Use of artificial intelligence for predicting different properties of concrete

Abstract

In the present study an effort has been made to predict different concrete properties using five different machine learning techniques - Support Vector Machine (SVM), Artificial Neural Network (ANN), Fuzzy Inference System (FIS), Adaptive Fuzzy Inference system (ANFIS) and Genetic Expression Programing (GEP). For this purpose, two hundred numbers of concrete mixtures data have been considered by varying the level of key ingredients- cement, water, fine aggregate, coarse aggregate and size of coarse aggregate along with their strength and workability. Among these eighty five percent data have been used for training purpose and fifteen percent data have been used for testing purpose in all the five machine learning methods. Furthermore, for validation of the five networks, experimental investigation have been carried out for fifteen numbers of different concrete mixes. It is to be mentioned that all the five methods yields satisfactory results for predicting compressive strength and slump value in both testing and validation. However, ANFIS yields best result among all the five methods for predicting the same in testing and validation. In addition to that, multi objective optimisation and robust optimisation have been carried out using the basis equations obtained from GEP to determine the proportions of concrete mixtures for maximum concrete strength and workability at lowest cost. The results corroborates that present methods can be used successfully for predicting concrete property using different ingredients. Moreover, the optimisation procedure and result can be used for obtaining accurate number of mix proportions with desired compressive strength, and workability at minimum cost.