Seismic Frequently Asked Questions
- Under what circumstances should we use a grid model of the structure rather than a stick model?
Stick models are the easiest to use. Most normal structures can utilize properly modeled stick models. For structures that have large skew, curvature, the stick model should be expanded to a 3-D model at the abutments and piers to ensure loads are properly applied to the substructure unit.
- Flared column with interlocking spirals - should the column be modeled as a constant cross-section (this assumes that the concrete outside the spirals spalls off)? Or should we model the column as a varying cross section?
To ensure a conservative design, you should look at both situations. In the case the details provided do not allow the flare to fall or if an event occurs that is smaller than expected, significant forces will be attracted to the flare. The flare also has significant mass to become excited during an event. However, if the flare falls, mass is removed and forces will be distributed to other substructure units.
- Given a two column bent with a stick superstructure model - should the bent model be modified (increased Iy?) to eliminate the mode that shows the cap bending in its weak axis due to the lumped mass of the superstructure being applied to the center of the cap?
The bent should be modeled with the proper properties. However, the superstructure should be expanded into a 3-D model with rigid elements and the girders be placed at the proper bearing locations. In reality, the bent cap is restrained by bridge deck or girders and can not move that easy. Even if the substructure was made very stiff, the mode that shows the cap bending about its weak axis will eventually show up. It will just require additional modes be calculated.
- Show how to use FEM output to check and modify the initial estimate of the spring constants. Are the models more sensitive to changing certain spring constants vs. others?
The actual initial stiffness of the springs is not important. Once an initial run is made, the loads should be placed in a soils program that properly models the foundation and soils. The deflections from the soils program then should be compared to the FEM. The spring stiffness should be modified until the deflections compare with each program.
- How do you know when a "good" model has been achieved?
There is no one "good" model. In reality, you must build several models to envelope the forces on each individual component designed. For example, create a soft abutment to drive the force into the piers for pier design. Create stiff abutments to drive the forces into the abutment for abutment design. When building your model, start simple to learn how the structure behaves. Slowly begin to add complex features such as soil springs, cracked section properties, etc.
When the results do not change with the addition of the new complexities, you have likely created the "best" model.
(1.) Model testing and debugging should be performed once model is done and inputs have been checked. (2.) Dead load reactions should be as expected. (3.) The fundamental period calculated by hand should be close to the model output.
- In low seismic zones no analysis of substructure is permitted. However, the connection of the superstructure to substructure is required to resist 0.1 x DL or 0.2 x DL. Shouldn't we check the substructure for these forces? Shouldn't we use these forces as a minimum in higher seismic zones as well, even when analysis shows smaller forces?
If you design the connection for a force, you must trace this force through the entire system until it dissipates into the soil. Yes, you must design the substructure for the force. For low seismic zones, for example Zone 1, the Code does not require bridge seismic analysis. For region where 0.0g<A< 0.025g, the maximum possible acceleration experienced by the bridge is 2.5 times 0.025g which equals to 0.06g and the Code requires to use 0.1g for the connection design (LRFD 188.8.131.52). Similarly, for region where 0.025g<A<0.09g, the maximum possible acceleration in the structure is 2.5 times 0.09g or 0.225g and the Code requires to use 0.2g for the connection design. It can be seen that both 0.1g and 0.2g experienced by the bridge are extreme values for the low seismic ground acceleration. Therefore, to apply R-factor (usually less than or equal to one at connection) to 0.1g or 0.2g for the connection design will be too conservative. For the same reason, if a seismic analysis is preformed, forces from analysis should supersede the 0.1g or 0.2g assumptions and R-factor should be applied to the analytical results regardless of low or high seismic zone.
- Explain the rationale of using half of the response modification factors in foundation design vs. factors used for column design in seismic zone 2. Response modification factors > 1.0 should only be used for ductile members, while foundations other than drilled shafts or pile bents are generally not ductile.
R-factor greater than 1.0 should only be used for ductile members for seismic elastic analysis. Therefore, foundation design has to ensure that the connection at pile to pile cap is stronger than pile capacity. For this reason, the pile to pile cap connection should behave elastically during seismic event and R-factor should be taken as 1.0. In addition, some other detailing requirement, such as confinement zone, must also be satisfied to prevent plastic hinge to occur at this interface. AASHTO LRFD C184.108.40.206 discusses this issue and states "it is recommended that for critical and essential bridges in Zone 2 consideration should be given to the use of the forces specified in Zone 3 and Zone 4" which requires R-factor to be taken as 1.0. Recommended Guidelines for the Seismic Design of Highway Bridges (NCHRP 12-49) also state that the R-factor to be 1.0 for the foundation design. The development of LFRD Guidelines for the Seismic Design of Highway Bridges, Version 2 is currently in progress under NCHRP 20-07, Task 193 and hopefully this issue will be address clearly in the new specification.