Calibration of the finite element model of a long span cantilever through truss bridge using artificial neural networks

Thesis Type: Postgraduate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Civil Engineering, Turkey

Approval Date: 2008




In recent years, Artificial Neural Networks (ANN) have become widely popular tools in various disciplines of engineering, including civil engineering. In this thesis, Multi-layer perceptron with back-propagation type of network is utilized in calibration of the finite element model of a long span cantilever through truss called Commodore Barry Bridge (CBB). The essence of calibration lies in the phenomena of comparing and correlating the structural response of an analytical model with experimental results as closely as possible. Since CBB is a very large structure having complex structural mechanisms, formulation of mathematical expressions representing the relation between dynamics of the structure and the structural parameters is very complicated. Furthermore, when the errors in the structural model and noise in the experimental data are taken into account, a calibration study becomes more tedious. At this point, ANNs are useful tools since they have the capability of learning with noisy data and ability to approximate functions. In this study, firstly sensitivity analyses are conducted such that variations in dynamic properties of the bridge are observed with the changes in its structural parameters. In the second part, inverse relation between sensitive structural parameters and modal frequencies of CBB is approximated by training of a neural network. This successfully trained network is then fed up with experimental frequencies to acquire the as-is structural parameters and model updating is achieved accordingly.