Economic and safe tunnelling requires a continuation of the detail design into the construction phase. Besides a prediction of the geological conditions ahead of the face a reliable prediction of the behaviour of the compound system of rock mass and support (system behaviour) is required. With the knowledge of the "normal" behaviour of the system the measured displacements can be compared to the predicted values in order to detect deviations in the behaviom in time. An additional benefit of such a prediction is the determination of the required over excavation in squeezing rock and the continuous control. To provide a reliable and accurate procedure for prediction of displacements the program "GeoFit" was developed. It considers several options, including the installation of support, sequential excavation, and non-steady tunnel advance.
Interpretation of displacement monitoring data in the past on most sites has been limited to "have a look" on displacement graphs. As shown in Rokahr et al. (2002), this visual examination of graphs has a number of shortcomings. One of the tasks of a geotechnical engineer on site is to do the final design of tunnel excavation and support on a day-to-day basis. The required decision-making is mainly supported by information from the geological documentation and the displacement monitoring data. The quality of the decision making process considerably can be improved by using advanced methods to interpret the monitored displacement data.To be able to continuously evaluate the stabilisation process the prediction of the "system behaviour" is required. Latter requires a special displacement prediction based on rock mass properties, stress field, installed support and excavation procedure. A recently developed procedure to predict displacements and thus system behaviour in tunnelling is shown in this paper.
Guenot et al. (1985) and Sulem et al. (1987) proposed a method based on analytical functions that describe displacements in a plane perpendicular to the tunnel axis as a function of time and the advancing face. Barlow (1986) and Sellner (2000) modified this approach. The displacement behaviour of the rock mass and support basically is represented by four function parameters. Two parameters are used to simulate time dependency and another two parameters to simulate the face advance effect. These parameters can be back calculated from case histories using curve fitting techniques. Artificial intelligence (neuronal networks) may be involved to identify dependencies between the function parameters and the geological and geotechnical conditions encountered. The system of these analytical functions was implemented in the program package called "GeoFit" (Sellner 2000). It provides easy-to-use tools for back calculating displacement monitoring data (curve fitting technique), for prediction of displacements and for handling the expert system. The application is acting interactively. Each change in the calculation assumptions is displayed on the screen immediately. Both, monitored and predicted results are shown. Trend lines, deflection lines, displacement plots and spatial displacement vector orientations can be evaluated and displayed on the basis of monitored, calculated and predicted data, allowing a continuous comparison of the actually measured and predicted data.