基于EMD-RBFNN的稀土原地浸矿边坡位移预测

受温差、冰霜、扰动等因素影响，原地浸矿在线监测系统采集的数据含有大量噪声和干扰信号，利用系统自带的温度补偿模块难以达到预定数据精度，使得后续预测预警工作出现误差。为此，对原始信号进行处理，EMD分解后，IMF分量可实现自由重构，去掉高频分量，能够较好地去除环境因素对在线监测位移数据的影响，低频分量能更好地反映实际位移值。借助EMD技术的自适应分解特性，提取真实监测数据，并利用RBFNN的最佳逼近效果，建立在线监测数据EMD-RBFNN预测模型。根据某稀土矿实测地表位移数据，进行预测检验，结果表明，EMD-RBFNN模型的地表位移预测数据相对误差仅0.12%，具有较好可靠性和预测精度。

Displacement Prediction of In-Situ Leach Mining Slope of Rare Earth Based on EMD-RBFNN Model

As data collected from the online monitoring system of in-situ leach mining would contain a lot of noise and interference signals due to some influencing factors including temperature difference，frost and disturbance，it is difficult to reach a scheduled data accuracy with the system′s temperature compensation module，resulting in inaccuracy in the follow-up forecasting and pre-warning work.Therefore，after the EMD(Empirical Mode Decomposition) decomposition，the IMF component can achieve freedom refactoring，and remove the high frequency component.It can well purify environment factors on the on-line monitoring displacement data，and the influence of low frequency component can better reflect the actual displacement.The dissertation establishes a forecasting model-EMD-RBFNN（Radical Basis Function neural network）for online monitoring data，levering the best approximation effect of RBFNN，after dealing with the original signal and having the real monitoring data extracted by taking advantage of the adaptive decomposition characteristics of EMD technology.The prediction test is carried out based on the actual measured earth surface displacement data of some rare earth mine.The results show that the surface displacement prediction data boasts more reliability and accuracy as its relative error is within only 0.12%.