Files

Abstract

Fatigue tests were carried out on welded circular hollow section K-joints typical to bridges. The tests specimens were large-scale (approximately 9 m long and 2 m high) trusses loaded in the plane of the truss. Measured member stresses showed that a significant proportion of the load in a truss member may be due to bending, underlining the importance of considering correctly this load case in the design of these structures. Measured hot-spot stresses in the joints were compared with hot-spot stresses calculated using the current design guidelines. It was found that the measured values are considerably lower than the calculated values, calling into question the applicability of the design guidelines to these types of (bridge) structures. The S–N fatigue results from the current study, on the other hand, showed that the fatigue resistance of the joints that were tested is lower than the corresponding S–N design curves. This means that when the considerably higher calculated hot-spot stress range is applied to the corresponding design curve, the predicted resistance is similar to the resistance predicted using the lower measured hot-spot stresses in combination with the lower measured S–N curve too. This has highlighted the importance of relating hot-spot stresses to the appropriate, corresponding S–N curves. Evidence from the fatigue tests has clearly demonstrated the effect of size on the fatigue strength of welded tubular joints. A comparison of fatigue S–N results from smaller and larger welded circular hollow section (CHS) joints has shown the same trend indicated in design specifications: a thicker failed member results in a lower fatigue strength. The size correction factor integrated into the S–N design curves of the specifications, however, does not seem to represent this significant effect justly. In light of the size effect results presented in this paper and the major influence of this effect on the design of welded CHS joints in general, it is recommended that a soundly based solution with targeted S–N curves and a representative size effect should be sought.

Details

Actions

Preview