GOALI: Machine Learning Enabled Multiscale Modeling Platform and Virtual Sensing Digital Twin for Fatigue in Metals Additive Manufacturing (MAM)

Metal additive manufacturing (MAM), particularly laser-based powder bed fusion (L-PBF), has demonstrated significant potential for producing complex, high-performance components, yet its widespread industrial adoption remains constrained by challenges in qualification and certification (Q&C). These challenges arise from uncertainties in fatigue performance caused by process-induced variability in microstructure and defect distributions, which limit confidence in the reliability of critical structural components. This Grant Opportunity for Academic Liaison with Industry (GOALI) project will develop a predictive framework for understanding, quantifying, and monitoring fatigue damage in additively manufactured metallic materials. The research has the potential to accelerate the qualification of MAM components for safety-critical applications in aerospace, energy, and defense by reducing uncertainty in structural integrity assessments and enabling reliable in-service health monitoring. The project will contribute to national prosperity and technological competitiveness by advancing predictive modeling and digital engineering capabilities for advanced manufacturing. In addition, the program will provide interdisciplinary training for graduate students and postdoctoral researchers in multiscale modeling, machine learning, uncertainty quantification, and digital twin technologies, while strengthening partnerships among academia, industry, and federal laboratories. This research leverages a physics-based, machine learning-enabled multiscale computational framework to establish direct relationships between process-induced microstructure, defect characteristics, and fatigue performance in additively manufactured metals. This collaborative effort between Johns Hopkins University and Pratt & Whitney will bring together Integrated Computational Materials Engineering (ICME), uncertainty-quantified constitutive-damage modeling, and advanced machine learning to develop predictive tools for fatigue crack nucleation and propagation with direct links to location-specific microstructural features. The project will further develop a Virtual Sensing Digital Twin (VSDT) that combines graph neural operator encoders and decoders with Fourier neural operator spectral layers to reconstruct full-field fatigue damage evolution from optimally placed surface-mounted sensors. Validation will be performed using L-PBF Inconel 718 through collaborations with the National Institute of Standards and Technology (NIST) and NASA. The resulting platform seeks to improve qualification and certification by efficiently predicting fatigue life as a function of both manufacturing process variability and local material state, while enabling scalable digital tools for industry adoption through a software portal for dissemination to industrial, government, and research partners. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.