MODELING OF HEAT TRANSFER AND FLUID FLOW IN KEYHOLE MODE WELDING

Open Access
- Author:
- Rai, Rohit
- Graduate Program:
- Materials Science and Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 22, 2008
- Committee Members:
- Tarasankar Debroy, Dissertation Advisor/Co-Advisor
Tarasankar Debroy, Committee Chair/Co-Chair
Ivica Smid, Committee Member
Long Qing Chen, Committee Member
Suzanne E Mohney, Committee Member
Todd Palmer, Committee Member - Keywords:
- genetic algorithm
numerical modeling
keyhole mode welding
electron beam
laser - Abstract:
- Since deep penetration welds can be made with keyhole mode laser and electron beam welding, they are widely used for joining thick sheets of metals and alloys. The weld geometry and microstructure depend on the temperature distribution and the cooling rates. Experimental determination of temperatures in the work-piece through the use of thermocouples, can provide data for a limited number of points and is time consuming and expensive. Numerical modeling of heat transfer and fluid flow in high energy density laser and electron beam welding can provide previously unavailable information about the temperature distribution and thermal cycles at all points in the computational domain in a relatively short time and at low cost. A critical review of the available literature indicates the following problems with the numerical models of keyhole mode welding. (1) There is no comprehensive three dimensional model of keyhole mode electron beam welding (EBW) available in literature. (2) While comprehensive models of keyhole mode laser beam welding (LBW) have been proposed, none has been tested for the welding of various materials under different process conditions. (3) None of the existing models of keyhole mode laser or electron beam welding contains a structural component designed to provide good agreement between the computed and experimental results. (4) None of the existing models can work backwards, i.e. provide a set of welding process variables that will result in desired weld characteristics. The goal of this thesis is to address these important issues. In this work, computationally efficient numerical models have been developed for linear keyhole mode LBW and EBW processes. The models combine an energy balance based model for keyhole geometry calculation with a well tested 3D heat transfer and fluid flow model. For LBW, keyhole wall temperatures are assumed to be equal to the boiling point of the alloy at 1 atm pressure. Keyhole wall temperatures in EBW are calculated from the equilibrium vapor pressure versus temperature relation for the work-piece material. The vapor pressure is, in turn, calculated from a force balance at the keyhole walls between the surface tension, vapor pressure and hydrostatic forces. A turbulence model is used to estimate the effective values of viscosity and thermal iv conductivity to account for the enhanced heat and mass transport in the turbulent weld pool due to the fluctuating components of velocities in both LBW and EBW. The proposed model for LBW has been tested for materials with wide ranging thermo-physical properties under varying input powers and welding speeds covering both partial and full penetration welds. The tested materials include Al 5754 alloy, A131 steel, 304L stainless steel, Ti-6Al-4V, tantalum, and vanadium. These materials vary significantly in their thermo-physical properties, including boiling point, thermal conductivity, and specific heat. The EBW model was tested for 21Cr-6Ni-9Mn steel, 304L stainless steel, and Ti-6Al-4V for different input powers and power density distributions. To improve the agreement between the calculated and experimental results, a methodology is presented to estimate the values of uncertain input parameters like absorption coefficient and beam radius using a genetic algorithm with the numerical model and limited amount of experimental data. Finally, a genetic algorithm is used with the numerical model to prescribe welding conditions that would result in a desired weld attribute. The computed weld cross-sectional geometries and thermal cycles agreed reasonably well with the experimental observations. The weld pool shapes depended on the convective heat transport within the weld pool. Convective heat transfer was more important for materials with low thermal diffusivity. The calculated solidification parameters showed that criterion for plane front stability was not satisfied for the alloys and the range of welding conditions considered in this work. Higher peak temperatures were found in the EBW of Ti-6Al-4V welds compared to similar locations in 21Cr-6Ni-9Mn stainless steel welds due to the higher boiling point and lower solid state thermal conductivity of the former. Non-dimensional analysis showed that convective heat transfer was very significant and Lorentz force was small compared to Marangoni force. Comparison of calculated weld geometries for electron beam and laser beam welds for similar process parameters showed that lower keyhole wall temperatures in EBW tend to make the welds deeper and narrower compared to laser beam welds. A genetic algorithm was used to optimize the values of absorption coefficient and beam radius based on limited volume of experimental data for 5182 Al-Mg alloy welds. The weld geometry calculated using the optimized values of absorption coefficient and beam radius was in v good agreement with experimental observations. The optimized values of absorption coefficient and beam radius were then used to prescribe sets of welding conditions to obtain specified weld geometry. These sets of welding conditions differed significantly but resulted in the same weld geometry. The results show that a widely applicable and computationally efficient 3D model of heat transfer and fluid flow can be developed by combining an energy balance based keyhole calculation sub model with a 3D convective heat transfer model. The modeling results can improve the understanding of the keyhole mode welding process. The results also show that by combining numerical models with an optimizing algorithm, the model results can be made more reliable. Finally, systematic tailoring of weld attributes via multiple pathways, each representing alternative welding parameter sets, is possible based on scientific principles.