SEHHO-COBL: An Enhanced Harris Hawks Optimization with History-Guided Adaptive Parameter Memory and Chaotic Opposition-Based Learning for Numerical and Engineering Optimization

Harris Hawks Optimization (HHO) offers a structurally elegant escape energy-driven exploration-exploitation transition, yet its underlying siege strategies exhibit weak directionality, rigid step-size control, and progressive diversity loss as dimensionality grows. This paper introduces SEHHO-COBL, which retains HHO's phase division skeleton but entirely replaces the siege mechanism with a history-guided adaptive differential evolution engine: a circular memory bank records successful scaling factors and crossover rates, enabling landscape-responsive parameter self-adaptation via DE/current-to-pbest/1 mutation. The framework is further strengthened by Tent-map chaotic opposition-based learning initialisation, an external archive that enriches differential-vector diversity, and L\'{e}vy flight perturbation during the exploration phase. Comparative experiments on the CEC2022 benchmark (12 functions, 10D and 20D, 30 independent runs, 8 competitors) demonstrate that SEHHO-COBL ranks first on 11 of 12 functions at 10D (Friedman 1.25) and on all 12 at 20D (Friedman 1.00; Wilcoxon 96/0/0). A systematic ablation study confirms that the adaptive parameter memory is the most critical module ($p = 0.002$). Application to five constrained engineering design problems ($D = 4$-$38$) yields a perfect first-rank record, validating the algorithm's practical effectiveness.