Abstract: Although the settlement deformation of high-speed railway tunnels is not high, it is possible that a settle? ment curve may not reflect actual settlement deformation due to random noise interference. An RLG denoising method combining a wavelet transform with a Kalman filter is put forward that has the functions of noise-related interference removal and multiple resolution analysis by wavelet transform, and also has the advantages of linear unbiased minimum variance estimation to unknown signals by Kalman filtering. By applying the GM(1,1) model to analyze the data after denoising, it is detemined that the GM(1,1) model combining the wavelet transform and Kalman filtering is higher in precision than that of the GM(1,1) model based only on Kalman filtering and can be used for settlement analysis in high-speed railway tunnels.