Publication Type : Journal Article
Source : IOP Conference Series: Materials Science and Engineering 872 (1), 012119, 2021
Url : https://www.sciencedirect.com/science/article/pii/S2214785322022738
Keywords : Geopolymer Mortar, Self-Compacting Geopolymer Mortar, Deep learning, Machine Learning, Strength Prediction, Durability tests
Campus : Coimbatore
School : School of Engineering
Department : Civil
Verified : No
Year : 2022
Abstract : One of the looming sustainable technology in the construction industry is geopolymer composites. Experimental and machine learning approach has been used for the study of fly ash – ground granulated blast furnace slag (GGBFS) based geopolymer mortar with ambient curing condition. Fresh properties of geopolymer mortar and self-compacting geopolymer mortar including setting time and workability characteristics have been studied. Harden property (compressive strength, split tensile strength, flexural strength) and durability studies (shrinkage, acid resistance, sulphate resistance, salt resistance, water absorption) are also conducted for selected geopolymer mixes. Self-compacting geopolymer mortar shows high workability and increased setting time than normal geopolymer mortar. But in the hardened property as well as in durability tests normal geopolymer mortar shows better performance.
Geopolymer mortar required a standard framework for mix design due to its complexity of different parameters such as water to solid ratio, alkali activators to binder ratio, sodium silicate solution to sodium hydroxide solution ratio etc. A multilayer ANN architecture is used for an effective prediction of compressive strength. Most of the research works were based on regression algorithms. The present study utilized the tensor flow approach developed by Google for the prediction of compressive strength. For training 150 data from different journal papers were used and the data generated by the authors experimentally in the lab is used for validation. Input independent variables considered in the proposed model were fly ash and GGBFS content, the molarity of sodium hydroxide solution, sodium silicate to sodium hydroxide ratio, and alkaline solution to binder ratio, fine aggregate to binder ratio. Compressive strength was considered as the output dependent variable. Model accuracy is checked quantitatively checked and obtained 3.6 MPa as Root mean squared error, 2.6 MPa as Mean absolute error, 8.09 % as Mean absolute percentage error and 0.6 as Root mean squared error.
Cite this Research Publication : RR Lakshan, AM Rosini, K Sathiyan, D Sathyan, KM Mini, "Comparison of different dosages of PCM incorporated wallpanels", IOP Conference Series: Materials Science and Engineering 872 (1), 012119, 2021