The results obtained from ABAQUS FE predictions are presented in Column 7 of Table 4。 From Table 4, it can be observed that there is very good agreement between prediction from Eq。 (2) and nu- merical results obtain from ABAQUS simulation (with all ABAQUS results repeating the same results as the theoretical prediction, except for cylinder, CY2_t1。0, with 2% difference between the two predictions)。 Also, there is a good visual comparison of deformed shape of ABAQUS simulation and experiment, as exemplified in Fig。 6 for cylinder CY1_t0。5。

For 0。5 mm cylinder, theoretical prediction overestimates the experimental result by 7%。 Also, for 1 mm cylinder, theoretical prediction underestimate the experimental collapse force (for cy- linder CY1_t1。0) by 2% and overestimate the experimental collapse force (for cylinder CY2_t1。0) by 4%。 Whereas, for 2 mm cylinder, theoretical prediction underestimate the experimental data by 1% (for cylinder CY1_t2。0)。

Also, theoretical prediction of collapse force given  by  Eq。 (2) was compared to past experiment on axially compressed cylinder as presented in Ref。 [18]。 Cylinder was machined using Computer Numerical  Controlled  (CNC)  machining,  with  integral  top    and

Fig。 6。  View of collapsed cylinder CY1_t0。5 (a), and its ABAQUS computed collapse shape (b)。

Fig。 7。 Photograph of CNC machined cylinder with flanges (a), and clamping arrangement with top and bottom covering plates prior to testing (b), adopted by Blachut [18]。

bottom flanges (see Fig。 7a)。 The axial length of the cylinder is said to be the same as that of the present work。 The cylinder is as- sumed to have a nominal thickness of 3 mm and nominal diameter of 169。22 mm。 Prior to testing, cylinder was covered with the top and bottom plates to create the desired boundary conditions as shown in Fig。 7b。 For wants of better comparison, average values of measured data were used (i。e。, average mid-surface diameter   of论文网

169。13 mm and average wall thickness of  2。90 mm)。  In addition, the average upper yield stress obtain from material testing of material  from  which  the  cylinder  was  made  is  assumed  to    be

230。6 MPa。

Result indicates that theoretical prediction underestimates the experimental collapse load by 17% (355。33 kN versus 430。0 kN)。 Apparently, there is close agreement between the theoretical prediction of this present work and the numerical prediction re- ported in (Table 6 of Ref。 [18]), about 1% difference (355。33 kN versus 359。46 kN)。 It is difficult to explain this phenomenon, but it is assumed that the large difference between  theoretical predic- tion and experimental buckling load can be attributed to the effect of the top and bottom plates used to cover the cylinder to create fixed  boundary  condition。

Overall, it can be seen from this piece of work that there is very good agreement between theoretical prediction, numerical pre- diction and experimental collapse load for all the cases of radius- to-thickness ratio, R/t, considered。 Also, results reveals that the higher the radius-to-thickness ratio, the higher the difference margin。 This is adjudge to be due to the manufacturing techniques adopted (i。e。, the bigger the thickness of the plate, the better the MIG welding process)。

5。Conclusions

Two sets of nominally identical cylinders (1 mm and 2 mm wall thickness) failed at almost the same magnitude of collapse force。 Hence, confirming repeatability of the experimental results。 In addition, theoretical reference buckling load based on Eq。 (2) and ABAQUS predictions for all tested cylinder gives a close result to the experimental value as shown in Table 4。 It is apparent that the results are in very good agreement as the reference buckling load predicts the same results as ABAQUS FE code and very close results to the experiment with difference ranging between — 7% and þ2%。 Hence, this study concludes that the manufacturing process can influence the theoretical prediction as compared to the experi- mental data, but this does not result in large differences。 Rather, it is suggested that the type of boundary condition (edge support) employed  during the experiments  must be carefully   considered。