Heliyon (Feb 2024)
Durability, thermo-physical characteristics, and mechanical strength prediction of green Portland cement matrix incorporating recycled soda-lime glass and lead glass
Abstract
The current study is concerned with acid and calcination durability, thermal and thermo-physical properties, and mechanical strength prediction of mortars containing soda-lime glass (PVS) and lead glass (PVP). It demonstrates that up to 30% of PVP (PVP30) and PVS (PVS30) enhancements lessen the consequences of acid attack. In both cases, 20% additions show the best acid resistance at 2 days, but mortars with 10% addition resist better at 28 days. Furthermore, sulfuric acid damages the formed mortars more aggressively than hydrochloric acid. According to the thermal study, the loss of mass owing to calcination is reduced with increasing glass addition. It falls from 22% to −19.5% for PVS30 and -18% for PVP30. The flexural strengths of the calcined mortars significantly drop after firing, although the compressive strengths are higher at 400 °C than at ambient temperature. However, at 600 °C, a 20% glass addition retains the mortar's fire resistance. However, around 800 °C, all formulations mechanically deteriorate. PVP20 has the best fire behavior with relative variations of 48.6% at 400 °C, 18.5% at 600 °C, and −45.8% at 800 °C, while PVS20 has 45.4% at 400 °C, 24.8% at 600 °C, and −33.1% at 800 °C. The hydrates found in the calcined mortars emphasize autoclave reactions that improve mechanical characteristics between 400 and 600 °C, whereas at 800 °C, advanced dehydration of the matrix results in a generalized decrease in resistance. Furthermore, the gradual inclusion of glass reduces the thermal conductivity of mortars correspondingly. The inclusion of 30% PVS results in a reduction of −38.99%, while 30% PVP results in a reduction of −49.95%. The other thermophysical parameters are calculated as a function of these values. The models developed in the area of mechanical strength prediction using the Multilayer Perceptron (MLP) method of Artificial Neural Network (ANN) allow for R2 correlation coefficients of 0.86–0.92 during training with the database and 0.77 to 0.90 during validation, with values of MAE ≤ 2.12 and RMSE ≤ 2.67 in all situations.
