Friction Stir Welding effects of defects in Glare

Abstract

Splicing technology is used to create fuselage panels due to the maximum width of the aluminium sheets of 60 inches as a result of the production process. A set of designing rules is required for the design of a spliced Glare fuselage resulting in a significant increase in complexity. Friction stir welding of the thin aluminium sheets in Glare results in larger Glare panels where the maximum dimensions are limited by the size of the autoclave. FSW Glare could result in reduced weight and production cost, and perhaps as well as having enhanced mechanical properties. But welding defects could occur which will compromise fatigue properties. Fatigue and damage tolerant properties of FSW Glare with welding defects are investigated, and recommendations are given in this thesis. Friction stir welding generally does not produce defects that are common to conventional welding technologies, such as warping and porosity. Instead unique defects can occur, such as the zig zag curve or lazy S defect, where the oxide particles of the aluminium are present in a concentrated form. Another typical form of defect is the tunnel defect, also know as the channel, cavity, or void defect. A physical channel is present in the cross section of the weld, which is not visible looking at the surface of the welded sheet. A similar defect is the groove defect where the absence of material is at the surface which makes it visible and tangible. Nevertheless, friction stir welding is a robust process where a wide range of welding parameters result in a defect free weld. Friction stir welding results in much more favourable fatigue properties compared to conventional welding methods, and S-N data lie well above the FAT40 design curve. Welding parameters, such as welding speed and tool pressure, as well as tool geometry play an important role in fatigue properties. Defects as a result of improper welding parameters have a large impact on fatigue properties. Surface properties of the weld also impact fatigue properties; post machining the rough surface results in improvements. Static properties are also favourable with friction stir welding. Two sets of FSW sheets with a welding defect are available for testing both in single sheet and FML form. Artificially damaged sheets are produced to model a welding defect and tested in both single sheet and FML form. This data is compared to standard 2024-T3 in single sheet form, and standard Glare in FML form. Both fatigue and static tests are performed. The Ccg and ncg coefficients derived from the single sheet 2024-T3 specimens deviate quite a lot from the Ccg and ncg coefficients derived from the Forman equation. Using the latter would have resulted in inaccurate comparisons of other single sheet FSW and Artificial Defect specimens. The single sheet FSW 0.3 – 0.4 mm specimens have a kissing bond defect, which resulted in degradation of fatigue performance. This defect was present, even though the welds on the sheets looked flawless. The single sheet FSW defect specimens show a significant drop in fatigue properties, but that is primarily due to the reduced cross section thickness of the weld. Correcting this data to the real thickness, or real stress, results in a substantial improvement, where some specimens even have the same fatigue properties as that of standard single sheet 2024-T3. The results are surprisingly favourable considering the visual state of the FSW sheets. The single sheet Artificial Defect groove specimens show the most reduction in fatigue properties. Standard Glare has a shorter crack initiation life than single sheet 2024-T3, but a much longer crack growth life resulting in an approximate doubled fatigue life. The worst case scenario is tested with the FML Artificial Defect severed weld specimens. Fatigue life is comparable to single sheet 2024-T3 at high stress amplitude, but fatigue properties are comparable to standard Glare at low stress amplitude. All other FML types perform between these two types of FMLs. The biggest degradation in fatigue occurs at high stress amplitude, whereas fatigue performance is comparable to standard Glare at low stress amplitude. The exception is FML 0.3 – 0.4 mm, which even performs slightly better than standard Glare. The reason is that the cross section of the centre FSW sheet was 0.35 millimetres, whereas it is considered as if it were only 0.3 mm. Fatigue properties of different types of material is much smaller in FML form than in single sheet form. The FMLgrow model predicts the fatigue life most accurately at low stress amplitude with a fatigue life that is 70% of the measured value. The least accurate prediction is at high stress amplitude with a fatigue life that is about 25% compared to the tested specimen. Crack initiation in the centre layer of the FML occurred only faster than in the outer aluminium layers with the Artificial Defect groove specimens, where the groove acts as a stress riser. The centre fatigue crack in this case is not longer than the fatigue crack in the outer aluminium layers when the latter reaches its final value of 2a = 50 millimetres. This is visible in the C-scan result where the delamination pattern doesn’t exceed the final crack length value in the outer layer. The outer crack, after initiation, catches up with the centre crack, and then all cracks continue to grow at the same rate. Crack growth rate in FML reduces after initiation to reach a minimum value before increasing again. Crack bridging is less efficient when cracks are small, and the discontinuity due to the notch are the reasons for this initial high crack growth rate before reducing to a minimum value. Exponential growth as a function of half crack length was observed in the FML specimens, resulting in the following expression: da/dN=C_exp·a Cexp is an intrinsic fatigue property of an FML as are the Ccg and ncg coefficients for single sheet. Different basic fatigue calculations can be performed. Modeling and understanding of crack growth in Glare can be facilitated knowing this relation, since exponential growth is an intrinsic fatigue property in an FML. Plotting the logarithmic Cexp as a function of linear Sa yields a relation similar to the S-N relation. Exploration of the FMLgrow model to account for the differences between its predictions and measured data resulted in the understanding that the Alderliesten model is flawed due to interdependency of the aluminium layer with the glass fibre in the (W-2a) region, resulting in an incorrect Kfarfield. Ktip?Kfarfield+Kbridging Glare is considered as a structure where the primary load is carried solely by the aluminium sheets. Hence the feasibility of FSW Glare is based on the fatigue properties of FSW single sheet. Single sheet FSW with a defect shows more degradation in fatigue performance than the FML variant. Therefore using single sheet performance to evaluate feasibility of FSW Glare result in a stricter selection of what defects can and cannot be tolerated. At the same time will the stricter selection not result in a heavier structure; only damage tolerance will benefit. 90% of the initiation and crack growth life of 2024-T3 is chosen as an arbitrary value for the requirements of the FSW sheets. This means that 90% of the fatigue life of 2024-T3 is required. None of the tested specimens fulfill this requirement. Zero defects are not feasible in reality; an example is the presence of air inclusions in Glare. Defects in FSW are continuous and occur only as a result of incorrect welding parameters. Therefore there is no validation in the discussion what type of FSW defect can or cannot be tolerated; FSW sheets should not have any welding defects at all for Glare application.