Abstract: In order to solve the problems of limited design parameters, over-reliance on manual adjustment and failure to make full use of the existing test data in the traditional proportion design, a novel intelligent optimization strategy for high performance concrete (HPC) was proposed, which integrates artificial intelligence algorithms and meta-heuristic search technology, and can realize the proportion design under the consideration of the multi-objective demand of HPC. Firstly, a database with a large amount of HPC proportion data and performance test information was constructed. Secondly, three machine learning algorithms, XGBoost, MLP and RF algorithm, were used to establish the nonlinear mapping relationship between compressive strength of concrete and the amount of material used, and the model with the best performance was selected as the strength objective function and constraints in the multi-objective optimization intelligent. Then, the Pareto optimal predicate was solved by establishing the mathematical model with the objectives of compressive strength, economy, and environmental protection, and the Pareto optimal predicate was solved by using the NSGA-II algorithm. Lastly, the unique optimal solution was obtained by combining with the ideal-point method. The proposed method was validated by practical engineering examples, and the results show that the proposed multi-objective intelligent optimization design method of high-performance concrete proportion takes into full consideration of interactions between variables, greatly reduces the number of tests, significantly improves the efficiency of the proportion design, and can provide a reference for the multi-objective intelligent proportion optimization design of HPC.

Key words: high performance concrete, strength prediction model, NSGA-II algorithm, multi-objective optimization, XGBoost algorithm