Performance Prediction And Mix Optimization Of High Performance Concrete Based On Evolutionary Algorithms

High-performance concrete (HPC) has since been used in many important concrete construction industries with its high strength, high durability and high workability. In addition to the basic ingredients in conventional concrete, i.e., Portland cement, fine and coarse aggregates, and water, the making of HPC needs to incorporate superplasticizer and chemical admixture. High-performance concrete is such a highly complex material that performance prediction model is a difficult task to be built. So far , there has not been any guide specification on the mix proportion of HPC. Required mix proportion is therefore obtained by trail and error method that is based on rough mix proportion by calculate and then does experiment to decide the final mix proportion. Such methods waste a lot of time, a great deal of labor and material resources. Thus, it is significant to find a method of HPC mix-proportion, which can predict the performance of HPC and get the mix meeting the needs of strength, durability and workability quickly and dependably.In this paper, on the basis of existing experimental data, genetic programming (GP) theory is applied to predict the performance of high performance concrete and the mix optimization of high performance concrete is also researched by applying the multi-objective genetic algorithms theory on such basis in order to solving the mix meeting the needs of strength, durability and workability quickly and dependably. The main contents of the dissertation are listed as follows:(1) Based on the computing platform of MATLAB,the procedure of HPC performance prediction is programmed by applying genetic programming (GP) theory. According to existing experimental data, a visible nonlinear mathematical expression that contains the complex relationship between strength or slump and respective influence factors is given by applying the method of GP. The performance of the proposed GP model is analysed and the computing result is compared with that of the linear regression and BP neural network model, which show that the GP model has higher fitting precision with the experiment data and better prediction effect. Thus, it can be used to predict performance of high performance concrete as a new method and it has a very good guide for mix optimization research of high performance concrete. (2) Based on multi-objective genetic algorithms theory,the model of HPC mix optimization is built and corresponding computer program is worked out, in which the nonlinear function expressions of prediction strength and slump that are got by GP model served as objective function and basic restriction of every variable used as constraint conditions. The results show that the algorithm can effectively solve the problem of HPC mix design, and a non-dominated solutions that contains optimal solution is obtained, which can make more favorable decision support for decision-makers.(3) Based on the above study, intelligent decision system of performance prediction and mix design of HPC is programmed with C#, which has such features as interface-friendly, convenient and easy to use and thus can provides decision support for performance prediction and mix design for decision-makers.