A Predictive Model Based on Singular Spectrum Analysis and ARIMA Model With Adaptive Orders for Shield Tunneling Machine Cutterhead Torque

Accurate prediction of cutter head torque is of great significance to ensure the safe tunneling of the shield tunneling machine. However, because geological conditions are complex and changeable, frequent outliers and time-varying nonstationary characteristics exist in the data of the cutterhead torque. Existing predictive models fixed their model structures and parameters, which can not catch the above characteristics. Thus, to solve this problem, we propose a predictive model based on the singular spectrum analysis (SSA) and the autoregressive integrated moving average (ARIMA) model with adaptive orders. First, SSA is utilized to decompose the original data, and the trend quantity of the original data is adopted to represent the original data to reduce the influence of outliers. Second, to catch the time-varying nonstationary characteristics of data, the augmented Dickey-Fuller (ADF) test and Bayesian information criterion (BIC) are adopted to construct the adaptive orders identification algorithm for the ARIMA model. Finally, sufficient results illustrate that the proposed model can accurately predict the data of cutterhead torque in real time under different geological conditions. Specifically, the average values of the coefficient of determination (R2) obtained by the proposed method in 1–5 steps are 0.842, 0.800, 0.758, 0.706, and 0.654, which outperform existing models.