Estimation of the compressive strength of mortars using microwave curing method and artificial neural networks

Concrete is the most widely used building material today. Compressive strength, which is the most important mechanical property of concrete, is determined by experiments conducted 28 days after its production. In the specifications, it is an important problem to learn the compressive strength that determines the quality and class of concrete so late. It is possible to estimate the compressive strength of concrete by using early strength values. In this study, it is aimed to evaluate the artificial neural network (ANN) model and its effectiveness which can predict the compressive strength of samples under normal curing conditions using experimental data obtained from mortar samples cured by microwave. Microwave and normal curing were applied to 46 mortar samples prepared in different mixing ratios, with or without mineral additives (fly ash and silica fume), and with or without chemical additives (plasticizers), using two different cement types. Microwave (MD) cured samples were subjected to the weight measurement, ultrasonic pulse velocity, bending strength and compressive strength tests. In the study of artificial neural networks, the data obtained from the MD cured test samples were arranged as the input data, and the 28 days compressive strengths obtained from normally cured samples were assigned as output data. In order to find the optimal network topology of ANN, training was carried out by increasing the number of neurons in the hidden layer from 1 to 50. Levenberg-Marquard (LM), Scaled Conjugate Gradient (SCG) and Bayesian Regulation (BR) backpropagation algorithms were used for training. The optimal ANN structure for each of the three training algorithms was investigated separately. It should be noted that the neural network with 3 neurons in its hidden layer, which was trained with the BR algorithm was identified as the most successful network with the 0.9149 correlation (R) value.