Additive manufacturing (AM) is revolutionizing manufacturing, allowing complex parts to be constructed that are not readily fabricated by traditional techniques. Although there has been significant interest and investment in AM, the fraction of this investment devoted to modeling and simulation is relatively small and is not focused on developing high-fidelity predictive models but instead on reduced-order models for industry use. The Exascale Additive Manufacturing project (ExaAM) represents a unique opportunity to use exascale simulation to enable the design of AM components with location-specific properties and the acceleration of performance certification.
Additive manufacturing, also known as 3D printing, plays a critical role in transforming the manufacturing sector by providing the ability to make parts with very complex geometries. While 3D printers for home use can easily make plastic components, industrial use of 3D printing often seeks to make metal components. For many applications, such as aerospace and automotive, the strength of these metal components is very important. The strength of a metal component is determined by the component’s microstructure, or the pattern that the metal crystals make as they solidify. Predicting the strength of a metal component – critical for industry that relies on them – is necessary and requires great computing power.
The biggest influence on metal microstructure is the thermal history that the component experiences as it is printed layer by layer, with a laser that melts the appropriate pattern of metal powder in each layer. Every location in one of these parts can have a different thermal history, so every location can have a different microstructure and different strength properties. Different component geometries can have very different thermal histories, even if the 3D printer settings were exactly the same. Thus, predicting the ultimate performance of a 3D printed metal part is extremely challenging.
ExaAM is an Exascale Computing Project effort aimed at producing a suite of simulation codes capable of predicting the strength properties of a 3D printed metal part. This is done by predicting the thermal history of every location in the part, and then predicting the effect that these thermal histories have on the microstructure and the resulting strength properties. Previous modeling efforts had successfully demonstrated similar abilities on a very small scale, such as a 1 millimeter section of part, but to enable this capability on a large scale of an entire part requires very large computational resources, such as the Frontier supercomputer at ORNL.
The ExaAM project spent significant time and effort making the various codes in the suite capable of running efficiently on GPUs, since the vast majority of the capabilities of modern supercomputers comes from GPUs. The resulting suite, which is freely available as open source software, successfully ran on more than 8,000 nodes on the Frontier supercomputer (more than 32,000 GPUs!) and successfully predicted the strength properties of a National Institute of Standards and Technology additive manufacturing benchmark component.
Though the ExaAM projected ended in December 2023, the key components of the suite have proven so valuable that other Department of Energy projects – such as the Digital Factory at ORNL’s Manufacturing Demonstration Facility and the Advanced Materials and Manufacturing Technology effort within DOE’s Nuclear Energy program – continue to fund the development of these simulation tools. Thus the ExaAM supercomputing project promises to move additive manufacturing forward, as it continues to pay benefits for the nation and world.
ExaAM aims to develop the Integrated Platform for Additive Manufacturing Simulation (IPAMS), a suite of exascale-optimized capabilities that directly incorporate microstructure evolution and the effects of microstructure within AM process simulation. In AM, a geometric description of the part is processed into 2D slices. A feedstock material is melted, and the part is built layer by layer. In metal AM, the feedstock is often in wire or powder form, and the energy source is a laser or electron beam. ExaAM focuses on powder bed processes in which each layer is approximately 50 µm. For example, a part that is 1 cm tall would require 200 layers, each requiring the spreading of new feedstock powder and one or more passes of the laser or electron beam to sinter and/or melt the powder in appropriate locations.
The physical processes involved in AM are similar to those of welding—a field with decades’ worth of experimental, modeling, simulation, and characterization research. Although calibrated and approaching predictive capability, the simulation tools developed for welding and other similar processes are unfortunately inadequate for AM processes, as demonstrated by the inability to predict the failure rate for new AM parts, which can be as high as 80%. This is believed to be largely because the process-structure-property-performance relationship is traditionally modeled in an uncoupled manner, relying on tabular databases that cannot adequately capture the implicit, dynamic, nonequilibrium nature of AM processes.
One goal of ExaAM is to remove those limitations by integrating high-fidelity mesoscale simulations within continuum process simulations to determine the microstructure and properties by using local conditions. Sub-mesoscale physics is upscaled from detailed science simulations. Typically, thermomechanical finite element models are employed at the macroscopic part scale; finite volume or finite element models are used at millimeter scales for fluid dynamics and heat transfer to capture the melt pool dynamics and solidification; mesoscale approaches (e.g., discrete elements, cellular automata, kinetic Monte Carlo, or phase-field models) are used at the micron scale to simulate melting, solidification, and microstructure formation; and polycrystal plasticity models are used to develop the microscale mechanical property relationships.
The ExaAM IPAMS suite is a collection of simulation capabilities for performing process-aware performance modeling of additively manufactured parts by using locally accurate properties predicted from microstructures that develop based on local processing conditions. ExaAM will demonstrate this capability by simulating the complex bridge structure developed for the 2018 National Institute of Standards and Technology AM-Bench Conference, known as AMB2018-01. The simulation will be performed where experimental observations were taken (e.g., “cut locations” for transverse and longitudinal scanning electron microscope [SEM] specimens for microstructure images).
This challenge problem will be demonstrated on the exascale computer as a workflow involving three stages. First, a continuum powder layer melt-refreeze model, AdditiveFOAM, is used to simulate the laser scan pattern for several layers at a particular location within the part. Next, the resulting thermal histories at this location are used by ExaCA to drive simulations of the microstructure evolution. These microstructures, unique to the AM process and requiring careful validation with experimental observations, are then used by ExaConstit, a finite element crystal plasticity model which resolves the grain structure and behavior of individual grains to predict localized, continuum-scale stress-strain responses. The demonstration of this process-to-structure-to-properties workflow is a key enabler for accelerating the certification of parts produced with an AM process.