engleski

Seismic capacity of infilled frames using neural networks

The applicability of neural networks, trained on compiled experimental database, for prediction of the seismic behavior and capacity of reinforced-concrete infilled frames is explored. The work was motivated due to a great deal of uncertainty in the estimation of the seismic capacity of infilled frames and evaluation of complex infill effect. Despite a big number of experimental studies on infilled frames there is still a lack of understanding between the behavior and the geometric parameters such as ratio of height to length of frame, reinforcement ratio of frame elements, material properties of frame (reinforced - concrete) and infill (masonry), loading histories, etc. An experimental database used in this study has been compiled from the available literature and includes data from the laboratory test carried out on one story, one bay reinforced - concrete frames infilled with unreinforced masonry infill. Neural networks simulate human's brain ability to classify patterns or to make predictions or decisions based upon past experience using data sets. The network consists of input layer, hidden neurons and output layer. The inputs of the created neural networks are geometrical and material properties of frame and infill, reinforcement ratios of frame elements and loading. Output variables, which have an important role in performance evaluation, are deformation capacity in terms of drift, shear strength and failure modes of infilled frames. The main goal of this paper is to make a contribution towards the quantitative determination of the performance capability of specific structural elements – reinforced/concrete infilled frames, whose behavior depends on many variables and is often unpredictable. Their performance, expressed in terms of shear strength and deformation capacity is of vital importance for the evaluation of the seismic performance of existing buildings as well as for the design of earthquake resistant reinforced concrete buildings.

r/c frames; masonry infill; experimental results; PBD; neural networks

nije evidentirano