Polyurethane materials are widely used in marine environments, mainly for military applications such as acoustic coatings of submarine hulls.

These highly-combustible materials are nevertheless a potential fire hazard. The purpose of this thesis is to limit this risk by determining a fireproofing solution of these materials compatible with the marine environment.

The mechanical, acoustic and physico-chemical properties of the elastomer must also be maintained. A promising flame-retardant system has therefore been developed through this thesis work, with no significant alteration to the other functional properties required for the application.

Fire tests conducted on materials formulated with fire retardant fillers have shown a significant improvement in the fire reaction of polyurethane. The weathering study of fireproofed materials raises hopes for a very stable system in marine environments, despite tests at temperatures around 70°C showing a possible alteration in the load/matrix interface.

This thesis is part of the R&D initiative conducted since 2013 to better understand the thermohydraulic behaviour of single-phase heat exchangers, condensers and evaporators and suggest optimised designs for them.

Two theses have already been submitted on these themes and a model coupling a CFD description of the tube exterior (equivalent environment type) with an analytical approach to the tube interior/exterior thermal coupling has been developed.

The description of the equivalent environment that forms the basis for the model involves laws of pressure drop and heat transfer calculated also for standard geometries (tubes-shell type). However, additive manufacturing techniques now make it possible to consider complex geometries where these laws are inappropriate.

The purpose of this thesis is thus to determine laws that can account for these new geometries.

"Wave by wave" swell predictions in real time can anticipate ship motion and thus operations can be envisaged in rougher seas with improved safety. This work is also targeting realistic physical modelling of the swell (incorporating certain non-linearities), still in real time.

Water height measurements taken around a ship (e.g. using a Lidar scanner) are used as the basis for preparing an initial state of the swell. Once propagated over time and space, this can predict the future location of the ship and feed a ship response calculation.

The work presented covers the development of an effective wave model to model certain non-linearities with few calculation costs and a method of identifying an optimum initial swell state from truncated data.

The impact of taking hydrodynamic non-linearities into account in the quality of the prediction is studied by varying the characteristic camber of the sea state and referring to modelling that incorporates all the non-linearities (HOS). The approach is lastly qualified using actual data from an experimental campaign in a wave basin.

The submarine hull assembly process impacts their mechanical properties. It is essential to understand and fully grasp this impact to ensure the performances.

This work aims therefore to assess the impact of the coated electrode multi-pass welding process on the mechanical behaviour of welded joints on high-yield strength steel.

The mechanical behaviour of the weld metal is linked to its microstructure which itself stems from the welding process. Two factors appear decisive in this process: the chemical composition of the filler product and the thermal log from the welding process.

After initial microstructural analyses of the weld metal, the study focused on the effect of the thermal log on the microstructure.

An experimental thermal test campaign (Gleeble tests), coupled with the numerical simulation of the welding, was conducted to help in identifying an optimum microstructure, ultimately to optimise the expected mechanical properties. A parametric study of welding conditions will be conducted later on and the accumulated effects of welding settings and the chemistry of the filler product will be investigated.

Despite the constantly-improving calculation power of computers, calculating and analysing the response of submerged structures remains lengthy and costly.

In a finite element approach, the size and complexity of our structures result, for example, in a need to process several million degrees of freedom. The design of our structures, including justifying their resistance to the diversity of loadings applied to them, requires huge numbers of simulations.

Their number is currently severely limited by the calculation times. Model reduction techniques reduce dramatically the calculation times without any real loss of precision. Parametric calculations are now becoming accessible.

The design basis for military vessels requires an analysis of the behaviour of structures in far-off underwater explosions.

During an underwater explosion, a shock wave with steep wavefront (rapid dynamic excitation) is released initially. The burned gases from the detonation then form a bubble with very high internal pressure. The bubble then oscillates between expansions and contractions, triggering a slow fluid movement that applies stress to the vessel in the low frequencies for longer.

"Our aim is to design, implement (under the HPC calculation) and validate a numerical simulation methodology for the fluid-structure interaction problem taken both phenomena into account. To achieve this, we are developing an accelerated boundary element method (BEM) to simulate effectively rapid 3D transient wave propagation problems in unbounded environment, which offers very favourable complexity: O(1) with respect to the temporal discretisation and O(N log N) with respect to the spatial discretisation.

Lastly, we introduce efficient FEM/BEM coupling strategies for the fluid-structure interaction phase of the shock wave (linear acoustics) and the gas bubble (non-compressible flows). The overall procedure, validated on academic problems, provides highly promising results in realistic industrial cases."

The acoustic behaviour of a submarine pressure hull is an important aspect of their operational performance. Fully understanding this behaviour from the first phases of a project nevertheless depends on access to efficient numerical simulation tools.

Naval Group has recently developed the ORCAA code for this purpose as part of the collaboration with LVA of INSA Lyon. Based on a sub-structuring approach, it can calculate the vibro-acoustic behaviour of submerged pressure hulls. The three important acoustic performances of an underwater vehicle (i.e. radiated acoustic noise, self-noise and target echo strength) can therefore be understood over a wide frequency range.

The work undertaken aims to develop a so-called "subtractive" modelling strategy, so that the vibro-acoustic behaviour of a system can be modelled by removing an element from a global system that is easier to model.

The formalism of the sub-structuring method must be extended for this. We shall present the method together with validation cases.

A complex system is a collection of sub-systems linked by structural constraints and interacting to provide the overall performance of the system.

Complex system optimisation is a topic explored widely for functions with continuous and scalar values, especially through decomposition and coordination approaches. These approaches alternate between distributed optimisation of sub-systems, made independent by fixing a state of their interaction, and updating the state of their interaction.

We study the case of functions with vector values, without a unique optimum, but with a set of non-dominated solutions. No optimum state of interaction can therefore exist between the sub-systems and traditional approaches cannot therefore be transposed.

Nevertheless, we show that the notion of decomposable optimisation is generalised to the multi-objective case, with the difficulty that combinations of non-dominated solutions for the sub-systems can be dominated for the global system. We suggest several algorithms to filter them effectively. For functions with Boolean variables, a solution to the original optimisation problem could be to enumerate all the possible states of interaction between sub-systems and resolve the associated decomposable problems.

In addition, we define, for a fairly wide range of complex system optimisation problems, notions of lower and upper bounds of all non-dominated solutions, bounds that can be obtained by decomposition.

Paint films used to protect naval structures from corrosion must be maintained appropriately for their condition.

It is especially important to know the state of the paint film for nuclear applications or in areas difficult to access, where corrosion can be critical (ballast tanks, bilges). Nowadays this is based on manual observations, adherence characterisation and electro-chemical analyses, but their performance cannot however be assessed up to the next scheduled maintenance.

The aim of this doctoral thesis work, conducted in conjunction with iRDL, is to prove the feasibility of monitoring the weathering of corrosion-proof paints. The principle is to drown Quantum Resistive Sensors (QRS) in the paint film to characterise its state. QRS prototypes adapted to the main corrosion-proof coatings used at Naval Group have been manufactured and characterised and will be installed inside these protective coatings.

The prototypes will be used to monitor the general state of the coating in real time during different accelerated weathering tests.

The QRS could thus be used for predictive maintenance purposes by warning of the onset of metal substrate corrosion. They could help to setting up as-required maintenance, thereby reducing servicing costs.