This paper aims at introducing a new method to determine the buckling coefficient kb of the stiffened plate girders under pure bending using deep learning, one of the most powerful algorithms in machine learning. Firstly, output data kb is generated from eigenvalue buckling analyses based on input data (various geometric dimensions of the girder). This procedure is implemented by using the Abaqus2Matlab toolbox, which allows the transfer of data between Matlab and Abaqus and vice versa. After that, 2,200 training data are used to build the model for predicting kb using deep learning. Finally, 200 test data are used to evaluate the accuracy of the model. The results obtained from this model are also compared with analogous results of previous works with a good agreement.
Prediction of buckling coefficient of stiffened plate girders using deep learning algorithm
Lecture Notes in Civil Engineering
2019-10-11
6 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Optimum design of stiffened composite girders
British Library Conference Proceedings | 1998
|Springer Verlag | 2013
|Buckling of curved tension-field girders
Engineering Index Backfile | 1938
|Ritz Approach for Buckling Prediction of Cracked-Stiffened Structures
Online Contents | 2013
|