Natural hazards in mountains pose threat to people, properties and infrastructures. Several protection structures are implemented in such areas aiming to reduce the induced risk and to protect vulnerable exposed issues. These structures constitute critical interdependent multi-component systems that deteriorate over time due to the harsh phenomena they are subjected to since their construction. The propagation of damage is influenced by the dependencies between different failure modes and between the structures themselves. Such cascading events reduce the efficacy of the overall system in providing protection. Hence, if not regularly maintained, the level of protection offered by these structures will be reduced. Due to limited budgetary resources, it is essential to identify the most critical types of failures and existing dependencies in order to capture the state of degradation at which preventive maintenance should be carried out. Consequently, performing a reliability analysis of such systems is highly important. This requires analyzing the dynamic deterioration of each structure and the local or distant effects of its partial or total unavailability on other structures involved in a multi-component system. The methodology presented in this project integrates physics-based and reliability-based models for monitoring the state evolution of protection structures and improving maintenance decision-making processes. The modeling approach proposed is based on (1) physics-based modeling for identifying the probabilistic laws of the transition times between the defined states of the structure depending on its behavior over time and (2) a decision aiding method based on Petri nets, which helps in choosing the best maintenance strategy while considering budgetary constraints. In this context, undertaken decisions always depend on the amount of available data, diverse sources, assumptions, etc. Consequently, this study also focuses on assessing the influence of uncertain inputs on the outputs obtained from deterioration and maintenance models. The developed models are applied on two different protection structures (check dams and retention systems) subjected to different types of natural phenomena (clear water floods and debris flows). The achieved results permit to analyze the time-dependent behavior of a structure and to prioritize several proposed maintenance strategies taking into consideration economic constraints and the efficiency of each strategy in increasing the availability time of the structure in non-critical states.
 

Keywords
Natural risk; protection structures; deterioration modeling, transition times; stochastic Petri nets; maintenance optimization; information imperfection

MEMBRES DU JURY

John ANDREWS - Professeur, University of Nottingham, Rapporteur
David BIGAUD - Professeur des Universités, Université d'Angers, Rapporteur
François PÉRÈS - Professeur des Universités, Ecole Nationale d'Ingénieurs de Tarbes, Rapporteur
Johannes HÜBL - Professeur, University of Natural Resources and Life Sciences, Examinateur
Didier GEORGES - Professeur des Universités, Université Grenoble Alpes, Examinateur
Aurélie TALON - Maitre de Conférences, Université Clermont Auvergne, Examinateur
Christophe BÉRENGUER - Professeur des Universités, Université Grenoble Alpes, Directeur de thèse
Jean-Marc TACNET - Ingénieur des Ponts, Eaux et Forêts (ICPEF, Dr.), INRAE, Co-encadrant de thèse

Discover her work