Prediction of early age compressive strength of concrete using machine learning

Abstract

This study introduces an innovative approach to predicting the early-age compressive strength (CS) of concrete by integrating Internet of Things (IoT) with Artificial Neural Networks (ANNs). Unlike traditional destructive testing methods, this technique employs a non-contact and non-destructive structural health monitoring (SHM) system that continuously tracks the hydration temperature of concrete using embedded LM35 temperature sensors. These sensors are connected to an Arduino microcontroller and an ESP8266 Wi-Fi module, enabling real-time data transmission to the ThingSpeak cloud platform. To calculate the temperature‒time factor (TTF), Concrete temperature, temperature sensors, and IoT techniques were used. Based on the maturity method, for the mix grades 20, 25, 30, 35, and 40, the graph was subsequently developed using the temperature time factor and the concrete compressive strength (CS). To determine the flexural strength (FS) of the concrete, it was validated experimentally using the developed graph. Further analysis performed using ANNs to validate the experimental results. The study investigated via the classification of neural networks (NNs), such as Neural Network-Levenberg-Marquardt (NN-LM) and Neural Network Gradient Descent (NN-GD). These two ANN models were compared with NN-LM, showing marginally better accuracy. The research employs a feed-forward back-propagation neural network. A comparative study reveals that the differences in prediction of flexural strengths range from 0.02 to 3.90 MPa for the maturity method (MM) and − 0.69 to -4.80 MPa for the ANN. Estimated strength via a sensor through the IoT, MM and ANN prediction models closely match the experimental findings. The proposed model can assist civil engineers in real-time formwork removal and quality control decisions without destructive testing.