Development of Nanomodified Graphene Concrete Using Machine Learning Methods

Enhancing the performance of cementitious composites using nanomaterials have gained a significant research interest over the last few decades. Many studies have explored the use of various nanomaterial additives such as Carbon Nanotubes (CNTs), Carbon Nanofibers (CNFs), nano-silica and Graphene Oxide (GO) for cementitious composites. GO has shown more promising results over other nanomaterial additives as GO is hydrophilic and improves many properties of cementitious composites such as the mechanical strength, durability, resistance to fire, etc. While many studies have reported the compressive strength improvement of GO induced cementitious composites, there is a notable variance between the increments. Based on parameters such as GO dosage and water/cement ratio, the strength increments vary from as little as 5% to more than 125%. This paper primarily investigates the relationship between water/binder ratio and GO dosage in cementitious composites. The study includes a series of experiments on mechanical performance of GO induced cementitious composites, an intensive data collection from previous studies, synthesis, and characterization of GO. Moreover, for the first time, this paper presents a machine learning (ML) regression model to identify the most important factors for strength improvements and to predict the compressive strength gain of GO induced cementitious composites. The ML model was developed using more than 200 datapoints including various inputs as training sets and testing tests. The ML model was tuned to achieve a reliable accuracy and the model was validated through experimental investigations on cementitious composites. Based on the outcomes of the ML model, the most important variables for strength enhancement were identified and then applied to develop a GO induced high performance concrete which showed significant strength improvements over reference samples.