MODIFIED CLOUD ANALYSIS: NON-LINEAR DYNAMIC ANALYSIS WITH UN-SCALED RECORDS

One of the most efficient non-linear dynamic analysis procedures for analytic fragility evaluation is based on simple regression of structural response versus seismic intensity for a set of registered records. This procedure, also known as the Cloud Method (Jalayer, 2003), has many advantages such as simplicity, use of un-scaled records, and relatively small computational effort. This method suffers draw-backs such as modeling of record-to-record variability with constant standard deviation. Nevertheless, the simplicity of the formulation makes it a quick and efficient analysis procedure for fragility assessment or safety-checking in the context of the SAC-FEMA formulation (Cornell et al., 2002). Record selection is quite critical for Cloud Analysis as the results are often sensitive to the suite of ground motion records selected. The criteria for record selection for use in Cloud Analysis can be summarized as (Jalayer et al., 2015 and 2017): 1. The records should span a wide range of the intensity measure (in order to reduce the error in the estimation of the regression slope) 2. A significant proportion of the records should lead to the exceedance of the limit state (in order to avoid extrapolationin the zone of interest) 3. The number of ground motion records from the same event should be limited (in order to avoid data set with high correlation)4. The soil type and the predominant focal mechanism (s) for the site should be kept in mind. Respecting these criteria for the ultimate limit states (e.g., life safety, collapse prevention) might lead to cases of structural collapse (loss of load-bearing capacity), global dynamic instability (very large deformations) or cases of numerical non-convergence, which are referred to generically as “collapse” cases. Clearly, the original Cloud Analysis based on simple regression cannot be applied directly to the “collapse”-inducing ground motion records. The Modified Cloud Analysis (MCA) is proposed as an alternative to the original Cloud Analysis for handling the collapse-inducing records. Based on the Total Probability Theorem, “collapse” and non-“collapse” can be treated as mutually exclusive and collectively exhaustive events. This way the original cloud analysis can be applied to the noncollapse portion of the records. The probability of collapse can be estimated, in the simplest case, as the ratio of the number of collapse-inducing records to the total number of records. The estimation of the collapse probability can be improved through using a generalized regression model (logistic regression). Comparisons with multiple-stripe analysis with variable set of conditional spectrum-compatible records indicates a very good agreement for the range of intensities, which matters for the sake of seismic risk assessment (Jalayer et al., 2017).

کلیدواژه ها:

Seismic risk ، Performance-based earthquake engineering ، Non-linear dynamic procedures ، Ground motion records