Intelligent mix design of steel fiber reinforced concrete using a particle swarm algorithm based on a multi-objective optimization model

Abstract

Steel fiber reinforced concrete (SFRC) is widely used in construction and is important for concrete-based applications. Nevertheless, compared with conventional concrete, it is difficult to efficiently and accurately design SFRC with a given mix ratio, because many factors affect the SFRC performance. This study proposes a multi-objective optimization model using numerical simulations and artificial intelligence to effectively derive the optimal SFRC mix proportion. For a quick and accurate relationship between the SFRC mix ratio and compressive strength, Latin hypercube sampling was used for sampling the random variables determining the mix ratio, yielding a set of random numbers. Subsequently, a finite-element simulation dataset was constructed, and a backpropagation neural network (BPNN) was used for predicting the complex nonlinear relationship between the raw material mix proportions and uniaxial compressive strength (UCS). The BPNN model was then utilized as the objective function in the multi-objective particle swarm optimization model, with the mix ratio parameters as the input variables and the compressive strength, unit production cost, and carbon dioxide emission as objective functions, which facilitated the search for the optimal mix proportion, yielding a Pareto-optimal solution set. Finally, based on the engineering preferences, the best solution was determined to be the recommended mix proportion.