EMI-machine learning based monitoring and prediction of compressive strength of graphene-oxide based concrete

The present study employs a real time, non-destructive electro-mechanical impedance (EMI) technique using embedded piezoelectric sensor (EPS) to monitor and predict the compressive strength of graphene oxide (GO) modified concrete composites via machine learning (ML) models. Concrete mixes were prepared: one as control, that is without GO and another with the addition of 0.01 to 0.05% weight of cement in concrete. EPS collects EMI signatures of a GO concrete composites specimen during the 95 days of strength development. From these responses, an equivalent stiffness parameter (ESP) was formed using an ideal mechanical representation of spring and damper. These ESP values were utilized along with the experimental destructive test results for strength and were employed to develop the ML models for predicting the compressive strength of GO modified concrete. Five-fold cross-validation method is used to for training and testing of the dataset. Experimental results revealed that GO based concrete composites are higher in strength compared to the control mix without GO. The addition of GO (0.03% by weight of cement) helped in the increment of 31% in compressive strength at 3 days and around 25% increase at 28 days compared to conventional concrete (CC). The ESP calculated non-destructively matches well with the compressive strength calculated destructively. Further, the developed ML models show accurate results in predicting the strength of GO based concrete composites. Hence, for assessing and monitoring the strength growth development of concrete composite at a nano level the proposed method can be effectively utilized.