混沌序列PSO-RBFNN耦合模型在滑坡监测中的应用-杨虎

滑坡位移系统的发展演化受到多种不确定性因素的影响，可能存在非线性特征。而同时包含了确定性和非确定性分析的混沌理论，能有效阐释滑坡位移序列的复杂的非线性过程。因此本文首先对滑坡位移序列进行混沌分析，揭示其内在演化机理；在相空间重构的基础上，再采用拟合和泛化能力较好的径向基网络（RBFNN）对其位移值进行实时动态预测，针对RBF网络存在参数选取困难的问题，运用粒子群算法（PSO）对RBF网络的参数进行优选。提出了基于混沌理论的PSO-RBFNN滑坡位移预测模型。经过实例验证，并与粒子群优化的BP神经网络（PSO-BP）和单独RBF网络进行对比，表明滑坡位移序列确实存在混沌特性且PSO-RBFNN模型预测精度更高、效果更好。

Application of PSO-RBFNN coupling model Based on Chaos Theory in Forecast displacement of landslide

Due to the nonlinear deformation evolution of landslide system, it is difficult to describe it with simple physical and mathematical model. Radial basis function neural network (RBFNN) is proposed to dynamically forecasting the landslide displacement for its good fitting and generalization ability, but the parameter of RBF neural network is difficult to select. Based on the chaos character analysis the time series of landslide displacement, the Particle Swarm Optimization (PSO) is used to training parameters of RBF neural network, and a PSO-RBFNN Coupling model based on Chaos theory is given for predicting displacement of landslide. It is testified by instance and compared with RBFNN model and BP neural network with Particle Swarm Optimization, the predicted results indicated RBF neural network based on PSO is more precise and has better performance in the prediction of landslide displacement.