Mix Design of Fly Ash Based Alkali Activated Concrete

• Autor(a) principal: Tashima, M. M. [UNESP]
• Data de Publicação: 2021
• Outros Autores: Payá, J., Borrachero, M. V., Monzó, J., Soriano, L.
• Tipo de documento: Capítulo de livro
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1016/B978-0-323-85469-6.00017-9; http://hdl.handle.net/11449/240543
• topic: Compressive strength; Fly ash alkali activated concrete; Machine learning methods; Mix design; Sustainability

Despite the widespread availability of research on fly ash alkali activated concrete and several proposed methodologies to calculate mix proportions, a universally applicable mix design process for the same is still predominantly reliant on the trial and error method. To address this deficit Artificial Neural Network (ANN) and Multivariate Adaptive Regression Spline (MARS) techniques have been utilized to compare the 28-day compressive strength predictions against the actual values. Prepared database was divided into training and testing in order to evaluate model performance. It is evident that MARS model performed more accurately than ANN model, predicting estimated compressive strength similar to the actual compressive strength values obtained through laboratory experiments. Contour plots were developed to represent the correlation between four key parameters and compressive strength. Expected compressive strengths at 28 days varied from 30 to 55. MPa were obtained, using the proposed mix design methodology. Hence, this mix design tool has the ability to deliver a novel approach for the design of fly ash alkali activated concrete mixes in order to obtain the expected compressive strength applicable to the requirement of the construction application.