https://sam.ensam.eu:443 The DSpace digital repository system captures, stores, indexes, preserves, and distributes digital research material. Fri, 15 May 2026 22:47:31 GMT 2026-05-15T22:47:31Z http://hdl.handle.net/10985/8542 A model comparison to predict heat transfer during spot GTA welding LE MASSON, Philippe; CARIN, Muriel; DAL, Morgan The present work deals with the estimation of the time evolution of the weld fusion boundary. This moving boundary is the result of a spot GTA welding process on a 316L stainless steel disk. The estimation is based on the iterative regularization method. Indeed, the three problems: direct, in variation and adjoint, classically associated with this method, are solved by the finite element method in a two-dimensional axisymmetric domain. The originality of this work is to treat an experimental estimation of a front motion using a model with a geometry including only the solid phase. In this model, the evolution of this solid domain during the fusion is set with the ALE moving mesh method (Arbitrary Lagrangian Eulerian). The numerical developments are realized with the commercial code Comsol Multiphysics® coupled with the software Matlab®. The estimation method has been validated in a previous work using theoretical data ([1]). The experimental data, used here for this identification are, temperatures measured by thermocouples in the solid phase, the temporal evolution of the melt pool boundary observed at the surface by a fast camera and the maximal dimensions of the melted zone measured on macrographs. These experimental data are also compared with numerical results obtained from a heat and fluid flow model taking into account surface tension effects, Lorentz forces and the deformation of the melt pool surface under arc pressure. Wed, 01 Jan 2014 00:00:00 GMT http://hdl.handle.net/10985/8542 2014-01-01T00:00:00Z LE MASSON, Philippe CARIN, Muriel DAL, Morgan The present work deals with the estimation of the time evolution of the weld fusion boundary. This moving boundary is the result of a spot GTA welding process on a 316L stainless steel disk. The estimation is based on the iterative regularization method. Indeed, the three problems: direct, in variation and adjoint, classically associated with this method, are solved by the finite element method in a two-dimensional axisymmetric domain. The originality of this work is to treat an experimental estimation of a front motion using a model with a geometry including only the solid phase. In this model, the evolution of this solid domain during the fusion is set with the ALE moving mesh method (Arbitrary Lagrangian Eulerian). The numerical developments are realized with the commercial code Comsol Multiphysics® coupled with the software Matlab®. The estimation method has been validated in a previous work using theoretical data ([1]). The experimental data, used here for this identification are, temperatures measured by thermocouples in the solid phase, the temporal evolution of the melt pool boundary observed at the surface by a fast camera and the maximal dimensions of the melted zone measured on macrographs. These experimental data are also compared with numerical results obtained from a heat and fluid flow model taking into account surface tension effects, Lorentz forces and the deformation of the melt pool surface under arc pressure. http://hdl.handle.net/10985/10174 Solving Stefan problem through C-NEM and level-set approach DAL, Morgan; MONTEIRO, Eric; LORONG, Philippe Numerical methods to solve problems involving discontinuities (jumps, kinks or singularities) on moving internal boundaries have received much attention over the last decade. Among them, the most suitable is probably the extended finite element method (XFEM) in tandem with the level-set technique due to its ability to take into account these discontinuities without matching meshes [1]. The present contribution aims to elaborate a numerical approach to model interfacial discontinu- ities within a meshless context. This approach couples the constrained natural element method (C-NEM) [2] and the level-set technique. In the former, the natural neighbours interpolation, based on a Voronoi diagram, is locally enriched through the partition of unity concept. This enrichment is built from level-set functions that represent and track implicitly discontinuities inside the domain [3]. Like in XFEM, key features of the proposed approach is (i) to determine the intersection between Voronoi cells and discontinuities and (ii) to integrate numerically the weak form over cells containing discontinuities. After testing the proposed method on classical benchmarks, both accuracy and efficiency are examined on the two phase Stefan problem that deals with heat flow involving a solid-liquid phase boundary on which a jump condition must be satisfied [4]. [1] Chessa J., Smolinski P. and Belytschko T. The extended finite element method (XFEM) for solidification problems. Int. J. Numer. Meth. Engng 53:1959–1977 (2002). [2] Yvonnet J., Chinesta F., Lorong P. and Ryckelynck D. The constrained natural element method (C-NEM) for treating thermal models involving moving interfaces. Int. J. Therm. Sci. 44:559–569 (2005). [3] LiuJ.T.,GuS.T.,MonteiroE.andHeQ.C.Aversatileinterfacemodelforthermalconduc- tion phenomena and its numerical implementation by XFEM. Comp. Mech. 53:825–843 (2014). [4] Carslaw H.S. and Jaeger J.C. Conduction of Heat in Solids. 2th Edition, Clarendon Press, (1959). Thu, 01 Jan 2015 00:00:00 GMT http://hdl.handle.net/10985/10174 2015-01-01T00:00:00Z DAL, Morgan MONTEIRO, Eric LORONG, Philippe Numerical methods to solve problems involving discontinuities (jumps, kinks or singularities) on moving internal boundaries have received much attention over the last decade. Among them, the most suitable is probably the extended finite element method (XFEM) in tandem with the level-set technique due to its ability to take into account these discontinuities without matching meshes [1]. The present contribution aims to elaborate a numerical approach to model interfacial discontinu- ities within a meshless context. This approach couples the constrained natural element method (C-NEM) [2] and the level-set technique. In the former, the natural neighbours interpolation, based on a Voronoi diagram, is locally enriched through the partition of unity concept. This enrichment is built from level-set functions that represent and track implicitly discontinuities inside the domain [3]. Like in XFEM, key features of the proposed approach is (i) to determine the intersection between Voronoi cells and discontinuities and (ii) to integrate numerically the weak form over cells containing discontinuities. After testing the proposed method on classical benchmarks, both accuracy and efficiency are examined on the two phase Stefan problem that deals with heat flow involving a solid-liquid phase boundary on which a jump condition must be satisfied [4]. [1] Chessa J., Smolinski P. and Belytschko T. The extended finite element method (XFEM) for solidification problems. Int. J. Numer. Meth. Engng 53:1959–1977 (2002). [2] Yvonnet J., Chinesta F., Lorong P. and Ryckelynck D. The constrained natural element method (C-NEM) for treating thermal models involving moving interfaces. Int. J. Therm. Sci. 44:559–569 (2005). [3] LiuJ.T.,GuS.T.,MonteiroE.andHeQ.C.Aversatileinterfacemodelforthermalconduc- tion phenomena and its numerical implementation by XFEM. Comp. Mech. 53:825–843 (2014). [4] Carslaw H.S. and Jaeger J.C. Conduction of Heat in Solids. 2th Edition, Clarendon Press, (1959). http://hdl.handle.net/10985/9752 Generation and characterization of T40/A5754 interfaces with lasers PEYRE, Patrice; BERTHE, Laurent; POUZET, Sébastien; SALLAMAND, Pierre; TOMASHCHUK, Iryna; DAL, Morgan Laser-induced reactive wetting and brazing of T40 titanium with A5754 aluminum alloy with 1.5 mm thickness was carried out in lap-joint configuration, with or without the use of Al5Si filler wire. A 2.4 mm diameter laser spot was positioned on the aluminum side to provoke spreading and wetting of the lower titanium sheet, with relatively low scanning speeds (0.1 to 0.6 m/min). Process conditions did not play a very significant role on mechanical strengths, which were shown to reach 250-300 N/mm on a large range of laser power and scanning speeds. In all cases considered, the fracture during tensile testing occurred next to the TiAl3 interface, but in the aluminum fusion zone. In a second step, we have investigated the interfacial resistance with the LASAT bond strength tester, based upon the generation and propagation of laser-induced shock waves. This allowed us estimating a uniaxial bond strength of 0.68 GPa for the T40/A5754 interface under dynamic loading conditions. Wed, 01 Jan 2014 00:00:00 GMT http://hdl.handle.net/10985/9752 2014-01-01T00:00:00Z PEYRE, Patrice BERTHE, Laurent POUZET, Sébastien SALLAMAND, Pierre TOMASHCHUK, Iryna DAL, Morgan Laser-induced reactive wetting and brazing of T40 titanium with A5754 aluminum alloy with 1.5 mm thickness was carried out in lap-joint configuration, with or without the use of Al5Si filler wire. A 2.4 mm diameter laser spot was positioned on the aluminum side to provoke spreading and wetting of the lower titanium sheet, with relatively low scanning speeds (0.1 to 0.6 m/min). Process conditions did not play a very significant role on mechanical strengths, which were shown to reach 250-300 N/mm on a large range of laser power and scanning speeds. In all cases considered, the fracture during tensile testing occurred next to the TiAl3 interface, but in the aluminum fusion zone. In a second step, we have investigated the interfacial resistance with the LASAT bond strength tester, based upon the generation and propagation of laser-induced shock waves. This allowed us estimating a uniaxial bond strength of 0.68 GPa for the T40/A5754 interface under dynamic loading conditions. http://hdl.handle.net/10985/12374 Analysis of laser–melt pool–powder bed interaction during the selective laser melting of a stainless steel GUNENTHIRAM, Valérie; PEYRE, Patrice; COSTE, Frédéric; FABBRO, Rémy; DAL, Morgan; SCHNEIDER, Matthieu The laser powder bed fusion (LPBF) or powder-bed additive layer manufacturing process is now recognized as a high-potential manufacturing process for complex metallic parts. However, many technical issues are still to overcome for making LPBF a fully viable manufacturing process. This is the case of surface finish and the systematic occurrence of porosities, which require postmachining steps. Up till now, the porosity origin remains unclear but is expected to be related to the stability of the process. As a LPBF part is made by the accumulation of hundreds of meters of small weld beads, it also appears to be important to understand all the phenomena that occur during the laser-powder-melt pool (MP) interaction for each single track. For this reason, in the first part of our study, using an instrumented LPBF setup and a fast camera analysis (>10 000 image/s), single tracks were fabricated and analyzed in real time and postmortem. Spatters ejections and powder denudation phenomena were observed together with variations of melt pool dimensions and melt-pool instabilities. In turn, the physical origin of this powder denudation and the dynamics of the MP were investigated and discussed. Sun, 01 Jan 2017 00:00:00 GMT http://hdl.handle.net/10985/12374 2017-01-01T00:00:00Z GUNENTHIRAM, Valérie PEYRE, Patrice COSTE, Frédéric FABBRO, Rémy DAL, Morgan SCHNEIDER, Matthieu The laser powder bed fusion (LPBF) or powder-bed additive layer manufacturing process is now recognized as a high-potential manufacturing process for complex metallic parts. However, many technical issues are still to overcome for making LPBF a fully viable manufacturing process. This is the case of surface finish and the systematic occurrence of porosities, which require postmachining steps. Up till now, the porosity origin remains unclear but is expected to be related to the stability of the process. As a LPBF part is made by the accumulation of hundreds of meters of small weld beads, it also appears to be important to understand all the phenomena that occur during the laser-powder-melt pool (MP) interaction for each single track. For this reason, in the first part of our study, using an instrumented LPBF setup and a fast camera analysis (>10 000 image/s), single tracks were fabricated and analyzed in real time and postmortem. Spatters ejections and powder denudation phenomena were observed together with variations of melt pool dimensions and melt-pool instabilities. In turn, the physical origin of this powder denudation and the dynamics of the MP were investigated and discussed. http://hdl.handle.net/10985/12175 Simplified numerical model for the laser metal deposition additive manufacturing process PEYRE, Patrice; POUZET, Sébastien; CASTELNAU, Olivier; DAL, Morgan The laser metal deposition (LMD) laser technique is a free-form metal deposition process, which allows generating near net-shape structures through the interaction of a powder stream and a laser beam. A simplified numerical model was carried out to predict layer heights together with temperature distributions induced by the (LMD) process on a titanium alloy, and a metal matrix composite. Compared with previously developed models, this simplified approach uses an arbitrary Lagrangian Eulerian free surface motion directly dependent on the powder mass feed rate Dm. Considering thin wall builds of Ti-6Al-4V titanium alloy, numerical results obtained with comsol 4.3 Multiphysics software were successfully compared with the experimental data such as geometrical properties of manufactured walls, fast camera molten pools measurements, and thermocouple temperature recordings in the substrate during the manufacturing of up to 10 LMD. Even if the model did not consider coupled hydraulic-thermal aspects, it provides a more realistic local geometrical description of additive layer manufacturing walls than simpler thermal models, with much shorter calculation times than more sophisticated approaches considering thermocapillary fluid flow. In a second step, microstructures (equiaxed or columnar) were predicted on Ti-6Al-4V walls using microstructural map available in the literature, and local thermal gradients G (K/m) and solidification rate R (m/s) provided by the FE calculation near the solidification front. © 2017 Laser Institute of America. Sun, 01 Jan 2017 00:00:00 GMT http://hdl.handle.net/10985/12175 2017-01-01T00:00:00Z PEYRE, Patrice POUZET, Sébastien CASTELNAU, Olivier DAL, Morgan The laser metal deposition (LMD) laser technique is a free-form metal deposition process, which allows generating near net-shape structures through the interaction of a powder stream and a laser beam. A simplified numerical model was carried out to predict layer heights together with temperature distributions induced by the (LMD) process on a titanium alloy, and a metal matrix composite. Compared with previously developed models, this simplified approach uses an arbitrary Lagrangian Eulerian free surface motion directly dependent on the powder mass feed rate Dm. Considering thin wall builds of Ti-6Al-4V titanium alloy, numerical results obtained with comsol 4.3 Multiphysics software were successfully compared with the experimental data such as geometrical properties of manufactured walls, fast camera molten pools measurements, and thermocouple temperature recordings in the substrate during the manufacturing of up to 10 LMD. Even if the model did not consider coupled hydraulic-thermal aspects, it provides a more realistic local geometrical description of additive layer manufacturing walls than simpler thermal models, with much shorter calculation times than more sophisticated approaches considering thermocapillary fluid flow. In a second step, microstructures (equiaxed or columnar) were predicted on Ti-6Al-4V walls using microstructural map available in the literature, and local thermal gradients G (K/m) and solidification rate R (m/s) provided by the FE calculation near the solidification front. © 2017 Laser Institute of America. http://hdl.handle.net/10985/12459 Multiphysics Simulation and Experimental Investigation of Aluminum Wettability on a Titanium Substrate for Laser Welding-Brazing Process PEYRE, Patrice; DAL, Morgan The control of metal wettability is a key-factor in the field of brazing or welding-brazing. The present paper deals with the numerical simulation of the whole phenomena occurring during the assembly of dissimilar alloys. The study is realized in the frame of potential applications for the aircraft industry, considering the case of the welding-brazing of aluminum Al5754 and quasi-pure titanium Ti40. The assembly configuration, presented here, is a simplification of the real experiment. We have reduced the three-dimensional overlap configuration to a bi-dimensional case. In the present case, an aluminum cylinder is fused onto a titanium substrate. The main physical phenomena which are considered here are: the heat transfers, the fluid flows with free boundaries and the mass transfer in terms of chemical species diffusion. The numerical problem is implemented with the commercial software Comsol Multiphysics™, by coupling heat equation, Navier-Stokes and continuity equations and the free boundary motion. The latter is treated with the Arbitrary Lagrangian Eulerian method, with a particular focus on the contact angle implementation. The comparison between numerical and experimental results shows a very satisfactory agreement in terms of droplet shape, thermal field and intermetallic layer thickness. The model validates our numerical approach. Sun, 01 Jan 2017 00:00:00 GMT http://hdl.handle.net/10985/12459 2017-01-01T00:00:00Z PEYRE, Patrice DAL, Morgan The control of metal wettability is a key-factor in the field of brazing or welding-brazing. The present paper deals with the numerical simulation of the whole phenomena occurring during the assembly of dissimilar alloys. The study is realized in the frame of potential applications for the aircraft industry, considering the case of the welding-brazing of aluminum Al5754 and quasi-pure titanium Ti40. The assembly configuration, presented here, is a simplification of the real experiment. We have reduced the three-dimensional overlap configuration to a bi-dimensional case. In the present case, an aluminum cylinder is fused onto a titanium substrate. The main physical phenomena which are considered here are: the heat transfers, the fluid flows with free boundaries and the mass transfer in terms of chemical species diffusion. The numerical problem is implemented with the commercial software Comsol Multiphysics™, by coupling heat equation, Navier-Stokes and continuity equations and the free boundary motion. The latter is treated with the Arbitrary Lagrangian Eulerian method, with a particular focus on the contact angle implementation. The comparison between numerical and experimental results shows a very satisfactory agreement in terms of droplet shape, thermal field and intermetallic layer thickness. The model validates our numerical approach. http://hdl.handle.net/10985/13274 Analysis and possible estimation of keyhole depths evolution, using laser operating parameters and material properties FABBRO, Rémy; PEYRE, Patrice; COSTE, Frédéric; GUNENTHIRAM, V; DAL, Morgan; SCHNEIDER, Matthieu The authors propose an analysis of the effect of various operating parameters on the keyhole depth during laser welding. The authors have developed a model that uses the analysis of the thermal field obtained in 2D geometry, which is mainly defined by the characteristic Peclet number. This allows us to show that the dependence of the aspect ratio R of the keyhole with the operating parameters of the process is a function of two parameters: a normalized aspect ratio R0, controlled by the incident laser power and the spot diameter, and a characteristic speed V0 related to the process of heat diffusion. The resulting general law R = f (R0, V/V0) appears to be very well verified by different experimental data and allows to define mean thermophysical parameters of the used materials. These data can then be used for keyhole depths prediction for any subsequent operating parameters of the same material. This model also allows us to define precisely a criterion for a keyhole threshold generation. The authors will apply the derived procedure to successfully analyze experiments on materials with very different thermophysical properties (such as steel alloys and copper), with various focal spots, incident laser powers, and welding speeds. Mon, 01 Jan 2018 00:00:00 GMT http://hdl.handle.net/10985/13274 2018-01-01T00:00:00Z FABBRO, Rémy PEYRE, Patrice COSTE, Frédéric GUNENTHIRAM, V DAL, Morgan SCHNEIDER, Matthieu The authors propose an analysis of the effect of various operating parameters on the keyhole depth during laser welding. The authors have developed a model that uses the analysis of the thermal field obtained in 2D geometry, which is mainly defined by the characteristic Peclet number. This allows us to show that the dependence of the aspect ratio R of the keyhole with the operating parameters of the process is a function of two parameters: a normalized aspect ratio R0, controlled by the incident laser power and the spot diameter, and a characteristic speed V0 related to the process of heat diffusion. The resulting general law R = f (R0, V/V0) appears to be very well verified by different experimental data and allows to define mean thermophysical parameters of the used materials. These data can then be used for keyhole depths prediction for any subsequent operating parameters of the same material. This model also allows us to define precisely a criterion for a keyhole threshold generation. The authors will apply the derived procedure to successfully analyze experiments on materials with very different thermophysical properties (such as steel alloys and copper), with various focal spots, incident laser powers, and welding speeds. http://hdl.handle.net/10985/16855 [INVITED] An overview of the state of art in laser welding simulation FABBRO, Rémy; DAL, Morgan The work presented in this paper deals with the laser welding simulation. Due to the rise of laser processing in industry, its simulation takes also more and more place. Nevertheless, the physical phenomena occurring are quite complex and, above all, very coupled. Thus, a state of art is necessary to summarize phenomena that have to be considered. Indeed, the electro-magnetic wave interacts with the material surface, heating the piece until the fusion and the vaporization. The vaporization induces a recoil pressure and deforms the liquid/vapor interface creating a vapor capillary. The heat diffused in the material produces thermal dilatation leading to mechanical stress and strain. As a complete simulation is too large to be computed with one model, the literature is composed by two kinds of models, the thermo-mechanical simulations and the multi-physical simulations. The first aims to find the mechanical stress and strain due to the welding. The model is usually simplified in order to reduce the simulation size. The second, compute the more accurately the thermal and the velocity fields. In that case authors usually search also the size of the weld bead and want to be totally self consistent. In this review, the major part of equations and assumptions needed to simulate laser welding are shown. Their effects on simulation results are illustrated for each simulation type. The paper aims to give sufficient knowledge and tools to allow a simulation of laser welding Fri, 01 Jan 2016 00:00:00 GMT http://hdl.handle.net/10985/16855 2016-01-01T00:00:00Z FABBRO, Rémy DAL, Morgan The work presented in this paper deals with the laser welding simulation. Due to the rise of laser processing in industry, its simulation takes also more and more place. Nevertheless, the physical phenomena occurring are quite complex and, above all, very coupled. Thus, a state of art is necessary to summarize phenomena that have to be considered. Indeed, the electro-magnetic wave interacts with the material surface, heating the piece until the fusion and the vaporization. The vaporization induces a recoil pressure and deforms the liquid/vapor interface creating a vapor capillary. The heat diffused in the material produces thermal dilatation leading to mechanical stress and strain. As a complete simulation is too large to be computed with one model, the literature is composed by two kinds of models, the thermo-mechanical simulations and the multi-physical simulations. The first aims to find the mechanical stress and strain due to the welding. The model is usually simplified in order to reduce the simulation size. The second, compute the more accurately the thermal and the velocity fields. In that case authors usually search also the size of the weld bead and want to be totally self consistent. In this review, the major part of equations and assumptions needed to simulate laser welding are shown. Their effects on simulation results are illustrated for each simulation type. The paper aims to give sufficient knowledge and tools to allow a simulation of laser welding http://hdl.handle.net/10985/13281 Experimental analysis of spatter generation and melt-pool behavior during the powder bed laser beam melting process GUNENTHIRAM, V; PEYRE, Patrice; SCHNEIDER, Matthieu; COSTE, Frédéric; FABBRO, Rémy; DAL, Morgan; KOUTIRI, Imade The experimental analysis of spatter formation was carried out on an instrumented SLM set-up allowing the quantification of spatter ejections and possible correlation with melt-pool behavior. Considering nearly similar SLM conditions than those carried out on SLM machines, an increase of large spatters (>80 μm) with volume energy density (VED) was clearly demonstrated on a 316L stainless steel, which was attributed to the recoil pressure applied on the melt-pool by the metal vaporization and the resulting high velocity vapor plume. In a second step, much lower spattering was shown on Al-12Si powder beds than on 316L ones. Fast camera analysis of powder beds indicated that droplet formation was mostly initiated in the powder-bed near the melt-pool interface. On Al-12 Si alloys, such droplets were directly incorporated in the MP without being ejected upwards as spatters like on 316L. Last, it was shown that a strong reduction of spattering was possible even on 316L, with the use of low VED combined with larger spots (≈0.5 mm), allowing to melt sufficiently deep layers in conduction regime and ensure adequate dilution between layers. Mon, 01 Jan 2018 00:00:00 GMT http://hdl.handle.net/10985/13281 2018-01-01T00:00:00Z GUNENTHIRAM, V PEYRE, Patrice SCHNEIDER, Matthieu COSTE, Frédéric FABBRO, Rémy DAL, Morgan KOUTIRI, Imade The experimental analysis of spatter formation was carried out on an instrumented SLM set-up allowing the quantification of spatter ejections and possible correlation with melt-pool behavior. Considering nearly similar SLM conditions than those carried out on SLM machines, an increase of large spatters (>80 μm) with volume energy density (VED) was clearly demonstrated on a 316L stainless steel, which was attributed to the recoil pressure applied on the melt-pool by the metal vaporization and the resulting high velocity vapor plume. In a second step, much lower spattering was shown on Al-12Si powder beds than on 316L ones. Fast camera analysis of powder beds indicated that droplet formation was mostly initiated in the powder-bed near the melt-pool interface. On Al-12 Si alloys, such droplets were directly incorporated in the MP without being ejected upwards as spatters like on 316L. Last, it was shown that a strong reduction of spattering was possible even on 316L, with the use of low VED combined with larger spots (≈0.5 mm), allowing to melt sufficiently deep layers in conduction regime and ensure adequate dilution between layers. http://hdl.handle.net/10985/15457 Prediction and sensitivity analysis of bubble dissolution time in 3D selective laser sintering using ensemble decision trees LY, Haibang; LE, Tien Thinh; LE, Vuongminh; PHAM, Binh Thai; DAL, Morgan; REGNIER, Gilles; MONTEIRO, Eric The presence of defects like gas bubble in fabricated parts is inherent in the selective laser sintering process and the prediction of bubble shrinkage dynamics is crucial. In this paper, two artificial intelligence (AI) models based on Decision Trees algorithm were constructed in order to predict bubble dissolution time, namely the Ensemble Bagged Trees (EDT Bagged) and Ensemble Boosted Trees (EDT Boosted). A metadata including 68644 data were generated with the help of our previously developed numerical tool. The AI models used the initial bubble size, external domain size, diffusion coefficient, surface tension, viscosity, initial concentration, and chamber pressure as input parameters, whereas bubble dissolution time was considered as output variable. Evaluation of the models' performance was achieved by criteria such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and coefficient of determination (R 2 ). The results showed that EDT Bagged outperformed EDT Boosted. Sensitivity analysis was then conducted thanks to the Monte Carlo approach and it was found that three most important inputs for the problem were the diffusion coefficient, initial concentration, and bubble initial size. This study might help in quick prediction of bubble dissolution time to improve the production quality from industry. Tue, 01 Jan 2019 00:00:00 GMT http://hdl.handle.net/10985/15457 2019-01-01T00:00:00Z LY, Haibang LE, Tien Thinh LE, Vuongminh PHAM, Binh Thai DAL, Morgan REGNIER, Gilles MONTEIRO, Eric The presence of defects like gas bubble in fabricated parts is inherent in the selective laser sintering process and the prediction of bubble shrinkage dynamics is crucial. In this paper, two artificial intelligence (AI) models based on Decision Trees algorithm were constructed in order to predict bubble dissolution time, namely the Ensemble Bagged Trees (EDT Bagged) and Ensemble Boosted Trees (EDT Boosted). A metadata including 68644 data were generated with the help of our previously developed numerical tool. The AI models used the initial bubble size, external domain size, diffusion coefficient, surface tension, viscosity, initial concentration, and chamber pressure as input parameters, whereas bubble dissolution time was considered as output variable. Evaluation of the models' performance was achieved by criteria such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and coefficient of determination (R 2 ). The results showed that EDT Bagged outperformed EDT Boosted. Sensitivity analysis was then conducted thanks to the Monte Carlo approach and it was found that three most important inputs for the problem were the diffusion coefficient, initial concentration, and bubble initial size. This study might help in quick prediction of bubble dissolution time to improve the production quality from industry.