Abstract: Improved dung beetle optimizer is proposed to solve the disadvantages of the dung beetle optimizer such as weak stability, and sudden decrease in population diversity during the later iteration. Firstly, by using the good point set strategy to generate the population, the distribution of population positions becomes more balanced, laying the foundation for global optimization. Besides, the butterfly optimization algorithm with random inertia weight is integrated to enhance the ability of population mutation, increase the exploration degree of dung beetles in unknown fields, and balance global exploration and local development. Then, the stealing beetle endowed with the ability of adaptive spiral theft helps the algorithm conduct more detailed search. Finally, the lens imaging learning strategy , the throw out cross strategy and the greedy strategy are used to improve the algorithm’s ability to resist stagnation and help the algorithm capture global optima. Through comparative experiments on 15 benchmark test functions and partial CEC2014 test functions, ablation experiments with different strategies, Wilcoxon test, it is shown that the new algorithm has stronger robustness. The effectiveness of the improvement strategy and the engineering practicality of the algorithm are further verified by applying the improved algorithm to tension/compression spring engineering problem, three bar truss engineering problem, bearing fault diagnosis optimization problem.

Key words: dung beetle optimizer, butterfly optimization algorithm, spiral theft, hybrid mutation, engineering problem

摘要： 针对蜣螂优化器存在稳定性不强，迭代后期种群多样性骤减等缺陷，提出了一种改进的蜣螂优化算法（improved dung beetle optimizer）。通过佳点集策略生成种群，种群位置的分布更加均衡，为全局寻优作铺垫；融合蝴蝶优化算法并加入随机惯性权重，增强种群变异能力，增加蜣螂对未知领域的探索程度，平衡全局探索与局部开发；赋予偷窃蜣螂自适应螺旋偷窃的能力，帮助算法进行更加细致的搜索；引入透镜成像学习和外抛交叉随机混合变异策略与贪心策略，提高算法的抗停滞能力，帮助算法捕获全局最优。通过对15个基准测试函数及部分CEC2014测试函数的对比实验，不同策略的消融性实验，Wilcoxon检验，结果表明该算法有着更强的鲁棒性。将改进算法应用到拉伸/压缩弹簧工程、三杆桁架工程及轴承故障诊断优化问题上，进一步验证改进策略的有效性及算法的工程实用性。

关键词: 蜣螂优化器, 蝴蝶优化算法, 螺旋偷窃, 混合变异, 工程问题