变形监测数据分析的群集智能模型

针对标准微粒群优化算法(PSO)存在早熟收敛和搜索精度低等缺陷,对其惯性权重因子进行改进,并将非均匀变异机制引入PSO,即在算法进行搜索时以一定变异概率对选中的粒子进行变异操作,同时对飞离搜索区域的粒子用新产生的粒子取代,由此得到改进的微粒群优化算法(IPSO)。然后将IPSO用于优化BP神经网络的连接权重和阈值,分析和建立变形监测数据处理的群集智能模型(IPSO-BP),为了进行比较,同时建立了回归分析模型、标准PSO-BP模型,并将3种模型分别应用于某大坝水平位移数据的分析预测,结果表明,IPSO-BP模型收敛速度更快、预报精度更高,是一种新的且有效、可靠的变形数据处理方法。

Swarm Intelligence Model for Data Processing of Deformation Monitoring

In view of the disadvantages of standard PSO: premature convergence,low searching accuracy,an improved inertia weight factor is adopted,and an inhomogeneous variation mechanism is introduced to improve PSO,which is named IPSO.When searched,mutation operation is conducted with certain mutation probability on selected particles,particles away from searching areas are substituted by mutated ones.Then IPSO is used to optimize the weight and threshold value of BPNN and a swarm intelligence model(IPSO-BP) for def...