Predicting the compressive strength of eco-friendly and normal concretes using hybridized fuzzy inference system and particle swarm optimization algorithm

• Published in: The Artificial intelligence review Vol. 56; no. 8; pp. 7965 - 7984
• Main Authors: Golafshani, Emadaldin Mohammadi, Behnood, Ali, Arashpour, Mehrdad
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.08.2023; Springer; Springer Nature B.V
• Subjects: Algorithms; Analysis; Artificial Intelligence; Blast furnace slags; Cement; Compressive strength; Computer Science; Concrete; Datasets; Fly ash; Inference; Machine learning; Mathematical optimization; Mixtures; Natural resources; Neural networks; Optimization algorithms; Particle swarm optimization; Prediction models; Silica fume; Swarm intelligence; Fuzzy inference system; Supplementary cementitious materials; Fuzzy system; Fuzzy type-2; Concrete compressive strength
• ISSN: 0269-2821; 1573-7462
• DOI: 10.1007/s10462-022-10373-4

The use of supplementary cementitious materials ( SCMs ) in the binder system of concrete mixtures can provide several environmental and technical benefits. Several previous studies have focused on evaluating the compressive strength ( CS ) of concretes containing SCMs using machine learning ( ML ) techniques. However, there have been a limited number of studies for modeling the CS of concretes using type-2 fuzzy inference system ( FIS ) in which the uncertainties in measuring the input variables and membership functions can be handled. This study serves the interval type-2 FIS ( IT2FIS ) to develop predictive models for the CS of concretes containing three types of SCMs , including blast furnace slag, fly ash, and silica fume. Particle swarm optimization ( PSO ) algorithm was also used to optimize the parameters of the IT2FIS . In addition, type-1 FIS ( T1FIS ) was served as the control ML technique. The dataset used in this study contains information on the mixture proportion and CS values of 3240 concrete mixtures. A total of 18 FIS models, including 6 T1FISs and 12 IT2FISs were developed. The results showed insignificant differences between the error metrics of the FIS models for the training and testing phases, which indicates the good generalization capabilities of the developed FIS models. To have more insight into the role of input variables on the CS of concrete, the relevancy factor ( RF ) analysis was carried out for the input variables of the best-developed FIS model. It was found that cement content had the most positive effect on the value of  CS .