000013325 001__ 13325
000013325 005__ 20161114160336.0
000013325 04107 $$aeng
000013325 046__ $$k2009-06-22
000013325 100__ $$aArangio, S.
000013325 24500 $$aBayesian neural networks to investigate the effects of corner radius on the performance of frp confined columns

000013325 24630 $$n2.$$pComputational Methods in Structural Dynamics and Earhquake Engineering
000013325 260__ $$bNational Technical University of Athens, 2009
000013325 506__ $$arestricted
000013325 520__ $$2eng$$aIn recent years, the use of fiber reinforced polymer (FRP) composites has become increasingly popular for retrofitting concrete elements. One important application regards the confinement of reinforced concrete columns: jacketing with FRP wraps results in a remarkable enhancement of their strength and ductility. The confinement action depends on the column shape: tests reported in the literature have recognized that the corner radius significantly affects the confinement effects. A significant amount of research has been devoted to circular columns retrofitted with FRP, but less is known about rectangular/square columns where the concrete is non-uniformly confined and the effectiveness of confinement is reduced. Some studies and structural codes suggest that a sharp corner offers no confinements. However, published results indicate that a jacket with sharp corners provides a certain degree of effective confinement. In the first part of the paper, some existent models for strength prediction in case of square confined sections are discussed. The second part of the work aims at developing a statistical model that can predict the confinement effect, given the corner radius and the concrete grade. For this purpose, a Bayesian neural network model is chosen because it can effectively deal with uncertain information, making it highly promising for handling prediction problems. Some published data from experimental tests are used for building and training a neural network model. Then, the proposed model is tested on other experimental samples that were not included in the training set. The model shows good approximation properties: the average percentage error is about 8.9%. The results obtained with the proposed model are also compared with those obtained by applying the models presented in the first part on the same data set.

000013325 540__ $$aText je chráněný podle autorského zákona č. 121/2000 Sb.
000013325 653__ $$aFRP confinement, corner radius effect, ultimate strength prediction, Bayesian neural networks Abstract. In recent years, the use of fiber reinforced polymer (FRP) composites has become increasingly popular for retrofitting concrete elements. One important application regards the confinement of reinforced concrete columns: jacketing with FRP wraps results in a remarkable enhancement of their strength and ductility. The confinement action depends on the column shape: tests reported in the literature have recognized that the corner radius significantly affects the confinement effects. A significant amount of research has been devoted to circular columns retrofitted with FRP, but less is known about rectangular/square columns where the concrete is non-uniformly confined and the effectiveness of confinement is reduced. Some studies and structural codes suggest that a sharp corner offers no confinements. However, published results indicate that a jacket with sharp corners provides a certain degree of effective confinement. In the first part of the paper, some existent models for strength prediction in case of square confined sections are discussed. The second part of the work aims at developing a statistical model that can predict the confinement effect, given the corner radius and the concrete grade. For this purpose, a Bayesian neural network model is chosen because it can effectively deal with uncertain information, making it highly promising for handling prediction problems. Some published data from experimental tests are used for building and training a neural network model. Then, the proposed model is tested on other experimental samples that were not included in the training set. The model shows good approximation properties: the average percentage error is about 8.9%. The results obtained with the proposed model are also compared with those obtained by applying the models presented in the first part on the same data set. 1

000013325 7112_ $$aCOMPDYN 2009 - 2nd International Thematic Conference$$cIsland of Rhodes (GR)$$d2009-06-22 / 2009-06-24$$gCOMPDYN2009
000013325 720__ $$aArangio, S.$$iBontempi, F.
000013325 8560_ $$ffischerc@itam.cas.cz
000013325 8564_ $$s278632$$uhttps://invenio.itam.cas.cz/record/13325/files/CD477.pdf$$yOriginal version of the author's contribution as presented on CD, section: Robust stochastic analysis, optimal design and model updating of engineering systems - i (MS).
000013325 962__ $$r13074
000013325 980__ $$aPAPER