Date

2021-2-15

Author

Güzel, Selçuk

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

204
views

113
downloads

Cite This

The sudden change in the load carrying capacityunder compresive loading, called buckling,may cause catastrophic failures. Therefore, determination of the first buckling and collapse loads of structural elements is essentialin preliminary design stages. Finite element (FE)analyses and structural testing are used to determine buckling characteristics of a structural element. However, in early design stages, FE analyses are time consuming and structural testing is costly. In this study, an artificial neural network tool(ANN)is used to reduce computational effort to determine buckling loads of integrally stiffened structural panels in early design stages. Reuslts of FE analyses are employed to train the ANN. Moreover, Latin Hypercube Sampling (LHS) methodology is used to reduce the number of required FE analyses to generate database that artificial neural network is based on. Finally, a Multi-fidelity samplingalgorithmthat uses FE models with different mesh resolutionsis implemented for generation of the ANN database in order to reduce computational time spent for finite element analyses. Mean errors and fit performance model results are compared to determine accuracy of the neural network results.

Subject Keywords

Structural optimization, Artificial neural network, Integrally stiffened structures, Multi fidelity sampling algorithm, Latin hypercube sampling

URI

https://hdl.handle.net/11511/89833

Collections

Graduate School of Natural and Applied Sciences, Thesis