Abstract: In order to inverse divisional mechanic parameters and overcome shortcomings of conventional Particle Swarm Optimization (PSO), improved PSO algorithms such as stretching objective function based PSO, Metropolis algorithm based PSO and adaptive algorithm based PSO were used and compared with Quantum Particle Swarm Optimization Algorithm and catastrophic PSO. The case study showed that improved PSO was more effective than conventional PSO, and fitted to inverse divisional mechanic parameters of dam.

摘要： 为反演大坝的力学参数，需克服粒子群算法不能保证收敛到全局最优解的缺点，本文应用了一种新的粒子群算法：采用了拉伸目标函数法、Metropolis算法和自适应权重。同时将上述改进算法同量子粒子群算法和灾变粒子群算法进行了比较，将实际反演结果同超声波检测和常规基于混合模型的反演结果进行了比较，解释了常规反演结果偏小的原因。工程实例表明，改进算法具有计算量小、全局收敛性能高的特点，与灾变粒子群算法和量子粒子群算法性能相当。