Prediction of Concrete Compressive Strength Using Fuzzy Logic and Model Tree

Abstract

Concrete is a popular construction material which is used in most of the construction processes as a major stress resistance material due to its strength characteristics. Designing a concrete mix is tough task which includes the appropriate proportion of the ingredients used in particular environment with the objective to produce concrete mix with specified strength, durability, workability and as economical as possible. Traditionally, a concrete mix is accepted or rejected depending on its destructive testing which demands time and materials. Recently, data-driven techniques like artificial neural networks, genetic programming, fuzzy logic, etc., are been used efficiently to predict 28-day strength of concrete. Fuzzy logic is efficiently used for predicting the strength of concrete but involves a tedious activity of defining the rules. Thus, the aim of the current paper is to predict 28-day strength of concrete using fuzzy logic algorithm, and an attempt is done to define the rules of the algorithm using model tree. Two different models were developed: with fuzzy logic and traditional method of rule formation and fuzzy logic and rule formation using model tree (termed as hybrid model). The inputs of the models are cement, fly ash, fine aggregate, 10 mm coarse aggregate, 20 mm coarse aggregate, water, admixture all in kg/m³ and output as 28 days strength of concrete in MPa. The results of this hybrid model display promising results as compared to the basic fuzzy model developed.