Artificial Intelligence-based Compressive Strength Prediction of Medium to High Strength Concrete

Predicting the compressive strength of concrete before casting is a requisite criterion. This research develops an artificial intelligence-based model for compressive strength prediction based on experimental data. A laboratory experimental program, covering the main variables that affect the test results, was conducted. Then, neural network (ANN) and neuro-fuzzy (ANFIS) models were developed based on four targets compressive strength (40, 60, 70, and 80 MPa), three testing ages (7, 28, and 90 days), three protocols of curing, two different specimens shapes (cylinder and cube), and three different specimens sizes. The model output, the predicted compressive strength, revealed a good agreement with the experimental test results for both ANN and ANFIS approaches. The correlation coefficients for ANN and ANFIS models are 0.976 and 0.989, respectively. The best results are obtained by ANFIS because it provides a slightly lower root mean squared error and a higher correlation coefficient.

Automated summary

This research developed an artificial intelligence-based model for predicting the compressive strength of concrete, and validated it with neural network and neuro-fuzzy models, achieving high correlation coefficients and low root mean squared errors in multiple aspects.