Metrology for AM Model Validation

Summary

Multi-physics and data-driven models are necessary to simulate, study, and optimize metal additive manufacturing (AM) processes, such as powder bed fusion (PBF) and directed energy deposition (DED). Before these models can be used to design manufacturing processes or qualify parts for medical and aerospace applications, they must first be validated. In fact, the ANSI Additive Manufacturing Standardization Collaborative (AMSC) specifically identified AM model verification and validation as a key gap (Gap D9) in the Standardization Roadmap for Additive Manufacturing (2023), and the NASA Vision 2040: A Roadmap for Integrated, Multiscale Modeling and Simulation of Materials and Systems specifically mentioned the need for “gold standard” reference datasets for this purpose.  Unfortunately, for many with the capability and expertise to model metal AM process, the expense and complexity of empirical studies to validate their models is prohibitive.  In these and other roadmap reports on the state-of-art, NIST is specifically called upon to provide such AM model validation data.  This project, along with a large number of collaborators across NIST and outside research organization, aims to provide such trusted measurement data for the purpose of AM model validation, primarily disseminated through the Additive Manufacturing Benchmark Test Series (AM-Bench).  This project will design and perform various measurements on metal AM processes and parts, disseminate the data through datasets, publications, and AM-Bench conference series, and work directly with AM modelers to explore the best approaches for deriving quantitative, comparable metrics from both measurement and model results, and to provide the statistical framework for validating these complex and multivariate data.

Description

Objective
To create and openly disseminate measurement datasets, using carefully calibrated, characterized, and newly developed measurement techniques designed for the development and validation of physics-based and data-driven computational AM models and simulations.

Technical Idea
The project will meet the needs for advancing AM model development and validation put forth by the AM modeling and simulation stakeholders and outlined in roadmaps by providing the following:

Publicly accessible measurement datasets to validate the following types of AM process models.  These measurements, and the experiments that incorporate them, will be coordinated among the various NIST AM project teams to ensure their utility to meet multiple project and program objectives:

• Powder packing, spreading, and distribution models aim to study or predict how metal powder spreading, placement, and dynamics affect the AM fabrication process quality.  Measurement data will include:
  •  Particle position and velocity field tracking in 2D using the NIST powder spreading testbed.
  • Characterization and quantification of powder denudation and its effect on solid material consolidation.
• Melt pool scale models may incorporate all the complex physics of the laser-materials interaction process, including everything from laser optical energy absorption, thermal heat conduction, metal vaporization and local gas dynamics, molten metal fluid flow, and nonequilibrium solidification dynamics.  Measurement will include: 
  • High-fidelity melt pool imaging and thermography to obtain solidification temperatures, gradients, and cooling rates of the melt pool.
  • Dynamic and directionally-resolved laser reflectance, absorption, and metal-vapor plume interaction.
  • Single melt track, multi-track, and 3D part surface topography.
  • Cross-sectional microscopy of melt pools.
• Part-scale models typically utilize various physical or computational assumptions or reduced-fidelity to enable simulation of various aspects of the full 3D fabricated AM part.  Target physics include dynamic and location-specific thermal heterogeneity, residual strain and part distortion, and microstructure evolution.  Measurement data will include: 
  • Meso-scale in-situ thermographic measurements (i.e., able to observe single melt pool dynamics throughout a 3D build).
  • Dynamic residual strain utilizing synchrotron-based X-ray diffraction.
  • Static residual strain and part distortion using synchrotron or neutron-based diffraction or contour method, and external part geometry (e.g., via coordinate measuring machine, CMM)
• Hybrid physical and data-driven models are a relatively new class of modeling that uses machine-learning approaches to synthesize large volumes of measurement data, as well as high-fidelity simulation results, into separate or parallel data-driven models that allow for more rapid, computationally-efficient simulations.  The powder, melt pool, and part-scale measurement data described above may all be used for this purpose, given the appropriate formatting and metadata description is provided.  Datasets will include all necessary metadata to integrate into data-driven models, including instrument and machine calibration/characterization and spatial registration metadata. 

‘Time above melt’ values processed from high speed, in-situ thermographic imaging of each layer a laser powder bed fusion 3D build and compiled into a digital twin of the part shape. Part of AM-Bench 2022 datasets.

Metrology and measurement analysis techniques to populate and interpret the datasets listed above:

• Further develop and expand unique measurement capabilities focusing on melt-pool scale complex physics, such as:
  •  Laser-material interaction metrology, including directionally-resolved and power-calibrated dynamic laser reflection/absorption.
  • Illuminated high-speed imaging used for powder particle dynamics or powder denudation.
  • Vapor plume characterization, including temperature, size, shape, and influence on laser scattering or from ambient gas flow conditions. 
• Develop and disseminate measurement system models that connect AM simulation outputs (e.g., temperature) to measurement system outputs (e.g., signal).
• Design and conduct measurements and instrument calibrations and characterizations for the purpose of developing simulation-to-measurement relationship models.
• Assist other NIST projects on the design, development, calibration, characterization, or use of measurement systems used in or related to this project. 
  • Assist with novel measurement tools such as calibration sources, high-intensity illumination, image processing, etc.
  • Develop methods to relate higher-fidelity (e.g., high speed, high resolution, multi-wavelength) measurements to those used for more industrially practical in-situ process monitoring.

Methods for quantitative and statistical comparison between measurement and modeling data, and how that comparison affects real-world engineering decisions: 

• Work directly with AM modeling collaborators to understand the modeling output data and its uncertainty, and likewise provide modelers better understanding of measurement data uncertainty.
• Develop and disseminate extractable, comparable, and relevant data features from projects measurement and modeling data sources, and evaluate statistics-based equivalency or comparison tests across different dimensional domains.
• Test and evaluate suite of complex, multidimensional data equivalency tests. (e.g., distribution equivalency tests such as Kolgomorov-Smirnov, cross-entropy tests, image or array cross-correlation, etc., and assess their relevance for describing model accuracy and real-world applications.

Research Plan
The Metrology for Multi-Physics Model Validation Project has three primary activities to meet the requirements set forth by the AMSC and NASA reports. The first activity is to provide reference data to validate models of metal AM processes. This will be accomplished in-part through the Additive Manufacturing Benchmark Test Series (AM-Bench). AM Bench allows modelers to test their simulations against rigorous, highly controlled AM benchmark test data, which is generated at NIST. This data includes in-situ temperature and cooling rate measurements during the PBF process and post-process distortion, stress, and microstructure characterization. In addition to the AM Bench, reference data (thermal history, stress, distortion, microstructure) will be generated from highly controlled fabrication tests of varying complexity: from single weld beads up to parts with complex, freeform geometry, and from high-purity metals to variable-composition high entropy alloys.

The second activity is to develop the metrology, analysis techniques, and standard guidelines necessary to measure various physical phenomena key to the AM fabrication quality and required to develop models that can predict the quality.  NIST has developed several world class and unique AM metrology testbeds that will be used to develop these methods.  These include NIST Additive Manufacturing Metrology Testbed (AMMT) will be used to conduct 3D builds with uniquely tailored laser-scan control and high-speed, in-situ thermographic and surface topographic measurements throughout the build.  The Fundamentals of Laser-Material Interaction (FLaMI) testbed (activity leader: Dr. David Deisenroth) provides focused study of laser-induced melt pool physics, using a flexible array of unique measurements such as dynamic, absolute calibrated, and directionally-resolved laser reflectometry, high-magnification, high-speed, multi-wavelength thermography, or illuminated high-speed imaging.  Finally, the laser-processing and diffraction testbed (LPDT, Activity Leader: Dr. Ho Yeung), will provide synchrotron-based x-ray diffraction (XRD) and imaging data to study dynamic phase evolution of alloys during laser processing. The powder spreading testbed (PST, Activity Leader: Dr. Vipin Tondare) will provide high-resolution mapping of powder flow dynamics and spreading behavior.  

The third activity focuses on progressing the mechanisms by which AM modeling or simulation data is used and accepted as part of the qualification and certification framework.  Currently, AM models are calibrated or validated against a diverse array of measurement types, and the statistical (or sometimes heuristic) approaches used in comparing measurement to model outputs is similarly diverse.  This project will explore and develop the important data features or metrics extractable from both models and measurements and provide the statistical analysis framework for calibrating or validating AM multiphysics models with complex measurement data, considering the uncertainty in both.  Since NIST are primarily metrology experts, this will be done by close collaboration with the AM modelling research community.  As certain benchmark measurements and the AM model types advance in their accuracy, utility, and acceptance, this project will help draft and/or advance the development of standards to more effectively disseminate these methods and propel the acceptance of AM modeling data in AM qualification and certification.  

Major Accomplishments
Publications, presentations, and datasets for this project are primarily disseminated through AM-Bench challenges.  In particular, readers are encouraged to review the AM Bench Data and Challenge Problems page.