Data Analysis of Fly Ash Geopolymer Compressive Strength Using Machine Learning Method

• Main Authors: Hissyam, Hazmi, Idawati, Ismail, Annisa, Jamali, Mohamad Nazim, Jambli
• Format: Proceeding
• Language: English
• Published: Springer 2022
• Subjects: TA Engineering (General). Civil engineering (General)
• Online Access: http://ir.unimas.my/id/eprint/42570/3/Data%20Analysis.pdf; http://ir.unimas.my/id/eprint/42570/

Geopolymer is an alternative material that is suitable to substitute Ordinary Portland Cement (OPC) to produce concrete. A mixture of geopolymer paste that binds coarse and fine aggregate and other unreacted materials together is called Geopolymer Concrete. Previous studies stated that alkaline activator molarity, water binder ratio, and type of activator played a significant role in the compressive strength of geopolymer concrete. Machine learning or artificial neural networks are particularly appropriate for modelling non-linear relationships, and they are characteristically used to accomplish pattern recognition and categorize objects or signals in vision, speech, and control systems. This research is to analyze compressive strength data sets of geopolymer concrete by using the machine learning method. The result comparison of compressive strength is divided into three parameters which are based on molarity, water binder ratio, and the type of activators in the ratio between sodium hydroxide (NaOH) and sodium silicate (Na2SiO3). The materials used for the preparation of geopolymer concrete in this study are fly ash as a binder, fine and coarse aggregates, water, sodium silicate sodium hydroxide, and (NaOH) (Na2SiO3) as activators. A total of 240 samples were cast and cured at 80 oC for 24 hours with 28 days age of maturity before it’s have been tested for the compressive strength. This study confirms that the molarity, water binding ratio, and the type of activator pointedly affect the compressive strength of geopolymer concrete. The compressive strength was further analyzed by MATLAB to observe the neural network and clustering of the compressive strength data. It is found that there are 9 clusters discovered. The clustering of the compressive strength shows that there is a likeness of material usage in the creation of Geopolymer Concrete.