摘要: |
【目的】针对土石坝坝体沉降存在多变量、强耦合、强干扰的复杂问题,建立基于KPCA-RVM的土石坝沉降预测模型。【方法】利用核主元分析(KPCA)对输入向量进行降维处理,以减少因子个数,随后利用相关向量机(RVM)模型对土石坝沉降进行预测,并以平均相对误差为指标对预测精度进行评价。【结果】 实例应用表明,KPCA-RVM模型将输入向量由14个降低到7个,预测结果的平均相对误差仅为0.9%,预测效果得到明显提升。【结论】利用KPCA-RVM模型对土石坝进行沉降预测,不仅可以减少输入向量个数,而且可以提高预测精度,可在实际工程中推广应用。 |
关键词: 土石坝 KPCA-RVM模型 沉降预测 核主元分析 相关向量机 |
DOI: |
分类号: |
基金项目:国家自然科学基金项目(51409205);陕西省重点科技创新团队项目(2013KCT 15);博士后自然科学基金项目(2015M572656XB);水文水资源与水利工程科学国家重点实验室开放研究基金项目(2014491011);西安理工大学水利水电学院青年科技创新团队项目(2016ZZKT 14) |
|
KPCA-RVM based prediction model for settlement of earth rockfill dam |
MA Chunhui,YANG Jie,CHENG Lin,et al
|
Abstract: |
【Objective】A KPCA-RVM based prediction model for settlement of earth-rockfill dam was established aiming at the complex characteristics of multi variables,strong coupling and strong interference in settlement of earth-rockfill dams.【Method】The kernel principal component analysis (KPCA) was used to reduce the number of the input vectors.Then,the settlement of earth-rockfill dam was predicted using the relevant vector machine (RVM) model,and the prediction accuracy was evaluated using average relative error.【Result】The number of input vectors was reduced from 14 to 7 by KPCA-RVM model.The average relative error of prediction results was only 0.9%,indicating the prediction was significantly improved.【Conclusion】Using KPCA-RVM model to predict settlement of earth dam not only reduced the number of input vectors,but also improved the prediction accuracy.The KPCA-RVM model has great application in practical projects. |
Key words: earth-rockfill dam KPCA-RVM model settlement prediction kernel principal component analysis relevance vector machine |