MAPLES: Multi-fidelity and Probabilistic Lifetime Estimation for Slender Marine Structures

The oceans, with their relentless waves, strong currents, and floating platforms, pose significant challenges for slender marine structures like risers and dynamic power cables. One critical issue is Vortex-Induced Vibrations (VIV), which can affect their safety and longevity. VIV occurs when water flows around the structures, forming vortices that cause rhythmic vibrations, potentially leading to fatigue and damage over time. Many of these structures, designed and installed before 2004, are nearing the end of their service lives, posing the challenge: how can their operational lifespan be extended safely without full replacement? Traditional approaches separate the effects of waves and VIV, resulting in uncertainty in safety factors. Overly conservative factors lead to costly designs, while underestimation can cause failures. These methods also fail to account for the combined effects of waves, currents, and floating platform motions, which often act simultaneously in real-world conditions. One advancement is the use of time-domain VIV prediction models, which simulate how structures behave under real-world conditions, providing a clearer picture of their performance over time. However, current technologies still have limitations, leading to higher-than-expected probabilities of failure. The MAPLES project aims to tackle these challenges through probabilistic and multi-fidelity modeling. By combining data from both laboratory experiments and real-world measurements, MAPLES seeks to improve the accuracy of lifetime predictions, reducing uncertainty and enabling safer, more cost-effective designs. This approach will enhance the understanding of the underlying physics, optimize safety factors, and make operations more efficient, reducing the need for expensive over-design.

Slender marine structures, such as marine risers and dynamic power cables, are exposed to complex environmental loads due to waves, currents and floater motions. The response to these loads in terms of Vortex Induced Vibrations (VIV) is associated to large uncertainties, representing a safety risk and a major design consideration, adding notable costs to all stages of the system development. In addition, over 300 top-tensioned risers and 100 steel catenary risers installed worldwide prior to 2004 are close to the end of their designed service lives. Potential lifetime extension relies on the accurate prediction of the accumulated fatigue damage by also considering the changes in design conditions over time. MAPLES intends to use the monitored data, time-domain formulations, artificial intelligence and multi-fidelity modelling to realize accurate algorithms for decision support related to the inspection, maintenance and repair of marine risers and dynamic power cables for future applications. With a limited number of sensors available due to cost considerations, active use of monitored data for model update and uncertainty estimations is considered mandatory for obtaining robust and cost-effective installations. The results will also improve the present design practice by considering the simultaneously acting wave and VIV loads. The extended application of hybrid models that combine ML with classical modelling provides more accurate modelling of risers and slender structures in general with an increased understanding of the underlying physics. Incorporating uncertainty measurements in ML models for physical systems is of high value, and something that is underexplored to date. Methodologies for multi-fidelity training of hybrid models can maximize the use of limited data. The ambition is to improve lifetime prediction of slender marine structures by reducing uncertainty for cost-effective and safe designs as well as operation/decision support.