N23A-T004   TITLE: DIGITAL ENGINEERING - Digital Twin-based Machine Control for Adaptive Additive Manufacturing Processing of Metallic Aerospace Components

OUSD (R&E) CRITICAL TECHNOLOGY AREA(S): Artificial Intelligence (AI)/Machine Learning (ML); Autonomy; General Warfighting Requirements (GWR)

OBJECTIVE: Develop a digital twin (DT)-based system that can autonomously tailor local microstructure, heal defects, and minimize residual stresses and surface roughness in near real-time to assure repeatable, reliable, and optimal fatigue performance for additive manufactured (AM) metallic aerospace components.

DESCRIPTION: Additive manufacturing (AM) has attained significant popularity in research and application fields such as aerospace, automobile, maritime, biomedical, and other industrial sectors. AM processes are capable of depositing near-net-shape complex geometries; but major drawbacks of the AM materials remain, including anisotropy, surface irregularities, residual stresses, and defects (such as porosity, microcracks, inclusions, dislocations, etc.). These drawbacks significantly influence the static and dynamic mechanical properties of a material. As a result, the AM parts still lack quality consistency and repeatability. While post-processing such as Hot Isotropic Pressing (HIP) could enhance performance for some alloys, it could lead to additional cost and production time.

The various defects in AM processing can arise due to many causes, and significant attention has been given to different strategies to ameliorate these defects. Two significant types of defects observed from AM processing are lack-of-fusion (LOF) defects and gas porosity. LOF occurs when an insufficient amount of energy is applied to melt a specific location of a powderbed. This is directly influenced by site-specific processing parameters, such as laser power, hatch spacing, and scan speed. A rescanning strategy in the region with the LOF may eliminate the LOF defects. Gas porosity is often spherical due to gas trapped in the raw metal powder particles or trapped inert gas during the AM processing. It has been shown that gas porosity can be minimized with increasing scan speed and appropriate power level.

Another challenge that causes build failures and poor part quality is large residual stresses due to the high cooling rates in AM processing. However, recent research has found that rescanning the top layer reduces residual stresses near that surface. In order to enable in-situ controlling the quality of the build, a monitoring and control system is necessary.

There are several in-situ monitoring methods that have been demonstrated. For example, in order to capture the porosity, a high-speed camera and a photodiode were used to measure the dimensions of the melt pool condition and the mean emitted radiation. A two-color pyrometer was used to relate the consolidation phenomena with the surface temperature and to understand the solidification process of the molten powder. However, the collected data could be enormous (~0.5 Terabytes per build) requiring a large amount of storage and fast computational algorithms. Furthermore, the level of accuracy to detect and classify defects/anomalies still needs significant improvement, especially minimizing false positives and negatives.

Once the undesirable state of the deposit is monitored and sensed, near real-time healing and tailoring of the deposit are needed. A close-loop feedback laser system, such as a laser with shaped beam profiles, may be used for in-situ treatment of the deposit, as well as preventing material spattering and defects. For example, it was reported that equiaxed grains occupied a larger area fraction, and texture was reduced in parts built using an elliptical beam, compared to those built using a Gaussian beam. Laser power and wavelength control could also tailor the cooling rate to prevent cracking and improve the mechanical properties of the part. The use of ring distributed power in welding can result in a decrease of laser penetration depth, along with huge spatter reduction observed on the deposited bead and surrounding area. However, a near real-time close-loop feedback system also requires robust reduced-order/surrogate modeling coupled with Artificial Intelligence (AI)/Machine Learning (ML) that link powders, process parameters, microstructure, and site-specific properties at the component level, including the effects of build orientation, laser-material interaction, and specific part geometry. These are the key elements of a DT-based system.

The Navy requires an integrated DT-based system that can provide near real-time machine control for fully autonomous adaptive AM processing of metallic aerospace components. The system should be able to: (a) locally re-scan and re-melt; (b) autonomously adjust and control deposited energy density including laser power/intensity, spot size, and beam shape/profile, (de)focusing, scanning speed and pattern, hatch spacing, layer thickness, and interlayer delay time; and (c) build time of a complete part should not exceed by more than 20% compared to the traditional fixed parameter pre-set method. It is envisioned that such an in-situ tailoring/healing system will not only significantly improve the fatigue life performance comparable to wrought alloys, but also assure the repeatability and reliability of AM structural parts.

PHASE I: Demonstrate feasibility of a feedback control concept that integrates with a beam shaping laser system, sensors, ML-enabled monitoring and control methodologies, and reduced-order modeling (ROM) for Laser Powder Bed Fusion (LPBF) system. Demonstrate the feasibility of healing and tailoring the AM deposit. A detailed plan should be laid out to perform the validation of the effectiveness of the concept with implementation for a metallic alloy such as AlSi10Mg aluminum. The concept should have the potential to be developed into a full-scale, near real-time, in-situ monitoring and control system of an AM process to improve fatigue properties in Phase II. The Phase I effort will include prototype plans to be developed under Phase II.

PHASE II: Develop, demonstrate, and validate the prototype system developed in Phase I. Fully develop and validate the ML-enabled monitoring and control system with reduced-order modeling for a laser-based AM process to efficiently and robustly heal and tailor the AM deposited material properties, based on the optimized AM process parameters for additional selected materials, such as Ti6Al4V, SS316L, and aluminum alloys. Demonstrate its capability of manufacturing aircraft components with complex geometry and tailored performance.

PHASE III DUAL USE APPLICATIONS: Fully develop the advanced monitoring and control system coupled with reduced order models for various laser-based AM processes to fabricate naval aircraft components that can be integrated into the fleet. Conduct final component-level testing to achieve the geometry and material property of AM components meeting the Navy�s needs.

The monitoring and control system will be directly applicable to a wide range of AM process methods and machines due to the high amount of usage of AM parts in the commercial aerospace and medical industries. The oil and gas, automotive, and shipping industries could also benefit from this developed technology.

REFERENCES:

1.       Liu, M., Fang, S., Dong, H., & Xu, C. (2021). Review of digital twin about concepts, technologies, and industrial applications. Journal of Manufacturing Systems, 58, 346-361. https://doi.org/10.1016/j.jmsy.2020.06.017

2.       Gunasegaram, D. R., Murphy, A. B., Matthews, M. J., & DebRoy, T. (2021). The case for digital twins in metal additive manufacturing. Journal of Physics: Materials, 4(4), 040401. https://iopscience.iop.org/article/10.1088/2515-7639/ac09fb

3.       Mukherjee, T., & DebRoy, T. (2019). A digital twin for rapid qualification of 3D printed metallic components. Applied Materials Today, 14, 59-65. https://doi.org/10.1016/j.apmt.2018.11.003

4.       Kim, F. H., & Moylan, S. P. (2018, May). NIST advanced manufacturing series 100-16: Literature review of metal additive manufacturing defects. U.S. Department of Commerce. https://doi.org/10.6028/NIST.AMS.100-16

5.       Roehling, T. T., Shi, R., Khairallah, S. A., Roehling, J. D., Guss, G. M., McKeown, J. T., & Matthews, M. J. (2020). Controlling grain nucleation and morphology by laser beam shaping in metal additive manufacturing. Materials & Design, 195, 109071. https://doi.org/10.1016/j.matdes.2020.109071

6.       Kim, J., Ji, S., Yun, Y. S., & Yeo, J. S. (2018). A review: melt pool analysis for selective laser melting with continuous wave and pulse width modulated lasers. Applied Science and Convergence Technology, 27(6), 113-119. https://doi.org/10.5757/ASCT.2018.27.6.113

7.       Gunasegaram, D. R., Murphy, A. B., Barnard, A., DebRoy, T., Matthews, M. J., Ladani, L., & Gu, D. (2021). Towards developing multiscale-multiphysics models and their surrogates for digital twins of metal additive manufacturing. Additive Manufacturing, 102089. https://doi.org/10.1016/j.addma.2021.102089

8.       Yadav, P., Rigo, O., Arvieu, C., Le Guen, E., & Lacoste, E. (2020). In situ monitoring systems of the SLM process: On the need to develop machine learning models for data processing. Crystals, 10(6), 524. https://doi.org/10.3390/cryst10060524

9.       Adnan, M., Lu, Y., Jones, A., Cheng, F.-T., & Yeung, H. (2020). A new architectural approach to monitoring and controlling AM processes. Applied Sciences, 10(18), 6616. https://doi.org/10.3390/app10186616

10.   Liu, C., Le Roux, L., Ji, Z., Kerfriden, P., Lacan, F., & Bigot, S. (2020). Machine Learning-enabled feedback loops for metal powder bed fusion additive manufacturing. Procedia Computer Science, 176, 2586-2595. https://doi.org/10.1016/j.procs.2020.09.314

11.   Vandone, A., Baraldo, S., & Valente, A. (2018). Multisensor data fusion for additive manufacturing process control. IEEE Robotics and Automation Letters, 3(4), 3279-3284. https://doi.org/10.1109/LRA.2018.2851792

12.   Zhu, Q., Liu, Z., & Yan, J. (2021). Machine learning for metal additive manufacturing: predicting temperature and melt pool fluid dynamics using physics-informed neural networks. Computational Mechanics, 1-17. https://doi.org/10.1007/s00466-020-01952-9

13.   McCann, R., Obeidi, M. A., Hughes, C., McCarthy, É., Egan, D. S., Vijayaraghavan, R. K., Joshi, A. M., Acinas Garzon, V., Dowling, D. P., McNally, P. J., & Brabazon, D. (2021). In-situ sensing, process monitoring and machine control in Laser Powder Bed Fusion: A review. Additive Manufacturing, 102058. https://doi.org/10.1016/j.addma.2021.102058

14.   Du, Y., Mukherjee, T., & DebRoy, T. (2021). Physics-informed machine learning and mechanistic modeling of additive manufacturing to reduce defects. Applied Materials Today, 24, 101123. https://doi.org/10.1016/j.apmt.2021.101123

15.   Pandita, P., Ghosh, S., Gupta, V. K., Meshkov, A., & Wang, L. (2021). Application of deep transfer learning and uncertainty quantification for process identification in powder bed fusion. ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part B: Mechanical Engineering, 8(1), 011106. https://doi.org/10.1115/1.4051748

 

KEYWORDS: Additive Manufacturing; Digital Twin; In-situ Monitoring; Closed-Loop Feedback Control; Artificial Intelligence/Machine Learning; Reduced-Order Modeling

---