基于CE-EMD-GWO-RVM-CT模型的碎石土滑坡变形预测研究

Study on deformation prediction of gravel soil landslides based on CE-EMD-GWO-RVM-CT model

摘要

【目的】为实现碎石土滑坡的高精度变形预测,【方法】基于滑坡勘查成果及变形监测成果,先利用互补式集合经验模态实现变形数据的多尺度变量识别,再利用灰狼优化算法、相关向量机及混沌理论构建变形预测模型,以期掌握滑坡变形特征,并指导其灾害防治。【结果】结果显示:互补式集合经验模态在构建过程中的优化处理能合理提高识别效果,能有效识别滑坡变形数据的多尺度变量。预测过程的各组合优化步骤能有效提高预测精度,且在4个监测点最终预测结果中,相对误差均值介于1.96%~2.20%,方差值介于0.002 0~0.003 5,充分说明预测模型具有极强的稳健性。【结论】结果表明:顾及变形数据多尺度变量识别的预测模型适用于碎石土滑坡变形预测,且据其外推预测结果对比,C1监测点的现有变形速率相对更大,其余3个监测点的预测速率相对更大,说明滑坡后缘稳定性将趋于稳定,但其前缘位置的稳定性将会趋于不利,建议尽快开展此滑坡的防治处理。

Abstract

[Objective] To achieve high-precision deformation prediction of crushed stone soil landslides, [Methods] based on the result of landslide exploration and deformation monitoring, multi-scale variable identification of deformation data is first achieved using complementary set empirical modes. Then, a deformation prediction model is constructed using Grey Wolf optimization algorithm, correlation vector machine, and chaos theory to grasp the deformation characteristics of landslides and guide their disaster prevention and control. [Results] The result show that the optimization of the complementary set empirical mode in the construction process can reasonably improve the recognition effect and effectively identify multi-scale variables of landslide deformation data. The combination optimization steps of the prediction process can effectively improve the prediction accuracy,and in the final prediction result of the four monitoring points, the average relative error ranges from 1. 96% to 2. 20%, and the variance value ranges from 0. 002 0 to 0. 003 5, fully demonstrating the strong robustness of the prediction model. [Conclusion] The result indicate that the prediction model considering multi-scale variable identification of deformation data is suitable for predicting the deformation of gravel soil landslides. According to the comparison of its extrapolation prediction result, the existing deformation rate of C1 monitoring point is relatively higher, and the prediction rate of the other three monitoring points is relatively higher. It indicates that the stability of the landslide's trailing edge will tend to be stable, but the stability of its leading edge position will tend to be unfavorable. It is recommended to carry out prevention and control measures for this landslide as soon as possible.

关键词

碎石土滑坡 / 多尺度变量 / 经验模态 / 变形预测 / 相关向量机 / 滑坡

Key words

gravel soil landslide / multi scale variables / empirical mode / deformation prediction / relevance vector machine / landslides

基金资助

国家自然科学基金项目(52278352)