31。17 0。114 25 0。27

31。93 0。121 27 0。295

32。1 0。130 29 0。335

32。12 0。144 30。13 0。385

32。07 0。165 30 0。405

31。82 0。197 29 0。465

31。31 0。244 28 0。495

30。4 0。314 26 0。515

29。11 0。418 23 0。565

27。15 0。575 19 0。665

24。51 0。814 14 0。925

21。3 1。184 12 1。115

and theoretical prediction based on average material data are shown in Table 4。 It can be seen that good repeatability of test results   were   obtained   for   1 mm   cylinders   (84。67 kN    versus

85。33 kN) and 2 mm cylinders (210。32 kN versus 208。56 kN), respectively。

4。Comparison of experimental and theoretical prediction

Five test data are shown in Table 4。 Comparisons of experi- mental results were made to the reference buckling load to cause yield of the cylinder as presented in Eq。 (2)。 Since the axial load is applied relative to the middle surface of the cylindrical shell, the diameter,  D,  in  Eq。  (2)  is  taken  as  the  mid-surface    diameter D (where D = Dinner + t , i。e。, the sum of the inner diameter of the cylinder at the point of load application and the thickness of the cylinder)。

Table 4 shows theoretical predictions of magnitude of collapse force obtained using Eq。 (2), in which the average wall thickness

and average mid-surface geometry (Table 2 and Table 3) were used。 In addition, the yield stress was taken as the average upper yield stress obtained from uniaxial tensile test (column 3 of Table 4)。

From Table 4, theoretical reference load predicts the magnitude of collapse force close to the experimental collapse load [(32。55 kN versus 30。13 kN) for CY1_t0。5, (83。36 kN versus 84。67 kN) for CY1_t1。0, (88。73 kN versus 85。33 kN) for CY2_t1。0, (207。97 kN versus 210。32 kN) for CY1_t2。0 and (208。53 kN versus 208。56 kN) for CY2_t2。0。 Furthermore, based on the measured value of the cylinder geometry and obtained material properties, ABAQUS FE code was used to compute the buckling load for the entire five cylinders。 Comparison between experimental and numerical (ABAQUS) plot of load versus compression extension, for cylinder CY1_t0。5, is shown in Fig。 5。 The corresponding magnitudes of load and compression extension are shown in Table 5。 Both plots follow a similar path up until collapse。 It can be seen here that the ABAQUS FE predictions of the slope underestimate the experi- mental slope。 The discrepancies between the slope of FE predic- tions and the experiment (i。e。, magnitude of the compression ex- tension) can be attributed to the boundary condition assumed for the FE predictions。