ABSTRACT
Modern bridge designs are expected to operate under ever increasing lifecycle durations. Increased lifecycles subject bridges to evolving and increasingly variable operational demands. The nature and severity of environmental changes, security-related risks, and a host of other demands may dramatically vary over a bridge’s lifecycle, all the while material degradation from fatigue, creep, and corrosion mechanisms may act to degrade the bridge’s performance. The combinations of evolving factors lead to high potential for variability in a bridge structure’s expected performance. Further challenging bridge designers, weight optimization efforts, which are critical in achieving cost-effective designs, often come at a cost to reliability in the face of variability. The need to explicitly address bridge reliability under variable conditions, though sensitivity analysis and/or uncertainty quantification, is greatly increased by today’s lifespan expectations. Many sources of uncertainty under extreme events such as blast cannot be readily quantified to utilize propagation of uncertainty. This paper demonstrates a new variability-based method for addressing these challenges while simultaneously assessing the structural design’s optimization and system reliability. Machine learning and stability indicators are utilized, and a new indicator called the Instability Index is proposed to predict relative system stability for a given loci of input variables. Both the structural optimization and reliability objectives are then balanced to find an optimal design that best achieves both objectives. The utility of the proposed method is demonstrated through its application to a realistic bridge design scenario that is subject to poorly defined sources of variability. The proposed methods are applied in the contexts of structural system reliability under undisturbed (operational) conditions and under structural failure scenarios (element removal).
