Neural network-aided prediction of post-cracking tensile strength of fibre-reinforced concrete
| Autor: | Tai Ikumi, R.D. López-Carreño, Pablo Pujadas, A. de la Fuente, Eduardo Galeote |
|---|---|
| Přispěvatelé: | Universitat Politècnica de Catalunya. Departament d'Enginyeria de Projectes i de la Construcció, Universitat Politècnica de Catalunya. Departament d'Enginyeria Civil i Ambiental, Universitat Politècnica de Catalunya. GRIC - Grup de Recerca i Innovació de la Construcció, Universitat Politècnica de Catalunya. EC - Enginyeria de la Construcció |
| Rok vydání: | 2021 |
| Předmět: |
Artificial neural network
Artificial intelligence Boosting (machine learning) Computer science Generalization Reinforced concrete construction Enginyeria dels materials [Àrees temàtiques de la UPC] Tensile strength Dimension (vector space) Informàtica [Àrees temàtiques de la UPC] Ultimate tensile strength Coherence (signal processing) General Materials Science Civil and Structural Engineering Construcció en formigó armat amb fibres business.industry Intel·ligència artificial Mechanical Engineering Fibre-reinforced concrete Structural engineering Computer Science Applications Residual strength Cracking Modeling and Simulation business |
| Zdroj: | UPCommons. Portal del coneixement obert de la UPC Universitat Politècnica de Catalunya (UPC) |
| ISSN: | 0045-7949 |
| DOI: | 10.1016/j.compstruc.2021.106640 |
| Popis: | Structural fibres are an effective method to improve concrete post-cracking tensile strength (fctR). Currently, the characterization of this property is mainly performed experimentally. This is a source of uncertainties at design stages, which hinders the development of new fibre type and/or optimization of those currently existing. This paper presents a multilayer perceptron neural network to predict fctR of fibre-reinforced concrete (FRC) subjected to the Barcelona Test. The optimal architecture of the predictor is obtained by evaluating 9216 configurations of input dimension and number of hidden layers and neurons. The generalization performance is assessed using repeated random sub-sampling validation with 50 iterations. The final model can predict with high accuracy the fctR of FRC for different cracking stages. A parametric analysis is performed to prove coherence between the results predicted by the model and the established understanding of the FRC behaviour. Finally, numerical expressions are provided as an alternative tool to traditional testing to predict the residual strength of the Barcelona Test for pre-design and quality control purposes based on fibre dosage, concrete strength, specimen type and height and fibre geometric characteristics. These type of approaches are found to be necessary for boosting the development of the FRC technology. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect