Compound intensity measures for improved seismic performance assessment in cross-fault hydraulic tunnels using partial least-squares methodology

Abstract

Seismic performance assessment of cross-fault hydraulic tunnels is typically expressed in terms of the probability of reaching or exceeding specific damage states related to earthquake intensity measures (IMs) and engineering demand parameters (EDPs). Obviously, this would require determining the appropriate earthquake IMs to reflect the seismological characteristics of ground motions and improving the predictive ability for the EDPs of cross-fault hydraulic tunnels. It is worth emphasising that an unreasonable IM may increase the ground motion uncertainty, which in turn introduces additional uncertainty to the EDP estimates of cross-fault hydraulic tunnels. Hence, this study presents a compound IM, integrating multiple earthquake IMs, to improve the reliability of the seismic fragility analysis of a cross-fault hydraulic tunnel based on the partial least squares methodology. To this end, seven scalar IMs are first determined from the 20 examined earthquake IMs as the fundamental element of the compound IM using four test criteria. Subsequently, a 1080 complete nonlinear dynamic time history analysis of the cross-fault hydraulic tunnel using the incremental dynamic analysis method is conducted to generate a large amount of data. The findings of this study highlight that the compound IM-based probabilistic seismic demand models of the cross-fault hydraulic tunnel incorporate various sources of uncertainties, which enhances the correlation, efficiency, practicality, and proficiency of probabilistic seismic demand analysis and eventually leads to a more reasonable estimation of probabilistic failure. Moreover, the potential of nonlinear polynomial regression of compound IM can further improve the correlation and efficiency of probabilistic seismic demand analysis. Relative to the classical fragility curves utilising a single optimum-IM peak ground velocity, the probability of damage to the cross-fault hydraulic tunnel under a particular damage state related to the compound IM is significantly larger than the corresponding single optimum-IM peak ground velocity. Similarly, the probability of damage to the cross-fault hydraulic tunnel under the Wenchuan seismic event is underestimated, derived from a single optimum-IM peak ground velocity. The abovementioned phenomena demonstrate that a compound IM based on the partial least squares methodology can be suggested for use in estimating the probability of structural damage in performance-based earthquake engineering.