Synergistic effect on the performance of ash-based bricks with glass wastes and granite tailings along with strength prediction by adopting machine learning approach
2022
Praburanganathan, Selvaraj | Chithra, Sarangapani | Simha reddy, Yeddula Bharath
The study proposes a novel and sustainable method to appropriately utilize wastes from granite as well as glass industries in brick manufacturing. An ecofriendly and low-cost manufacturing process of ash-based bricks pertaining to the Indian standard codal provisions that can be adopted on the commercial scale is deliberated. The research also recommends the method for predicting the strength of the ash-based bricks using machine learning algorithms like random forests and decision trees. For positive synergy in the performance, both the granite tailings and glass waste must be used together. Using the granite tailings and glass waste together led to a significant reduction of 75% in the fly ash requirement without compromising the brick’s performance. The addition of the granite tailings and glass waste in the mix could increase the strength of the brick by 90.5% and 37.7%, respectively. Beyond 30% dosage of granite, tailings are not recommended as they may lead to the poor gradation of particles and weak bonding in the microstructure. The glass waste in the mixture should not be more than 15% as it causes the dilution of pozzolanic reactions thereby forming fewer hydrated compounds. Brick’s durability is known after exposing the specimens for 1 year to sewers and biogenic corrosion environment, marine environment, and saline soil environment, respectively. The inclusion of the industrial wastes significantly reduced the specimen damage in the extreme environmental conditions along with the least absorption rates. The dosage of ash, granite tailings, and glass waste has to be maintained around 15%, 30%, and 15%, respectively for attaining the optimum performance. Out of the generated machine learning algorithms, only random forests could be able to predict the values accurately with R² values at 0.90 and with comparatively lesser errors.
显示更多 [+] 显示较少 [-]