RESEARCH ON OPTIMIZATION OF PROPORTIONS FOR HIGHLY DURABLE CONCRETE MIX BASED ON RF-NSGA-Ⅱ ALGORITHM

The problem of insufficient durability for concrete has always been a serious in the complex environment of high cold areas. It is of great significance to design the proportion of concrete mix to ensure the good working performances and durability of concrete. A RF-NSGA-II algorithm was adopted. The multi-objective model of optimization design for concrete mix was constructed, in which the two important durability indexes of concrete:frost resistance and impermeability were taken as the objectives, so as to achieve high-precision prediction of the relative elastic moduli, chloride ion permeability coefficient and optimal proportion of concrete mix. Firstly, based on the data of in real time concrete material and mix proportions, the prediction models of random forest (RF) for frost resistance and impermeability of concrete was constructed respectively. Then,the regression prediction function was taken as the objective function, and the constraint range of factors for mix proportions was determinal according to Specification for Mix Proportion Design or Ordinary Concrete(JGJ 55-2011) and engineering requirements. Finally,the NSGA-Ⅱ algorithm was used for multi-objective optimization to obtain the optimal mix proportion. The results showed that the prediction model for relative elastic moduli in time and chloride ion permeability coefficients of concrete was very good, the error of φ_RMSE was 0.04, the coefficient of determination R² was in time as 0.985. Through optimization, the optimization target value of the proportion concrete mix met the requirements of durability and mechanical tests, which was consistent with the actual situation of the project.