Q & A Session

We had an interesting and lively Q&A session during the "Some Failures & Studies With Lessons Learned - Buckling, Instability & Collapse" webinar and we wanted to share with you the questions that we didn't have the time to address during the event.

Q: Does buckling always lead to failure?

A: As you will have seen from the examples in the webinar, buckling can involve stable and unstable responses. With most structures, buckling is a response to be avoided, as damage and collapse can occur. However there are also components that are designed to buckle many times during their life … but, even in this case, there may be fatigue damage accumulating. So, buckling may not always result in immediate failure, but may so eventually. The outcome depends on the type of buckling, the structure, material and its response. It is certainly something that a designer will always want to know about.

Q: Do I need a nonlinear FEA system to carry out accurate buckling calculations?

A: If you are interested in determining an accurate response after buckling has occurred, you will need a physical test or a good nonlinear post-buckling simulation (arguably both, with uncertainty quantification, for validation purposes). If you are only interested in predicting an accurate or safe buckling load, it is possible to use experimentally derived factors in conjunction with an approximate eigenvalue buckling load, to accurately predict the onset of buckling.A nonlinear analysis, with realistic imperfections will normally provide a more accurate prediction and also has the potential to provide the post-buckling response.

Q: How do I know if I need to do a buckling analysis?

A: The simple answer is that the code of practice you may be using will generally require you to! If you are not being guided by a code of practice and you need to anticipate whether buckling is possible in the load range, you can use a combination of simulation and test to establish the possibility.If buckling is a possibility it would always be wise to establish the design margin on buckling for all loading cases and combinations.

In general, FEA systems don’t automatically flag the possibility of buckling and don’t advise on the many possible pitfalls either… it requires competent engineers to anticipate the possibility (paying attention to long slender components under compressive load, or large unsupported areas of shells under compression etc). Constraints on expansion can also be a cause. Time-dependent phenomena such as creep, corrosion, erosion etc should also be borne in mind.

Q: Is buckling a static or dynamic phenomena?

A: If applied loading is very slow (quasi static), inertial / dynamic effects can possibly be neglected and the analysis treated as static linear or static nonlinear. If there is likely to be significant inertia or acceleration in the response, dynamic effects can be significant and hence should be included.

The significance of dynamics will also depend in many cases whether you are interested in simply predicting the response onset, or whether you are trying to determine the post-buckling response - which can involve sudden releases of energy, sudden movements, damage & collapse. In this case you may well require to model the dynamic response along with all the various nonlinearities

Q: Is it possible to have applied case on marine application?

A: The main marine example that I use in my DBA course material is the Alexander Kielland disaster. There are quite a few interesting lessons in this case and I would recommend a read of the material available on the web. The lessons from Kielland cover quite a few of the recognized causes of disasters listed on my slides, including both fatigue and instability. This failure also led to changes in design codes of practice. Poor material and manufacture also came into play. I am sure there will be quite a few more marine examples in the texts I listed at the start of my presentation. The SS Thresher submarine loss also springs to mind as does the Ro-Ro ferry capsizes eg Herald of Free Enterprise and others. Welded ship failures, including the Liberty ships of WW2 is another good failure example with lessons. Not all of these are examples of buckling, but involve general instability and design flaws – that led to capsizing, unstable crack growth .

In the early days of the deployment of Kevlar ropes a phenomena called “bird caging” was identified, which caused failure of a 200T, 64mm diameter rope on a mooring buoy. The bobbing of the buoy in the sea caused axial compression and repeated twisting, flexing and buckling of Kevlar fibres … which in turn caused failure of the braided rope. There were moves made after this to use torque-free ropes, keeping ropes in tension, avoiding tight jackets and increase fibre freedom to allow smooth buckling rather than sharp kinks in the fibres.

Q: Any suggestions how to cope with slow degradation of polymers due to UV eg sunlight?

A: My experience with plastics and this well-known phenomena is confined to the use of exposure test specimen data in design calculations. In this case, modified stiffness and strength data would be used for a particular design life / exposure time. Once again, getting such quality data for a particular plastic type and grade may be problematic as this is another time-dependent phenomena. As with fatigue and creep, accelerated testing may be problematic.

Actually modelling the slow degradation is another challenge and something that I have not been involved in. I would be surprised if someone has not attempted to simulate this process and the basic modelling andnumerical procedures, will probably be in place. Modifying modulus and strength in a time-stepping analysis, according to a known degradation rate/form will be a different challenge to actually trying to model a variation of UV across a structure over time (eg using a ray tracing algorithm as used in radiation heat transfer) and trying to predict change to the polymer molecular structure.

A quick search on Elsevier’s “Engineering Village” reveals that there would appear to be useful references eg “UV degradation model for polymers and polymer matrix composites” (Open Access), Lu, T. et al, Polymer Degradation and Stability, v154, p203-2110, August 2018.

Q: Is it tension load in the Kevlar umbilical test?

A: Yes.

Q: Generally, we try to avoid the cases that thin-walled structures are under in-plane compressive loading in designs, correct?

A: Yes. This is in essence also a conclusion from the thoughts on preventing buckling instability in the penultimate slide of the presentation.

Q: Regarding accelerated creep and strain rate testing on plastics, are you familiar with time-temperature-superposition and if so, what's your opinion about it?

A: I am aware of time-temperature-superposition to accelerate creep testing as being a long-standing method. A search on Elsevier’s “Engineering Village” of the terms polymer creep time temperature superposition reveals 329 records in Compendex, with 12 in 2019 and 13 in 2020 so far. Clearly still an active area of study.

The idea is very attractive as providing comprehensive creep data for all variables of interest for plastics arguably remains an impractical task in a reasonable real time-scale. Replacing a successful “sitting tenant” material, for promised increases in performance and profit, is however well recognized as carrying risk. Use of any such techniques would obviously still benefit from correlation with actual long-term behaviour in the design environment and it would be interesting to see if anyone has deployed Uncertainty Quantification in their studies.

You will gather that I am somewhat sceptical about the validity of accelerated testing methods for polymers in general (including fatigue). That being said, I do recognize the need for generating good quality long-term material data in shorter timescales. Such research and new method development by polymer scientists and engineers will be critical for such developments.

In the case of the tank failure in my presentation, neither the tank manufacturer nor their material supplier were able to provide the creep data that I was looking for in the failure study and no doubt the claimed manufacturability and cost benefits were significant enough to consider such a change of material.

The failure study also had a focus on the effect of different support configurations.

Q: Does non-linear analysis post buckling calculation have a good correlation with tests? What is the best way to use the nonlinear approach to calculate the critical loads?

A: In general a nonlinear post-buckling analysis (as demonstrated in the CFRP panel buckling with cut-out examples), will provide better correlation with reality … not only with establishing the buckling load, but also the response of the structure after buckling has occurred. Large displacement can be combined with other nonlinearities such as contact, plasticity etc, if necessary.

I believe the example I showed in my presentation is indicative of the method and can be appropriate in design. There are other variables not addressed in the studies reported however that should be addressed depending on circumstances and goals. Dynamic effects could be important in some circumstances, if inertia effects, energies, impact and time dependencies are significant ie a static approach is not appropriate … this would be appropriate if the focus had been to model damage and collapse.

Q: We find that lower bound/small displacements problem in ASME VIII.2 Chap.5 they talk about that, e.g., a pipe elbow under in-plane bending, shall be evaluated using 5.2.4.

A: Yes indeed, Part 5 of ASME VIII Div2 defines the DBA requirements and the 3 approaches allowed (Elastic, Limit & Plastic Collapse [elastic-plastic]). Furthermore 5.2.3 indicates that the limit load option, as defined, with elastic-perfectly plastic material model, strain-displacement relations that are those of small displacement theory and equilibrium satisfied in the undeformed configuration, provides a lower-bound to the limit load of a structure. In 5.2.3.3 the point is made that for components that experience a reduction in stiffness with deformation eg a pipe bend under IPB, the elastic-plastic method in 5.2.4 shall be used. The onus is on the user to recognize this and the pipe bend is but one example however. Section 5.2.4 also makes the point that the elastic-plastic approach provides a more accurate assessment of protection because the actual structural behaviour is more closely approximated … partly because the deformation characteristics of the component are considered directly in the analysis. The elastic-plastic approach includes non-linear geometry and can also include hardening or softening post-yield behaviour.

Q: Do you mean Limit Load should take large displacement into account, not small displacements as in 5.2.3.1?

A: Not quite … the pressure vessel DBA codes provide choice in how design checks are carried out. This typically will allow a choice between linear analysis and non-linear analysis. As indicated in the previous answer,the nonlinear options will allow further choice between limit and plastic collapse (small displacement + elastic-perfectly plastic constitutive OR small displacement plus a more representative constitutive behaviour). If a weakening mechanism is present (see presentation slide for a list of some examples) then the codes will expect you to use a large displacement /nonlinear geometry solution.

With simple stress analysis, calculations of stresses are based on the undeformed shape. Large displacements / nonlinear geometry allow the deformation of the structure to provide more accurate, displacements (also strains and stresses). This has the overhead of using a step-iterative solver, where loading is stepped and matrices are re-evaluated, iterated and solved at each step. This overhead in computational resource can be a problem, but with todays solver technology and available computational resource, this is less of a problem. In addition, if matrices are already being re-evaluated due to material plasticity effects ie requiring a step-iterative solver anyway, the extra overhead of large displacements / nonlinear geometry is increasingly acceptable. If large displacements are likely then including this in analyses would arguably be a safe choice. If it transpires that large displacement effects are not significant, in general you will only have increased solve time.

Q: Are the presentations slides available for this webinar?

A: Yes, they are available from ResearchGate - simply browse to the ResearchGate page ( https://www.researchgate.net/), select Browse Publications from this page using using James Wood Strathclyde University. Click on the appropriate search result and that should take you to all relevant publications. Browse down and the webinar presentation should be near the top. You can download this and others from there.