基于变量筛选的岩质高边坡锚杆轴力预测

Anchor Bar Stress Prediction of High Rock Slope Based on Variable Selection

锚杆轴力的演变和发展规律对评价边坡稳定性极为重要.将平均影响值MIV与传统BP模型相结合,对锚杆轴力的影响因素进行筛选.同时引入粒子群算法与遗传算法对BP模型进行优化,在筛选时间、温度和开挖等变量的基础上,建立多个锚杆轴力预测模型.研究表明,MIV变量筛选能够提高预测模型的精度和减小误差.基于MIV变量筛选的PSO-BP模型在预测精度、模型稳定性上表现最好,在边坡锚杆轴力预测中具有一定的工程应用价值.

Evolution and development law of anchor bar stress are very important for the evaluation of the stability of slope. Mean impact value is combined with the traditional BP model,in which case effects of anchor bar stress can be filtrated. Meanwhile the particle swarm algorithm and genetic algorithms are introduced to optimize the error back propagation. Based on filtrating the variables of time,temperature and slope excavation and so on,several anchor bar stress prediction models are built. The research shows that,variable selection by MIV can improve the precision of prediction models and reduce errors. The PSO-BP model based on MIV variable selection has the best performance in prediction accuracy and model stability. It has certain engineering application value in the anchor bar stress prediction of slope.