关键词: 隧道边坡, 变形预测, PSO-LSSVM预测, ARMA预测, MC误差修正

Abstract:

The deformation rules of slope of tunnel portal should be learned in time and the safety of tunnel portal construction should be guaranteed. Hence, the error information of the deformation sequence is eliminated by wavelet transform and the original sequence is decomposed into trend term sequence and error term sequence. And then the trend term and error term are predicted by PSOLSSVM model and ARMA model, respectively; the comprehensive deformation prediction values of slope are obtained by superimposing trend term and error term; and the correction model of prediction error is established by Markov chain to further improve the prediction accuracy. Finally, a case study is carried out. The study results show that: (1) The denoising effects of sym9 wavelet function, heuristic threshold standard, hard threshold selection criteria and 10layer wavelet decomposition are better. (2) The average value of relative error and variance value of comprehensive prediction results are 1.03% and 0.042 6, respectively, which indicates the high prediction accuracy, stability and feasibility of the prediction model.

Key words: tunnel slope, deformation prediction, PSOLSSVM prediction, ARMA prediction, MC error correction