Prediction of compressive strength of concrete using Adaptive neuro-fuzzy inference system

Evaluating the in situ concrete compressive strength is difficult due to having the non-linear relationship between parameters, different mix designs, methods of mixing, and concrete properties. In this paper, considering the experimental results based on in-situ tests, adaptive neuro-fuzzy inference system (ANFIS) is established for predicting the 28 days’ compressive strength of concrete. The database of actual concrete properties is collected based on the earlier researches and used for training and testing the ANFIS network. The dataset includes water to cement ratio, aggregate fine materials, and aggregate coarse materials to determine the concrete compressive strength. For this purpose, 90% (180) data is considered for training the network and the remaining 20 data (10%) is divided for testing and evaluating the proposed model. The results of model and loss function values showed that trained network can be considered as powerful model to predict concrete compressive strength.

Prediction of compressive strength of concrete using Adaptive neuro-fuzzy inference system Keywords:

ANFIS , 28 days’ compressive strength , concrete , artificial neural network