An advanced option allows the user to specify the method of calculation of the error and the spectral variable to be used (Sa, Sv, Sd, etc) and whether the spectral value of its Log are to be used in the calculation (Linear/Log). The fitting error measure is computed for entire suite while performing the Ground Motion Suite Selection, and it is also calculated for individual records when fitting records within the Search Panel.
The following error measure calculation options are available :
•Absolute Difference Sum
•Absolute Difference Integral
•Square Sum (SSE)
•Square Integral
For suite selection, the Match Objective can be set to one of the following:
•Suite Mean: the resulting Mean of a suite of records will be matched to the target
•Suite Mean + Dispersion: the resulting Mean of a suite and Dispersion will be matched to the target (recommended)
•Top Records: match each record individually to match the target
Users can select the “Top Records” feature to match top records (records with smallest errors) individually to the defined target spectrum (without optimizing the overall suite statistics). This feature is similar to matching using Search Panel, but while matching the additional constraints defined in the GM Suite Selection dialog, and may be used to select record seed candidates for further modification (such as spectral matching).
Suite Mean and Suite Mean + Dispersion options optimizes the resulting mean of the suite to match the target spectrum, where the error is measured as the difference between the resulting mean spectrum of the suite and target spectra. However, the Top Records option optimizes each record individually to match the defined target spectra, and the error measure is calculated as the difference between each record and the target spectrum and averaged by the sum of errors of the records in the suite.
The mean of the generated records using the Top Records option should not necessary have a good fit compared to the target spectrum; however, individually the generated records are the best. On the contrary, the records generated using the Suite Mean and Suite Mean + Dispersion have the best resulting mean spectrum.
Users are able to choose the Suite Mean Calculation method: either arithmetically or using the Geomean method.
If the Suite Mean + Dispersion option was selected in the Match Objective, then the Suite Dispersion Calculation method may be selected as Standard Deviation or LogSigma.
After setting the Suite Dispersion Calculation method, users may set the Dispersion Objective using one of the following mathematical equality or inequalities:
•Equals Target Standard Deviation
•Greater or Equals Target Standard Deviation
•Lesser or Equal Target Standard Deviation
It is also possible to specify the weight for the Standard Deviation Error term relative to Mean Error term for a specific Suite(s).
As an example if the weight is set to 1, then mean error and dispersion error are given equal weight when optimizing the suite record selection and scaling.
|