基于粒子群与支持向量机的隧道变形预测模型  

Tunnel deformation prediction model based on support vector machine with particle swarm optimization algorithm

作　　者：范思遐[1] 周奇才[1] 熊肖磊[1] 赵炯[1] 

机构地区：[1]同济大学机械与能源工程学院,上海201804

出　　处：《计算机工程与应用》2014年第5期6-10,15,共6页Computer Engineering and Applications

基　　金：国家自然科学基金(No.51375345);上海科学技术委员会项目(No.08201202103)

摘　　要：针对粒子群算法易早熟且在算法后期易在全局最优解附近产生振荡现象,提出一种自适应调整惯性权重的优化粒子群算法。该算法引入双曲线正切函数的非线性变化思想,使惯性权重随着迭代次数的增加产生自适应调整,有利于增强粒子搜索能力及收敛速度,不易陷入局部极值点。将该算法应用于基于支持向量机的隧道变形预测模型中,对预测模型的超参数进行优化,并利用稳态与非稳态两组实测工况数据对组合算法进行工程测试,结果表明采用SaωPSO+SVM算法可有效提高预测模型的计算精度,增强其鲁棒性,有助于隧道变形的工程建模。Aiming at solving the precocious convergence problem and the oscillation phenomena around global optimal solution of Particle Swarm Optimization（PSO）algorithm, a self-adaption inertia weight method based on PSO（SaωPSO） is proposed. The nonlinear change idea of the hyperbolic tangent function is introduced to the new algorithm. Additionally, the inertia weight with the increasing iterations produces adaptive adjustment. Moreover, this would enhance the searching ability and convergence speed of particle, and avoid falling into local extremum. The SaωPSO with Support Vector Machine （SVM）is applied to build the tunnel deformation prediction model. In the case analyses, the new method is used for parameters optimization. The effectiveness of the hybrid algorithm is tested by the two performance measured data in steady and non-steady states. The result shows that SaωPSO＋SVM could improve the forecasting precision, enhance its robustness, and also contribute the deformation prediction model.

关 键 词：支持向量机 粒子群 隧道变形 预测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]