PREDICTION OF BEARING FAILURE WITH SURROGATE MODELING

Isolation, which comes at an increased design and construction cost, is sold as a return on investment due to the decreased damage to both the structure and its contents under typical earthquakes. Given this, and due to the higher upfront cost of isolated structures, building owners would assume an increased level of safety under large earthquakes. Worrisomely, recent studies have shown that code-permitted isolation designs do not outperform typical buildings in terms of collapse probabilities. Researchers have used the FEMA P695 methodology to ascertain these poor collapse probabilities. Furthermore, isolated buildings are used primarily for their high-performance, and designers should be able to adjust the targeted collapse probability based on their clients’ needs. However, running detailed structural models to collapse is extremely computationally inefficient and achieving a specific collapse probability requires iterative analysis. This study looks at the ability of surrogate models to predict the performance of double friction pendulum bearings (DFPs) under large events. The data on which the surrogate model is built is derived from in-plane analysis of the DFP modeled with a rigid body model including inertia for each of the bearing components and a non-linear viscoelastic impact element to simulate the impact between bearing components. This model can simulate bearing component impact and uplift and thus, is capable of simulating bearing failure. As isolation systems are particularly vulnerable to long period excitations, analytical pulses are used as input excitations. The pulse parameters and design parameters of the DFP are used as input parameters for the surrogate model. Discrete outcomes of no-impact (building most likely functional), impact (large forces imparted to structure), and failure (significant damage or failure of structure expected) are predicted given a new bearing design and input. This is a first step in facilitating rapid initial design of bearings considering desired collapse margins.