Xingyu Gao

Dendrobium polysaccharide is an important indicator for evaluating the quality of Dendrobium huoshanense. To improve the prediction accuracy of polysaccharide content, this study proposes an improved Artificial Lemming Algorithm (IALA) optimized BP neural network model. In IALA, a periodic mutation strategy and a fast hybrid opposition learning strategy (FHOBL) are introduced to enhance population diversity, improve global search ability, and avoid premature convergence. The proposed IALA was first evaluated on CEC2017 and CEC2020 benchmark functions. Experimental results show that IALA achieves better or competitive performance compared with seven other algorithms in terms of mean fitness, best fitness, and standard deviation. Statistical tests, including Wilcoxon rank-sum and Friedman tests, further verify the significant superiority and robustness of IALA. Then, IALA was used to optimize the initial weights and thresholds of BP neural networks for Dendrobium polysaccharide content prediction. The results show that IALA-BP achieves the best overall prediction performance, with an R² of 0.8731, RMSE of 2.1581, and MSE of 4.6683. Compared with standard BP and other optimized BP models, IALA-BP provides more accurate and stable prediction results. Therefore, the proposed IALA-BP model is effective for rapid prediction of Dendrobium polysaccharide content.

The Kangaroo Escape Optimization Technique (KET) combines escape and safe-area searches, but its fixed stage allocation, restricted guidance range, limited refinement of low-ranked individuals, and insufficient use of population-state information can reduce its performance on complex problems. This study develops a Multi-Strategy Kangaroo Escape Optimization Technique (MSKET) through iteration-dependent stage switching, population-proportion-based candidate guidance, selective greedy DE/rand-to-best/1 refinement, and beta-distribution opposition-based global-best guidance. The contribution lies in assigning established mechanisms to specific KET limitations and coordinating them across different stages and population subsets. MSKET was evaluated through 30 independent runs on the 100-dimensional CEC2017 and CEC2020 suites. It ranked first on 22 of 29 CEC2017 functions and 8 of 10 CEC2020 functions, with the best average rank on both suites. Ablation, diversity, and nonparametric statistical analyses further supported the observed performance gains. MSKET also achieved the best overall repeated-run results on the piston–lever and three-bar truss design problems. For near-infrared prediction, MSKET-BP obtained an R² of 0.86440, an RMSE of 2.1377, and a MAPE of 5.0083% on the testing set. These results indicate improved search and prediction performance within the examined tasks, although at a higher computational cost than KET.

To address the problems of the Electromagnetic Wave Propagation Algorithm (EMWPA)—insufficient initial-population coverage, an imbalance between exploration and exploitation, and a tendency to fall into local optima—in high-dimensional complex optimization problems, this paper proposes a boosted electromagnetic wave propagation optimization algorithm, BEMWPA. First, a cubic chaotic map is introduced in the population-initialization stage to enhance the uniformity of the initial-solution distribution and the search-space coverage. Second, nonlinear phase modulation is applied to the electric- and magnetic-field driving terms, and a differentiated probabilistic switching mechanism is constructed to improve the dynamic coordination between global exploration and local exploitation. Furthermore, a Beta-distribution opposition-based learning strategy is introduced to enhance the algorithm’s ability to escape local optima by generating high-quality opposite candidate solutions. To verify the effectiveness of the proposed algorithm, systematic comparative experiments are conducted on the CEC2017 benchmark function set, and BEMWPA is combined with rapidly-exploring random tree (RRT) and applied to path planning of a welding manipulator in complex multi-workpiece scenarios. For a three-dimensional welding scenario containing 12 workpieces, 12 closed weld seams, and multiple obstacle constraints, BEMWPA-RRT reduces the initial inter-seam transfer path length of RRT from 586.00 mm to 479.11 mm, representing a relative reduction of 18.24%, and the complete end-effector path length is reduced from 2974.00 mm to 2867.11 mm, representing a relative reduction of 3.59%. Meanwhile, the optimized transfer path length is only 1.59 mm longer than the obstacle-free ideal transfer length of 477.52 mm, indicating that the proposed method can approach the geometric lower bound of this scenario while satisfying the obstacle-avoidance constraints. Kinematic verification on a seven-degrees-of-freedom welding manipulator further shows that the optimized Cartesian-space path can be converted into a continuously executable joint-space trajectory, providing an effective method for offline welding path planning of complex multi-workpiece tasks.

To overcome challenges associated with Cloud Drift Optimization (CDO), primarily involving early stagnation and the imbalance between global exploration and local exploitation in tackling high-dimensional, complex optimization tasks, an improved variant named Chaotic Random Cloud Drift Optimization (CRCDO) is proposed in this study. First, a Kent chaotic initialization mapping is adopted to generate diverse initial solutions. Second, the parameter sensitivity is adjusted to appropriately strengthen the local exploitation capability of the algorithm. Third, a centroid opposition-based learning strategy is introduced to generate mutated clouds, which enables the algorithm to escape local optima. Owing to the cooperative effect of these three strategies, CRCDO achieves faster convergence and higher computational accuracy while maintaining dynamic stability. To evaluate its performance, CRCDO is compared with eleven well-established optimization algorithms on the CEC2017 suites. The experimental results demonstrate that CRCDO exhibits superior optimization performance on a wide range of complex problems. Furthermore, the proposed algorithm has been successfully applied to three representative engineering design problems, including robot gripper design, welded beam design, and compression spring design, as well as a regression task for predicting praeruptorin A and praeruptorin B in Peucedanum praeruptorum. The results show that CRCDO achieves faster convergence, higher solution accuracy, and greater stability on the benchmark problems, while its engineering and prediction applications demonstrate robustness and practical applicability across different problem classes.