A total probabilistic seismic loss assessment framework of structures by the KDE-based nonparametric method

In this paper, a total kernel density estimation (KDE)-based nonparametric probabilistic seismic loss assessment framework for structures is proposed, and the cloud-point strategy and stochastic nonstationary excitation are well incorporated during the procedure. The methodology framework and implementation steps are first introduced in detail, and then an application example of a reinforced concrete frame is adopted to verify the feasibility of the total KDE-based nonparametric framework. A comparison is performed via the classic parametric least-square regression (LSR) approach, and lognormal distribution is also adopted during the analysis. Moreover, the classic Monte Carlo Simulation (MCS) approach is adopted as the benchmark for validation. In general, the KDE-based framework is nonparametric without predefined assumptions of generated curves, making it an effective nonparametric method for probabilistic seismic loss estimation and statistical prediction of structures. Use of the cloud-point strategy greatly improves the calculation efficiency, and the stochastic nonstationary excitation effectively guarantees the real situation of load input. From the application example, the high fidelity and great efficiency of the KDE-based framework (fragility, damage, loss, and other parameters) are suggested in probabilistic seismic analysis. Thus, the total KDE-based nonparametric framework provides a new perspective for further development and in-depth investigation in probabilistic seismic loss research.