Prediction of Mechanical Properties of Polymer Concretes Under 3-Point Bending Loading Using Artificial Neural Networks
摘要
The bending response of a polyester resin reinforced by marble powder, silica, and sand grains was analyzed. For this purpose, 18 formulations (F1-F18) of polymer concrete (PC) composite specimens were prepared in a special mold according to ASTM C580-02. A model using artificial neural networks (ANN) was also proposed to predict the combinations of each constituent that lead to maximizing the mechanical properties of PCs. The results showed that all formulations of PCs involving 20 wt% resin have higher mechanical performance than those containing 14 wt% resin. PCs containing 20% resin, 10% sand, 30% silica, and 40% marble powder (F11) and 20% resin, 20% sand, 10% silica, and 50% marble powder (F13) have the highest bending stress (25.69 MPa) and bending modulus (14.63 GPa), respectively. On the other hand, the highest bending strain (0.37%) is obtained for F7 containing 14% resin, 30% sand, 10% silica, and 46% marble powder. Finally, it can be concluded that marble powder can replace silica, but there is a limit value beyond which the mechanical properties of PCs are greatly affected.