1 p.m. – 1.05 p.m.

Dr. Torsten Hermanns


1.05 p.m. – 1.35 p.m.

Keno Kruse
Business Development Manager Additive Manufacturing
Hannover Germany

Simulation as an elementary component of the digital additive process chain

1.35 p.m. –2.05 p.m.

Dr. Markus Nießen
Senior Scientist
Fraunhofer ILT
Aachen Germany

Simulation activities of the Fraunhofer ILT for laser-based additive manufacturing

2.05 p.m. –2.35 p.m.

Paree Allu
Senior CFD Engineer
Flow Science Inc.
Santa Fe New Mexico USA

CFD modelling advances in additive manufacturing

2.35. p.m. – 3.05 p.m.

Prof. Wolfgang Schulz
Head of Instruction and Research Department for Nonlinear Dynamics of Laser Manufacturing Processes
RWTH Aachen Germany

Modelling, model reduction and simulation of non-linear dynamics in laser processing:

Learn the Inverse Solution comparing Automatic Differentiation AD and Radial Basis Functions RBF-networks

3.05 p.m.

Open discussion


Simulation as an elementary component of the digital additive process chain

Additive manufacturing is a key factor for functional integration, lightweight construction and individualised products. It overcomes a lot of traditional manufacturing limitations, but at the same time introduces new ones. Dimensional accuracy of parts is a serious challenge. An optimised build strategy is the result of process simulation as well as geometry compensation that takes into account the deformation caused by cyclic heating, cooling, melting and solidification. To achieve good accuracy, a detailed simulation approach based on the real pressure patterns is recommended. In addition to predicting the stresses and distortions, a process simulation allows the determination of the ideal scan speed and laser power, keyhole formation, balling-up and varnish-of-fusion. The simulation can cover the melt pool size and shape and allows prediction of grain size, microstructure orientation and porosity. Just as important as the determination of these parameters is the systematic organisation of this information around the material. With the help of all this knowledge, even complex additive projects can be realised systematically.

In this presentation, we will show the status quo of Ansys additive solutions and give an insight into the pipeline of additive developments at Ansys.

Keno Kruse, CADFEM GmbH

Simulation activities of the Fraunhofer ILTfor laser-based additive manufacturing

Laser based Additive Manufacturing technologies such as powder bed based selective laser melting allow an almost unlimited geometrical freedom in the production of metallic functional parts. Due to existing process deficits like residual stresses and part distortion this immense potential for the production of functional parts of high complexity can be exploited in a limited way only. This is caused by the currently insufficient understanding of the involved physical processes both locally in the interaction or process zone (e.g. beam propagation in the powder layer, evaporation) and globally in the part (thermomechanics). The objective of simulations is to provide specific recommendations for a stress- and distortion-minimizing process control for specific components or materials.

In this presentation, work of the Fraunhofer ILT on the simulation of the LPBF as well as the LMD process will be presented.

Dr. Markus Nießen, Fraunhofer ILT

CFD modelling advances in additive manufacturing

Laser processing technology has contributed to the success of metal additive manufacturing (AM) processes such as laser powder bed fusion and direct metal deposition. Although AM has been generating significant interest, challenges remain towards a more widespread adoption of this technology. These challenges include defects such as porosity and spatially non-uniform micro-structures that occur because of insufficient knowledge in process control. Computational fluid dynamics (CFD) modelling can help understand the effects of process parameters such as laser power, beam shapes and scan paths on the underlying physical phenomena such as laser-powder interaction, melt pool dynamics, phase change and solidification. With experimental studies successfully capturing melt pool temperatures and weld bead dimensions, it is possible to calibrate numerical models to the experimental data. These numerical models, which are based on a rigorous solution of the conservation equations, can provide further insights into fluid convection in the melt pool, temperature gradients, solidification rates and microstructure predictions. In this presentation, case studies from industry and academia highlighting the successful use of CFD and numerical models in understanding powder bed fusion and direct energy deposition processes are discussed in detail. It is shown how process parameter optimization is used to control porosity formation, balling defects and microstructure evolution for several alloys. Furthermore, these high fidelity, multiphysics CFD models provide a framework to better understand AM processes at the particle and melt pool scales. Ultimately, this information can be used to accurately model additional aspects of AM processes such as thermal and residual stresses and distortions at the part scale.

Paree Allu, Flow Science Inc.

Modelling, model reduction and simulation of non-linear dynamics in laser processing: Learn the Inverse Solution comparing Automatic Differentiation AD and Radial Basis Functions RBF-networks.

A modern research focus is learning from experience and data, like the way our brain works. Learning means finding a model that will generate the data. Experience means including some physical models that allow predictions to be verified. In fact, we are used to train everyday actions through combining experience and practice, just through trial and error, without knowing more comprehensive physical reason for it.

Sir Arthur Conan Doyle let say his Sherlock Holmes: “There are few people, however that if you told them a result, would be able to evolve from their own inner consciousness what the steps were that led to that result. This power is what I mean when I talk of reasoning backward.”

Finally, we understand inferring properties of physical systems from data means solving the inverse task. Let us face the challenge of learning how the inverse problem can be solved with the help of data that were initially generated by established reduced models that are more easy to control. Then, step by step, the multiple difficulties of larger models or even machine data can be considered, like finite amount of data, data uncertainties, non-uniqueness, etc.

Two approaches solving the inverse problem applied to the AsymptoticDrill Model are discussed, namely so-called Automatic Differentiation and Radial Basis Function Networks.

Prof. Wolfgang Schulz, RWTH Aachen