Application of Adaptive Neuro-Fuzzy Inference System in High Strength Concrete

Abstract

Adaptive Neuro-Fuzzy Inference System is growing to predict nonlinear behaviour of construction materials. However due to wide variety of parameters in this type of artificial intelligent machine, selecting the proper optimization methods together with the best fitting membership functions strongly affect the accuracy of prediction. In this study the non-linear relation between splitting tensile strength and modulus of elasticity with compressive strength of high strength concrete is modelled and the effect of different effective parameters of Adaptive Neuro-Fuzzy Inference System is investigated on these models. To specify the best arrangements of parameters in the System to utilize in high strength concrete properties, different combinations of optimization methods and membership functions in the Sugeno system have been applied on more than 300 previously conducted experimental datasets. Both the grid partition and sub-clustering methods have been applied to models and compared to get the best combination of parameters.