ABSTRACT
To assure high quality of tunnel, controlling shield attitude is crucial issue while tunneling. Current control models mainly rely on the expert experience rules or geometrical axis fitting and neglect the important fact that shield attitude control strategies not only impact the direction of tunnel axis but also impact the disturbances of surrounding soil, so it is not reasonable to consider only a single control target. This paper proposes a new big data-driven Shield Attitude Control method (DSAC) to optimize shield posture Adjustment strategy, which consider tunnel axis control and settlement control comprehensively. DSAC mine historical engineering data from different projects by machine learning to establish two sub-models, one for setting shield attitude control target, the other for embodying the relationship between control variables (shield oil pressures) and control objectives (shield attitude). We used 550 thousand data records in Shanghai Metro line as training sample to create DSAC model and applied it to control other shield attitude of Shanghai Metro Line 18, The engineering application show the performance of tunnel axis and settlement are improved compared with the traditional control strategy method.
