project

Application future failure frequency prediction method

Expert(s):
Karel van Laarhoven PhD MSc, Yuki Fujita PhD, Bas Wols PhD MSc, Andreas Moerman MSc

  • Start date
    01 Jan 2018
  • End date
    31 Dec 2019
  • Principal
    Bedrijfstakonderzoek
  • collaborating partners
    Brabant Water

The prediction of failure frequencies is important in determining pipe replacement volumes and replacement prioritisation. The predictions are used to prioritise pipe replacements, and are a factor in estimating the evolution of operational expenses (OPEX) and CML. Uncertainties in the predictions have a big impact on water utilities’ investment policies and in determining which replacement are next in line and the replacement timing.

This project will be further elaborating the failure prediction method. This will be done by adding physical models to the comparison with statistical models, and by studying the method’s possibilities in a case study.

Supported degradation curves

The prediction of failure frequencies is usually done by selecting a statistical model and by fitting the model to the failure data. It is difficult to make a well-founded choice of a model, even if the choice will mean significant differences in the results. In this project a method will be elaborated in which the Comsima stress model is used to make a well-founded choice of a statistical model. In addition, the researchers will compare the most common models in the literature with the failure frequencies, after which they will determine the set of models to be calculated. This will provide an improved prediction of future failure frequencies as well as a reliability margin regarding the results.

Application in a case

The method will be applied in a case study to test its possibilities. In this case study, the failure registrations will be analysed – e.g., failure frequency distribution as a function of installation location, cohort layout and trends. The project will also compare the results with USTORE data. Among other things, this will produce differences in trends, interpretations and recommendations so as to complement USTORE with desirable parameters. Subsequently the method will be run. The project will interpret and describe the results and make recommendations for improvement.