×

Neural networks: new results and prospects of applications in structural engineering. (English) Zbl 1325.74170

Mota Soares, Carlos A. (ed.) et al., Computational mechanics. Solids, structures and coupled problems. Plenary and keynote lectures presented at the III European conference on computational mechanics: solids, structures and coupled problems in engineering, (ECCM-2006, Lisbon, Portugal, June 5–8, 2006. Dordrecht: Springer (ISBN 1-4020-4978-1/hbk). Computational Methods in Applied Sciences (Springer) 6, 555-576 (2006).
Summary: Applications of BPNNs (Back-Propagation Neural Networks) to four problems of structural engineering are discussed basing on laboratory tests or maesuremets on real buildings. The simulation problem of soil-structure interaction is considered as the transmission of response spectra from ground to the basement level inside prefabricated buildings subjected to paraseismic excitations. The analysis is performed by BPNN learned by Kalman filtering. Identification of placement of an attached mass to a steel plate is examined on the base of dynamic responses of a laboratory model. Reliability of a cylindrical panel under concentrated load is analyzed by means of a hybrid Monte Carlo method. In this approach the training and testing patterns are computed by the COSMOS/M finite element program and the trained BPNN is explored as a rapid simulator of MC trials. A hybrid system for the FE model updating is presented in which BPNN serves for calibrating of control parameters.
For the entire collection see [Zbl 1097.74003].

MSC:

74S30 Other numerical methods in solid mechanics (MSC2010)
74K20 Plates
92B20 Neural networks for/in biological studies, artificial life and related topics
74S05 Finite element methods applied to problems in solid mechanics

Software:

COSMOS/M
PDFBibTeX XMLCite