Abstract
Concrete mix proportion affects concrete performance and cost. Therefore, it arouses an increase of interest in the research on performance improvements in concrete through the optimization of the raw material mix proportion, especially in the high cold, high-salinity and other complex environment. In this paper, an intelligent multi-objective optimization framework is constructed based on Bayesian optimization Random Forest (BO-RF) algorithm and fast non-dominated sorting genetic algorithm (NSGA-III). The key factors affecting concrete performance are taken as the input indexes, the freezing resistance, impermeability, carbonization depth and cost are taken as the evaluation indexes of concrete. Based on 400 groups of freeze–thaw cycle of salt invasion experiments in cold environment, the orthogonal experiment method is used to obtain the index system of the original data sets, then, the RF prediction algorithm is adopted to establish the key factors and performance index of a nonlinear mapping relationship and the Bayesian optimization random forest is used to establish a high-precision fitness function. Ultimately, the BO-RF-NSGA-III model is applied for multiobjective optimization. The results show that (1) Based on the proposed framework, the rapid prediction and optimization of concrete performance can be realized, which can well meet the needs of the complex alpine environment (2) Accurate model prediction can be made through BO-RF. The R2 values obtained at the time of prediction range from 0.944 to 0.998, the mean absolute error (MAE) range from 0.001 to 0.09. (3) Based on BO-RF-NSGA-III and grey correlation method, the comprehensive optimal mix proportion of concrete is obtained.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have