ANN based predictive mimicker for mechanical and rheological properties of eco-friendly geopolymer concrete. (December 2022)

• Record Type: Journal Article
• Authors: Rehman, Fazal; Khokhar, Sikandar Ali; Khushnood, Rao Arsalan
• Is Part Of: Case studies in construction materials. Volume 17(2022)
• Publication Date: 2022-12
• Subjects: FA Fly ash -- GGBFS Ground Granulated blast furnace slag -- OPC Ordinary Portland Cement -- NZ Natural Zeolite -- ANN Artificial Neural Network -- GEP Gene expression programming -- ML Machine learning -- DT Decision Tree -- SVM Support Vector machine -- RF Random Forecast -- RHA Rice Husk ash -- NaOH Sodium Hydroxide -- Na2SiO3 Sodium Silicate; Geopolymer -- Machine learning -- ANN -- Predictive model -- Mechanical properties -- Rheological properties; Building materials -- Case studies -- Periodicals; 691.05
• DOI: 10.1016/j.cscm.2022.e01536
• Languages: English
• ISSNs: 2214-5095

Abstract

Abstract: Due to an increase in global warming, the construction industry, like the rest of the world is turning towards sustainable solutions. The construction industry is the major contributor to global warming primarily due to the use of cement. Geopolymer is an eco-friendly construction material that utilizes zero cement for its production. However, the major issue that limits its commercial implementation is its complex mix design, which is not as straightforward as conventional concrete. As geopolymer contains more elements than conventional concrete, its mix design process is more challenging. Alongside there are no defined guidelines for material designing of geopolymer concrete, which makes the task of designing it quite time-consuming, uneconomical, and iterative. The objective of this research is to develop a machine learning model that can predict the mechanical and rheological properties of geopolymer concrete. An Artificial Neural Network-based model was developed, which takes the input of the mix's constituents and predicts both mechanical and rheological properties as a result. MAE (Mean square error) for compressive strength, elastic modulus, flexural strength, and slump value for a training set were 2.53, 0.72, 0.121, and 8.9, respectively, while MAE for the testing set was 4.32, 1.5, 0.65, and 19.7. These performance results of MAE seem excellent to be used for prediction. This paper will help in the effective design of geopolymer concrete with limited experimentation. Graphical Abstract: ga1 Highlights: ANN based technique was used to predict the mechanical and rheological properties of geopolymer concrete. The accuracy of the model was improved by using hyperparameter tuning. This predictive model can help to find the optimum mix of geopolymer without extensive iterations.