Modified Stiffness for seismic?

[Utve]

Ductility has been assumed to get the seismic input. However, in my model for analysis, can I not change the stiffness of all the columns and walls and design to these original stiffness as well?
I find this approach might not be conservative for some walls because let's say I have wall A and wall B and once wall A cracks and stiffness is thus reduced, the lateral load will thus redistribute to wall B, so wall B will get more loads than the unmodified stiffness model. But at the same time, wall B should be allowed to crack and design to cracked section although the lateral load distributed to it increases.
I understand codes give information on how much percentage of the stiffness can be reduced but I also believe each wall and column should be examined to decide how much stiffness they will lost instead of blindly apply same modified stiffness ratio to them as a whole. But this seems to me a onerous iteration: wall A cracks - reduce A stiffness - load goes more to B, B cracks, loads goes to A or C and then this circle keeps going... I believe this is not what "SIMPLIFIED" method that the code wants it to be? (a non-linear approach seems more 'easy' to this simplified elastic analysis if this is the case)
Anyway, I end up coming up with this question, can I simply not change stiffness of all walls and columns? How do you usually deal with the wall and column stiffness?

 

[hardbutmild]

I believe you can not get the "real" values no matter what you do. Because of that, it seems unnecessary to do anything more complex than the code suggests.

You can't know which vertical load will actually be applied at the time of an earthquake, the variability of concrete material is such that you can't really know which element has cracked (you can assume they all have the same tensile strength which statistically is impossible, the placing of concrete is not uniform), especially when you consider the long term effects. On top of that, stiffness depends heavily on the reinforcement so you'd have to iterate the whole procedure, not just the first part. Also, you'd need to know the loading history because that can influence what is cracked (e.g. under load case 1 a wall B will crack, but under load case 2 a wall C will crack. Which wall is cracked at the time of an EQ? It depends on the loading history and you can't know that since you don't know when the EQ will come, or how exactly will the structure be used. Think like in plate design, where you do different combinations to get the "worst" one for ULS... how do you know which of these combinations will come before and which after the EQ? Or will they actually come at all). In addition, the stiffness changes (and the load travels) during an earthquake and doing elastic analysis (or even if you're doing inelastic static analysis) you need to assume an equivalent stiffness depending on the maximum expected displacement and maximum expected internal force in the element under that specific EQ, which is always extremely approximate. Depending on the shape of the EQ wave and the direction the walls will crack in a different order and to a different extent. The soil is not of uniform quality... the list can go on for a long time.

In short, anything more complex than the code provisions is only fooling you into a false sense of security (unless you actually do a ton of testing... even then it's not really an exact science).

 

[Mohanlal0488]

Instead of of changing the stiffness why don't you just use direct integration methods (transient analysis) and include non-linear geometry and material as this will allow redistribution.

I have seen a number of codes which have a clause which allows one to perform analysis using this method.

 

[Blackstar123]

I believe this is not what "SIMPLIFIED" method that the code wants it to be?

I believe the opposite is true. Code is allowing you to design for the final cracked state, that is, when A to C all walls are cracked. This is what you described in the post right?

As I understand, the idea behind modifying the properties of a member is
1. To have some sort of consistency between a linear static analysis and strength based design approach.
2. To have realistic estimates of deformation at design load level.

Ideally, the analysis should be based on the stiffness that takes into account the degree of cracking expected to occur in each member during a seismic event. However, as hardbutmild have explained that calculating different stiffnesses for each member will be very complex and it would make the analyses inefficient.

As far as I know, the effective moment of inertia values given in code are not arbitrary but based on test results and is a simple way to take into account the structure non-linearity.