Transactions on Engineering Sciences vol 2, 1993 WIT Press, ISSN

Transcription

1 The effect of grinding and peening on the fatigue strength of welded T-joints P.J. Haagensen Department of Civil Engineering, The Norwegian Institute of Technology, University of Trondheim, Norway ABSTRACT The fatigue strength properties of welded T-joints improved by toe grinding and a combination of toe grinding and hammer peening are described. Constant amplitude fatigue tests were performed in four point bending on 30 mm thick specimens. Reference S-N curves for as-welded specimens were also established. The steel was a low carbon micro-alloyed steel of 360 MPa yield strength that is commonly used in offshore structures. Information on fatigue lives in terms of cycles to the initiation of a small crack, and total fatigue endurance was obtained in the test program. The results are presented in S-N diagrams and compared with published data from similar tests. INTRODUCTION Grinding and peening are postweld methods that are used frequently to increase the fatigue strength of welded joints. Grinding to remove cracks is also an accepted method for extending the life of structures that experience fatigue damage during their service life [1]. However, grinding out a crack leaves a groove which reduces the cross section and creates a stress concentration, thereby limiting the depth of crack that can be repaired. By combining grinding with a subsequent mechanical surface stressing process, e.g. by hammer or needle peening, or shot peening, a residual stress field is introduced in the groove. This local compressive stress delays crack initiation and reduces crack growth rate, giving a substantial increase in fatigue life. Deeper cracks can therefore be repaired successfully than for the case when grinding is used alone. The tests described in this paper are part of a program aimed at establishing guidelines for the depth of cracks that can be repaired by grinding

2 40 Surface Treatment Effects and subsequent hammer peening. The limited test program described herein was designed to quantify the effects of weld toe surface grinding and the combination of grinding and hammer peening. EXPERIMENTAL PROGRAM Material The specimens were fabricated from a normalized low carbon micro alloyed steel. Chemical composition and mechanical properties of the base material are shown in Tables 1 and 2, respectively. Table 2. Mechanical properties. Yield strength, MPa Tensile strength, MPa Elongation, % 35.6 Table 1. Chemical composition, weight % c 0.10 Si 0.41 Mn 1.32 p S N Cr 0.13 Ni 0.39 Al Cu 0.16 Mo 0.02 Nb V Ti Zr <0.01 Co B.0002 W 0.01 As Sn Carbon equivalents: Cgq = 0.39 PQ^ = 0.20 Specimen Design and Fabrication Specimen geometry The test specimens were T-joints withfilletwelds joining the attachment to the horizontal plate as shown in Figure 1. Specimen dimensions are also shown in Figure 1. Welding and specimen machining The attachment was welded manually to a 2.5 m long plate in the horizontal 2F position, using a gas shielded flux cored wire (G-FCAW) process, with a wire diameter of 1.2 mm. The preheat minimum temperature was 30 C. Heat input was 1.27 kj/mm. No post weld heat treatment was applied. The welds had a throat thickness of 15 mm and nearly flat weld flanks with a toe angle of approximately 45. A length of the plate corresponding to the number of specimens in each test series was cut from the plate. After each weld toe treatment the plate segments were cut into specimens by cold sawing. The side edges of the specimens at the welds were deburred and rounded slightly by a file to avoid premature crack initiation from the edges.

3 Surface Treatment Effects 41 Weld toe grinding For specimens tested in the ground condition dressing was performed with a rotary burr of 12 mm diameter using an electric grinder at to rpm. The procedure was in accordance with the UK Department of Energy design guidance [1] which prescribes grinding to a depth of 0.5 to 1.0 mm below any undercuts. Only the two weld toes at the main plate were dressed. Weld toe hammering A series of weld toe ground specimens were subjected to hammer peening. Four runs were made with an pneumatic hammer, together with a solid tool having a 12 mm diameter hemisphericaltip.the air operating pressure was approximately 6 bars. Fatigue Testing The specimens were tested in 4-point bending of the base plate in a servohydraulic fatigue test frame equipped with an actuator of 40 kn load capacity. The specimens supported on roller supports to minimize load errors due to friction. All tests were run in load control. The correlation between bending stress in the specimen and actuator load signal was verified by strain gage measurements on one specimen. Testing was performed at a room temperature of approximately 22 C, with a relative humidity of 50 ± 10 %. Test frequency was 4 to 6 Hz. The applied load ranges were selected to produce fatigue lives in the range 10^ to 10 cycles. Fatigue life to crack initiation was taken as the number of cycles at which a surface crack could be visually observed. Application of a solvent to the toe area resulted in bubble formation at the crack which was found to facilitate crack detection. The surface length of the crack at this stage was 1 to 1.5 mm. Failure was defined to have taken place when increased specimen compliance gave an actuator stroke of approximately 5 mm. Generally this would occur when cracking had progressed through approximately 35% of the plate thickness. RESULTS The as-welded specimens always failed at the weld toe. Cracking also generally started in the weld toe region of the improved specimens, in most cases in the groove at the weld toe, but in two cases cracking initiated at the notches between thefirstand second bead. A few tests were terminated at long lives (minimum 2 mill, cycles) before failure had occurred. Examination of the fracture surfaces indicated that multiple cracking had occurred at an early stage, followed by rapid coalescence into an single, straight-fronted crack. Linear regression analysis was used to calculate best fit mean life S-N curves for the failed specimens. The test results for each test series are plotted

4 42 Surface Treatment Effects in double logarithmic S-N diagrams in Figures 2 to 5. The stress range plotted in the diagrams is the nominal stress based on the output from the actuator load cell. The mean life S-N curves for the as-welded, ground, and ground/hammer peened specimens are shown in Figures 2, 3, and 4, respectively. All mean life S-N curves are plotted together in Figure 4. The mean line for the Class F curve, adjusted for 30 mm thickness according to the DEn thickness correction rule [1], is plotted in the diagrams for comparison. The mean line for Class F ground joints, which is equal to the mean Class F curve for as-welded plus 30% on stress is shown in Figures 3 and 4. The mean life S-N curve for ground, axially loaded transverse joints derived in an examination of published data by Booth [3] is also shown in Figure 3. This curve which is almost parallel to the Class F curves, is 64 % higher in stress range at 2x10* cycles than the mean as-welded Class F curve. The mean curves for the fatigue lives to failure for all three test series are plotted in Figure 5. The equations for these S-N curves obtained from the regression analysis are given in Table 3. Table 3. Details of S-N curves obtained from linear regression of test results. Test series S-N curve: AWf= C m C (m/cycle, MPaVm) Standard deviation of log N Stress range at 2x10* cycles Improvement % As-welded x10^ Toe ground x10* Toe ground + hammer peened x10^ DISCUSSION In each of the three test series from 6 to 8 specimens were tested. While larger samples would have provided a better statistical basis for the observed trends, the specimens were tested at two stress levels spaced as far apart as possible within the endurance range aimed for, in order to obtain maximum confidence in the estimates of the two parameters that define the S-N curve. The mean S-N curves are therefore considered to indicate valid trends and therefore allow comparisons to be made with existing design rules and other experimental data. However, the sample size is insufficient to establish safe lower bounds for the variability that are found in a large population and therefore the data alone cannot be used for the derivation of design information.

5 Surface Treatment Effects 43 As-welded specimens The results for the as-welded specimens, shown in Figure 2a), agree reasonably well with the Class F mean life curve, however, as is frequently found in laboratory tests on small scale specimens, the test data are somewhat higher than the design rules curve, especially in the high endurance region. One reason for the discrepancy is that the design rules curves are adjusted to take into account the detrimental effects of the high tensile residual stresses that exist in full scale structures, while such stresses are low in the type of specimens used in this investigation [2]. Effect of weld toe grinding The S-N curve for ground specimens obtained in the present investigation, shown in Figure 2b), indicates a significant increase in fatigue strength, particularly in the low stress, high cycle region. The effect of grinding is less in the high stress, low cycle regions as the test mean life curve and the Class F curve for ground specimens converge at around cycles. The increase in stress range at two million cycles is 94% (Table 3), which is considerably higher than the 30 percent increase allowed in the UK design rules, and also somewhat higher than the 56 % increase in stress for ground joints found by Booth in his study [3]. With the exception of one specimen all data points are above Booth's mean line, which incidentally is almost parallel to the DEn Class F as-welded mean curve. Some axial loading tests were made in an early investigation by Gurney [4] on transverse joints in the as-welded, fully ground, and hammer peened conditions in a mild steel with a yield strength of approximately 300 MPa and two higher strength steels with yield strengths of 346 to 370 MPa. These tests gave improvements in fatigue strength at 2x10^ cycles for the fully ground transverse welds of approximately 100 % for the low strength steel and 70 % for the two higher strength steel. It was also found that the S-N curves for ground and as-welded specimens were nearly parallel. In another, more recent investigation [5], fatigue tests were made in 3-point bending on 30 mm plate specimen of similar geometry to those used in the present investigation. The material was also a low carbon micro-alloyed steel of similar chemical composition and mechanical properties. The as-welded reference S-N curve nearly coincided with the DEn Class F mean curve for 30 mm plate thickness [5]. For the toe ground specimens a fatigue strength of 153 MPa was obtained, representing a 50 % increase at 2x10^ cycles. Also in this case the S-N curves for ground welds were parallel to the as-welded curves. Effect of weld toe grinding combined with hammer peening Weld toe grinding with subsequent hammer peening produces a significant additional increase in fatigue strength over that due to grinding alone, as shown in Figures 2c) and 3. At 2x10^ cycles the fatigue strength was 304 MPa, representing an increase over the as-welded fatigue strength of 142 %, or 25 %

6 44 Surface Treatment Effects higher than the fatigue strength for ground specimens. While tests on the base material were not performed in the present investigation, three point bending test on a nearly identical material in the same thickness [5], gave a fatigue strength at 2x10^ cycles of 290 MPa, indicating that the fatigue strength obtained in the present tests on toe ground welds with subsequent hammer peening is of the same magnitude as that of the unwelded plate material. Very few data are available in the literature on combinations of improvement methods, however, in the investigation by Gurney [4] quoted above, some axial loading tests were made on transverse joints in the aswelded, fully ground, and ground and hammer peened conditions in a mild steel with a yield strength of approximately 300 MPa. The fatigue strengths for aswelded, fully ground and ground plus peened specimens at 2x10^ cycles were 104, 209 and 247 MPa, corresponding to increases in fatigue strength of 96 % and 138 % for grinding and combined grinding and peening, respectively. The fatigue strength of the unwelded base material in the Gurney tests was 247 MPa, ie. of the same magnitude as the fatigue strength of the ground plus hammered joints [4]. In the previously mentioned investigation [5] of a steel with almost identical chemical composition and mechanical strength, the fatigue strength of parent plate material was determined to be 290 MPa at 2x10^ cycles. From this it can be inferred that in the present investigation the fatigue strength obtained for weld toe grinding followed by hammer peening is as high as that of the unwelded plate material. The scatter for ground welds appears to be larger than for the as-welded and ground plus hammer peened tests series. However, the number of specimens in each series is to small to enable any firm conclusions to be made with regard to the relative magnitude of the scatter between the tests series. CONCLUSIONS From the investigation of the fatigue strength of 30 mm thick T-joints with fillet welds tested in bending at R = 0.1 in the as-welded, toe ground and toe ground with subsequent hammer peening the following main conclusions are made. 1. Weld toe grinding resulted in an increase in fatigue strength at 2x10^ cycles of 94 % compared with as-welded joints. This agrees well with the 50 to 100 % increases reported in the literature for similar tests. The results are well above the 30 % increase specified in the UK Department of Energy guidance for ground joints.

7 Surface Treatment Effects Weld toe grinding combined with hammer peening gave a fatigue strength at 2x10^ cycles of about 300 MPa, representing an improvement of 142 % in comparison with the as-welded fatigue strength. This fatigue strength is of similar magnitude as that of unwelded plate specimens tested in the as-rolled condition. This result is also in agreement with published data from similar tests on welded specimens improved by grinding and peening. ACKNOWLEDGEMENTS The support from Aker Verdal who supplied the welded plates is gratefully acknowledged. REFERENCES 1. Department of Energy, * Off shore Installations - Guidance on Design and Construction', 4th edition, HMSO, London, Gurney, T.R. and Maddox, S.J. 'A Re-Analysis of Fatigue Data for Welded Joints in Steel', Welding Institute keport No. E/44/72, Booth, G.S. 'Improving the Fatigue Strength of Welded Joints by Grinding - Techniques and Benefits,' Metal Construction, Vol.18, pp , Gurney, T.R. 'Effect of Peening and Grinding on the Fatigue strength of Fillet Welded Joints', British Welding Journal, pp , December Haagensen, P. J., Slind, T. Lian, B. and Gunleiksrud, A., 'Prediction of the improvement in fatigue life of welded joints due to grinding, TIG dressing, weld shape control and shot peening', in SIMS '87 (Ed. C. Noordhoek and J. de Back), pp , Proceeding of the 3rd Int ECSC Conf. on Steel in Marine Structures, Delft, Netherlands, Elsevier, Amsterdam, 1987.

8 46 Surface Treatment Effects 30-*) [* Fig. 1 I Geometry of fatigue test specimen As-welded 1 CO i L A [ j! I Class F mean, 30 mm Fig " 10 ^ 10 NO. OF CYCLES N Fatigue test results for as-welded specimens. ^5 2 I Booth [3]' Class F mean, ground Toe ground o jji.v : g 6 cc Fig. 3 1 A 1 =3 ~^ Class F mean, As-welded, 30 mm 10 * 1 0 * NO, OF CYCLES N Fatigue test results for toe ground specimens with Class F mean lines and Booth's mean line for specimens with transverse gussets [3].

9 Surface Treatment Effects i 3-2 s* >V ~~~_ " - ^ Toe ground ~~. ^ ^4- Hammer peened Class F mean, ground Z/-_ ^ ^ 8" 1 Co a A 1 ~^^"-_ ' I t " ^ ~ _ p. 0.1 Class F mean, As-welded, 30 mm ^ NO. OF CYCLES Fig. 4 Fatigue test results for toe ground and hammer peened specimens Toe ground + Hammer peened Toe ground As-welded 1 0 * ivo. CYCLES 1 N Fig. 5 Comparison of S-N curves for as-welded, toe ground, and ground and hammer peened specimens.