were carried electric accelerometer. The calibration of accelero out at a room temperature of about 20°C in a dry, meter, see Appendix, was made according to ref (6) nop-corrosiv·e atmosphere, The 7 fatigLJ.e limits at where more details on testing apparature are given. 10 loading cycles were determined according Impact. to intensity is the accelerometer signal, uni·" stair-case method (7). The fatigue limit is is m/s2 de , In agreement with published data, no units fined as impact intensity or tensile stress ampli were specified in the present paper. The testing tude at which 50 %of the specimens failed equipment 7 within was stopped by optical fracture detector 10 cycles, The experimental scope comprised 30 when a small fragment was torn off. Some construc specimens. To plot S-N curves, additional 50 speci tional ;nodif.ications were made to enable the con mens were tested. tinuous monitoring of the impact intensity,

116 S-N DIAGRAl"!S

The impact testing results of UI-IB 20 w1d UHB Stain- less 716 are shown in Fig, 4 and 5. In both cases the fracture_probability for 10, 50 and 90% are 1 given. At 10 impact cycles S-N curves become asymptotic, The stainless steel exhibited higher It impact fatigue strength compared with UHB 20, ex should be emphasized thac tested valve steels hibited a low scatter, li/r IMPACT INTENSITY a PROBABILITY OF FRACTURE 3

2 ~ 1- 90% ~ 50 10

Fig. 2 - Broken impact speci Fig. 3 - Tensile mens, .An early stage (left) fatigue specim8n. w1d late stage of impact fracture (right). NUMBER OF IMPACT CYCLES 1'11\.TERIALS 2 Fig, 4 - S-N curves for URi3 20, UTS 1 930 'Jil'l.lijm loading, Two stanJ.ard hardened and tempered valve steels (280 ksi), impact were used, UHB 20 and UHB Stainless 716, strip thickness 0,)8 Dm (0.015 in). The strip materials originating from two different heatp were tested H1PACT I NTES ITY PROBABILITY OF FRACTURE is given a for each grade, The chemical composition 3 below:

Table 1 - Chemical composition, 2 Grade U1ill wt % c Si Mn Cr Mo ' ~ """-..... ;o% 18 20 heat A 1,01 0 25 .45 .14 "" """' B 1,00 ,20 .39 .16 stainless 716 A .38 .33 .45 13.3 1.02 1 B -37 -41 .48 13.5 1.00

Static tensile properties and hardness of test ma-terials: lol NUMBER OF IMPACT CYCLES Table 2 - Tensile properties and hardness. Fig. 5 - S-N curves for UHB Stainless 716, UTS 2 Grade U1ill Tensile Yield Elong. HV 1 860 MlJ/m (270 ksi), impact loading. strength strength gauge 100 N 2 "MN/m2 10 mm To make sure that there is a significant difference MlJ/m of % between the measured impact fatigue ·intensity URB 20 and UHB Stainless 716, the results were by 20 1220 1200 7.0 385 treated statistically. The method proposed 5.5 440 Welch-Aspin adopted for the stair-case procedure 1430 1370 show 1600 4.5 505 was used (s). The results of this calculation 1680 values 1930 1750 4.0 560 that there is a difference between the mean 1850 3.5 565 the confidence level of 99.5 %. It was there 1960 on of 2240 2050 2,0 635 fore concluded that the impact fatigue strength than that for TJHB 20 Stainl. 716 1860 1500 6.5 570 UHB Stainless 716 is higher 1920 1510 6,0 570 on the significance level of 0.5 %. Many practical applications proved that UHB Stainless 716 ex-

117 hibited superior impact fatigue strength compared certain stress level will be more with the c~bon steel. retarded in this material thru1 in UHB 20. This is the pro bable explanation to the superior impact fatigue tt is known that crack initiation stage takes the main strength eYJlibited by valve steel UHB stainless part of the total fatigue life in high 716. strength materials exposed to dynamic loading (1, 9, 10). The initiation stage depends on stress EFFECT field intensity OF TENSILE STRENGTH ON IMPACT AND TENSILE at the tip of the crack or defect. FATIGUE LIJ11IT In practice, whenever the stresses exceed the yield strength of the material a region of plasticity The effect of mechanical properties is developed near the crack tip. on valve An estimate of the material fatigue response was size of this zone, Fig. 6, for plane examined. Simul stress con taneously with impact fatigue experiment, dition prevailing in valve strip material the (1,9) may tensile fatiguing was performed on U1lll 20. The be obtained from (10) 7 number of loading cycles was 10 at both loading modes. The results are 1 presented in Fig. 7. r y ""-2n eq. 1

ESTIMATED MAXIMUM plastic zone radius IMPACT IMPACT STRESS INTENSr]-TY-,---~--r--~-..,..--,....---r-, stress intensity factor MN/m2 KSi yield strength I 1800 260 IMPACT rAi!Gl!E r-t---j----j----j-----t-::-T---t=...q---t-11600_J_

220 IV~~-""" " 11100 I 1.2 ~4-----~--~--~F----r----r----T~ I \ tVl 1zoo l8o \ 1.0 11 ·-~. Ol.lO UTS +0. S2 mmE STAND DEY 0.012 (T, ~~~~7~~ __K_ AMPLITUDE' CORR CO OFF 0, 996 y-y2n, MN/m2 1-"'T""--r--...,..--t---t---+----r~ STRESS MNfm2KSI 700!700 ~+----t~----~r---~~-t-----T----r--1 l ~00 200

000!600 TENs I lE FAT.,]G::_::"~t-E --~-,.+---:__...-=fUt-u.!=:::::::J=,_l+--ir-1 180 --- I Ar- ll! 1 lzoo J .... V 16o soo!5oo ~-+--+---t>_._L_._.r __ r------r--~r-----r------t-----j looo l~o Fig. 6 - 800 120 Onset of plastic deformation at the ~00:!:(100 ll 00 1500 crack tip. 1900 2300 MN/m' 150 200 The effective crack 20>0 300 350 KSI length will be the initial TENSILE STRENOTH UTS crack length plus the plastic zone radius. The samples of both materials possess the similar external (flatness, straigthness, surface ruld edge Fig. 7 - UHB 20, finish) and internal properties (carbide size, impact and tensile fatigue (R = 7 :nicroinclus ion size and content, tensile 0) limit for 10 loading cycles vs sh•ength). teiisile The difference in yield strength can be roughly strength. The scatter corresponds to attributed to the differences in some structural the standard deviation. parameters. It should be noted that at tensile testing of the fatigued tensile samples (107 load ing cycles, stress amplitude 2 600 ± 600 HN/m , The tensile 87 ± 87 ksi) the changes in fatigue li~it increases linearly with tensile and yield the increased strength were smaller than 5 compared tensile strength up to ~ 1 700 MN/m2 % with the (250 ksi) while for impact initial values, see Table 2. Furthermore, no fa-bigue limit the structural changes linear relationship was documented over the whole were found due to impact load investigated ing of high-strength strip materials tensile strength range, 1 400 - (6). Using 2 300 2 K -values determined for MN/m (200 - 330 ksi). This difference in UHB 20 and UHB Stain- fatigue behaviour lW~ 716 w~th UTS ~ 1 900 2 is believed to be dependent 3 2 MN/m , 87 and mainly 131 MN m- / respectively (9), on the difference in the critically stressed the plastic zone volume. By critically radius will be 0.4 mm in carbon steel stressed volume is m1der and 1.2 mm stood the volume of the i" "'tainless steel. Conaidering that material subjected to at the "true least 95 %of the maximum defect size" is the same, one can expect, stress (11). The criti that cally stressed volume is due to significantly larger plastic zone smaller in the impact in UHB samples thrul in tensile fatigued Stain2.ess 716 the fatigue crack initiation at samples, compare Fig. 3 ru1d 12. Assuming, that potential crack

118 The appearance of the edge after frag sites are random cl.istributed, the pro this stage. initiation off is visible in Fig. Bd, It is impor P to find one or more initiation sites in ment torn bability to know, that the fragment or chip forma~'on stressed volume yields the Poisson distribu tant the in this study is attributed to the final fracture tion (12) of and not to the ini.tiation stage, The mixturt~ and edge tearing gives P = 1 - e-NV eq. 2 the radial crack g:r·m;th the late fracture stage, exhibited by broken fig. Be, f. The initiation of the pri or potential initiation sites in the samples, N rrlllll.ber mary fatigue cracks is visible on both fracto stressed volume. &,'Taphs. No stress raisers Wei'S detected, which indicates the necessity of the high stresses V stressed vol~~e the impact fracture, The location of governing out the primary fatigue cracks was on the surface It is well established that the higher tensile specimen and seat, the side the contact ring between strength gives the higher notch effect for existing defects, Consequently, with increased ma tensile strength the critical defect size in terial decreases and number of potential initia P, tion sites N increases. For a given probability the smaller stressed volume V allows the higher tensile strength wtich can explain the different ten behaviour of valve steel material at repeated sile and impact loading. and Presented relationships between fatigue limit strength are in a good agreement with tensile be established knowledge concerning the fatigue haviour of the high-strength materials. Compared with the published results (6) exhibited tested large material a very low scatter, Generally, a a low experimental scatter in fatigue data gives statistical significance and makes often the pro reasonable evaluation of the basic material perties such as composition, tensile strength, mate surface condition etc, with respect to the rial fatigue response impossible • impact fa .An attempt was done to determine the 220 tigue strength for strip material with UTS 1 2 min, ( 177 ksi) , HV 385, which is lower than 1'1l!!/m The tensile strength for standard valve steels. groove formation in the sample surfaces pronmmced groove contacting the valve seat was detected. The formation was attributed to the plastic deforma and wear, The cracking started at deep wear tion. of the marks, It was stated, that fatigue behaviour cause soft material, where impact induced stresses ex predominantly the plastic deformation of the posed volurne, differs strongly from the valve 300 material with tensile strength of 1 400 2 - 1~/m2 (200- 330 ksi), see below. FRACTOGRAPHY Fig, 2, A large number of fractured specimens, see in a scanning electron microscope. was examined is, A typical appearance of impact fatigue fracture edges. that small fragments are torn off from the The primary fatigue cracks were radially oriented. near In all cases the initiation on the surface were the edges wa~ found, The initiated cracks by circumferential (tmLgential) stresses created was indicating the opening mode, The initiation both on the upper and impact surface. observed Fig, 8 - (a) Initial fracture stage. (b) Fragment plastic de illu immediately before tearing, note the The subsequent impact fracture formation is Fragment torn stage is formation of the surface (c). (d) strated in Fig, 8, The initial fracture fracture stage, the off from the edge, (e,f ) A late shown in Fig. Sa. Fig. Bb, c demonstrates the primary the pro the arrows show the initiation of fragment immediately before tearing, Note ring. during fatigue cracks outside the contact nounced plastic deformation of the surface

119 It can be seen that in this zone no mechanical It is well mown that tigh->:>trength materials are loads due to the impact can be expected. This in notch sensitive. Moreover, the fractographic dicates, that initiated cracks were created by observations show clearly that at impa.ct loadintr stress waves i~duced by impact. The fracture the surface and edge are exposed to the highestu occurence is similar to the fractographic ana stresses. This empha:::izes the importance of tho lysis of failed compressor valves (1). surface conditions. An attempt with coarse ground specimens showed tha~ fatigue cracking at the The initiation stage takes the main part of the ini"tial stage was strongly influenced by surface fatigue life. The growth of the initiated cracks stress raisers. Fig. 11. The initiated crack is very fast and leads to the sudden final frac propagates along the sw~face defects. At fraJto ture. Depending on the stress system exposed to graphic analysis of broken impact samples of UHB the samples due to the flexural and torsional 20 and UHB Stainless 7"16 no evidence of fatigue vibrations resulting in the oblique impact on the crack initiation belo·w the surface was fouH·3_, It seat the growing crack can propagate in two direc was stated that initiation stage is si.'llHar to tions, namely in the radial direction, which is fatigue behaviour of hig-h-strength material at the initial direction and perpendicular to it, other loading modes. Fig, Sf, 9. The pictures indicate that near the surface the circumfex·ential (tangential) stress component is dominating. With increasing depth Fig. 11 -Initiated below the surface the stress component in thick fatigue crack pro ness direction can ir~luence the crack growth, see pagates the a~ong- the "stair-case pattern", Fig. 6• The resulting surface defects. fracture is oriented approx. 45 to the surface.

The crack extension during one loading cycle was estimated to 10 - 20 f,l.m, Fig. 9, 1-rhich is very large compared with conventio_:,a1 stable crack growth. A detailed fractograph show that the crack propagates due to the microvoid coalescence, Fig. 10. The duration of the stress pulses resulting from the stress-e wave propagation can be appro:x:i- CALCULAriON OF MAXIMUM IMPACT STRESSES mated to 10 sec, see ref (1) where more details- are given. From a statistical viewpoint, the larger the volume of material experiencing maximum stress, the greater the probability of finding a wem~ point that would lead to more rapid failure, see even eq. 2. It was demonstrated by \veibull ( 13), that the size effect in fatigue at egual fracture probability levels can be expressed by

crm. V = canst eq. 3 o fatigue limit m material constant V critically stressed volu~e

There are only a few determinations of m of current interest, since most investigations concern less sophisticated materials Fig. 9 - A late fractm-e stage, note than high-strength valve the steel. Available data "stair-case pattern" of the growing crack show, that a value of m bet in two directions. ween 20 - 30 is valid for this type of steel (11 14). ' The ratio between the fatigue liQit of two speci mens with different stressed volumes derived from eq. 3 is

V 1/m "'

The critically stressed volume for the tensile fatigue specimen is 70 =\ Fig. 3. Based on fracto graphic analysis the stressed volume of impact loaded specimen was assessed 3 to be between 1.5 - 4.5 = , see Fig. 12. Hence Fig. 10 - Crack growth, micro void coalescence. eq. 5

120 7 '-"1 maxim1un impact stress, 10 le>ading cycles, CONCLUSIONS 7 a limit, maximum stress, 10 - The valve steel UHB Stainless 716 '~xhibited 02 tensile fatigue than ·JHB 20, loading cycles, Fig, 7, higher impact fatigue strength Both materials showed a very low scatter in 2 For compressor valve values for oz above UTS 1 680 MN/m (245 ksi) presented fatigue data, The severe impact loading is were calculated by extrapolation from the linear applications, where the Stainless 716 sho1.~ld be relationship given in Fig, 7, The estimated expected, the UHB impact stresses corresponding to the 50 % pro selected. bability of fractill'e are summarized in Table 3, - The differences between tested grades cru1 be fatigue crack initiation Table 3 - Estimated maximum impact stress attributed to the stage, This stage takes the main part of the loaded high FaHgue limit Impact Estimated total fatigue life of dynamically Tensile zone size strength Max.tensile stress intensity Max impact strength materials, The plastic stress occill'ring at the crack tip was much larger in which leads to the retarded ini p 50 ~b p 50% p 50 % stainless steel Measured Extrapol, tiation stage, 2 2 MN/m MN/m MN/nl- - The impact .fatigue strength is strongly dependent mechanical properties of valve 1050 1.09 1190±63 on the static 1430 The linear relationship was documented 1650 1160 1,20 1320±70 steel, 2 the range of 1 400 - 2 300 MN/m (200 _, 330 1930 1220 1309 1. 31 1480±79 for 1960 1180 1319 1,32 1500±79 ksi), 2240 1210 1463 1.41 1660±88 - Extensive fractographic studies showed, that impact fracture initiates entirely on the sur Estimated maximum impact stresses are high com face, The identical behavio"L-;r was detected at pared with material ultimate tensile strength. analysis of service compressor valve failures. Even fractographic studies documented the ne The crack initiation below the surface was not cessity of high stresses responsible for the found, initiation of the fa"bigue cracks since no stress raisers were detected. • The primary fatigue cracks were radially oriented and located outside the contact ring with the seat where no mechanical loads due to the impact can be expected, This leads to an assumption that detected fatigue cracks were created by propagating stress waves,

- Based on the fractographic observations, the duration of the stress pulses governing the crack initiation and crack propagation in laboratory testing equipment was approximated to 10-6 sec, The similar results were found on service valve failures,

-Using the Weibull's concept of critically stressed volume, the maximum impact stresses were calculated, The determined stresses are 20 high compared with the valve material tensile strength. This is in accordance with the fractographic studies indicating the necessity Fig, 12 - Schematic representation of critically of the high stresses in order to initiate stressed volume in impact specimen, fatigue fracture since no stress raisers were detected in impact loaded specimens, The very short stress pulse duration approx 10-e sec, resulting from the complex random~load APPENDIX history, transient dynamic effects, stress inter impacts ference effects etc, when the valve reed The accelerometer was calibrated with a steel enormous the seat or valve stop, gives raise to ball falling freely from different heights. The experimental difficulties when the stresses momentum p, immediately before the first seat governing the valve fatigue failure are detev contact is mined by for instance strain gauge measill'ement·s, Even matema·bical analysis, such as finite elemeht p = vm eq, 6 performance of thoi! method, deseri"bing the dynamic of steel ball 2,04 g the complex where m mass compressor valve can hardly define velocity/2gh by operating · v impact random stress system, experienced g acceleration of gravity valve, h height

121 The result of calibration is shown in Fig. 13, where impact intensity a measured by Uddeholm IMPACT INTENSITY and IMPACT VELOCITY m/s impact intensity a f (6) is plotted. Comparison a ref between a and a re1s illustrated in Fig. 14, 3 r----r----.----r----.----r~v v2 vhere the strai~fi{ line was fitted by the least square method 10 100

aref = 0.44 a + 0.42 eq. 7 2 80 60 stand dev 0.045 aref"' 0,411 ' a + corr coef 0.997 1 0.42 40 STAND. IMPACT INTENSITY DEV. 0.045 5 CORR. COEFF, 0.997 20 4 0 0 1 2 3 4 5

MEASURED IMPACT INTENSITY a ~ig. 14_- Imp~ct intensity aref versus measured 1mpact 1ntens1ty a.

IMPACT INTENSITY a 2 4

1 3

2 3 4 5 2 MOMENTUM P1 • 103 k!JIIt"s a = 1.61 aD + 0.45 iig. 13 - Calibration of accelerometer with STAND. DEY. steel ball. 0.047 CORR. COEFF. 0.997 In Fig. 14, the impact velocity of the specimen 1 immediately before contact with the seat vs im pact intensity is also shown. 0, 4 0 •8 l. 2 l. 6 2, 0 aD Fig. 15 - Measured impact The continuous monitoring of the impact intensity a versus ~· during i~tensity fatigue was made by recording the damped ACKNOWLEDGEMENTS accelerometer signal an· The relation between a and ~was determined by simultaneous measure The authors are indebted to Dr l1lents. Fo.r each ~ value the H. Nordberg for corresponding a was valuable comments. Thanks recorded by the peak detector, Fig. also are due to 15. The rela Uddeholms AB, Steel Research tion between a and ~ Dept., for permission was found to publish this paper. a= 1.61 ~ + 0.45 eq. 8 REFERENCES

stand dev 0.047 (1) R. Dusil & B. Johansson: corr coef 0.997 Fatigue Fracture BehavioUL' of Impact Loaded Eq. 7 and 8 give Compressor Valves. Proceedings of the 1978 Purdue Compressor Technology Conf. West Lafayette Ind. USA. aref = O. 71 ~ + 0.62 eq. 9 (2) R. Dusil: Studies of Faults in Used Valves. Case Studies. Proceedings of the 1976 Purdue Compj'essor Technology Conf. pp 99-105.

122 (3) A. Futakawa et al: Fracture Surface Observation of Refrigerating Compressor Valve Using Electron Microscope. Mitsubishi Electric Corp. Report DMN 1539, 197 4. (4) A. Futakawa et al: Dynamic Stress of Ring TyPe Discharge Valve in Refrigerating Compressor. Mitsubishi Electric Corp. Report DMN 1538, i974.

(5) 11l. Soedel: On Dynamic Stresses in Compressor Valve Reeds or Plates During Colinear.Impact on Valve Seats. Proceedings of the 1974 Purdue Com pressor Technology Conf. pp 319-328.

( 6) M. Svenzon: Impact Fatigue of 'valve Steel. Proceedings of the 1976 Purdue Compressor Technology Conf. pp 65-73·

(7) A Guide for Fatigue Testing and Statistical Analysis of Fatigue Data, ASTM STP 91-A, 1963. (8) E.L. Crow: Statistics Manual, Dover Publications Inc. New York, 1960 .• (9) R. Dusil & B. Appell; Fatigue and Fracture Mechanics Properties of Valve steels. Proceedings of the 1976 Purdue Compressor Technology Conf. pp 82-90.

(10) R. W. Hertzberg; Deformation and Fracture Mechanics of Engineering Materials, J. Wiley, Inc. 1976 •

(11) R. Kuguel: A Relation between Theoretical Stress Concen tration Factor and Fatigue Notch Factor Deduced from the Concept of Highly Stressed Volume. American Society for Testing and Materials, Proceedings, Vol. 61, 1961.

(12) H. Nordberg: The Effect of Notches and Non-Metallic In clusions on the Fatigue Properties of High Strength Steel. Uddeholm Research Report FM75-015, 1975 (13) W. Weibull: A Statistical Theory of the Strength of Materials. DTA Handl. (in Swedish)p 151, 1939 (14) F. de Kazinczy: Size Effect in Fatigue of Cast Steel. Journal of the Iron and Steel Institute, Vol. 206, 1969.

123