Robust Design Strategies to Mitigate Uncertainty in Additive Manufacturing

Robustness in design is crucial because additive manufacturing naturally introduces variability and uncertainty. It helps manage uncertainties from process fluctuations, geometric deviations, material anisotropy and defects, post-processing differences, and end-use load variations, ensuring performance, safety, and cost efficiency.
Parts often suffer from distortions, warping, and shrinkage caused by residual stresses. Surface roughness and stair-stepping effects create unexpected geometric variations from the design model. Mechanical properties depend on build orientation and layer bonding. Microstructural defects (porosity, lack of fusion, inclusions) cannot be eliminated. Real-world conditions (thermal cycling, dynamic loads, vibration) may not match the simulation loading exactly. The geometric deviations in additive manufacturing are highly stochastic. To quantify these deviations systematically, engineers use uncertainty quantification techniques that combine measurement, statistics, and computational modeling.
In this presentation, techniques for statistical shape analysis that compare 3D scans to nominal CAD models will be demonstrated, focusing on quantifying distributions of deviations such as the mean and standard deviation of thickness or diameter of lattices. Advanced sampling techniques can be utilized to train a response surface approximation model. This model can manage a large number of probabilistic input parameters. Stochastic finite element analysis can then use the response surface approximation to address the randomness in geometric parameters, boundary conditions, material properties, or structural/thermal loads, thereby predicting the stochastic performance requirements.
An example of a heat exchanger with beam lattices will illustrate the process. The beams are categorized based on their orientation to the horizontal plane, and their mean and standard deviation of diameters are determined from experimental results. The mean and standard deviation of the maximum temperature are computed, and using the upper specification limit, the sigma quality level of the heat exchangers is assessed.