Probabilistic calibration framework for finite element thermal process modeling of metallic additive manufacturing. Application to promote certification/qualification of load critical aerospace flight parts.
Description
Application of active learning paradigm to efficiently develop Gaussian Process Regression surrogate where run time for the target finite element thermal process model is significant. Fully trained surrogate is then used by monte carlo process to develop probabilistic distribution of finite element model calibration variables which result in finite element model predictions that are in line with input empirical measurement distributions.
Detailed example
probabilistic distribution of finite element model calibration variables
AI / analytics pattern
Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relationships.
Automation level / stage
a) Pre-deployment – The use case is in a development or acquisition status.
Expected benefit
In short, this technique enables the robust and efficient calibration of a model simulating the manufacture of metallic parts by additive manufacturing. This is important because pure simulation of these processes is not possible without calibration due to large uncertainties in fundamental physical and material properties currently available due to the nature of the manufacturing process. The inherent variation in mechanical properties of parts produced using metallic additive manufacturing result in significant challenges to certification/qualification for flight. The goal is to alleviate these challenges with better probabilistic quantification of the fundamentals resulting in this inherent variation, thus promoting further adoption of this fledgling manufacturing technique by industry within the aeronautics industry.
Controls / human review
ATO: Not reported; PIA: Not published