Inflows into deep rock tunnels are controlled by the highly variable nature of the fractures. The tunnel may be dry in most places, damp in others and subject to significant inflows at localized positions. It is likely that the highest inflow encountered along the tunnel comes from a fracture that was not encountered in the borehole investigation. Infrastructure projects in urban areas include water treatment and pumping facilities, which need to be designed and sized, despite the technical difficulties of predicting the total inflow.
The most common type of information that is provided to designers are the results of water pressure tests carried out between packers over various intervals in the boreholes, called "packer" or "Lugeon" tests. The packer test results are also dependent on the variable nature of the fractures and vary widely. There are some important pre-conditions for fitting packer test data to a distribution and these are described in the paper. Raymer (2001 and 2003) describes the analysis of packer test data, based on the assumption that the data is log-normally distributed and provides a methodology for predicting tunnel inflows.
It should be expected that a data set of 3m long packers should be able to be merged into a data set of 6m long packers without a change in the mean value. However, the log-normal distribution is not mathematically set up for such a transformation. Other authors have considered alternative distributions, and the paper reviews these alternatives.
This paper describes the possible choices of distributions for packer test data, and using a large database of real packer test data, attempts to fit the variously proposed distributions. The paper reviews the results and discusses the possibilities - is there evidence that a different distribution is suitable, or do the difficulties with the log-normal distribution result from an inherent issue within the problem. The paper provides conclusions and suggestions for further research.