Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Regression versus artificial neural networks: Predicting pile setup from empirical data
Abstract Piles have been used as a deep foundation for both inland and offshore structures. After installation, pile capacity may increase with time. This time dependent capacity increase is known as setup, and was first mentioned in the literature in 1900 by Wendel. When accounted for accurately during the design stages, the integration of pile setup can lead to more cost-effective pile design as it will reduce pile length, pile section, and size of driving equipment. In this paper, Both Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN) models were developed for predicting pile setup for three pile types (pipe, concrete, and H-pile) using 169 dynamic load tests obtained from the published literature and the authors’ files. In addition, the paper discusses the choice of variables that were examined to obtain the optimum model. Furthermore, the paper compares the predictions obtained by the developed MLR and the ANN models with those given by four traditional empirical formulae. It is concluded that the ANN model outperforms both the MLR model and the examined empirical formulae in predicting the measured pile setup. Finally, static load test data was used to further verify the developed models. It’s noted that the optimal ANN model overestimated pile capacity by 17% to 21% for the H-piles, and underestimated pile capacity by 12% to 17% for the pipe piles.
Regression versus artificial neural networks: Predicting pile setup from empirical data
Abstract Piles have been used as a deep foundation for both inland and offshore structures. After installation, pile capacity may increase with time. This time dependent capacity increase is known as setup, and was first mentioned in the literature in 1900 by Wendel. When accounted for accurately during the design stages, the integration of pile setup can lead to more cost-effective pile design as it will reduce pile length, pile section, and size of driving equipment. In this paper, Both Multiple Linear Regression (MLR) and Artificial Neural Networks (ANN) models were developed for predicting pile setup for three pile types (pipe, concrete, and H-pile) using 169 dynamic load tests obtained from the published literature and the authors’ files. In addition, the paper discusses the choice of variables that were examined to obtain the optimum model. Furthermore, the paper compares the predictions obtained by the developed MLR and the ANN models with those given by four traditional empirical formulae. It is concluded that the ANN model outperforms both the MLR model and the examined empirical formulae in predicting the measured pile setup. Finally, static load test data was used to further verify the developed models. It’s noted that the optimal ANN model overestimated pile capacity by 17% to 21% for the H-piles, and underestimated pile capacity by 12% to 17% for the pipe piles.
Regression versus artificial neural networks: Predicting pile setup from empirical data
Tarawneh, Bashar (Autor:in) / Imam, Rana (Autor:in)
KSCE Journal of Civil Engineering ; 18 ; 1018-1027
25.04.2014
10 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Regression versus artificial neural networks: Predicting pile setup from empirical data
| Online Contents | 2014
Predicting Cost Deviation in Reconstruction Projects: Artificial Neural Networks versus Regression
| British Library Online Contents | 2003
Predicting Cost Deviation in Reconstruction Projects: Artificial Neural Networks versus Regression
| Online Contents | 2003
Application of Artificial Neural Networks in Predicting Pile Settlements
| British Library Conference Proceedings | 2000
Pipe pile setup: Database and prediction model using artificial neural network
| British Library Online Contents | 2013