Risk Minimization for a Portfolio of Buildings Considering Risk Aversion

摘要

This paper uses probabilistic models and minimization of certain risk measures to expose risk aversion and optimal retrofit strategies for an existing building portfolio. A case study is presented addressing 622 buildings on the Vancouver campus of the University of British Columbia, Canada, a location subjected to multiple sources of seismicity. An array of probabilistic models is used to predict ground-motion intensities, structural responses, and consequent cost of damage due to earthquakes in the next 50years. A key result is the exceedance probability curve for repair costs discounted to present value. A model is also postulated for the cost of retrofit, i.e.,strengthening each building. Because both costs depend on the level of retrofit, the total expected cost and other properly defined risk measures are convex functions of the retrofit level. As a result, the conducted optimization analyses yield unique optimal retrofit levels. A computer program is employed to coordinate the multimodel analyses, which consist of 281 random variables in 4,389 instances of 14 different model types. The analyses reveal substantial risk aversion in the current designs when the direct cost of damage is considered. If indirect costs eight times the direct cost of damage are included, then current designs are optimal from an expected cost perspective.