Goal: Numerical convergence to each solved equation at each milestone and will be evaluated by plotting quantities of interest (such as temperature, flow velocity, solid or liquid state at a particular location) against spatial mesh and timestep size, with a goal of demonstrating second order convergence in space and first order convergence in time. Industry relevant sample1 for experimental validation. Size comparable to 30cm*30cm*30cm. Parallel performance will be evaluated by demonstrating scalability for at least 64 CPUs with a goal maximum problem size of at least 50 million degrees of freedom (DOFs). Parallel performance will also be demonstrated on GPU hardware. Complete weak scaling study, with >75% efficiency when the code is scaled to at least 64 CPUs. Simulation of a normal size industrial-relevant part2 should be completed by a computer cluster within 1 day. Thermal modeling accuracy will be evaluated by showing quantitative agreement between predicted and measured temperature fields within 10%, except near peak temperatures which are sensitive to complex evaporative cooling effects. For material property modeling, quantitative agreement will be sought between experiments and simulations that examine microstructure (in the form of secondary dendrite spacing) as a function of process parameters (toolpath and energy input), and specifically, the dependence on local thermal history (cooling rate and temperature gradient). Quantitative agreement in yield strength, microhardness, hardness, and ultimate tensile strength is sought to within 15%, with very strong agreement sought in trends between process parameters and material properties.3 For goals specified in 2, 3, 4, 5, and 6, an industry relevant case is needed for validation experiment (1-2 models from collaborators to be used in validation). Simulation results should be within error envelope.
|Effective start/end date
|9/1/18 → 5/31/20
- Beijing Institute of Collaborative Innovation (20183405)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.