The Study On The Parameter Optimization Of Shield Tunneling

The shield tunneling method is more and more applied into metro projects. But the study on the regularity of tunneling parameters has lagged. It is necessary to study the regularity of parameters according to some geological condition and alignment condition, to forecast the proper parameters, and to diagnose typical failures that are inclined to occur in the process, so as to assure the process of security and lower the cost and the adverse effect of metro projects.In order to obtain the regularity of parameters, that is, to acquire the distribution of parameters on some conditions, to find out functional relationship and influencing factors for parameters, mathematical statistics is an effective way. The statistic conclusions go as follows: all the parameters appear approximately normal distribution on the conditions appointed; comparing with conclusions derived from model tests, all the parameters measured have a too weak quantitative relation to describe by explicit formulation; the geological conditions will have greater effects on all the parameters than the alignment conditions will do.According to the statistic conclusions above, the study on the regularity of tunneling parameters needs other means. The means of Artificial Neural Networks (ANN) is competent for disposal of problems with fuzzy information. The tunneling parameters can be grouped in two: one is input variable as outer conditions, and the other is output variable as system response. Then, the problem on the parameter regularity can be converted into the problem on forecasting the output variable by the input variable. The study conclusion is that, the parameter-forecasting method for shield tunneling is feasible, especially for the testing advance under complicated geological conditions.At present, comparing with the study on the state of stratum and structure for risk estimation for shield tunneling, less attention has been paid on the shield system, particularly on the tunneling parameters which reflect working state. Some kinds of typical failures can be described by state vectors composed of tunneling parameters after typical failures are classified. So, the problem of diagnosing of failures in the tunneling process can be converted into that of pattern discrimination of complex systems. By means of ANN pattern discrimination techniques, the typical failures in constructions can be diagnosed. By testing the diagnosis results, ANN can diagnose some typical failures well, and can offer a good guidance for the shield tunneling.