| dc.description.abstract | The use of unbounded tendons as a prestressing technique in the design and rehabilitation of concrete slabs / beams has in recent years been preferred due to its simple construction design, cost efficiency and maintenance possibilities. When implementing an analytical calculation of members with unbound prestressing tendons, it must be considered that the concrete and prestressing tendons are deformed independently of each other. This consideration creates a necessity in defining the global compatibility requirement to finally be able to determine the member's moment capacity.
The existing design guidelines / codes and literature provide different calculation models for estimating the tendon stresses in unbound post-tensioned concrete members in the case of ultimate limit state (ULS). Most of the methods are based on theoretical models (e.g., Collapse mechanism model, band reduction model) and statistically based empirical models, where there are no or few models based on ANN. This study presents an ANN-based model to be able to predict the stress in unbound tendons at ultimate limit state based on a database of 251 experiments on simply supported beam from 19 different research groups and 43 experiments on continuous beam over two spans from 5 different research groups. ANN-based model predictions show a very good agreement with experimental results given in the literature during training, testing and validation. A sensitivity analysis has been performed to quantify the degree of influence of input variables used in the developed neural network model. The results of the analysis show that predictions of tendons stress for both fully and partially prestressed beams over one and two spans using neural networks are more accurate than the results obtained using the models given design guidelines and literature. | |
