Reception & registration

Abstract: Numerous preventive maintenance policies have been published in the reliability and maintenance literature. Nevertheless, few examples on the application of preventive maintenance policies have been reported. The reasons are various, mainly because it is notoriously difficult to collect failure data. As a result, many developed maintenance policies divorce from the ground truth and are therefore inapplicable. This talk discusses some possible methods to shed some light on such problems. We tackle the challenge from a perspective of estimating system reliability based on sparse data and integrating various uncertainties in optimisation of maintenance policies. The uncertainties are stemmed from the uncertainty of parameter estimation on samples of small size, model specification uncertainty, and cost information uncertainty.

Abstract : Infrastructure managers face crucial investment decisions in the context of the ecological and energy transition, as well as the aging of assets installed decades ago. They are increasingly constrained by budgets, workforce, skills, planned outages, supply chain tensions, and financing. Risk science and reliability theory methods provide a solid foundation for informing investment decisions for asset renewal and the development of new infrastructures. However, adopting these methods also requires organizational transformations, adequate data, and specific skills that are not part of the core business of infrastructure managers. This lecture presents the current state of industrial practices, the main scientific challenges, knowledge transfer issues, and the role of standardization in supporting these transformations.

Abstract : Over the past 30 years, numerous papers have been published in the maintenance literature focusing on the maintenance of continuously deteriorating systems. A wide range of models have been proposed and developed to optimize maintenance decision-making for condition-based, predictive, or prescriptive maintenance. Recently, AI-based approaches have gained significant popularity. These methods show great promise for addressing a broad spectrum of problems and do not require theoretical hypotheses. In contrast, some maintenance models, particularly those dealing with continuous degradation spaces, rely on analytical developments grounded in the stochastic description of maintained systems. This talk focuses on this latter category of models, providing illustrative examples and exploring the advantages and drawbacks of this modeling framework for new developments. It discusses the contexts in which it can prove indispensable, in complementarity with other modeling frameworks e.g. in relation to sample sizes and data quality.