The short answer to your question,
How can we compute the Equivalent Static Lateral Forces, the overturning moment and other quantities in an irregular building/structural discretized by finite element?
is, simply, you can't. ASCE7-10 speaks directly to this in Section 12.6 when it describes the conditions under which Equivalent Lateral Force (ELF) can and cannot be used. If your structure is located in an area with moderate to high seismic activity (SDC C-F), you cannot use ELF if your structure has "structural irregularities." See Table 12.6-1 in ASCE7-10 for details, I've simplified a bit. There are certain structural irregularities that do not preclude ELF.
That said, there are a couple different ways you can go about calculating equivalent static seismic forces for highly irregular structures using FEM. These forces are not directly analogous to the forces you'd get using ELF method from ASCE 7, but they are forces obtained via rational analysis and commonly used/accepted in order to simplify the seismic analysis of irregular structures. They are:
- Perform a (fixed base) time history analysis on your structure. Use the results to generate acceleration response spectra at each level of your structure. Use the accelerations from all of your RS to design the structure. (Most conservative)
- Perform a (fixed base) time history analysis on your structure. Pull the peak acceleration in all 3 directions across all your time steps for each node. Use these accelerations to design the structure.
- Perform a (fixed base) time history analysis on your structure. Use the accelerations at each time step to design your structure (at each time step). The most conservative design controls. (Least conservative)
Time History to RS Method (#1)
This method is the most conservative of the three. The general procedure would be to apply the peak acceleration from your response spectra to the structural elements at each level. In my industry, when using this method, we even multiply the peak acceleration by 1.5 to ensure that we envelope the actual acceleration including any multi-modal effects. This is generally only done for designing components within the structure or modifications to an old structure when time & budget do not allow a more accurate (\$$$$) analysis. It is considered less conservative for component design because you, theoretically, know the natural frequency of the piece of equipment you are working with, and therefore can pull a more accurate acceleration off the response spectra.
Max Time History Acceleration Method (#2)
This method is fairly straightforward. After performing your time history analysis, you can extract acceleration time histories for every node in your FE model. For each node in your model, pull the highest acceleration in the X, Y, and Z directions. Multiply these accelerations with your tributary nodal mass to obtain an equivalent static seismic nodal force in all three directions. This is still considered (highly) conservative because you are pulling accelerations from different time steps and combining them.
There are several variations of this method related to exactly which accelerations/time steps you design to. The tricky part is in convincing yourself and/or the body having jurisdiction over your design that you are enveloping reality.
Time step Time History Acceleration Method (#3)
This method is similar to #2 except you will design your structure/run your analysis for the seismic accelerations at each time step. At each time step, for each node in your model, pull the acceleration in the X, Y, and Z directions. Multiply these accelerations with your tributary nodal mass to obtain the equivalent static seismic nodal force in all three directions.
This method is the least conservative but, by far, most computationally expensive approach and is mostly unrealistic. Your time histories will likely have upwards of 10,000-100,000 time steps. It is incredibly expensive to chug through the analysis.
I've personally used all 3 methods at one time or another in my career. #2 seems to hit the sweet spot with respect to the accuracy:expense ratio.