DISTRIBUTION OF PEAK HORIZONTAL FLOOR ACCELERATION FOR ESTIMATING NONSTRUCTURAL ELEMENT VULNERABILITY

Peak horizontal floor acceleration (PHFA) is now widely used by the engineering community for estimating the vulnerability of both attached and unattached acceleration sensitive nonstructural elements. Recent performance-based earthquake engineering approaches have cast this estimation in a probabilistic form, where seismic fragility curves are used to represent the probability that a specific damage measure (DM) will occur, given an earthquake of a specified intensity. Since both attached and unattached nonstructural elements are generally placed at various levels of a building structure, PHFA is taken as the engineering demand parameter (EDP) and placed as the abscissa in these curves. To estimate the force imposed on attached nonstructural elements such as architectural, mechanical and electrical components, PHFA is assumed to vary with the height of the building. Since the peak horizontal floor acceleration at a particular level of the building depends on the dynamic characteristics of the building, the level of nonlinearity induced, and the ground excitation, PHFA cannot be generalized as a function of height without considering these aspects. In this study, the distribution of absolute acceleration amplification Ω (PHFA normalized by peak ground acceleration) along the height of buildings with different dynamic characteristics is developed through nonlinear regression analysis. Numerical models of a total of eight moment-resisting steel frame buildings (flexible and rigid), representing actual buildings on the West Coast of the U.S., are constructed. An ensemble of thirty-two different ground motions, representing hazard levels of 2, 10, and 50% probability of exceedance are used as input to the building models and nonlinear dynamic analysis conducted. Resulting distributions are compared with code-recommendations and a simplified distribution of Ω is proposed, based on assumed physical contributions to the behavior and regression through the dynamic analyses results. Although simplified, the suggested distribution of Ω will lead to more reliable estimates of the vulnerability of acceleration sensitive nonstructural components.