https://sam.ensam.eu:443 The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. Sun, 10 May 2026 09:11:32 GMT 2026-05-10T09:11:32Z http://hdl.handle.net/10985/8688 Numerical study of a Flexible Sail Plan submitted to pitching : Hysteresis phenomenon and effect of rig Adjustments AUGIER, Benoit; HAUVILLE, Frederic; BOT, Patrick; AUBIN, Nicolas; DURAND, Mathieu A numerical investigation of the dynamic Fluid Structure Interaction (FSI) of a yacht sail plan submitted to harmonic pitching is presented to analyse the system's dynamic behaviour and the effects of motion simplifications and rigging adjustments on aerodynamic forces. It is shown that the dynamic behaviour of a sail plan subject to yacht motion clearly deviates from the quasi-steady theory. The aerodynamic forces presented as a function of the instantaneous apparent wind angle show hysteresis loops. It is shown that the hysteresis phenomenon dissipates some energy and that the dissipated energy increases strongly with the pitching reduced frequency and amplitude. The effect of reducing the real pitching motion to a simpler surge motion is investigated. Results show significant discrepancies with underestimated aerodynamic forces and no more hysteresis when a surge motion is considered. However, the superposition assumption consisting in a decomposition of the surge into two translations normal and collinear to the apparent wind is verified. Then, simulations with different dock tunes and backstay loads highlight the importance of rig adjustments on the aerodynamic forces and the dynamic behaviour of a sail plan. The energy dissipated by the hysteresis is higher for looser shrouds and a tighter backstay.  Yacht sails dynamic fluid structure interaction is simulated in harmonic pitching  Aerodynamic forces show hysteresis associated to energy dissipation  Dissipated energy increases with pitching frequency and amplitude  Hysteresis is cancelled and forces underestimated when motion is reduced to surge  Looser shrouds and tighter backstay increase dissipated energy Wed, 01 Jan 2014 00:00:00 GMT http://hdl.handle.net/10985/8688 2014-01-01T00:00:00Z AUGIER, Benoit HAUVILLE, Frederic BOT, Patrick AUBIN, Nicolas DURAND, Mathieu A numerical investigation of the dynamic Fluid Structure Interaction (FSI) of a yacht sail plan submitted to harmonic pitching is presented to analyse the system's dynamic behaviour and the effects of motion simplifications and rigging adjustments on aerodynamic forces. It is shown that the dynamic behaviour of a sail plan subject to yacht motion clearly deviates from the quasi-steady theory. The aerodynamic forces presented as a function of the instantaneous apparent wind angle show hysteresis loops. It is shown that the hysteresis phenomenon dissipates some energy and that the dissipated energy increases strongly with the pitching reduced frequency and amplitude. The effect of reducing the real pitching motion to a simpler surge motion is investigated. Results show significant discrepancies with underestimated aerodynamic forces and no more hysteresis when a surge motion is considered. However, the superposition assumption consisting in a decomposition of the surge into two translations normal and collinear to the apparent wind is verified. Then, simulations with different dock tunes and backstay loads highlight the importance of rig adjustments on the aerodynamic forces and the dynamic behaviour of a sail plan. The energy dissipated by the hysteresis is higher for looser shrouds and a tighter backstay. http://hdl.handle.net/10985/14895 Experimental validation of unsteady models for wind / sails / rigging fluid structure interaction AUGIER, Benoit; BOT, Patrick; HAUVILLE, Frederic; DURAND, Mathieu The aim of this paper is to present the work of experimental validation elements of the aero elastic and unsteady model ARAVANTI. Numerical and Experimental results comparison is made on the rigging and sails of a J80 sail boat. Yacht modelling demands to consider unsteady phenomena resulting from the sea state, variations of wind speed and direction, yacht motion or trimming by the crew. A dedicated instrumentation is developed to measure the loads in shrouds and tension points of the sail, the apparent wind, the yacht motion, the sails flying shape and the navigation data. A special effort is made on sensors calibration, physical measurement comprehension and data synchronisation. Comparison with numerical results shows that the loads and flying shapes are well predicted by the model. Fri, 01 Jan 2010 00:00:00 GMT http://hdl.handle.net/10985/14895 2010-01-01T00:00:00Z AUGIER, Benoit BOT, Patrick HAUVILLE, Frederic DURAND, Mathieu The aim of this paper is to present the work of experimental validation elements of the aero elastic and unsteady model ARAVANTI. Numerical and Experimental results comparison is made on the rigging and sails of a J80 sail boat. Yacht modelling demands to consider unsteady phenomena resulting from the sea state, variations of wind speed and direction, yacht motion or trimming by the crew. A dedicated instrumentation is developed to measure the loads in shrouds and tension points of the sail, the apparent wind, the yacht motion, the sails flying shape and the navigation data. A special effort is made on sensors calibration, physical measurement comprehension and data synchronisation. Comparison with numerical results shows that the loads and flying shapes are well predicted by the model. http://hdl.handle.net/10985/15097 Experimental and numerical trimming optimizations for a mainsail in upwind conditions SACHER, Matthieu; HAUVILLE, Frederic; DUVIGNEAU, Régis; LE MAITRE, Olivier; AUBIN, Nicolas; DURAND, Mathieu This paper investigates the use of meta-models for optimizing sails trimming. A Gaussian process is used to robustly approximate the dependence of the performance with the trimming parameters to be optimized. The Gaussian process construction uses a limited number of performance observations at carefully selected trimming points, potentially enabling the optimization of complex sail systems with multiple trimming parameters. We test the optimization procedure on the (two parameters) trimming of a scaled IMOCA mainsail in upwind conditions. To assess the robustness of the Gaussian process approach, in particular its sensitivity to error and noise in the performance estimation, we contrast the direct optimization of the physical system with the optimization of its numerical model. For the physical system, the optimization procedure was fed with wind tunnel measurements, while the numerical modeling relied on a fully non-linear Fluid-Structure Interaction solver. The results show a correct agreement of the optimized trimming parameters for the physical and numerical models, despite the inherent errors in the numerical model and the measurement uncertainties. In addition, the number of performance estimations was found to be affordable and comparable in the two cases, demonstrating the effectiveness of the approach. Fri, 01 Jan 2016 00:00:00 GMT http://hdl.handle.net/10985/15097 2016-01-01T00:00:00Z SACHER, Matthieu HAUVILLE, Frederic DUVIGNEAU, Régis LE MAITRE, Olivier AUBIN, Nicolas DURAND, Mathieu This paper investigates the use of meta-models for optimizing sails trimming. A Gaussian process is used to robustly approximate the dependence of the performance with the trimming parameters to be optimized. The Gaussian process construction uses a limited number of performance observations at carefully selected trimming points, potentially enabling the optimization of complex sail systems with multiple trimming parameters. We test the optimization procedure on the (two parameters) trimming of a scaled IMOCA mainsail in upwind conditions. To assess the robustness of the Gaussian process approach, in particular its sensitivity to error and noise in the performance estimation, we contrast the direct optimization of the physical system with the optimization of its numerical model. For the physical system, the optimization procedure was fed with wind tunnel measurements, while the numerical modeling relied on a fully non-linear Fluid-Structure Interaction solver. The results show a correct agreement of the optimized trimming parameters for the physical and numerical models, despite the inherent errors in the numerical model and the measurement uncertainties. In addition, the number of performance estimations was found to be affordable and comparable in the two cases, demonstrating the effectiveness of the approach. http://hdl.handle.net/10985/15076 Efficient optimization procedure in non-linear fluid-structure interaction problem: Application to mainsail trimming in upwind conditions SACHER, Matthieu; HAUVILLE, Frederic; DUVIGNEAU, Régis; LE MAÎTRE, Olivier; AUBIN, Nicolas; DURAND, Mathieu This paper investigates the use of Gaussian processes to solve sail trimming optimization problems. The Gaussian process, used to model the dependence of the performance with the trimming parameters, is constructed from a limited number of performance estimations at carefully selected trimming points, potentially enabling the optimization of complex sail systems with multiple trimming parameters. The proposed approach is tested on a two-parameter trimming for a scaled IMOCA mainsail in upwind sailing conditions. We focus on the robustness of the proposed approach and study especially the sensitivity of the results to noise and model error in the point estimations of the performance. In particular, we contrast the optimization performed on a real physical model set in a wind tunnel with a fully non-linear numerical fluidstructure interaction model of the same experiments. For this problem with a limited number of trimming parameters, the numerical optimization was affordable and found to require a comparable amount of performance estimation as for the experimental case. The results reveal a satisfactory agreement for the numerical and experimental optimal trimming parameters, considering the inherent sources of errors and uncertainties in both numerical and experimental approaches. Sensitivity analyses have been eventually performed in the numerical optimization problem to determine the dominant source of uncertainties and characterize the robustness of the optima. Sun, 01 Jan 2017 00:00:00 GMT http://hdl.handle.net/10985/15076 2017-01-01T00:00:00Z SACHER, Matthieu HAUVILLE, Frederic DUVIGNEAU, Régis LE MAÎTRE, Olivier AUBIN, Nicolas DURAND, Mathieu This paper investigates the use of Gaussian processes to solve sail trimming optimization problems. The Gaussian process, used to model the dependence of the performance with the trimming parameters, is constructed from a limited number of performance estimations at carefully selected trimming points, potentially enabling the optimization of complex sail systems with multiple trimming parameters. The proposed approach is tested on a two-parameter trimming for a scaled IMOCA mainsail in upwind sailing conditions. We focus on the robustness of the proposed approach and study especially the sensitivity of the results to noise and model error in the point estimations of the performance. In particular, we contrast the optimization performed on a real physical model set in a wind tunnel with a fully non-linear numerical fluidstructure interaction model of the same experiments. For this problem with a limited number of trimming parameters, the numerical optimization was affordable and found to require a comparable amount of performance estimation as for the experimental case. The results reveal a satisfactory agreement for the numerical and experimental optimal trimming parameters, considering the inherent sources of errors and uncertainties in both numerical and experimental approaches. Sensitivity analyses have been eventually performed in the numerical optimization problem to determine the dominant source of uncertainties and characterize the robustness of the optima. http://hdl.handle.net/10985/14918 Numerical study of a flexible sail plan : effect of pitching decomposition and adjustments AUGIER, Benoit; DEPARDAY, Julien; DURAND, Mathieu; BOT, Patrick; HAUVILLE, Frederic A numerical investigation of the dynamic Fluid Structure Interaction (FSI) of a yacht sail plan submitted to harmonic pitching is presented to analyse the effects of motion simplifications and rigging adjustments on aerodynamic forces. It is shown that the dynamic behaviour of a sail plan subject to yacht motion clearly deviates from the quasi-steady theory. The aerodynamic forces presented as a function of the instantaneous apparent wind angle show hysteresis loops. These hysteresis phenomena do not result from a simple phase shift between forces and motion. Plotting the hysteresis loops in the appropriate coordinate system enables the associated energy to be determined. This amount of exchanged energy is shown to increase almost linearly with the pitching reduced frequency and to increase almost quadratically with the pitching amplitude in the investigated ranges. The effect of reducing the real pitching motion to a simpler surge motion is investigated. Results show significant discrepancies on the aerodynamic forces amplitude and the hysteresis phenomenon between pitching and surge motion. However, the superposition assumption consisting in a decomposition of the surge into two translations normal and collinear to the apparent wind is verified. Then, simulations with different dock tunes and backstay loads highlight the importance of rig adjustments on the aerodynamic forces and the dynamic behaviour of a sail plan. Tue, 01 Jan 2013 00:00:00 GMT http://hdl.handle.net/10985/14918 2013-01-01T00:00:00Z AUGIER, Benoit DEPARDAY, Julien DURAND, Mathieu BOT, Patrick HAUVILLE, Frederic A numerical investigation of the dynamic Fluid Structure Interaction (FSI) of a yacht sail plan submitted to harmonic pitching is presented to analyse the effects of motion simplifications and rigging adjustments on aerodynamic forces. It is shown that the dynamic behaviour of a sail plan subject to yacht motion clearly deviates from the quasi-steady theory. The aerodynamic forces presented as a function of the instantaneous apparent wind angle show hysteresis loops. These hysteresis phenomena do not result from a simple phase shift between forces and motion. Plotting the hysteresis loops in the appropriate coordinate system enables the associated energy to be determined. This amount of exchanged energy is shown to increase almost linearly with the pitching reduced frequency and to increase almost quadratically with the pitching amplitude in the investigated ranges. The effect of reducing the real pitching motion to a simpler surge motion is investigated. Results show significant discrepancies on the aerodynamic forces amplitude and the hysteresis phenomenon between pitching and surge motion. However, the superposition assumption consisting in a decomposition of the surge into two translations normal and collinear to the apparent wind is verified. Then, simulations with different dock tunes and backstay loads highlight the importance of rig adjustments on the aerodynamic forces and the dynamic behaviour of a sail plan. http://hdl.handle.net/10985/15136 Sail trimming FSI simulation - Comparison of viscous and inviscid flow models to optimise upwind sails trim SACHER, Matthieu; HAUVILLE, Frederic; BOT, Patrick; DURAND, Mathieu A numerical comparison between two FSI models, based on inviscid and viscous flow solvers, is presented in this paper. The differences between aerodynamic coefficients, sail flying shape and pressures computed by both FSI tools are investigated for medium wind conditions. These differences are evaluated for different values of the main sheet length. The study has shown very close results when the main sheet is not over trimmed for medium true wind speed, but discrepancies increase when flow separation becomes significant. Then, an optimisation procedure based on inviscid FSI is performed to optimise the main sheet and car trims, in order to maximise an objective function based on the driving and side forces, in a case of low true wind speed. Limitations of the inviscid flow hypothesis are highlighted and the difficulties to use inviscid FSI models in an optimisation procedure, for a case of low true wind speed, are shown. Thu, 01 Jan 2015 00:00:00 GMT http://hdl.handle.net/10985/15136 2015-01-01T00:00:00Z SACHER, Matthieu HAUVILLE, Frederic BOT, Patrick DURAND, Mathieu A numerical comparison between two FSI models, based on inviscid and viscous flow solvers, is presented in this paper. The differences between aerodynamic coefficients, sail flying shape and pressures computed by both FSI tools are investigated for medium wind conditions. These differences are evaluated for different values of the main sheet length. The study has shown very close results when the main sheet is not over trimmed for medium true wind speed, but discrepancies increase when flow separation becomes significant. Then, an optimisation procedure based on inviscid FSI is performed to optimise the main sheet and car trims, in order to maximise an objective function based on the driving and side forces, in a case of low true wind speed. Limitations of the inviscid flow hypothesis are highlighted and the difficulties to use inviscid FSI models in an optimisation procedure, for a case of low true wind speed, are shown. http://hdl.handle.net/10985/14920 Flexible hydrofoil optimization for the 35th america's cup constrained ego method SACHER, Matthieu; DURAND, Mathieu; BERRINI, Elisa; HAUVILLE, Frederic; DUVIGNEAU, Régis; LE MAITRE, O This paper investigates the use of constrained surrogate models to solve the multi-design optimization problem of a flexible hydrofoil. The surrogate-based optimization (EGO) substitutes the complex objective function of the problem by an easily evaluable model, constructed from a limited number of computations at carefully selected design points. Associated with ad-hoc statistical strategies to propose optimum candidates within the estimated feasible domain, EGO enables the resolution of complex optimization problems. In this work, we rely on Gaussian processes (GP) to model the objective function and adopt a probabilistic classification method to treat non-explicit inequality constraints and non-explicit representation of the feasible domain. This procedure is applied to the design of the shape and the elastic characteristics of a hydrofoil equipped with deformable elements providing flexibility to the trailing edge. The optimization concerns the minimization of the hydrofoil drag while ensuring a non-cavitating flow, at selected sailing conditions (boat speed and lifting force). The drag value and cavitation criterion are determined by solving a two-dimensional nonlinear fluid-structure interaction problem, based on a static vortex lattice method with viscous boundary layer equations, for the flow, and a nonlinear elasticity solver for the deformations of the elastic components of the foil. We compare the optimized flexible hydrofoil with a rigid foil geometrically optimized for the same sailing conditions. This comparison highlights the hydrodynamical advantages brought by the flexibility: a reduction of the drag over a large range of boat speeds, less susceptibility to cavitation and a smaller angle of attack tuning range. Sun, 01 Jan 2017 00:00:00 GMT http://hdl.handle.net/10985/14920 2017-01-01T00:00:00Z SACHER, Matthieu DURAND, Mathieu BERRINI, Elisa HAUVILLE, Frederic DUVIGNEAU, Régis LE MAITRE, O This paper investigates the use of constrained surrogate models to solve the multi-design optimization problem of a flexible hydrofoil. The surrogate-based optimization (EGO) substitutes the complex objective function of the problem by an easily evaluable model, constructed from a limited number of computations at carefully selected design points. Associated with ad-hoc statistical strategies to propose optimum candidates within the estimated feasible domain, EGO enables the resolution of complex optimization problems. In this work, we rely on Gaussian processes (GP) to model the objective function and adopt a probabilistic classification method to treat non-explicit inequality constraints and non-explicit representation of the feasible domain. This procedure is applied to the design of the shape and the elastic characteristics of a hydrofoil equipped with deformable elements providing flexibility to the trailing edge. The optimization concerns the minimization of the hydrofoil drag while ensuring a non-cavitating flow, at selected sailing conditions (boat speed and lifting force). The drag value and cavitation criterion are determined by solving a two-dimensional nonlinear fluid-structure interaction problem, based on a static vortex lattice method with viscous boundary layer equations, for the flow, and a nonlinear elasticity solver for the deformations of the elastic components of the foil. We compare the optimized flexible hydrofoil with a rigid foil geometrically optimized for the same sailing conditions. This comparison highlights the hydrodynamical advantages brought by the flexibility: a reduction of the drag over a large range of boat speeds, less susceptibility to cavitation and a smaller angle of attack tuning range. http://hdl.handle.net/10985/14896 Unsteady numerical simulations of downwind sails DURAND, Mathieu; HAUVILLE, Frederic; BOT, Patrick; AUGIER, Benoit; ROUX, Yann; LEROYER, Alban; VISONNEAU, Michel Modelling the wind, sail and rig interactions on a sailing yacht is a complex subject, because the quality of simulation depends on the accuracy of both structural and fluid simulations which strongly interact. Moreover, the sails are submitted to highly unsteady sollicitations due to waves, wind variations, course changes or trimming for example, but sometimes also due to the unsteadiness of the flow itself (vortex shedding,…). The problem for downwind sails is even more complex because the flow is often detached from the sails, and the sails are subject to large shape changes. A specific dynamic coupling has been developed between a RANSE code from Ecole Centrale de Nantes for the aerodynamics (ISIS-CFD) and a FEM code from K-Epsilon for the structure (ARA) specialized to simulate the aeroelastic problem of yacht sails and rig. In this paper, the particular issues of coupling, remeshing and transfer of forces from one code to the other are detailed. An experimental comparison is made on a well controlled test case with an original experiment developed by IRENav Ecole Navale, consisting in a square of spinnaker fabric mounted on two carbon battens which are moved in a forced oscillation. The good agreement of numerical results with experimentals results permits to be confident to go ahead and investigate a full example of application on a racing yacht spinnaker. Fri, 01 Jan 2010 00:00:00 GMT http://hdl.handle.net/10985/14896 2010-01-01T00:00:00Z DURAND, Mathieu HAUVILLE, Frederic BOT, Patrick AUGIER, Benoit ROUX, Yann LEROYER, Alban VISONNEAU, Michel Modelling the wind, sail and rig interactions on a sailing yacht is a complex subject, because the quality of simulation depends on the accuracy of both structural and fluid simulations which strongly interact. Moreover, the sails are submitted to highly unsteady sollicitations due to waves, wind variations, course changes or trimming for example, but sometimes also due to the unsteadiness of the flow itself (vortex shedding,…). The problem for downwind sails is even more complex because the flow is often detached from the sails, and the sails are subject to large shape changes. A specific dynamic coupling has been developed between a RANSE code from Ecole Centrale de Nantes for the aerodynamics (ISIS-CFD) and a FEM code from K-Epsilon for the structure (ARA) specialized to simulate the aeroelastic problem of yacht sails and rig. In this paper, the particular issues of coupling, remeshing and transfer of forces from one code to the other are detailed. An experimental comparison is made on a well controlled test case with an original experiment developed by IRENav Ecole Navale, consisting in a square of spinnaker fabric mounted on two carbon battens which are moved in a forced oscillation. The good agreement of numerical results with experimentals results permits to be confident to go ahead and investigate a full example of application on a racing yacht spinnaker. http://hdl.handle.net/10985/14917 FSI Investigation on Stability of Downwind Sails with an Automatic Dynamic Trimming DURAND, Mathieu; LOTHODE, Corentin; HAUVILLE, Frederic; LEROYER, Alban; VISONNEAU, Michel; FLOCH, Ronan; GUILLAUME, Laurent Gennakers are lightweight and flexible sails, used for downwind sailing configurations. Qualities sought for this kind of sail are propulsive force and dynamic stability. To simulate accurately the flow around such a sail, several problems need to be solved. Firstly, the structural code has to take into account cloth behavior, orientation and reinforcements. Flexibility is obtained by modeling wrinkles. Secondly, the fluid code needs to reproduce the atmospheric boundary layer as an input boundary condition, and be able to simulate separation. Thirdly, fluid-structure interaction (FSI) is strong due to the lightness and the flexibility of the structure. The added mass is three orders of magnitude greater than the mass of the sail, and large structural displacement occurs, which makes the coupling between the two solvers difficult to achieve. Finally, the problem is unsteady, and dynamic trimming is important to the simulation of spinnakers [4]. The main objective is to use numerical simulations to model spinnakers, in order to predict both propulsive force and sail dynamic stability. Recent developments [2] are used to solve these problems, using a finite element program dedicated to sails and rig simulations coupled with a RANSE solver. The FSI coupling is done through a quasi-monolithic method. An ALE formulation is used, hence the fluid mesh follows the structural deformation while keeping the same topology. The fluid mesh deformation is carried out with a fast, robust and parallelized method based on the propagation of the deformation state of the sail boundary fluid faces [3]. Tests are realized on a complete production chain: a sail designer from Incidences has designed two different shapes of an IMOCA60 spinnaker with the SailPack software. An automatic procedure was developed to transfer data from Sailpack to a structure input file taking into account the orientation of sailcloth and reinforcements. The same automatic procedure is used for both spinnakers, in order to compare dynamic stability and propulsion forces. Then a new method is developed to quantify the stability of a downwind sail. Tue, 01 Jan 2013 00:00:00 GMT http://hdl.handle.net/10985/14917 2013-01-01T00:00:00Z DURAND, Mathieu LOTHODE, Corentin HAUVILLE, Frederic LEROYER, Alban VISONNEAU, Michel FLOCH, Ronan GUILLAUME, Laurent Gennakers are lightweight and flexible sails, used for downwind sailing configurations. Qualities sought for this kind of sail are propulsive force and dynamic stability. To simulate accurately the flow around such a sail, several problems need to be solved. Firstly, the structural code has to take into account cloth behavior, orientation and reinforcements. Flexibility is obtained by modeling wrinkles. Secondly, the fluid code needs to reproduce the atmospheric boundary layer as an input boundary condition, and be able to simulate separation. Thirdly, fluid-structure interaction (FSI) is strong due to the lightness and the flexibility of the structure. The added mass is three orders of magnitude greater than the mass of the sail, and large structural displacement occurs, which makes the coupling between the two solvers difficult to achieve. Finally, the problem is unsteady, and dynamic trimming is important to the simulation of spinnakers [4]. The main objective is to use numerical simulations to model spinnakers, in order to predict both propulsive force and sail dynamic stability. Recent developments [2] are used to solve these problems, using a finite element program dedicated to sails and rig simulations coupled with a RANSE solver. The FSI coupling is done through a quasi-monolithic method. An ALE formulation is used, hence the fluid mesh follows the structural deformation while keeping the same topology. The fluid mesh deformation is carried out with a fast, robust and parallelized method based on the propagation of the deformation state of the sail boundary fluid faces [3]. Tests are realized on a complete production chain: a sail designer from Incidences has designed two different shapes of an IMOCA60 spinnaker with the SailPack software. An automatic procedure was developed to transfer data from Sailpack to a structure input file taking into account the orientation of sailcloth and reinforcements. The same automatic procedure is used for both spinnakers, in order to compare dynamic stability and propulsion forces. Then a new method is developed to quantify the stability of a downwind sail. http://hdl.handle.net/10985/15139 A classification approach to efficient global optimization in presence of non-computable domains SACHER, Matthieu; DUVIGNEAU, Régis; LE MAÎTRE, Olivier; DURAND, Mathieu; BERRINI, Elisa; HAUVILLE, Frederic; ASTOLFI, Jacques Andre Gaussian-Process based optimization methods have become very popular in recent years for the global optimization of complex systems with high computational costs. These methods rely on the sequential construction of a statistical surrogate model, using a training set of computed objective function values, which is refined according to a prescribed infilling strategy. However, this sequential optimization procedure can stop prematurely if the objective function cannot be computed at a proposed point. Such a situation can occur when the search space encompasses design points corresponding to an unphysical configuration, an ill-posed problem, or a non-computable problem due to the limitation of numerical solvers. To avoid such a premature stop in the optimization procedure, we propose to use a classification model to learn non-computable areas and to adapt the infilling strategy accordingly. Specifically, the proposed method splits the training set into two subsets composed of computable and non-computable points. A surrogate model for the objective function is built using the training set of computable points, only, whereas a probabilistic classification model is built using the union of the computable and non-computable training sets. The classifier is then incorporated in the surrogate-based optimization procedure to avoid proposing new points in the non-computable domain while improving the classification uncertainty if needed. The method has the advantage to automatically adapt both the surrogate of the objective function and the classifier during the iterative optimization process. Therefore, non-computable areas do not need to be a priori known. The proposed method is applied to several analytical problems presenting different types of difficulty, and to the optimization of a fully nonlinear fluid-structure interaction system. The latter problem concerns the drag minimization of a flexible hydrofoil with cavitation constraints. The efficiency of the proposed method compared favorably to a reference evolutionary algorithm, except for situations where the feasible domain is a small portion of the design space. Mon, 01 Jan 2018 00:00:00 GMT http://hdl.handle.net/10985/15139 2018-01-01T00:00:00Z SACHER, Matthieu DUVIGNEAU, Régis LE MAÎTRE, Olivier DURAND, Mathieu BERRINI, Elisa HAUVILLE, Frederic ASTOLFI, Jacques Andre Gaussian-Process based optimization methods have become very popular in recent years for the global optimization of complex systems with high computational costs. These methods rely on the sequential construction of a statistical surrogate model, using a training set of computed objective function values, which is refined according to a prescribed infilling strategy. However, this sequential optimization procedure can stop prematurely if the objective function cannot be computed at a proposed point. Such a situation can occur when the search space encompasses design points corresponding to an unphysical configuration, an ill-posed problem, or a non-computable problem due to the limitation of numerical solvers. To avoid such a premature stop in the optimization procedure, we propose to use a classification model to learn non-computable areas and to adapt the infilling strategy accordingly. Specifically, the proposed method splits the training set into two subsets composed of computable and non-computable points. A surrogate model for the objective function is built using the training set of computable points, only, whereas a probabilistic classification model is built using the union of the computable and non-computable training sets. The classifier is then incorporated in the surrogate-based optimization procedure to avoid proposing new points in the non-computable domain while improving the classification uncertainty if needed. The method has the advantage to automatically adapt both the surrogate of the objective function and the classifier during the iterative optimization process. Therefore, non-computable areas do not need to be a priori known. The proposed method is applied to several analytical problems presenting different types of difficulty, and to the optimization of a fully nonlinear fluid-structure interaction system. The latter problem concerns the drag minimization of a flexible hydrofoil with cavitation constraints. The efficiency of the proposed method compared favorably to a reference evolutionary algorithm, except for situations where the feasible domain is a small portion of the design space.