Final Report Summary - INLADE (Integrated numerical modelling of laser additive processes)
Executive Summary:
Additive laser manufacturing is an innovative production technology able to manufacture an unrivalled range of part geometries and multi-material structures without any fixtures directly from CAD drawings. It is capable of producing the new generation of high-value, multi-functional products increasingly demanded by industry. The Multiple Layer Laser Direct Metal Deposition (DMD) process is one such method – suitable for manufacturing with metallic materials and also for adding features to existing parts or cladding surface coatings. What has been limiting the development of the method is its unpredictability. Internal defects, poor surface finish, high residual stresses or poor material properties can be produced by incorrect combinations of process parameters. To fully exploit the potential of the method it is essential to be able to choose the parameter combinations to optimise final properties. This project aimed to further the process and the industrial application of it via the development of a comprehensive model of the process capable of running on a desktop computer.
At the start of the project in 2009, numerical models were already the most realistic way of modelling this process, but tended to break the process down into parts in order to deal with the many process parameters in DMD. Probably the most advanced ‘whole-process’ numerical models were these from the University of Michigan [1, 2] or GERAILP-LALP in France [3]. However, in reality these were only analytical-numerical models, with the powder stream used to supply material to the build process portrayed in an analytical, quasi-stationary way. Additionally, the former model was only ever shown to simulate a few millimetres of deposition track.
This project was the first to publish work on a model that considered the complete process from inside the deposition nozzle to the final deposit in a single domain. A small number of other models currently use a single domain method, most notably that of Purdue University [4], but the comprehensive nature of this model has not been surpassed.
The model was assembled as a series of FORTRAN subroutines within ESI’s CFD-ACE+ multiphysics code and SYSWELD welding-dedicated FE code. The model is described below. Different components of the model were seamlessly integrated.
1. Flow within the nozzle was considered as stable and modelled in steady state mode. The dynamic behaviour of the powder particles was determined by the drag force of the surrounding gas, the influence of gravity, and trajectory deflections caused by collisions with the solid walls of the powder passages.
2. Flow between the nozzle exit and substrate was considered unsteady and modelled in transient mode. A Lagrangian approach was used meaning the locations and trajectories of particles were freely determined in the entire mesh space. Within the laser beam close attention was paid to interaction of the particles and beam. Particles were treated as lumped capacity elements, absorbed radiation and also lost it to the surroundings. Their attenuating effect on the laser beam was also accounted for.
3. On reaching the substrate, powder particles hitting the surface outside the melt pool were modelled as bouncing off according to the surface restitution coefficient, while those falling within the melt pool were made to stick. The Lambert law was used to describe reflection of the laser beam from the surface and reflected radiation could then again interact with the flying particles.
4. Within the melt pool, Buoyancy and Marangoni Flows were modelled. Simulations showed that the addition of powder mass slightly modified the flows as the newly added material was redistributed across the molten area. Both upper surface waves and lateral undulations could be found in a simulated clad.
5. After the CFD-ACE+ simulation, the calculated deposition geometry was replicated in SYSWELD by considering the solid mass fraction in each cell and, where relevant, the surface normal vector. The results of CFD-ACE+ became the input parameters for SYSWELD and its code was used to compute final part properties such as phase distributions and residual stresses. The coupled models can be run in sequential mode (the CFD simulation is completed, and then the SYSWELD simulation is run) or coupled mode (CFD-SYSWELD simulations are done on an alternate step basis).
Experimental work was performed to verify all stages of the model. We used a Laserline LDL 160-1500 high-power diode laser fitted with a coaxial nozzle (manufactured in house so we had exact CAD details for the model) and 2-axis CNC motion system. Powder stream concentration distributions were measured using a well-known ‘light-plane’ technique, which matched powder concentration at any position in the stream to the luminance of light reflected from it. Temperatures distributions in the powder stream were measured using a thermal camera. The track shape that is formed after deposition is the easiest property to measure but arguably the most difficult to predict and the project included extensive matching of predicted and modelled track and intersecting multiple tracks. Late in the project we concentrated on comparing measured and modelled final part properties: residual stresses in a series of 316L steel parts and phase distributions in a series of M2 steel test parts.
The single-domain simulation in CFD-ACE+ within this project was extensively verified and papers describing it peer reviewed. It is currently one of the most advanced in the world at simulating what geometry will be deposited from initial process parameters. Specific findings for future modelling of the process are:
• The trajectory of powder particles is mainly determined by wall collisions inside the nozzle cavity and drag forces from the assistive gases. Accurate trajectory predictions cannot be achieved solely by the use of trigonometry and geometrical methods as has usually been assumed in analytical and analytical-numerical models to date
• Use of ‘free-steam’ models, as has been the norm since the beginning of the modelling of direct metal deposition, is not always appropriate. The influence of recoiled powder and reflected radiation from the substrate into the region above the melt pool must be considered.
• State of the art melt pool shape calculations account for modification due to Marangoni flow induced by laser irradiation. However, Marangoni flows are also influences by incoming mass - incoming powder has an effect on flows within the pool.
Calculating the residual stress and phase distributions within the model is something that had been achieved by the end of the four years, fulfilling a major aim of the project. The model was completed very close to the end of the project due to time spent overcoming computer resource issues in the project leaving no time to apply it to the process. A conclusion that can be drawn from this phase of the work is that implementing a model such as this on a desktop computer requires considerable attention to resource efficiency and efficient parallelisation.
Overall, we conclude that the project met its technical objectives in terms of advancing the science of modelling laser direct metal deposition and similar processes and by attention to resource efficiency managed to achieve this with desktop computer power. Work in this area is continuing to fully exploit the CFD-FE interface and stress and phase distribution calculations this enables.
It is probably too early to judge the direct economic impact of this work. It will feed into society via the European manufacturing industry - improving the ability to produce the high-value, multi-functional products mentioned initially and thus keep the industry competitive. To this end we have ensured that the project has received industrial as well as academic exposure during its course via an industrial conference and workshops aimed at combined industrial and academic audiences. As the first single domain model of its type we believe it has acted as a catalyst for other fully numerical models that are beginning to emerge in academia.
References
[1] X. He, G. Yu, and J. Mazumder, "Temperature and composition profile during double-track laser cladding of H13 tool steel," Journal of Physics D-Applied Physics, vol. 43, pp. 1-9, 2010.
[2] H. Qi, J. Mazumder, and H. Ki, "Numerical simulation of heat transfer and fluid flow in coaxial laser cladding process for direct metal deposition," Journal of Applied Physics, vol. 100, p. 024903, 2006.
[3] P. Peyre, P. Aubry, R. Fabbro, R. Neveu, and A. Longuet, "Analytical and numerical modelling of the direct metal deposition laser process," Journal of Physics D: Applied Physics, vol. 41, p. 025403, 2008.
[4] S. Y. Wen and Y. C. Shin, "Comprehensive predictive modeling and parametric analysis of multitrack direct laser deposition processes," Journal of Laser Applications, vol. 23, pp. 1-7, May 2011.