Welded hemispherical or spherical shells in practice have initial geometric imperfections in them that are random in nature. These imperfections determine the buckling resistance of a shell to external pressure but their magnitudes will not be known until after the shell has been built. If suitable simplified, but realistic, imperfection shapes can be found, then a reasonably accurate theoretical prediction of a spherical shell's bucklinglcollapse pressure should be possible at the design stage.The main aim of the present paper is to show that the test results obtained at the David Taylor Model Basin (DTMB) on 28 welded hemispherical shells (having diameters of 0.75 and 1.68 m) can be predicted quite well using such simplified shape imperfections. This was done in two ways. In the jirst, equations for determining the theoretical collapse pressures of externally pressurized imperfect spherical shells were utilized. The only imperfection parameter used in these equations is 6,, the amplitude of the inward radial deviation of the pole of the shell. Two values for 6 , were studied but the best overall agreement between test and theory was found using 6, = 0.05J(Rt). This produced ratios of experimental to numerical collapse pressures in the range 0.98-1.30 (in most cases the test result was the higher). The second approach also used simplified imperfection shapes, but in conjunction with the shell buckling program BOSOR 5. The arc length of the imperfection was taken as simp = kJ(Rt) (with k = 3.0 or 3.5) and its amplitude as 6 , = O.O5J (Rt)
. Using this procedure on the 28 D T M B shells gave satisfactory agreement between the experimental and the computer predictions (in the range 0.92-1.20). These results are very encouraging. The foregoing method is, however, only ajirst step in the computerized buckling design of welded spherical shells and it needs to be checked against spherical shells having other values of Rlt. I n addition, more experimental information on the initial geometric imperfections in welded spherical shells (and how they vary with Rlt) is desirable.A comparison is also given in the paper of the collapse pressures of spherical shells, as obtained from codes, with those predicted by computer analyses when the maximum shape deviations allowed by the codes are employed in the computer programs. The computed collapse pressures arefrequently higher than the values given by the buckling strength curves in the codes. On the other hand, some amplitudes of imperfections studied in the paper give acceptable results. It would be helpful to designers i f agreement could be reached on an imperfection shape (amplitude and arc length) that was generally acceptable. Residual stresses are not considered in this paper. They might be expected to decrease a spherical shell's buckling resistance to external pressure. However, experimentally, this does not always happen.{ 12(1 -v ' ) }~/~~J ( R /~) z 1.82a,/(R/t) for v = 0.3 slenderness parameter = J(pyp/pcr) = 1.285J{(R/t)(ay,,/E)} for a spherical shell Poisson ratio...