Pontificia Universidad Católica de Chile Pontificia Universidad Católica de Chile
Sepúlveda C., de la Llera J.C. and Jacobsen A. (2012)

An empirical model for preliminary seismic response estimation of free-plan nominally symmetric buildings using ANFIS. http://dx.doi.org/10.1016/j.engstruct.2011.11.035

Revista : Engineering Structures
Volumen : 37
Páginas : 36-49
Tipo de publicación : ISI Ir a publicación


The good seismic performance of free-plan buildings during the last M-w = 8.8, February 27, 2010 Chile earthquake, deserves special attention in order to understand the clues for such behavior. Also, and because of economic reasons, the development of these building projects are always subjected to a fast track development, which makes difficult to iterate much on the structural configuration of the building unless quick and reliable responses can be provided to architects and owners. Motivated by these facts, this work proposes a new simplified methodology capable of predicting a reasonably accurate response of free-plan nominally symmetric buildings. Because it is well known that the flexural effect of the slab has a strong influence in the response of these buildings, the methodology is based on the decomposition of the building behavior in two simple extreme cases, zero and infinite flexural stiffness between vertical elements, and horizontal ones. The final building behavior lies in between these cases, and is predicted by producing a linear combination of the models through a single model-parameter a that has been calibrated by using a large training database and an adaptive neural network with a fuzzy inference system, ANFIS. Predicted errors for the displacement and forces of new building configurations are usually less than 10% and 15%, respectively, and in many cases are better than that. Besides its accuracy, the numerical efficiency of this model is usually one to two orders of magnitude larger than that of conventional finite element model of the building. (C) 2011 Elsevier Ltd. All rights reserved.