SBIR-STTR Award

Additive manufacturing (AM) will reduce the delay times required in producing Naval parts that are no longer stocked. However, rapid qualification of parts is still a challenge when a limited number of components are required. To fully exploit the potential advantages of AM, a means of accurately addressing the reliability of AM components is required. By simulating the entire design-build-operation lifecycle of a part (e.g. design optimization, AM build process, operation, and repair/replacement) and enabling the outputs and their uncertainties of prior analyses to be propagated as inputs to later analyses, the entire probabilistic lifetime reliability of an AM component can be predicted. Phase I of this proposal will focus on the development of ICME tools effected through machine learning predictive capabilities trained on the high fidelity data included in the Senvol Indexes. The target application under consideration is to qualify flight critical stainless steel (316L) aerospace parts, where fatigue is the primary mode of failure. Data-driven techniques will be developed to predict part quality and performance based on input process parameters and to quantify allowable bounds on those parameters. Components will be fabricated and fatigue tested to validate the analysis and identify additional data to reduce performance uncertainties.

Benefit:
Additive manufacturing (AM) is quickly maturing as a mainstream production method. AM will enable the engineering of highly complex, design-optimized components, where the production can be driven by the design, and not the other way around. Applications span defense, aerospace, automotive, marine, oil & gas, and manufacturing industries, for example. However, in order to fully exploit the benefits and possibilities of AM, the processes and challenges associated with the new opportunities need to be fully understood, and the reliability of the components included as part of the design process. This work presents the groundbreaking opportunity for a methodology to rapidly qualify AM parts, enabling their widespread industrial use.

Keywords:
ICME, ICME, Probabilistic Lifetime Reliability, additive manufacturing, Predictive Data Analytics, sensitivity analysis, fatigue life, Uncertainty Quantification, Machine Learning

The objective in this project is to implement and validate a probabilistic qualification framework that will enable additive manufacturing (AM) materials and part qualification through the use of a data-driven predictive model within a statistical framework. Senvol seeks to develop and validate a data-driven ICME probabilistic framework for assisting qualification of AM materials and parts. Phase II focuses on 4 main Thrust Areas: (1) Validation of the ICME probabilistic framework, (2) Extension quantification capability, (3) Data collection protocol development, and (4) Software improvement. The objective of the Phase II Base is to validate the probabilistic framework through a rigorous experimental program on a representative structural component, and implement a capability that would quantify the accuracy of extending previously trained ICME predictive models for use in predicting the ICME relationships of a new dataset. In the Phase II Option, the project team plans on demonstrating and validating an approach for using the predictive model, capabilities, and data collection protocol for the case where a new dataset is no longer accurately described by a previously trained model. This demonstration will show how to gather additional test data to re-qualify the updated process.

Benefit:
The technology that is being developed in this STTR is a data-driven machine learning algorithm with key capabilities that will enable a user to reduce the time, cost and resources required to characterize additive manufacturing (AM) materials and AM processes. This technology will be productized as an ICME software tool and will be commercialized by Senvol. The overall intention is for the algorithm to be AM material, machine and process agonistic and for the algorithm to learn from previous data sets and apply those learnings to new data sets. The proposed ICME tool can be used to assist the Navy in developing statistically substantiated material properties (in lieu of relying on conventional material characterization methods that could be used to develop design allowables). In addition to uses within the Navy, it is expected that the probabilistic framework will be used by the AM industry at large, and in particular by aerospace and defense contractors and the Navys supply chain at large.

Keywords:
additive manufacturing, fatigue life, sensitivity analysis, Probabilistic Lifetime Reliability, ICME, Uncertainty Quantification, Machine Learning, Predictive Data Analysis