| Abstract |
Additive manufacturing (AM) is emerging as a revolutionary manufacturing process that can push the limits of conventional techniques. The market of AM is expected to increase exponentially during the next years, reaching 16 billion USD by 2018. However, there are some critical issues being faced now by the AM industry. The first problem is geometrical precision. The high temperature gradients generated during AM induce distortions on the fabricated structure, posing difficulties for later assembling processes and generating high contact pressures. Further, residual stresses must be accurately predicted in order to standardize AM-produced components. Nowadays, expensive trial-and-error procedures, based on the knowledge and experience coming from previous similar designs, are the standard industrial practice. But this approach is expensive and slow. The solution to these problems, as in more conventional processes, is to virtualize the manufacturing process and to optimize it. Unfortunately, there is a tremendous lack of AM simulation tools. The physical processes related to AM are very complicated, involving, e.g., multi-phase materials, high thermal gradients, and complex (visco-)plastic constitutive models. Further, the incremental production by thin layers (tens of microns) poses another problem to the simulations. In order to accurately follow the scanning sequences, the number of time steps in a whole simulation of an small piece can easily require tens of thousands of time steps and extremely large meshes. Useful industrial simulations are out of reach now, because there is a reduced number of AM simulation tools, and none of them efficiently run on large scale HPC platforms. Only the sinergy of advanced numerical techniques, including a multiscale framework and mesh adaptivity techniques in space and time, and the efficient exploitation of parallel platforms can provide a short term answer to these problems. This is the objective of this project |
