Predicting part distortion field in additive manufacturing: a data-driven framework

The underlying complicated spatiotemporal thermo-mechanical processes in additive manufacturing (AM) technology pose challenges in predicting and optimizing the as-built part quality for production use. Physics-based simulations are being developed to provide reliable predictions such as part distortions, residual stresses/strains, microstructure contents, and grain morphology, which can guide the product design and manufacturing process for improved part quality. However, due to the complexity of the problem, e.g., multi-physics and multi-scale, physics-based models need the expertise to build an extensive simulation time preventing its use in real-time monitoring and control. This study proposes an inductive data-driven framework to simulate the part distortion field for additively manufactured components. The part distortion field data were collected from different AM build strategies using finite element (FE) simulations. The FE model consists of multi-layer part distortion fields as outputs constituted by AM process parameters as inputs. The surrogate model consists of two stages: (1) self-organizing map to project the high-dimensional spatial field of the part distortion into a likelihood estimator, (2) hybrid self-organizing methods to predict the extracted features and reconstruct the part distortion field. The data-driven model evaluates different build scenarios' effects on the distortion field for additively manufactured parts. The results correlated well with FE simulations and established a prediction-based compensation strategy to reduce part distortions. It should be noted that while the framework is currently applied to FE simulation data, the integrated data-driven methods can also be deployed on experimentally measured 3D cloud data or other high throughput field measurements.