Innovative model for accurate prediction of the transfer length of prestressing strands based on artificial neural networks: Case study

This study provides a case study showing the effectiveness of utilizing Artificial Neural Networks (ANNs) in prestressed concrete. The effectiveness of prestressing is crucially linked to the transfer length (TL) of the prestressing strands. The ANN technique was used for accurate prediction of TL based on more than 458 data points collected from various literature works. The ANN technique allowed for investigating the effect of various key parameters classified into major categories including: strand characteristics, concrete properties, geometric details, and manufacturing method. The MATLAB software was utilized to build, train, and test the ANN using 19 input variables and one targeted output. Sensitivity analysis identified the main parameters influencing the TL determination, which lead to the development of ANN-based analytical model with high prediction capability and low mean square error. The effectiveness and accuracy of the proposed model were verified through comparisons with the AASHTO and ACI codes models as well as with some proposed models in literature. The developed ANN-based prediction model combined the effects of most significant parameters influencing the TL with many important parameters never been considered before. For simplicity, the required coefficients for predicting the TL using the developed model were provided in clear and easy to use charts taking into consideration the beam cross-sectional shape.