A comparative study of performance prediction models used in hard rock tunneling - A case study from the new Ulriken tunnel
MetadataShow full item record
Full face boring became a popular way of excavation in Norway at the beginning of the 70`s to the beginning of the 90`s. The excavation method was mainly used in hydro power- and sewage projects, due to the given profile. However, the use of TBM stopped for almost 25 years. In the later years has the method become popular again, with three projects that have started up within the last 5 years. The overall purpose of this thesis is to compare several performance prediction models for hard rock tunnel boring based on data collected within a 3000-meters long tunnel segment at the new Ulriken tunnel. The aim is to find the most promising and accurate models. The performance prediction models used is listed below. NTNU model by Macias (2016) QTBM model by Barton (1999) CSM model by Rostami (1997) MCSM model by Yagiz (2002) Gehring model by Gehring (1995) Alpine model by Wilfing (2016) Model by Hassanpour et al. (2011) Model by Farrokh et al. (2012) The models require multiple input parameters consisting of geological and machine data, gathered mainly by the contractor and client of the project. Due to the absence of BTS testing has the parameter been converted from UCS by using a conversion formula by Altindag and Guney (2010). Further was the fracture spacing measured as apparent fracture spacing. It was converted to true fracture spacing using the strike and dip of the fracture sets. Lastly was RQD values of areas that had been applied with shotcrete not available, therefore is these values not accounted in the average values. Despite the lack of geological information from some sections and other potential errors linked with the geological and machine parameters, do several of the used models show good correlation with achieved net penetration. The comparisons show that the Alpine model, the NTNU model, and the MCSM model is closest to the TBM performance. All of these models contain fracture properties, which from the thesis and other articles is found to be the most influent parameter on net penetration rate. The results show that the predicted values of each section vary more than the weighted average of the entire tunnel segment. The models share many of the same input parameters, so it is recommended to use more than one model when estimating the net penetration rate to ensure a reliable result.