AFSD-Nets: A Physics-Informed Machine Learning Model for Predicting the Temperature Evolution During Additive Friction Stir Deposition

Recent developments in additive friction stir deposition (AFSD)

Additive friction stir deposition of multi-layer aluminum-silicon carbide metal matrix composites

Additive friction stir deposition (AFSD) is a novel additive manufacturing (AM) technology in which solid-state, friction-stirred metal is deposited layer-by-layer to build three-dimensional parts. Unlike the mainstream fusion-based AM methods, AFSD does not cause metal melting and solidification. Therefore, AFSD can eliminate defects such as lack-of-fusion, key-holing, and large residual stress. Currently, the understanding of AFSD is based on the friction stir welding (FSW) process that has been widely studied for the past two decades. However, the material feeding and spreading in AFSD is essentially different from FSW and can complicate the thermal field and material flow. In this work, a computational fluid dynamics (CFD) model is created to simulate the temperature and fluid flow for AFSD of aluminum 7050 alloys. The predicted temperature is validated against both literature and thermocouple measurements. The current work lays the foundation for a quantitative understanding of AFSD process physics and the simulation-guided process design to tailor the thermal-mechanical field.

In recent years, additive manufacturing (AM) has gained prominence in rapid prototyping and production of structural components with complex geometries. Magnesium alloys, which have a strength-to-weight ratio that is superior compared with steel and aluminum alloys, have shown potential in lightweighting applications. However, commercial beam-based AM technologies have limited success with magnesium alloys due to vaporization and hot cracking. Therefore, as an alternative approach, we propose the use of a near net-shape solid-state additive manufacturing process, additive friction stir deposition (AFSD), to fabricate magnesium alloys in bulk. In this study, a parametric investigation was performed to quantify the effect of process parameters on AFSD build quality including volumetric defects and surface quality in magnesium alloy AZ31B. In order to understand the effect of the AFSD process on structural integrity in the magnesium alloy AZ31B, in-depth microstructure and mechanical property characterization was conducted on a bulk AFSD build fabricated with a set of acceptable process parameters. Results of the microstructure analysis of the as-deposited AFSD build revealed bulk microstructure similar to wrought magnesium alloy AZ31 plate. Additionally, similar hardness measurements were found in AFSD build compared with control wrought specimens. While tensile test results of the as-deposited AFSD build exhibited a 20% drop in yield strength (YS), nearly identical ultimate strength was observed compared with the wrought control. The experimental results of this study illustrate the potential of using the AFSD process to additively manufacture Mg alloys for load bearing structural components with achieving wrought-like microstructure and mechanical properties.

As an alternative to current additive manufacturing (AM) techniques, additive friction stir deposition (AFSD) is a solid-state process that has been recently explored and does not require melting or remelting of the feedstock, which is a challenge in fusion AM processes. Once its yield point is reached, the material is deposited under elevated temperatures—a similar mechanism to that of friction stir welding (FSW), producing fully dense parts with more equiaxed and finer grain structures, potentially not requiring postheat treatment. Compared with direct energy deposition (DED) technologies, it benefits from reduced heat input and high build rates. It also allows for open-air deposition of reactive metals such as aluminum alloys. These alloys are used extensively in the aerospace industry, and demand is expected to double over the next decade. The 7XXX series is the hardest and strongest commercial grade among the aluminum alloys, making them useful for aircrafts, high-speed trains, and parts under high stress. Although research on AFSD development has mostly focused on aluminum alloys, advancements using 7XXX series have been either limited or nonexistent. In this paper, we analyze 7XXX series aluminum parts produced through AFSD by the introduction of different combinations of torque, deposition rate, and tool speed. Tensile and hardness tests are performed in different directions, including the interface between the baseplate and deposited material, assessing the overall strength of all AFSD parts carried out. Despite AFSD's potential, it is demonstrated that there is an opportunity for improvements, and further work (e.g., Charpy test, microstructural characterization, heat treatment) is required to comprehend the technologies’ impact and benefits.

Process planning for hybrid manufacturing using additive friction stir deposition

Additive friction stir deposition (AFSD) is attractive for its ability to create freeform and fully dense structures without melting and solidification. Hence, additive friction stir deposition is an alternative to fusion-based additive manufacturing technology. In this study, the influence of the AFSD parameter (i.e., rotational speed and feed rate) on the mechanical properties and microstructure of 6061 aluminum alloy is investigated. The as-deposited 6061 aluminum alloy exhibits a relatively homogeneous microstructure with extensive equiaxed grains. Compared to the feedstock material, the ultimate tensile strength of the as-deposited 6061 aluminum alloy decreased to 65% from 320 to 210 MPa. The results of electron backscatter diffraction indicate that continuous dynamic recrystallization occurs during the AFSD process. Furthermore, it is evidence that the grain size and ultimate tensile strength are positively correlated with feed rate and rotational speed, whereas the elongation at break decreases with the increase in feed rate and rotational speed.

Quasi-static and high-rate tensile experiments were used to examine the strain rate sensitivity of laser-directed energy deposition (L-DED)- and additive friction stir deposition (AFSD)-formed AerMet 100 ultrahigh-strength steel-additive manufactured builds. Electron backscattered diffraction (EBSD) revealed similar as-deposited grain sizes between the two AM processes at approximately 24 μm and 17 μm for the L-DED and AFSD samples, respectively. The strain hardening rate, θ, revealed little change in the overall hardening observed in the L-DED and AFSD materials, with a consistent hardening in the quasi-static samples and three identifiable regions in that of the high-rate tested materials. The LDED deposited materials displayed average ultimate tensile strength values of 1835 and 2902 MPa for the 0.001 s−1 and 2500 s−1 strain rates, respectively and the AFSD deposited materials displayed ultimate tensile strength values of 1928 and 3080 MPa for the 0.001 s−1 and 2500 s−1 strain rates, respectively. Overall, the strength for both processes displayed a positive strain rate sensitivity, with increases in strength of ~1000 MPa for both processes. Fractography revealed significant solidification voids in the laser DED material and poor layer adhesion in the AFSD material.

Predicting performance of in-situ microbial enhanced oil recovery process and screening of suitable microbe-nutrient combination from limited experimental data using physics informed machine learning approach

In this study, a finite volume simulation framework was developed, validated, and employed for the first time in a new solid-state additive manufacturing and repair process, Additive Friction Stir Deposition (AFSD). The open-source computational fluid dynamics (CFD) code openfoam was used to simulate the deposition of a single layer of Aluminum Alloy 6061 feedstock onto a substrate, using a viscoplastic model to predict the flow behavior of the material. Conjugate heat transfer was considered between the build layer, the surrounding atmosphere, and the substrate, and the resulting temperatures were validated against experimental data recorded for three processing cases. Excellent agreement between simulated and measured temperature data was obtained, as well as a good qualitative prediction of overall build layer morphology. Further analysis of the temperature field was conducted to reveal the variation of temperature in the build direction, an analysis not possible with previous experimental or numerical methods, as well as a global heat transfer analysis to determine the relative importance of various modes of heat input and cooling. Tool heating was found to be the primary heat input to the system, representing 73% of energy input, while conduction to the substrate was the main mode of part cooling, representing 73% of heat loss from the build layer.

Self-induced stochastic resonance (SISR) is the emergence of coherent oscillations in slow-fast excitable systems driven solely by noise, without external periodic forcing or proximity to a bifurcation. This work presents a physics-informed machine learning framework for modeling and predicting SISR in the stochastic FitzHugh–Nagumo neuron. We embed the governing stochastic differential equations and SISR-asymptotic timescale-matching constraints directly into a Physics-Informed Neural Network (PINN) based on a Noise-Augmented State Predictor architecture. The composite loss integrates data fidelity, dynamical residuals, and barrier-based physical constraints derived from Kramers’ escape theory. The trained PINN accurately predicts the dependence of spike-train coherence on noise intensity, excitability, and timescale separation, matching results from direct stochastic simulations with substantial improvements in accuracy and generalization compared with purely data-driven methods, requiring significantly less computation. The framework provides a data-efficient and interpretable surrogate model for simulating and analyzing noise-induced coherence in multiscale stochastic systems. • A physics-informed neural network that learns stochastic dynamics directly. • Noise-Augmented State Predictor (NASP) for data-efficient multiscale modeling. • Predicts noise-driven oscillations and coherence from short trajectory data. • Shown on self-induced stochastic resonance in FitzHugh–Nagumo neuron model.

Chapter 5 - Effects of tool geometry

In this paper, the phase structure, composition distribution, grain morphology, and hardness of Al6061 alloy samples made with additive friction stir deposition (AFS-D) were examined. A nearly symmetrical layer-by-layer structure was observed in the cross section (vertical with respect to the fabrication-tool traversing direction) of the as-deposited Al6061 alloy samples made with a back-and-forth AFS-D strategy. Equiaxed grains were observed in the region underneath the fabrication tool, while elongated grains were seen in the “flash region” along the mass flow direction. No clear grain size variance was discovered along the AFS-D build direction except for the last deposited layer. Grains were significantly refined from the feedstock (~163.5 µm) to as-deposited Al6061 alloy parts (~8.5 µm). The hardness of the as-fabricated Al6061 alloy was lower than those of the feedstock and their heat-treated counterparts, which was ascribed to the decreased precipitate content and enlarged precipitate size.

Microstructure and mechanical properties of Ti6Al4V alloys fabricated by additive friction stir deposition

An AA6061-7075 composite with a heterogeneous structure was fabricated via the additive friction stir deposition (AFSD) method, and in situ processing data were monitored during the manufacturing process. The results show that the cross-section of the composite subjected to AFSD exhibits a lower degree of plastic deformation behavior compared to the surface and side of the composite, owing to serious heat accumulation during the layer-by-layer stacking process. The denser, heterogeneous structure, consisting of finer (softer) and coarser (harder) grains, which correspond to AA6061 and AA7075, was formed according to transmission electron microscopy (TEM) analysis. Furthermore, the obtained composite subjected to AFSD in this work presents outstanding mechanical properties compared to other as-fabricated AA6061/AA7075 depositions acquired by other additive manufacturing methods along the horizontal building direction, with the ultimate tensile strength (266 MPa) being 89% of that of AA6061-T6 and the elongation 1.1 times that of AA7075-T6. The findings provide useful guidelines for the in situ preparation of Al-based composites and offer ideas for manufacturing high-strength heterostructures for large-scale practical engineering applications.