Keywords: artificial neural network, finite element, structural analysis, prediction, elasto-plastic plates.

The structural behaviour of plates is affected by many factors such as material properties, geometric properties, load and boundary conditions. Mathematical models have been used to describe aspects of this behaviour, but they fall short in considering a large number of variables simultaneously. This paper investigates the use of Artificial Neural Networks (ANNs) as a preliminary alternative to mathematical modelling or experimental testing for quick prediction of the structural behaviour of elasto-plastic plates. Such predictions could be utilized by a structural engineer on a preliminary basis to determine the initial suitability of a particular plate analysis and design.

ANNs are computational models that adopt a training mechanism to extract the relationships that link a set of causal input parameters to their resulting conclusions. Once neural networks are trained, they can predict the results for an unknown case (not used in training) if provided with the input parameters alone.

This study illustrates the application of the artificial neural networks for building analysis models of plates with uniformly distributed load and different ductile material properties, plate size and boundary conditions. The data were collected using 2240 runs of a nonlinear finite element analysis programme with the Huber-Mises failure criterion. Three forward analysis models were employed through this study. The models were used for predicting deflection and moment patterns. The testing of the models using reserved training data shows that the absolute error for all models was almost less than 10% and the correlation coefficients are more than 0.9. The parametric study shows that the boundary condition and geometry are the most significant factors affecting the output of all the three models, on the other hand the relative importance of material properties were insignificant in the output of models.

The neural network approach does not require a new model to be developed for each new problem; all the user has to do is input a few parameters describing the specific problem to be solved. In addition, a neural network model can solve simultaneously for a batch of problems in almost negligible time.

The results show that artificial neural networks have a strong potential as a feasible tool for predicting the structural properties of elasto-plastic plates within the range of input parameters considered in the training.

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 £62 +P&P)