Theatre B, Richard Berry Building
The University of Melbourne
CSIRO Mathematical and Information Sciences
This talk will discuss some large consulting projects dealing with failures in distribution networks for electricity and water. In each case, the traditional approach is to model failures as a non-homogeneous Poisson process, with a variety of covariates. There were major problems marrying up huge data sets from different sources before any thought could be given to the analysis. Modelling of the data then needed to take account of factors which influenced failure rates, but also factors to allow for biases arising from the data sources.