TECHNOLOGY AREA(S): Air Platform, Materials/Processes

ACQUISITION PROGRAM: JSF Joint Strike Fighter

OBJECTIVE: Develop an advanced machine learning (ML) tool capable of optimizing process parameters for metal laser powder-based additively manufactured components to achieve enhanced fatigue performance for aircraft components.

DESCRIPTION: Laser powder bed and powder feed additive manufacturing (AM) technologies have proven to produce complicated parts from high-performance alloys such as titanium, Inconel, and tool steel [Ref 1]. Many processes are currently able to consistently produce intricate geometries and meet standard geometric tolerances. However, achieving predictable part performance, including static (e.g., strength) and dynamic (e.g., high cycle and low cycle fatigue) behaviors remains a significant challenge. In order to attain satisfactory part performance, pre- and post-processing parameters are tuned using expensive trial and error approaches. Perhaps the use of various sensors integrated with simulation and modeling tools that leverage data analytics, data fusion, and machine learning (ML) techniques may improve fatigue performance of AM parts, potentially without any post-processing required.

Due to the multi-scale and multi-physics phenomena associated with processing and post processing of various metallic alloys, it is necessary to adopt an integrated computational materials engineering (ICME) framework [Ref 2] for efficient linkages between processing and performance. Furthermore, the addition of ML methodologies and data-fusion methods should provide increased throughput and fidelity in linking AM process parameters to fatigue performance under various loading conditions.

ML involves the scientific application of computational models that can predict systemic performance using data representing input-output tuples encapsulated in a computer system [Refs 3-6]. A crucial element of ML is that the performance of a ML system can progressively use available sensor data to improve a specific system performance prediction. Thus, this topic seeks novel ML methodologies and techniques for laser powder-based AM processes (e.g., laser powder bed fusion, direct energy deposition) to yield desired aircraft part fatigue performance.

PHASE I: Develop an initial computational concept for a ML ICME-based toolset for a laser powder metal AM process under the assumption of in-situ and/or ex-situ sensor data to link AM process parameters and/or state variables to the fatigue performance of the part. Ensure that the concept methodology demonstrates both its ability for sensor fusion and its ability to learn from trial runs to predict the final part geometry, associated material properties, and final part performance. Demonstrate the feasibility of the methodology using actual AM coupons, testing (e.g., ASTM E466, ASTM E606) [Refs 7, 8], and analyses for a single material. The computational prototype of the proposed advanced ML ICME tool should have the potential for development into a full-scale ML ICME system for integrating with AM machines to enable designer to optimize fatigue life in Phase II. The Phase I effort will include prototype plans to be developed under Phase II.

PHASE II: Fully develop, verify, and validate a prototype ML system for a laser powder-based metal AM process to perform geometry control and material property control during AM processing. Demonstrate its ability to manufacture aircraft components with complex geometry and tailored performance using additional metal alloys.

PHASE III DUAL USE APPLICATIONS: Further develop and refine an advanced ML ICME system for various powder-based AM processes to fabricate specific naval aircraft components for integration into the Fleet. Conduct final component-level testing to demonstrate the geometry and material property control of AM components meeting the Navy�s specifications.

The process will be directly applicable to a wide range of AM process applications due to the high amount of anticipated AM part usage in the commercial/private aerospace industry. The proposed toolset will allow the aerospace industry to apply the benefits of AM technology to many critical aircraft components.

REFERENCES:

1. Frazier, W.E. �Metal Additive Manufacturing: A Review�. Journal of Materials Engineering and Performance, 23 (6), 2014, pp. 1917-1928. DOI: 10.1007/s11665-014-0958-z. https://link.springer.com/article/10.1007/s11665-014-0958-z

2. "Integrated Computational Materials Engineering: A Transformational Discipline for Improved Competitiveness and National Security." National Research Council, The National Academies Press, Washington, DC, 2008. DOI: 10.17226/12199

3. Russell, S. and Norvig, P. �Artificial Intelligence: A Modern Approach.� Prentice Hall: Upper Saddle River, NJ, 2009. ISBN-10: 0136042597. https://www.cin.ufpe.br/~tfl2/artificial-intelligence-modern-approach.9780131038059.25368.pdf

4. Vandone, A., Baraldo, S., and Valente, A. �Multisensor Data Fusion for Additive Manufacturing Process Control.� IEEE Robotics and Automation Letters, 3 (4), 32018, pp. 279-3284. DOI: 10.1109/LRA.2018.2851792

5. Zhu, Z., Anwar, N., Huang, Q., and Mathieu, L. �Machine learning in tolerancing for additive manufacturing.� CIRP Annuals, 67 (1), 2018, pp. 157-160. DOI: 10.1016/j.cirp.2018.04.119

6. Garanger, K., Feron, E., Garoche, P., Rimoli, J. J., Berrigan, J. D., Grover, M., and Hobbs, K. �Foundations of Intelligent Additive Manufacturing.�� https://arxiv.org/pdf/1705.00960.pdf

7. ASTM E466 � 15 Standard Practice for Conducting Force Controlled Constant Amplitude Axial Fatigue Tests of Metallic Materials
https://www.astm.org/Standards/E466.htm

8. ASTM E606 / E606M � 12 Standard Test Method for Strain-Controlled Fatigue Testing� https://www.astm.org/Standards/E606

KEYWORDS: Machine Learning; ML; Sensor Fusion, Fatigue; Metal Additive Manufacturing; AM; Laser Powder Bed Fusion; Powder Feed; ICME; Material Property Control