Multi-strategy enhanced black-winged kite algorithm and its engineering applications

The Black-winged Kite Algorithm (BKA) is a metaheuristic algorithm inspired by the bird's migratory and predatory behaviors. While demonstrating strong adaptability, few parameters, and high convergence precision, the BKA algorithm tends to converge prematurely and exhibits slow convergence speed when solving complex optimization problems. To address these issues, this paper proposes a Multi-strategy Enhanced Black-winged Kite Algorithm (MSEBKA) incorporating four key improvements: First, in the attack phase, an adaptive inertia weight (AIW) strategy is introduced to reduce sensitivity to the quality of random numbers, enhance search stability, and effectively balance global exploration and local exploitation. Second, embedding an elite pool (EP) strategy during the migration phase enhances population diversity by integrating information from multiple elite individuals. Next, after the migration phase is completed, a dynamic Gaussian random walk (DGRW) strategy is adopted to perturb the black-winged kite population, promoting the generation of new individuals and helping the algorithm escape from local optima and accelerate convergence. In addition, lens imaging opposition-based learning (LIOBL) strategy is introduced after each round of population updates to further enhance the ability to escape local optima by generating opposite solution for the current best search individual. To evaluate the performance of the proposed MSEBKA algorithm, three different experiments are carried out. The experimental results demonstrate that, (1) the proposed MSEBKA algorithm not only outperforms the BKA algorithm and its four ablated variants on 6 out of 10 of the CEC2019 benchmark functions regarding mean fitness value and standard deviation, but also shows better convergence performance on all functions, (2) Evaluated on CEC2019, 23 classic and CEC2021 benchmark functions, the proposed MSEBKA algorithm shows superior overall performance, attaining notably better Friedman mean ranks (1.250, 1.804 and 1.550) compared to the BKA algorithm (4.950, 3.000 and 2.550) and five other mainstream metaheuristic algorithms, (3) furthermore, the proposed MSEBKA algorithm is successfully applied to five typical constrained optimization problems and performs better than the BKA algorithm on the design problems of pressure vessel design, tension/compression spring design, welded beam design and speed reducer design, with respective performance improvements of 0.0345%, 0.0375%, 1.7164%, and 0.0016%.