Identification of heat transfer coefficients in continuous casting by a GPU-based improved comprehensive learning particle swarm optimization algorithm

Abstract

The heat transfer coefficients (HTCs) of the secondary cooling zone (SCZ) are a critical boundary condition for numerical modeling of solidification heat transfer in the continuous casting process. However, it is difficult to identify accurately these HTCs. Due to the high computational complexity, the computation time in the identification process is not negligible, especially during the process of optimization and heat transfer model solution. To solve those problems, we first developed an improved comprehensive learning particle swarm optimization algorithm (ICL-PSO) to identify more accurate HTCs in the SCZ of continuous casting. Secondly, a two-layer parallel algorithm that implements ICL-PSO on a graphics processing unit (GPU) was then designed and tested to reduce the computation time of identifying these HTCs. Finally, the experimental results show that the accuracy of the proposed algorithm for HTC identification and the efficiency of GPU-accelerated computation have been improved.