中文摘要：

针对微粒群优化算法易陷入局部最优、出现早熟等不足,从作用力规则和种群拓扑结构两方面进行研究。提出一种混合作用力微粒群优化（Hybrid force PSO,HFPSO）算法,将算法的搜索过程划分为前期和后期两个阶段,分别构造引斥力规则和双引力规则,使算法搜索前期具有良好种群多样性、搜索后期有较高寻优精度。进一步将生物趋利避害的行为选择机制融入HFPSO算法,提出有向动态拓扑混合作用力微粒群优化算法,赋予微粒主观能动性使其靠近适应值较好微粒、远离适应值较差微粒,提出适应值驱动边变化的有向动态拓扑（Fitness-driven edge-changing unidirectional dynamic topology,FEUDT）结构,并将FEUDT结构与HFPSO算法以结构演化和算法进化同步进行的方式结合,进一步提升算法的优化性能。利用Benchmark函数对所提算法与标准PSO、搜索后期斥力增强型混合引斥力微粒群优化（LRPSO）算法进行性能对比测试,结果表明,所提算法具有较好的寻优能力和较快的收敛速度。通过桥式系统可靠性优化实例和供应商参与的某汽车产品子系统可靠性设计优化实例,验证了所提算法求解实际复杂优化问题的有效性。

英文摘要：

To overcome the defections of easy getting trapped in local optimum and premature convergence, the particle swarm optimization （PSO） algorithm is studied from two aspects, namely force rules and population topology. A hybrid force PSO （HFPSO） algorithm is proposed, the search process of the algorithm is divided into earlier period and later period two stages, attractive and repulsive force rule and double attractive force rule are constructed respectively, which can maintain good population diversity in earlier stage and improve the search accuracy in later stage. The unidirectional dynamic topology HFPSO algorithm is proposed, the biological behavior selection mechanism that biological individuals are willing to interact with better ones is integrated into HFPSO algorithm, a fitness-driven edge-changing unidirectional dynamic topology （FEUDT） is put forward, the FEUDT and HFPSO algorithm are combined by simultaneously evolving of both structure and algorithm, which can further improve the searching capability of the algorithm. Benchmark functions are used to compare the performance of the proposed algorithms with standard PSO and later-stage repulsion-enhanced hybrid attraction and repulsion PSO （LRPSO） algorithms, the results show that the proposed algorithms present better search capability of optimal solution and faster convergence speed. The proposed algorithms are applied in the reliability optimization of bridge network system and the automobile product subsystem in which the suppliers are involved, the effectiveness of the proposed algorithms to solve complex engineering optimization problems are further verified.