Download as pdf or txt
Download as pdf or txt
You are on page 1of 2

FURTHER NOTES ON THE K-METHOD 1. 2. FEA assumes the shape of the structure is constant during the analysis.

This shape is the input for the run. The external forces act upon this shape to produce deflections. The deflections computed by the FEA program are independent quantities. They indicate the movement of the structure under the loads applied onto the structure defined by the input coordinates. However, the shape of the structure remains unchanged throughout the analysis, even though there are deflections. Because the FEA assumes that the shape is unchanged by the analysis, there is always an error due to the shape change. This is what causes second order deflections. For the K-method, here are 3 concepts to understand :- the original shape, the input shape and the deflected shape. The original shape is the starting shape of the structure before loads are applied. The original shape is used for the first iteration. For each subsequent iteration, the input shape is adjusted and input into the FEA. The deflected shape is then defined by the deflections computed by the FEA added to the original shape. A first-order FEA achieves equilibrium based on forces acting on the input shape. The K-method achieves a structural equilibrium for the final shape of the structure. This is done by iterating until the final shape is the input shape. The calculated deflections are measured from the original shape rather than from the input shape. The difference between the original shape and the input shape is an indication of the strain energy stored in the system. Initial deflection is made up of D + 1, where D is the rigid body translation, and 1 is the distortion due to loads The initial distortion 1 acted upon by the axial load P causes an additional deflection 2 = K. 1, where K is the ratio P/Pcr. Subsequent additional deflection 2 = K. 1, 3 = K. 2 etc. Total deflection = 1/(1-K), assuming that K does not change. Classical buckling is based on sine curve deflection, due to sine curve moments from the P- effect. The shape of the moment curve does not change between iterations. If the moment curve is not a sine curve, the classic assumption does not hold, and K may vary, since the shape of the moment curve changes from iteration to iteration.

3.

1.

2.

1. 2. 3. 4. 5.

The paper by Albert Loh is acceptable but he may like to have the following comments. The directions of imperfections are required to be in the buckling mode and this hand method becomes difficult to apply. The author's approach is based on constant load Newton Raphson method and it will diverge for highly nonlinear problems, apart from coplication associated with 3-d structures and under many load cases. To remain competitive, engineers should be asked to adopt a newer technology. The paper should be accepted in the conference for presentation and publication as it contains a better method for design. Thank you for the comments on my paper. A lateral deflection is required to initiate the second order buckling effect. The method will work as long as there is a lateral deflection to start the buckling effect. This usually is no problem as there are lateral deflections in most analyses of anything except the simplest of structures. The starting lateral deflection does not need to come from imperfections; it can also come from the applied loads or from roundoff errors. (As a practising engineer, I do not like the latest trend toward modeling imperfections. What is needed is good engineering judgement instead of blind analysis. There are just too many imperfections and permutations of imperfections in the real world to ever calculate completely.) The proposed method yields very similar results to the Newton-Raphson type iteration used in many commercial programs. It will diverge for cases where the axial load exceeds the critical buckling load. For axial compression loads exceeding the critical load, the structure has reached buckling failure, so a divergent analysis correctly portrays the real situation. However, we have found a second case of divergence which is when the tension axial load exceeds the critical buckling load in magnitude. This leads to an oscillating divergence in the deflections, and obviously has no physical meaning. The exact same divergence was also present when we ran a conventional Newton-Raphson type analysis on a commercial program. This is a weakness of the linear formulation used in the structural stiffness matrix. Most experienced engineers are aware that slender tension members can yield erroneous results when loaded in the transverse direction. The workaround is to input an artificial additional stiffness due to axial tension in the slender tension members. Obviously non-linear cable elements are even better. Very often, an engineer may not have access to a non-linear program, and the proposed method allows him to make a second order analysis using a linear program with a little more effort. Although the procedure described in the paper appears to be tedious, it is actually quite easy to apply, as the deflection data can be easily cut and pasted to a spreadsheet for convenient number processing. The second order analysis is done only for critical load cases, and is quite practical to use, even for very complicated structures. We have used it on two different and very unusual roof designs with several hundred members, and the results correlate well with checks made using the conventional NewtonRaphson type analysis on a commercial program. An understanding of the structural behaviour is very important, and this manual iterative method helps in the comprehension of the second order behaviour, as it shows how the second order effect is influencing the structure at each successive iteration. Besides buckling, there are also other second order effects, such as axial shortening due to lateral loads, which will be interesting to research.

You might also like