A Markovian model for maintenance of floating wind turbines considering degradation and random shocks

Résumé

Floating offshore wind turbines (FOWTs) are exposed to complex degradation 
mechanisms and uncertain external events that make operation and maintenance 
activities particularly challenging. This paper proposes a Markov decision process 
(MDP) environment for representing the evolution of degradation of some critical 
components and subsystems of FOWTs considering two possible mechanisms: a 
baseline degradation due to normal operation loads an ageing process and a random 
abrupt degradation caused by exogenous factors such as environmental conditions, 
operational activities or external interactions. The probabilities of these degradation 
mechanisms are defined by degradation kernels. In addition, a set of maintenance 
actions, i.e. replacements, imperfect maintenance or no intervention, are integrated in 
the MDP to consider the effect of maintenance on the component degradation state. In 
order to validate the consistency of the proposed framework, we conduct a numerical 
case study. In this study, we estimate the transition matrix from annual failure rates 
derived from the literature on FOWTs. We propose two maintenance policies: first, a 
purely corrective policy to compare the simulated empirical failure probabilities with 
the theoretical annual failure probabilities reported in the literature; second, a random 
maintenance policy to determine the impact of exogenous stress factors that lead to 
random shocks. In all the cases, the model has provided good results, reproducing 
coherent degradation patterns and differentiated impact of random shocks depending 
on the type of component.  The proposed approach is intended to serve as a bridge 
between degradation modelling and intelligent maintenance policy generation.