Reliability-based assessment/design of floor isolation systems

• Reliability-based design of floor isolation systems is considered.
• Seismic hazard is described through stochastic ground motion models.
• Near-fault effects are explicitly considered and impact on optimal design assessed.
• Stochastic simulation is adopted for evaluating reliability performance.
• Probabilistic sensitivity analysis is integrated to identify importance of different risk factors.

Floor isolation systems have been becoming increasingly popular as a protective measure for critical structural contents such as computer servers or museum artifacts. Supplemental dampers, working in tandem with the isolation system, are also frequently considered in this context for reducing the isolated floor displacement or enhancing vibration suppression. This paper discusses a reliability-based optimization approach for this kind of applications that adequately addresses at the design stage the variability related to the earthquake hazard as well as the nonlinear dynamics of the coupled structure/isolation system. The floor isolation system is optimized based on reliability criteria, where the reliability of the system is quantified by the plausibility that the acceleration of the protected contents will not exceed an acceptable performance bound, and is calculated using stochastic simulation. The latter facilitates the adoption of complex numerical models for the coupled system. A stochastic ground motion model is utilized to characterize the seismic hazard, and an efficient stochastic optimization approach, called non-parametric stochastic subset optimization, is adopted for performing the associated design optimization. Near-fault directivity pulses are explicitly addressed within this modeling context and their effect on the optimal design is investigated in detail. Also, a global sensitivity analysis is integrated within the framework to investigate the importance of the different uncertain model parameters (risk factors) towards the system failure probability. For demonstrating the proposed framework, the protection of a computer server placed at different floors within a four-story structure is considered.