No abstract

No abstract

The Authors have presented a lucid description of buckling tests on nominally identical specimens at four different slenderness ratios. With the current move towards semiprobabilistic methods for assessing safety" it will become necessary to conduct series of tests of this nature where the maximum control is kept on some of the more obvious objective variables involved in order to assess the subjective uncertainties which reflect man's imperfect knowledge.la Only then will designers have the best bases for choosing characteristic values and partial safety factors for strength.46. It can be seen from Fig. 6(c) that the buckling strength coefficients of variation from the Author's tests (Fig. 5 ) and those of the European group* show a peak close to 10% at l/r= 78 which is the slenderness value closest to (f/r)T. However, it seems that it is at this slenderness value that the minimum value of do occurred. It would therefore appear that most of the variation in strength must be due to something other than geometrical imperfections which are k n o q to exert their greatest effect in the (I/r)T region. Likewise the excellence of the test arrangements would seem to discount the possibility that end constraints or eccentricities have any significant influence. If this is true then the uncertainties illustrated by Fig. 6(c) (Which average about 4%) are more likely to be subjective than objective. Can the Authors confirm these inferences and explain this scatter?47. In the same vein Fig. 11 may be of interest. It shows the result of compression tests on thirteen nominally identical uniaxially tee bar stiffened plate panels designed in the buckling-yield transition.18 The experimental conditions were carefully controlled to minimize their uncertainties. The panels were made from mild steel selected and constructed under normal shipyard conditions. The effect of stiffener shape imperfections was negligible but the effects of residual welding stresses and plate deflexions were apparent but moderate. The objective random uncertainties, which were dominated by variations in yield stress, had a coefficient of variation of about 6%. Subtracting this from the experimentally found total uncertainty of 10.9%, and assuming statistical independence, suggested a typical subjective random uncertainty for such structures of 9-10%. This is similar to that obtained by the Authors, although somewhat greater overall as one would expect for a more complex fabricated structure.48. I am surprised at the cusped nature of some of the figures at the transition slenderness. Of course many theoretical approaches show greatest sensitivity to shape and material imperfections at this transition, but it is my experience and belief that such discontinuities seldom appear in practical structures during test.