Keywords: seismic probabilistic analysis, slope stability, fragility curves, geostructures, artificial neural networks, Monte Carlo simulation.

The majority of earthquake engineering applications, especially the geotechnical ones, can be considered as "imprecise" problems, due to various (geotechnical, seismic, mechanical and geometrical) inherent uncertainties that characterize them. Nevertheless, geotechnical engineering practice compromises with the use of deterministic simplifications due to their low computational cost and their minimal complexity. However, nowadays, probabilistic methods are gaining popularity due to the advances in computational resources and numerical methods, since they offer a more realistic way to determine the seismic vulnerability of structures and, or geostructures. This paper presents an efficient methodology for evaluating seismic vulnerability of large-scale geostructures under seismic loading conditions. In an effort to increase the computational efficiency of the process, effective numerical predictions are used, utilizing advanced artificial neural network (ANNs) models, rather than solving the problem conventionally.

Artificial neural networks, expert and fuzzy systems as well as evolutionary methods are the most popular soft computing techniques. These so-called "artificial intelligence" (AI) methods are used either to reduce the computational cost, or when the complexity and, or the size of the problem forbids the use of conventional techniques. ANNs especially have been widely used in many fields of science and technology, as well as, into an increasing number of problems in earthquake engineering |citegeorgopo:1. From among general problems that can be analyzed by means of AI techniques simulation, inverse simulation and identification problems are the most popular paradigms. In this study, ANNs are used for simulating of the seismic response, i.e. for known inputs and characteristics of the examined geostructure the unknown outputs (seismic response of the system) are given.

Typically, the seismic design of embankments is performed utilizing pseudostatic slope stability procedures. In addition, seismic vulnerability (i.e. the possibility of geostructure's or structure's failure), is very commonly assessed by establishing its fragility curves, most often using the "general purpose" Monte Carlo simulation (MCS) technique. For large-scale geostructures, such as embankments and waste landfills, under seismic loading conditions, the fragility curves are the outcome of the probabilistic pseudostatic slope stability analysis of the geostructure. By incorporating efficient ANN-based models into the time-consuming process of fragility curves construction, it is feasible to increase the MCS sample size and predict more accurately the possibility of failure of a geostructure within a fraction of time compared to the conventional procedure. Thus, ANNs offer a precise and efficient way to determine a geo-structure's performance and the evaluation of its seismic vulnerability for multiple hazard levels and multiple limit states, in the viewpoint of the state-of-the-art performance-based earthquake engineering (PBEE).

1

N.D. Lagaros, Y. Tsompanakis, (Eds), "Intelligent computational paradigms in earthquake engineering", Idea Publishers, 2006.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £62 +P&P)