The Prediction Model of Dam Uplift Pressure Based on Random Forest

2017 2nd International Conference on Advanced Materials Research and Manufacturing Technologies

材料科学;机械制造

Li, Xing^1,2 ; Su, Huaizhi^1,2 ; Hu, Jiang^3

College of Water Conservancy and Hydropower Engineering, Hohai University, Nanjing

210098, China^1

State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Hohai University, Nanjing

210098, China^2

State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering, Nanjing Hydraulic Research Institute, Nanjing

210029, China^3

关键词: Evaluation factor;    Faster convergence;    Importance functions;    Prediction accuracy;    Prediction precision;    Pressure predictions;    Random forest algorithm;    Random forest modeling;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/229/1/012025/pdf
DOI  :  10.1088/1757-899X/229/1/012025

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

The prediction of the dam uplift pressure is of great significance in the dam safety monitoring. Based on the comprehensive consideration of various factors, 18 parameters are selected as the main factors affecting the prediction of uplift pressure, use the actual monitoring data of uplift pressure as the evaluation factors for the prediction model, based on the random forest algorithm and support vector machine to build the dam uplift pressure prediction model to predict the uplift pressure of the dam, and the predict performance of the two models were compared and analyzed. At the same time, based on the established random forest prediction model, the significance of each factor is analyzed, and the importance of each factor of the prediction model is calculated by the importance function. Results showed that: (1) RF prediction model can quickly and accurately predict the uplift pressure value according to the influence factors, the average prediction accuracy is above 96%, compared with the support vector machine (SVM) model, random forest model has better robustness, better prediction precision and faster convergence speed, and the random forest model is more robust to missing data and unbalanced data. (2) The effect of water level on uplift pressure is the largest, and the influence of rainfall on the uplift pressure is the smallest compared with other factors.