Modeling of marshall test flow value of modified bituminous mixtures by using fuzzy logic and regression methods

Transportation structures have become an indispensable part of our lives. Various types of pavement coatings are used in the design of these structures. In our country, flexible pavements are mostly preferred type of pavement due to their comfort, low noise level and low cost. In this thesis, Marshall stability test experiments, which require a laborious and costly work process, were applied and flow values were tried to be estimated using parameters such as viscosity, penetration, softening point, weight of the sample in air, weight in water, voids in mineral aggregates, void ratio, practical specific gravity. Flow data estimation models were made with six different regression analysis methods and two different methods of fuzzy logic, which is one of the artificial intelligence techniques. A total of 2190 prediction models were developed for ten different techniques by making 219 models for each technique. In addition, SMRGT method, which is a new technique, has been applied to the best models. By comparing the results obtained, the best estimation model was reached. As a result of comparison with real data, it was observed that the data obtained from fuzzy logic methods gave much better results than regression methods. Eventually, it was determined that the Marshall test flow value model created by the SMRGT method gave the most successful results.