How much longer can our metallic structures last? - Probabilistic fatigue analysis of welded metallic structures

Lock gates, penstocks, valves, turbines and storage tanks: many hydromechanical systems rely on metallic structures that are subjected to repeated loading year after year. These movements do not cause visible damage in the short term. Yet, after millions of loading cycles, steel gradually weakens. Cracks may initiate, propagate over time and eventually compromise the reliability and safety of the structure.

In the hydraulic, energy and civil engineering sectors, many infrastructures built after World War II are still in operation today. Asset managers are therefore faced with a critical question: How can the remaining useful life of these structures be assessed reliably in order to anticipate maintenance, inspection and replacement needs?

This is the challenge addressed by the PhD research of Kamal Harb, conducted within the Medelia Chair, with the objective of developing new analysis methods to improve the durability, safety and performance of metallic hydromechanical structures.

Fatigue: wear that cannot be seen

Bend and unbend a paper clip repeatedly. Eventually, it breaks, even though none of the individual movements would have been sufficient to cause failure. Hydromechanical steel structures are subjected to a similar phenomenon. Exposed to repeated loads over several decades, they may develop cracks that appear progressively, often in welded joints. Engineers refer to this process as fatigue degradation.

This degradation is not distributed uniformly throughout the structure. It tends to concentrate at structural discontinuities, particularly at welds. While essential for joining steel plates, welds also create geometric variations, residual stresses and localized material modifications. These areas are therefore common sites for crack initiation, with defects evolving from imperceptible to detectable and eventually critical.

An inspection can identify a crack once it exists. However, it cannot accurately predict when the next crack will appear, nor how quickly it will propagate.

Uncertainty at the heart of the analysis

Two structures that are identical on paper never age in exactly the same way. The mechanical properties of steel vary from one plate to another, no weld is perfectly identical to another, and loading conditions depend on operating regimes, vibrations and hydrodynamic effects. These parameters evolve over time and remain only partially known, particularly after several decades of service.

Traditional approaches often respond to this complexity with a single value: an estimated service life. Yet this figure gives an impression of certainty that does not fully reflect the actual behaviour of the structure.

Probabilistic fatigue analysis adopts a different perspective. Instead of defining a single end-of-life date, it evaluates multiple scenarios and produces risk indicators such as probability of failure, probability of survival, risk level and reliability index. The method explicitly incorporates uncertainties related to materials, loading conditions and the actual condition of the structure.

For asset managers, the difference is significant: rather than relying on a single number, they can assess risk levels and make informed decisions regarding inspection, strengthening, replacement or continued operation.

A scientific approachth

This PhD research develops a probabilistic fatigue assessment methodology for welded structures, with a particular focus on hydromechanical gate components. Its objective is to better account for the uncertainties that influence damage evolution and structural service life.

The approach is built on three complementary pillars:

  • Fracture mechanics: understanding how cracks initiate and propagate in welded joints.
  • Probabilistic modelling: integrating uncertainties related to materials, geometry, weld quality and loading conditions to represent the actual behaviour of structures more realistically.
  • Stochastic finite element methods: quantifying the effect of these uncertainties on structural behaviour and estimating remaining service life through probabilistic scenarios.

The methodology is validated through case studies based on real structures provided by the project's industrial partners. This field validation ensures that the developed tools address practical needs related to inspection, maintenance and decision support.

Expected impacts

For operators and asset managers :

  • More accurate assessment of the remaining service life of hydromechanical structures.
  • Decision-support tools for developing and comparing maintenance strategies.
  • Integration of uncertainties related to materials, loading conditions and operating environments.
  • Improved fatigue justification of ageing metallic structures.
  • Better prioritization of interventions on critical equipment.

For society:

  • Enhance the safety of water and energy infrastructure.
  • Extension of the service life of existing assets instead of replacing them prematurely.
  • Reduction of unnecessary replacement of material- and energy-intensive equipment.
  • More sustainable management of industrial infrastructure assets.
  • Preservation of the reliability of systems essential for energy production, navigation and water management.

Research grounded in real-world challenges

This project is conducted in close collaboration with several industrial and academic partners: Artelia / SPRETEC, la Compagnie Nationale du Rhône (CNR), EDFEDF Hydro, TEC21, as well as the laboratories 3SR and SIMaP.

These collaborations make it possible to confront scientific developments with real operational cases using field data, feedback from operators and practical engineering challenges. They ensure that the methodologies developed meet the concrete needs of infrastructure owners and operators in the areas of inspection, maintenance and decision support.

Making research accessible: Kamal Harb, PhD Candidate within the medelia chair

Kamal Harb was awarded first prize in the “My Thesis in 180 Seconds” competition during the HydroES 2025 conference. This distinction recognizes both the scientific quality of his research and his ability to make complex topics understandable to a broader audience.

“These structures were originally designed to last a few decades, yet many are still in service today. My work aims to assess their ability to continue operating safely by explicitly accounting for the uncertainties associated with ageing.”
– Kamal Harb, PhD Candidate, Medelia Chair

The Team Involved

PhD student : Kamal Harb

PhD Supervisors: : Rafael Estevez (Supervisor) Julien Baroth (Co-supervisor)

Industrial supervisors: : Arnaud Isaac and Vincent Michaud (SPRETEC, Groupe Artelia)

Conferences and scientific contributions

2023 :

29–30 nov.10th Fatigue Design Conference CETIM, Senlis, FranceParticipation

2024 :

28–29 maiFatigue Committee – 40th Spring Meeting – FIAP Jean Monnet, Paris, FranceParticipation
17–19 juinJFMS 2024 – 13th Reliability of Materials and Structures Conference – Rouen, FranceOral Presentation
19–20 juinCFBR Hydromechanical Gates Conference 2024 – Aix-les-Bains, FranceOral Presentation
14–15 oct.10th Business Hydro Meeting – Hydropower & Modernization – Alpexpo, Grenoble, FranceOral Presentation
2024I-MEP2 Doctoral School Day Grenoble, FrancePoster
2024Artelia Group / SPRETEC PhD Day – Grenoble, FranceOral Presentation

2025 :

janvierMECAMAT National Conference – Homogenization of the Mechanical Behaviour of Heterogeneous Materials – Aussois, FrancePoster
20–21 marsIIW Commission XIII – Sub-commission A: Fatigue Testing and Evaluation of Data for Design – Darmstadt, AllemagneOral Presentation
17–19 sep.HydroES Conference – “My Thesis in 180 Seconds” Competition – Grenoble, France1st Prize
19–20 nov.11th Fatigue Design International Conference – CETIM, Senlis, FranceOral Presentation

Contact

For any questions or collaboration opportunities:

NamePhD Candidate, Medelia Chair
E-mailharbkamal@outlook.com
Phone number+33 7 69 68 54 92
LinkedInwww.linkedin.com/in/kamal-harb
Web page:Coming soon
Share on facebook
Facebook
Share on twitter
Twitter
Share on linkedin
LinkedIn