Predicting the Shear Strength of RC Beams Without Stirrups Using Bayesian Neural Network (bibtex)

by Osimen Iruansi, Maurizio Guadagnini, Kypros Pilakoutas, Kyriacos Neocleous

Abstract:

Advances in neural computing have shown that a neural learning approach that uses Bayesian inference can essentially eliminate the problem of over fitting, which is common with conventional back propagation Neural Networks. In addition, Bayesian Neural Network can provide the confidence (error) associated with its prediction. This paper presents the application of Bayesian learning to train a multi layer perceptron network on experimental test on Reinforced Concrete (RC) beams without stirrups failing in shear. The trained network was found to provide good estimate of shear strength when the input variables (i.e. shear parameters) are within the range in the experimental database used for training. Within the Bayesian framework, a process known as the Automatic Relevance Determination is employed to assess the relative importance of different input variables on the output (i.e. shear strength). Finally the network is utilised to simulate typical RC beams failing in shear.

Reference:

Predicting the Shear Strength of RC Beams Without Stirrups Using Bayesian Neural Network (Osimen Iruansi, Maurizio Guadagnini, Kypros Pilakoutas, Kyriacos Neocleous), In Proceedings of the 4th International Workshop on Reliable Engineering Computing, Robust Design - Coping with Hazards, Risk and Uncertainty (R.L. Muhana M. Beer, R. L. Mullen, eds.), 2010. (Research Scientific)

Bibtex Entry:

@INPROCEEDINGS{Iruansi-etal:2010, author = {Osimen Iruansi and Maurizio Guadagnini and Kypros Pilakoutas and Kyriacos Neocleous}, title = {Predicting the Shear Strength of RC Beams Without Stirrups Using Bayesian Neural Network}, booktitle = {Proceedings of the 4th International Workshop on Reliable Engineering Computing, Robust Design - Coping with Hazards, Risk and Uncertainty}, year = {2010}, editor = {M. Beer, R.L. Muhana and R. L. Mullen}, pages = {597--613}, address = {Singapore}, month = {March}, note = {Research Scientific}, abstract = {Advances in neural computing have shown that a neural learning approach that uses Bayesian inference can essentially eliminate the problem of over fitting, which is common with conventional back propagation Neural Networks. In addition, Bayesian Neural Network can provide the confidence (error) associated with its prediction. This paper presents the application of Bayesian learning to train a multi layer perceptron network on experimental test on Reinforced Concrete (RC) beams without stirrups failing in shear. The trained network was found to provide good estimate of shear strength when the input variables (i.e. shear parameters) are within the range in the experimental database used for training. Within the Bayesian framework, a process known as the Automatic Relevance Determination is employed to assess the relative importance of different input variables on the output (i.e. shear strength). Finally the network is utilised to simulate typical RC beams failing in shear.}, doi = {10.3850/978-981-08-5118-7_067}, isbn = {978-981-08-5118-7}, keywords = {Bayesian learning, Neural networks, Reinforced concrete, Shear, Uncertainty modelling} }

Powered by bibtexbrowser