Research on the dynamic properties of piled structures using the neural networks and the support vector machines
|
Bibliografische Angaben
Autor(en): |
Haoyun Deng
(Department of Civil Engineering, Tongji University, Shanghai, China)
Xinyang Jin (Institute of Building Structures, China Academy of Building Research, Beijing, China) Ming Gu (Department of Civil Engineering, Tongji University, Shanghai, China) |
||||
---|---|---|---|---|---|
Medium: | Tagungsbeitrag | ||||
Sprache(n): | Englisch | ||||
Tagung: | IABSE Symposium: Engineering the Future, Vancouver, Canada, 21-23 September 2017 | ||||
Veröffentlicht in: | IABSE Symposium Vancouver 2017 | ||||
|
|||||
Seite(n): | 2149-2154 | ||||
Anzahl der Seiten (im PDF): | 6 | ||||
Jahr: | 2017 | ||||
DOI: | 10.2749/vancouver.2017.2149 | ||||
Abstrakt: |
A simplified method is proposed to analyse the dynamic properties of piled structures. Superstructures are modelled as generalized single degree of freedom systems using the virtual displacement principle, and the impedance functions of pile groups are solved based on the thin- layer method. Six dimensionless parameters are selected to characterize the soil-pile-structure systems and extensive parametric analyses are performed. A mathematical model for the seismic analysis of soil-pile-structure system is built in the neural network based on the outcomes of parametric analyses. The data of the analyses are divided into three different parts which are used for training, testing and validating of the artificial neural network(ANN) model. In order to validate the accuracy of ANN model, another analysis technique of the support vector machine is used. The outcomes show that the model can predict the dynamic properties of the soil-pile-structure system with good accuracy and less time which contribute to solve the dynamic characteristics of piled structures without performing complex analysis. |