نوع مقاله : پژوهشی
نویسندگان
دانشکدهی فنی مهندسی، دانشگاه اراک
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
The assessment of structural seismic response is often done by selecting ground motion records that conform to the seismic hazard conditions of the objective site, and which can be obtained based on probabilistic seismic hazard analysis (PSHA). A common recordselection practice suggests selecting seven records that are compatible with the dominant earthquake scenario in a given site. The selected records are then scaled (if necessary) to match the design level of the uniform hazard spectrum (UHS). Many research results have shown that using the UHS leads to a significant bias in structural response assessment. The conditional mean spectrum (CMS) has been recently proposed as an alternative to the uniform hazard spectrum (UHS) for employment as a target spectrum in ground motion record selection. The CMS provides the expected response spectrum، conditional to the occurrence of a target spectral acceleration value in the period of interest. The correlation of $\varepsilon$ values in different periods is considered in CMS development, but conventional regression analysis has been applied to measure the degree of correlation of $\varepsilon$ values in different periods, and the influence of outlier data has not been studied. Outliers are sample values that cause surprises in relation to the majority of samples. The main objective of this paper is to reveal an important drawback in the procedure for calculation of CMS. The authors believe that the developed procedure for CMS leads to a spectral shaped anomaly that is not consistent enough with real ground motion. A robust regression analysis is proposed in this paper to improve the current CMS, which is based on a conventional regression analysis. Robust regression is an important tool for analyzing data that are contaminated with outliers. Robust regression analysis works by assigning a weight to each data point. Weighting is done automatically and iteratively using a process called iteratively reweighted least squares. The results show that the proposed robust CMS significantly differs from the conventional CMS, especially for higher periods of interest. The shape of the robust CMS represents rare ground motion in a more reliable manner, compared with conventional CMS.
)