Research on LMD-BA-ELM-based model for nonlinear prediction of slope deformation time-series
Han Yong-lian · Zhongguo anquan kexue xuebao · 2015
For effective analysis and accurate prediction of slope deformation,a model was built for nonlinear prediction of time-series of slope deformation. It was based on the basic principle of LMD,BA and ELM. LMD was used to decompose the time series of the slope deformation into multiple groups of relatively stable components. The BA was used to search out the global optimization solution of the parameters in the ELM,then a BA-ELM model was built. The components were predicted rollingly. Through superimposing and reconstructing the BA-ELM model components,eventual slope deformation prediction results were obtained. Taking a certain open-pit slope deformation monitoring data as an example,a prediction result comparison was made between the model built by the authors and the other three models. The results show that the prediction average absolute error of the model based LMD-BA-ELM is 0. 151 0 mm,the average relative error is 1. 287 3%,the running time is 7. 614 3 s,this model can fully discover the internal rules of slope deformation,effectively reduce the nonlinear features,and that it also has higher forecast precision and operational efficiency.