ICF10B

(ii) Or a crack started close to the inner or outer transverse stiffener, namely near that portion of the stiffener into which the diagonal tension band in the web anchored, then propagated and rather at a slow pace – along this stiffener. Later on, as a rule, after several tens or hundreds of thousands of loading cycles (this depending on the magnitude of loading), the crack turned inside the web sheet. (iii) Or a crack initiated in the close vicinity of the fillet weld connecting the upper flange with the web sheet, then propagated (at first in both directions) along the weld so that on one side it reached the adjacent web corner. Then it turned down along the vertical fillet weld joining the web sheet with the final stiffener, so that in the close the whole upper outer web corner tore away from the web peripheral frame. It was observed that the tearing off of the web sheet occurred in that portion of the web in which the diagonal tension bend developed and was anchored into the boundary frame. (iv) In some cases, phenomenon (iii) interacted with process (ii), ie with a crack advancing (with a certain delay as far as the first crack) along the transverse stiffener. An examination of strain gauge measurements indicated that the cracks initiated in those regions (near the transverse stiffeners or the upper flange) where the ranges of the principal surface stresses in the web sheet were maximum. The number of load cycles to failure, ie the life of the test girder, was a function of load range, and varied from tens of thousands of load cycles when the load diapason was large to millions of load cycles when the load range was small. The propagation of the fatigue cracks was a more or less continuous process. Never during the tests did we notice any sudden increase in the rate of crack propagation, such as to indicate an unstable crack growth and to announce a critical length phenomenon known from Fracture Mechanics. The collapse of the girders was presented in a typical shear failure mode, well marked plastic buckles constituting along the tension diagonal of the web panels and plastic hinges developing in the flanges. In the end Stage of their fatigue lives, all the test girders conducted themselves like ones with an opening in their webs, the opening being represented by the developed fatigue crack. The effect of web slenderness was studied by way of one series of experiments performed at the Institute of Theoretical and Applied Mechanics in Prague, namely via tests on girders of group (ii), described above, the main aim being to check whether there existed a maximum slenderness λn,b max such that for λ ≤ λ n,b max all effects of web breathing could be neglected. In this regard it was found out that the problem was more complex and that it was impossible to define λn,b max independently of other factors, principally of the intensity of repeated loading and the magnitude of the initial imperfections of web panels. The main conclusions of the experiments are dual: (A) The webs of all the girders being part of this group (ii) exhibited pronounced breathing. (B) Referring to the initiation and growth of fatigue cracks in them, it was observed that just in part of the experiments no cracks in the breathing webs were discovered. On the other hand, it was concluded that when (a) the maximum value Fmax of the cycling load F was high and (b) the initial curvatures of the breathings webs were not small, the mechanism of fatigue crack growth was very like to that which was observed in the tests on group (i) test girders with web slenderness of 250. It is normal practice in the fatigue analysis of civil engineering steel structures that this analysis is based on stress range only, the effect of the mean value of cyclic stress being completely disregarded. This is why ones of us wanted to verify whether this idea could also be applied for the fatigue assessment of breathing webs. 18 tests of this kind have hitherto been carried out at the Institute of Theoretical and Applied Mechanics. Although the scatter of the results obtained was rather large (but this is the case with all web breathing tests), it was rendered that at least in some of the experiments, the influence of the mean value was quite marked, see Table 1. But still, more proofs in this principle is indispensable. TEST GIRDER MEAN VALUE OF LOAD F [kN] LOAD RANGE [kN] NUMBER OF LOAD CYCLES TO FAILURE BTG 52 120 60 1 132 780 BTG 55 120 60 572 700 BTG 51 40 60 4 970 350 Table 1. Effect of the mean value of cyclic loading

RkJQdWJsaXNoZXIy MjM0NDE=