Review of

L.F. Hart, T.S. Oo, F. Kal, Shell Structures, Bend & Break, Report for course CT3280 Shell Roofs, Delft University of Technology, Department of Civil Engineering and Geosciences, 2024, online: phoogenboom.nl/B&B_schaal_report_5.pdf

by dr.ir. P.C.J. Hoogenboom, Delft, 21 February 2024

 1. The reports has a clear structure. The report should have an abstract. The report should have a preface explaining the context of the work, The report should have a literature list, Few language errors. Handed in on time (19 Jan. 2024).

 2. Page 2. The introduction is unusually short. Almost perfect. Cut a few more words. Seriously.

 3. Page 3. Good that you came up with an original shape. There need to be a drawing with dimensions. After all, others need to be able to reproduce the experiment from this report.

 4. Page 5. The construction process is described on just one page. Nonetheless, it is complete.

 5. Page 7. Some of the hand calculations are okay. Some of the hand calculations go round and round instead of straight forward. This is not a good way of presenting calculations. It is uncheckable.

 6. Page 8. The 212.22 kg hand calculation is obviously ridiculous. How can the complete collapse load on the roof be the buckling normal force in one lath? Please, do not become a professional structural designer. Moreover, apparently, nobody checked the calculation. There are three people in a team and there should be some checking.

 7. Page 9. Good job, making the computer model. The supports are too complicated. No need for this.

 8. Conclusion of the study on page 12 and 13: For the small model self-weight is negligible. For the large model self-weight is not negligible. Self-weight needs to be switched off otherwise we do not know the magnitude of the load on the roof. When we switch of self-weight, we do know what load gives what stress and what deflection. This relation is linear, provided the load is not very large.

 9. Page 14. There should have been shape imperfections in the nonlinear analysis. Surely, the built shell was not perfect. It was demonstrated in the practical lecture how to include imperfections. It was demonstrated that imperfections can reduce the collapse load significantly. Hypar roofs (yours) are a known exception to imperfection sensitivity. Nonetheless, it would have been smart to check.

10. The same line load of 1 kN/m was applied to the small model and the large model. This is wrong. The distributed load [kN/m2] should have been the same. So, the line load on the large model should have been 100 kN/m. In addition, self-weight should have been switched off (see comment 8). Then the small and the large model would have produced the exact same buckling load factor of 0.18 (see scaling rules). Conclusion: we can use small physical models to determine the strength of big structures.

11. Good that those computations were made. In fact, this is playing with the software. Did you learn from it or did you get confused? It is simple; if you give the program the correct input, the program gives the correct output. Why did you not follow the example in the practical lecture (handout 6) which everybody did? By the way, this playing around part should have been put in an appendix.

12. Page 21. I do not agree that SCIA model was not accurate. It predicted the failure mode well (Fig. 29, 34, 35). I cannot tell whether it predicted the failure load correctly too because this result of the computations is not in the report. It should have been.

13. Page 22. The test set-up is clear.

14. Did you check the strengths of the ropes and the distribution woods? This is not in the report.

15. Page 23. What is the unit of the measured displacement? Why aren't decimals recorded?

16. Page 23. The collapse mode shows that the rope was attached to the bottom lath, not to the node. Is there a picture of this? So, it could have been a shear failure of the lath, and not buckling.

17. Page 24. Solid analysis of the collapse.

18. What about the real size shell? This objective is also formulated in the introduction. How large can it safely be? How will the real size edge beams look? How can it be constructed? How to make the joints? 

19. Aren't there additional conclusions to draw? What do you think about designing by model testing? Does it work? What do you think about the finite element analyses? Are these difficult? Are the results reliable? What do you think about the hand calculations? Are these useful? Do the experimental results match those of the computer calculations and those of the hand calculations? Engineers who read your report, would want to know this.