Defence of dissertation in the field of water and environmental engineering, Tuija Laakso, M.Sc. (Tech.)
Water and wastewater networks in Finland are aging and large-scale renovations will be needed at all utilities of the country. The aim of the dissertation was to study, how renovations can be targeted appropriately.
Research shows that risk-based asset management can be supported by integration and analysis of data on the networks and their environment. In this study pipe-level failure consequences are estimated for water and wastewater networks and sewer condition and life span are modelled with such data.
The predictive ability of machine learning methods applied (random forest and random survival forest) was similar or slightly better than that of the traditional statistical methods (logistic regression and Weibull regression), but the difference was small. Data availability and the purpose of creating the model also affect the choice of the method.
The effect of explanatory factors on sewer condition was studied using partial dependence plots, which enable examining the interconnections between different variable values and pipe condition. Results showed that that in order to benefit from life span models, the condition inspections need to be targeted differently than before. Pipes need to be selected randomly for inspections and the selected pipes need to be re-inspected regularly in order to gain reliable estimates.
The opponent is Professor, Dr Zoran Kapelan from TU Delft, Netherlands.
The custos is Professor Riku Vahala from the Department of Built Environment, Aalto University.
Electronic dissertation: http://urn.fi/URN:ISBN:978-952-60-3853-7
Doctoral candidate’s contact information:
Tel. 050 521 5968
A doctoral dissertation is a public document and shall be available at Aalto University, School of Engineering’s notice board in Otakaari 4, Espoo at the latest 10 days prior to public defence.