Seismic Response Reconstruction and Evaluation of Nonlinear Static Analysis Procedures Using Data from Instrumented Buildings

In Nonlinear Static Procedures (NSP) the maximum structural responses are predicted from the result of pushover analyses in which the structure reaches a "target displacement" determined from the response of an" equivalent" SDOF system to the seismic loading, defined as an elastic spectrum. This study evaluates the predictive capability of NSP using seismic data recorded on instrumented buildings and examines the issue of seismic response reconstruction. In connection with the second item it is shown that commonly used interpolation schemes are tantamount to fitting the response into a basis whose span is dictated by the number and location of sensors. A scheme, designated as the Minimum Norm Response Corrector (MIRC), which estimates the response as that of a nominal model subjected to the seismic excitation plus a set of pseudo forces that enforce the measurements is presented. It is found that MIRC provides estimates of the response that are significantly more accurate than that from interpolation schemes, especially for shears and drifts.;The single and multi-mode NSPs are evaluated by contrasting the predictions with the "true" values inferred from the measurements using MIRC. It is shown that the single-mode procedures provide reasonably accurate estimates of all responses in the short building but poor prediction of responses when the contribution from higher modes is substantial, i.e., shears and drifts in the tall buildings. An energy consistent (EC) SDOF is developed and is shown to result in more accurate predictions of roof (target) displacement than those used in the typical NSP. Incorporation of the EC SDOF into single and multi-mode procedures is presented as an alternative NSP and is evaluated. It is shown that while multi-mode procedures significantly improve the predictions in the tall buildings, a single-mode procedure using the EC SDOF and two load patterns, namely, one from the maximum linear story shears and the other from the code loading, gives prediction that are on par with those from multi-mode procedures.