Multiobjective Optimization of Long Irregular RC Bridges’ Piers Subjected to Strong Motions and Definition of Classification Tree Surrogate Models

Long irregular bridges can be difficult structures to design and analyze for earthquake actions due to the inherent complexity in predicting their nonlinear dynamic behavior. This stems from the possible influence of higher modes in the overall seismic behavior, among other factors (Kappos et al., in Seismic design and assessment of bridges, inelastic methods of analysis and case studies, Springer, New York, 2012), making the analysis in many cases dependent on the use of nonlinear dynamic analysis methods (NDAs). Multiobjective optimization by means of evolutionary algorithms, MOEAs, are good in dealing with constrained nonlinear non-convex optimization problems with continuous and discrete variables (Deb in Multi-objective optimization using evolutionary algorithms, Wiley, New York, 2001), such as earthquake design of long irregular RC bridges. In this study, a framework that applies one such algorithm, NSGA-II (Deb et al. in IEEE Trans Evol Comput 6:182–197, 2002), is defined and applied to optimize the pier design of a long irregular RC bridge case-study. As a result of the optimization search it is possible to compare the effects of different variables in the dynamic behavior of a bridge case-study. The optimization results were the input for the second phase of the methodology with the training of a classification tree for assessing seismic performance. The large-scale computation efforts were made feasible by parallel computation between 32-threads, and by performing the analysis on the OpenSEESMP platform (McKenna and Fenves in OpenSEES—Open System for Earthquake Engineering Simulation, The Regents of the University of California, Berkeley, CA, 1999).