Artificial neural networks in prediction of mechanical behavior of composite at different temperature – preliminary study
Referat wygłoszony
| Status: | |
| Autorzy: | Falkowicz Katarzyna, Karpiński Robert |
| Rok wydania: | 2025 |
| URL do źródła | LINK |
| Język: | angielski |
| Źródło: | X International Conference of Computational Methods in Engineering Science – CMES’25 |
| Miasto wystąpienia: | Cedzyna |
| Państwo wystąpienia: | POLSKA |
| Efekt badań statutowych | NIE |
| Abstrakty: | angielski |
| This paper discusses the results of experimental tests conducted to estimate the influence of temperature on the tensile and compressive strength of composite samples, as well as on changes in the mechanical properties of the tested material. The experiments were carried out at ambient temperature and at elevated temperatures of 40°C, 60°C, and 80°C, during which the specimens were subjected to tensile and compressive loading until failure. The obtained experimental data were then used to train and validate artificial neural network (ANN) models aimed at predicting the temperature-dependent mechanical behavior of the composite. The input parameters included the test temperature and selected material characteristics, while the outputs targeted the prediction of temperature- dependent properties at unseen intermediate temperatures. Several network architectures were evaluated to determine the optimal configuration providing the best correlation between predicted and measured data. The preliminary results confirmed the high potential of ANN-based approaches in capturing complex, nonlinear relationships between temperature and the mechanical response of composites. The developed models demonstrated satisfactory predictive capability even for temperatures not included in the training dataset and can serve as a foundation for future intelligent tools supporting the design and optimization of composite structures operating under variable thermal conditions. |
