组合最小二乘支持向量机与粒子群优化算法研究黄土湿陷性

A Study of Loess Collapsibility by Combining Least Squares Support Vector Machines with Particle Swarm Optimization Algorithm

通过静力触探试验指标结合扰动黄土试样的液限、塑限及含水量等指标用最小二乘支持向量机方法进行建模,提出了静力触探试验指标和湿陷系数的非线性关系模型,并引入粒子群优化算法进行模型反演分析,确定最优参数。通过6个对比勘探点的50组试样实例应用分析,显示了最小二乘支持向量机是一种较为有效的非线性建模方法,粒子群优化算法进行模型参数优化能够保证全局最优。验证结果表明模型的精度较高,有一定的实用价值。

Through the static contact probing test indexes in combining with the indexes of liquid limit, plastic limit and water contents of disturbed loess samples, the mathematical model is established using the least square support vector machine method. The non-linear relation model between the static contact probing test indexes and loess collapsibility coefficients is suggested. Also, the particle swarm optimization algorithm is introduced to carry out the model inverse analysis so as to determine optimal parameters. The real sample application analysis of so groups from 6 comparative surveying points indicates that the least square support vector machine is a kind of effective non-linear model establishment method, and that the particle swarm algorithm to optimize model parameters is able to guarantee the whole optimization. The testing results show that the model is high in accuracy and practical in use.