Calibration of a geometric computer model for planar stochastic fibre networks with application to tissue papers

A stochastic morphology model for constructing random fibre networks can be used to gain insight into tissue paper, more specifically laboratory-made paper sheets mimicking it. Geometry based models of heterogeneous materials pose unique challenges of their own that need to be considered in the respective calibration procedure. This means that detailed accounts would be quite valuable for the identification and evaluation of potential solutions which can be of general interest. We describe a calibration method for a single parameter which is a rational number between \numlist{0; 1} that specifies the flexibility of individual fibres in a network. Of particular interest here is an alternative to least squares estimation to eliminate the influence of constant bias. Errors resulting from measurement and model bias were corrected using the so-called discrepancy function making use of a frequentist two-step procedure for computer model calibration. The procedure is validated by cross-validation.