The concrete mix proportion of 502groups was collected as training data. A concrete formula design model which can be used to control concrete cost and formula optimization was constructed based on BP neural network, genetic algorithm and particle swarm optimization algorithm. In this model the cost of raw materials of concrete and the several key factors influencing the compressive strength of concrete were taken into consideration, the penalty function was introduced to the particle swarm algorithm for the fitness value of the objective function of punishment. They were used to solve the concrete formula in the design of nonlinear constraints discrete variables and continuous variables, and may be used to control concrete cost and optimize the mix proportion. The 27 sets of concrete mix proportions after cost reduction were obtained, and the compressive strength test was conducted by the established model. The test results show that the combined ratio cost to the target cost is close to 97%, when the concrete cost per cubic meter is reduced by 5 yuan, 10 yuan and 15 yuan, the performance of mix proportion exported by the model meets the strength requirements.