Influence of Surface Finishing on Fatigue Properties of Valve Steel

The fatigue strength of flapper valves depends among other things on surface properties of the steel. This paper deals with the effect of a tumbling method*) on bending and tensile fatigue properties of test bars. For carbon steel SANDVIK 20C an improvement was observed in several test series, particularly in the transverse direction of the strip, whereas there was little effect for stainless steel SANDVIK 7C27Mo2. The results are discussed mainly in terms of surface roughness and residual stresses. INTRODUCTION Compressor valves are exposed to high stresses during a large number of load cycles. Therefore it is necessary that the strip steel used in the valves has a high fatigue strength. As previously reported (1, 2, 3) there are several factors, such as tensile strength, metallurgical cleanness and surface condition, influencing this property. The present study deals with the effect of surface finishing and, simultaneously, edge preparation on the fatigue strength, determined under reversed bending or fluctuating tensile stress. Valves in actual operation are exposed to bending but hardly to tensile fatigue. The reason why also the latter type of load was used in the testing was the well defined state of stress in tension. Two types of surface finish were tested, viz. polished or in one case ground sur-* face, and subsequently specially tumbled ) surface. *) !so-finishing, a proprietary process by DeStaCo Division of Dover Corporation 98 EXPERIMENTAL Two valve steels, continuously hardened and tempered in strip form, were used, SANDVIK 20C and 7C27Mo2, corresponding to AISI 1095 and a modified version of AISI 420. The compositions of the heats are given in table 1. Material from four different strips, here designated I-IV, was used. Strips no I, III and IV were polished in the rolling direction after heat treatment, whereas no II was fine ground. Part of strip no I had a rather deep longitudinal scratch. The strip thickness was 0.508 mm (0.020") or 0.381 mm (0.015"). The static strength was determined along and, for strip no I, also across the rolling direction, table"2. The microstructure of the steels comprised essentially tempered martensite and small carbides, undissolved during the hardening process. There was no surface decarburization. The content of non-metallic inclusions, i.e. sulphides and oxides, was very low. Six series of test pieces for bending fatigue testing and two series for tensile fatigue testing, table 3, were blanked from the strips and ground to the desired shape, fig. 1. At least 60 specimens were made in each case. Every second test piece in each series was polished along the edge of the test length. The resulting edge shape is illustrated in fig. 2. The remaining test pieces were tumbled. Figure 3 shows the round edge of one of those specimens. The fatigue testing followed the staircase procedure to determine the fatigue strength at 2·10 6 cycles using 50% probability of fracture as a criterion. All reversed bending fatigue testing was performed in the same machine, type UMG. The loading frequency was 25 Hz. A 20 kN Amsler high frequency pulsator operating at about 75 H~ was used for the fluctuating tensile fatigue testing (R=O). The testing temperature was 20+1.5 C and the relative humidity 40-70%. The surface finish of polished and of tumbled parts was checked by measuring the surface roughness and by e~amining the surfaces in optical and scanning electron microscopes. Table 3 gives surface roughness data, and figures 4 and 5 illustrate the surface appearance. Residual stresses in the specimen surfaces were determined with x-ray diffraction technique. The reported values, table 4, are mean values from measurements on both sides of at least two test pieces. stresses were determined along and across the edge for test series no 3. In this case about 20 test pieces were mounted together to give a sufficiently large test area. FATIGUE TESTING RESULT Series no 1 and 2 give a comparison of the bending and tensile fatigue strength along and across the rolling direction. Table 5 indicates about 8% lower values for transverse than for longitudinal specimens in the polished condition. In all cases the strength was improved by tumbling, and the influence of test direction·was much reduced. This applies in particular to tensile fatigue data. Series no 3 was made to study the influence of a scratch on the bending fatigue properties. The test pieces were blanked so that the defect ran across the middle of their trapezoidal part. The depth of the longitudinal scratch.was 2.5 1 um on the polished surface and reduced to 1.5 1 um after t,umbling. Figures 6 and 7 illustrate the scratch as seen in the scanning electron microscope. A fairly low fatigue strength was obtained for the polished surface. However, tumbling restored it to the same level as for series 2, table 5. The fine ground surface of tbe strip in series no 4 gave a very good bending fatigue strength, as determined in the rolling direction, table 5. The tumbling process did not improve the mean value of the fatigue strength, but there was some reduction of the already low fatigue strength variation between test pieces. Series no 5 was added to see if the positive influence of tumbling as demonstrated in series no 1 would appear also on this thinner polished gauge. Table 5 shows that this way was the case. As previously shown (1, 3) the hardened and tempered stainless grade, modified AISI 420, is superior to the carbon steel AISI 99 1095 with respect to fatigue strength. Test series no 6 again proved this fact, table 5, particularly with polished test pieces. Tumbling did in fact give a slight decrease of the faligue strength, measured in the strip rolling direction. FRACTOGRAPHIC OBSERVATIONS It was previously (3) shown that bending fatigue cracks fre~uently started near one edge of the test piece. Table 6 reveals that about 80% of the fractures in series 1-4 in this investigation also started at the specimen surface within 2 mm from the edge. This applied to polished as well as to tumbled test pieces. The remaining bending fatigue cracks als·o started from the surface, although further from the edge. As a rule it was no possible to exactly define the crack initiation site, since the crack surfaces rubbed against each other for many cycles prior to the final bending fatigue fracture. In some cases tiny surface defects were identified as the starting points. The distribution of the initiation sites was not quite uniform along the trapezoidal waist, but with some concentration towards the narrowest section. This observation was also valid for series no 3 where one would have expected the scratch to be the serious defect. In fact, in only one of the polished test pieces did the crack start from the scratch. None of the fatigue fractures in the tumbled specimens was associated with the scratch. In the polished test pieces of series no 2 and 3, i.e. transverse specimens, the fatigue crack often followed for some distance, and on one surface at a time, the straight shallow marks from the strip polishing operation. The longitudinal bending test pieces and the tumbled specimens did of course not show any similar straight crack part. The crack started at the surface also in the polished and in the tumbled tensile fatigue test pieces. Only in three transverse, tumbled specimens was the fracture initiated at an interior position. Somewhat surprisingly, the starting point was close to the edges in a majority of cases, table 6. One reason might be a slight misalignment of the test pieces. DISCUSSION Since practically all fatigue fractures in reversed bending as well as in fluctuating tension started at the specimen surface or edge it is obvious that the surface properties and the edge preparation are of vital importance for the fatigue life of valves. There are two essential factors to conside~. the geometry and the residual stress system. The manufacturing process of strip, involving a final polishing operation, gives a product with longitudinal marks that cause a somewhat lower fatigue strength in the transverse direction than in the rolling direction. Tumbling removes the oriented polishing marks, and the transverse properties are clearly improved even if the resulting surface has less good roughness value than the original one. The rounding of the edges resulting from the tumbling is of course very important, since the probability of fracture initiation at the edge evidently is high. Compressive surface stresses improve the fatigue strength. This can be seen when comparing fatigue data for polished and tumbled carbon steel specimens taken in the strip rolling direction. In this case the improvement can hardly be explained only by the change in geometry brought about by tumbling. Compressive stresses probably also contribute significantly to the improvement of the fatigue strength in the transverse direction. It is interesting to notice that tumbling gave a higher compressive stress along the test piece edge than across it, and also higher than on the flat surface, table 4. Very local high compressive stress around the transverse scratch on specimens in series no 3 might explain why the scratch did not cause more than one fatigue failure. Scratches of this magnitude are not permitted on regular Sandvik valve steel, however. For series 4, carbon steel with fine ground surface, and series 6, stainless steel, tumbling gave no improvement of the fatigue strength in the rolling direction in spite of an increase in the compressive surface stress. It should be remembered, however, that the compressive stress from grinding of carbon steel and polishing of stainless steel, respectively, penetrates deeper than stresses from polishing of carbon steel strips. The effect of the additional stress caused by the tumbling seemci to have been counterbalanced here by the coarsening of the surface finish. Nevertheless, tumbling still is recommended since the transverse fatigue properties, which were not measured on the strips, are expected to be positively influenced. It is also worth mentioning that the comparison is made against specimens with carefully polished edges and not against blanked parts with sharp edges, 100 blanking burrs and deleterious tensile stresses in the edge area. It should be borne in mind that fatigue loading gradually reduces the residual stresses (1). At the moment of fracture the compressive surface stresses were certainly lower than the values given in table 4. The enhanced frequency of bending fractures in the edge region and towards the narrowest section of the trapezoidal test area demonstrates the difficulty in obtaining a uniform bending stress along a specimen. The test piece geometry and the clamping method are of great importance and will be discussed separately. Hence, all bending fatigue data must be regarded with some care, particularly when results from different sources are compared. Tensile fatigue, on the other hand, gives a well defined state of stress, and more reliable data are obtalnaule. The drawback is that this type of stress does not very well represent the stress pattern in operating valv<;'s. CONCLUSIONS 1. The fatigue strength across the rolling direction was somewhat lower than along the strip due to polishing marks. This difference was greatly reduced by a tumbling operation that removed the longitudinal marks. 2. Tumbling generally improved the fatigue data also in the longitudinal direction. The beneficial effect was in this case mainly the incr-eased compressive stress in the specimen surface. 3. A favourable, round edge was obtained by the tumbling process. 4. Nearly all bending and tensile fatigue cracks started at the test piece surface, very often close to the edges. 5. The superiority of the stainless valve steel grade, modified AISI 420, over carbon steel, AISI 1095, with regard to fatigue strength was confirmed. ACKNOWLEDGEMENT I would wish to express my gratitude to Mr. J. Thullen at DeStaCo Division of Dover Corporation for his cooperation. REFERENCES 1. R. Johansson, G. Persson. Influence of testing and material factors on the fatigue strength of valve steel. 1976 Purdue Compressor Technology Conference. 2. G. Persson. Fractures in springs of steel strip. Proc. 1st National Conference on Fracture, Univ. of Witwatersrand, Johannesburg, 7-9 Nov. 1979. 3. J-0. Nilsson, G. Persson. Bending fatigue failures in valve steel. 1980 Purdue Compressor Technology Conference. Table 1 SANDVIK AISI Heat Strip Chemical composition, % No No c Si Mn p s cr Mo 20C 1095 325196 r, II 1. 01 0.33 0.40 0.016 0.008 0.16 20C 1095 456301 III 0.99 0.28 0.42 0.009 0.007 0. 16 7C27Mo2 420 mod. 411825 IV 0.40 0.45 0.63 0.019 0.005 14.0 1. 01