Computational & Technology Resources
an online resource for computational,
engineering & technology publications
Civil-Comp Proceedings
ISSN 1759-3433
CCP: 92
PROCEEDINGS OF THE FIRST INTERNATIONAL CONFERENCE ON SOFT COMPUTING TECHNOLOGY IN CIVIL, STRUCTURAL AND ENVIRONMENTAL ENGINEERING
Edited by: B.H.V. Topping and Y. Tsompanakis
Paper 33

Hybrid Mechanical-Neural Modelling Framework of Beam-to-Column Connections

J.H. Kim, J. Ghaboussi and A.S. Elnashai

Department of Civil and Environmental Engineering, University of Illinois at Urbana-Champaign, United States of America

Full Bibliographic Reference for this paper
J.H. Kim, J. Ghaboussi, A.S. Elnashai, "Hybrid Mechanical-Neural Modelling Framework of Beam-to-Column Connections", in B.H.V. Topping, Y. Tsompanakis, (Editors), "Proceedings of the First International Conference on Soft Computing Technology in Civil, Structural and Environmental Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 33, 2009. doi:10.4203/ccp.92.33
Keywords: hybrid, mechanical, neural network, modelling, semi-rigid connection, component-based, information-based.

Summary
The analysis of steel and composite frames has traditionally been carried out by idealizing beam-to-column connections as either rigid or pinned. Although some advanced analysis methods have been proposed to account for semi-rigid connection behaviour, the performance of these methods strongly depends on proper modelling of the connection behaviour. The primary impediments to modelling bolted beam-to-column connections are their highly inelastic response and instantaneous variability in stiffness, strength and ductility. In this paper, a new hybrid mechanical-neural model is proposed to exhibit the complex hysteretic behaviour of bolted beam-to-column connections.

The conventional modelling is based on the mathematical equations that have been developed to represent the observation of the natural behaviour of the connections. The mathematical representations are formulated by using mechanical properties such as material and geometric properties. This is what we call mechanical modelling. In modelling a top-and-seat angle connection, there are components of the deformation that are not suited to mechanical representations. This may be because (i) the underlying theory is not available or not sufficiently developed, or because (ii) the existing theory is too complex and it therefore not suitable for modelling within the building frame analysis. An example of such a component of deformation is slippage, which is closely associated with pinching effects. This is most suitable for a neural network model because the primary benefit of neural networks lies in the fact that they are capable of inferring a rule from the data with greater efficiency than developing a function by hand, which in some cases may be entirely impractical. The information about the underlying mechanics is extracted from the observed data and stored in neural networks. A trained neural network cooperates with the mechanical model.

In the hybrid model, therefore, the conventional mechanical model is complemented by the neural network technique. The role of the neural network is to model the complicated aspects that the simple mechanical model leaves out. The moment-rotation curves predicted by the proposed hybrid model are compared with the experimental results. It is observed that the hybrid model is capable of representing the pinched responses as well as the overall behaviour. In addition, the proposed hybrid modelling and simulation has potential applications in a wide range of fields in computational mechanics.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £78 +P&P)