Study On Dam Deformation Prediction Model Based On Wavelet Packet And Gene Expression Programming

Affected by lots of factors such as environment factor in the dam deformationmonitoring,there must be lots of noise in monitoring data. If we use the dam deformationmonitoring data which have lots of noise to build model,the predictive results of the model willbe unreliable.Moreover,traditional methods of dam deformation have limitations on datapretreatment and forecast model,so it is necessary to research dam deformation using waveletpacket and gene expression programming.Firstly,taking the shortage of wavelet packet whose hard threshold function is notdeviation into account，this paper proposes new threshold function of wavelet packet.Secondly,the value of traditional wavelet packet threshold function remainunchanging,conflicting with the fact that noise attenuate while filter stages increase. this paperproposes new value of threshold function of wavelet packet.Thirdly,as too much factors in dam deformation is hard to extract，this paper use VC++tobuild a n-GEP forecast model according to gene expression programming principle and modelingprocess.Fourthly,this paper use Matlab and VC++to build an improved wavelet and n-GEPcombined forecast model considering the limitation in traditional dam deformation monitoring.Lastly,this paper use Matlab as a platform to make comparison in de-noising effect usingwavelet,wavelet packet and improved wavelet packet mothod and forecast effect usingMetabolism Grey Model,BP Neural Network and improved wavelet and n-GEP combinedforecast model,and analyze the results.These results demonstrate:firstly，the de-noising effect of improved wavelet packet is betterthan wavelet and wavelet packet.secondly, On the five evaluation criterion such as sum ofsquares for Error,standard deviation for error,the average relative error,the variance ratio of errorto original value and root-mean-square error.the predicting results in improved wavelet andn-GEP combined forecast model are less than Metabolism Grey Model corresponding about10、3.1、4.2、3.1、3.2multiple and less than BP Neural Network about15、3.3、37.5、3.3、3.9multiple.So this model has a good predicting effect and the results are stable.Therefore,improved wavelet and n-GEP combined forecast model is worth popularizing in damdeformation monitoring.