A hierarchical sparrow search algorithm to solve numerical optimization and estimate parameters of carbon fiber drawing process

The sparrow search algorithm (SSA) is an efficient swarm-intelligence-based algorithm and has been widely studied in recent years. Nevertheless, as with other swarm intelligence optimization approaches, the SSA is prone to fall into local solutions, which weakens the exploration ability. In order to cope with this problem, in this paper, a novel hierarchical SSA, named the HSSA, is proposed. Specifically, by introducing the virtual individual strategy in each iteration, each scrounger can obtain more effective guidance information in the developed HSSA, thus contributing to a thorough search in the entire problem space. On the other hand, the proposed hierarchical strategy is employed to realize the information interaction among individuals according to the division layers (including top, medium and bottom layers) with the purpose of enhancing the diversity of the original SSA. In addition, the life cycle mechanism is introduced in order to avoid the current scrounger individuals being trapped in local optima positions as much as possible, which can further improve the solution accuracy of the traditional SSA. The developed HSSA is verified on a series of benchmark functions (i.e., IEEE CEC2022) and the parameter optimization problem of the carbon fiber drawing process. The proposed HSSA is compared with three classes of existing swarm-intelligence-based approaches: (1) CLSSA, C-SSA, and CSSA as popular SSA-based methods; (2) BWO, HGS, GJO, and DBO as state-of-the-art optimization techniques; and (3) GWO, WOA, HHO, MPA, WSO, and POA as highly-cited swarm intelligence algorithms. Results demonstrate that the HSSA exhibits the best performance among twelve test functions and one practical application in terms of solving accuracy and convergence speed. Finally, the overall experimental results indicate that the developed HSSA is an effective method to deal with the defects of the basic SSA and can obtain satisfactory solutions for different optimization problems.