Seismological criteria for selecting and scaling real accelerograms for use in engineering analysis and design

Student: Ana Beatriz Acevedo
Supervisor: Dr. J.J. Bommer



When dynamic analysis is needed, seismic actions need to be defined in terms of accelerograms. There are three fundamental types of accelerograms available: artificial records compatible with the design response spectrum, synthetic records obtained from seismological models and real accelerograms recorded in earthquakes. This work focuses on the most accessible solution for engineering applications: the selection, and when needed scaling, of real accelerograms.

The first issue addressed is the criteria used to select real records. The two basic options are to select records that match some specified feature of the ground motion, such as the elastic response spectrum, or to select on the basis of an earthquake scenario. The latter is shown to be a preferable approach, with the minimum parameters being the magnitude, distance and site classification.

Once the records are selected, it will generally be necessary to scale them to match the elastic response spectrum of acceleration. Matching criteria must be established to judge the degree to which the scaled records are compatible with the design criteria, for which it is judged that at least two of the following should be considered: divergence, over a given period range, between the mean of the scaled records and the design spectrum; covariance of the scaled spectra; maximum residual between a scaled record and the design spectrum.

When the criteria for matching are chosen, the final issue is to choose a scaling technique to achieve the required adjustment to the real records. Based on the premise that the scaled records should retian the characteristics of real records, options based on spectral amplitudes and intensities are explored to derive linear scaling factors to be applied to amplitudes; scaling of the time axis is discouraged because it changes the frequency content and duration without altering the number of cycles and is hence unlikely to result in realistic motions. An important issue that is addressed is the maximum scaling factor that can be applied, since the limit of two established by Vanmarcke (1979) has been widely adopted in practice without any rigorous re-evaluation. The conclusion is that provided the selection criteria outlined above are used to obtain the suite of real accelerograms, scaling factors significantly outside the range of 0.5 to 2.0 can be applied to obtain acceptable input to dynamic analyses.

You may download a digital version of this MSc dissertation here.