Research On Soil-Shield Interaction And Driving Force Planning

Shield tunneling method has been widely used in the construction of the urbanunderground railway, cross-river channels, underground pipelines and other tunnel project.However, due to the uncertainty of geological circumstance and the complexity ofoperating conditions, the shield tunneling encounters the major technical problems duringexcavation, such as safety, efficiency, reliability and geological adaptability. Therefore, it isnecessary to investigate the soil-shield interaction and propose a rational driving forceplanning strategy. Against that background, particular attention will be focused on thetunnel stability, the driving force model and control problems in this paper. Throughin-depth study, the models of soil-shield interaction and driving force planning aredeveloped, and the validities are discussed by comparing with numerical simulation andengineering projects. The main works can be described as follows:①Tunnel stability analysisMuch works have been dedicated to tunnel face stability with a constant tunnelpressure by pioneering researchers. However, the support pressure increases in-depth with agradient, especially in the super large diameter tunnel, which affects the face stabilitygreatly. Numerical simulations confirm the trues that the gradient of support pressure notonly affects the limit support pressure, but also determines the active failure mechanism:global or partial failure. Furthermore, Terzaghi theory is commonly used to predict theloose earth pressure when the active failure is triggered in the vertical direction of tunnel.However, the centrifuge tests have shown that Terzaghi loose earth pressure is onlyconvenient for the shallow tunnel. Therefore, we present a multi-layer-blocks failuremechanism (MLBFM) for active failure of tunnel face and a limit analysis model for looseearth pressure based on upper bound approach. The results of examples confirm that theMLBFM is reasonable to predict the limit support pressure corresponding to supportpressure gradient; and the accuracy of limit analysis model is higher than that of Terzaghitheory.②Driving forces model for predicting onlineDue to the uncertainty factors during excavation, the force acting on the shieldperiphery and the force acting at the face are the predominant random resistances, whichgreatly influence the shield behavior. This paper takes into account the mutual coupling effect between the ground properties, shield behavior, structural parameters and operatingparameters, and presents a back-analysis model for dynamically predicting the driving force.The back-analysis loads model are composed of two parts: the force acting at the face andforce acting on the shield periphery. The former is obtained by the back-analysis of theearth pressure in the chamber recorded online, so that the dynamic characteristics, such asrandom fluctuation, can be represented immediately. In the latter, the ground reactioncurves are adopted to calculate the earth pressure normal to the shield periphery.Particularly, the ground displacement is considered to be induced by two kinds of motion:tunneling-induced ground loss and postural changes of shield. The model is validated bythe numerical simulation and the Line9in Tianjin, it confirms that: at the smooth stage, thepredicting results of the driving force show the same trend with the measured data, andachieve a higher prediction accuracy; but at the mutation stage, the predicting results have asignificant lag with respect to the measured data, and the error is obvious.③Driving force planning based on MDPIn shield tunneling, the control system needs very reliable capability of deviationrectifying in order to ensure that the tunnel trajectory meets the permissible criterion. Tothis goal, we present an approach that adopts Markov decision process (MDP) theory toplan the driving force with explicit representation of the uncertainty during excavation. Weconsider the case that the transition probability varies in a given domain estimated by fielddata, and discuss the optimal policy based on the interval arithmetic. The validity of theapproach is discussed by comparing the driving force planning with the actual operatingdata from the field records of Line9in Tianjin. It is proved that: at the smooth stage, theforce planning results agree well with the measured data, and the planning accuracy ishigher than predicting accuracy obtained by the dynamic predicting model of driving force;but at the mutation stage, the planning results have a significant error.④Driving force planning based on POMDPDriving force planning belongs to the multi-stage decision problem in the extremelycomplex environment. In order to comply with random mutation loads during excavation,the information of tunneling loads (the force acting on the shield periphery and the forceacting at the face) should be considered, so that the sequence decision-making ability canbe enhanced continuously with the interaction between the shield and environment.Therefore, we take advantage of the POMDP theory, dynamic predicting model andplanning model above mentioned to plan the driving force. In the POMDP, The shield operation information (such as shield attitude, tunneling speed, and earth pressure in thechamber) and the tunneling loads of possible world are scattered as the state set and actionset, respectively. Furthermore, the deviations of shield attitude from planned alignment andthe stationary of the loads acting on the shield are considered in the calculation of the valuefunction. In order to solve ultra-large-scale POMDP problem, a stochastic tournament valueiteration (STVI) is proposed with considering of the uncertainty of the belief state and theuniformity of belief distribution. The experimental results confirm that the driving forceplanning based on POMDP is reasonable, and have certain ability of compliance at themutation conditions.