Data Management and Fusion for AM Industrialization

Summary

The maturation of additive manufacturing (AM) into an industrialization (wide-scale production) technology requires an expanded notion of integration of both AM systems and AM data.  AM data integration and analytics need to scale up as well to automate workflows and improve decision-making across the AM supply chain.

This project focuses on the development of innovative methods, models, standards, tools, and a testbed to enable the use of AM data for industrialization. The proposed work expands the scope of earlier domain-specific efforts on AM information modeling, data registration and data fusion to include production system and supply chain integration. This new AM data infrastructure will allow 1) automated AM data/metadata flows from shop floor to enterprise, 2) seamless integration and management of data in AM supply chain, and 3) streamlined data co-processing to improve AM lifecycle and supply chain decision makings. More specifically, this project will address existing measurement science barriers to AM data interoperability, data security and quality, and develop advanced AM analytics including data registration, data fusion and digital twin.

Research will be conducted through close collaborations with the AM community and the results will provide the foundation for new AM standards development.  Use cases in AM industrialization will be identified and analyzed for requirements engineering in AM data integration, management, and fusion. Common data models and data fusion algorithms will be developed for those use cases. Meanwhile, an AM data integration testbed will be built to emulate AM production environment and evaluate and validate various data models, data integration standards and data fusion methods. Research data used and generated in this process will be published. In addition, software tools will be provided to the AM community to improve the research results transferability and standards adoption.

Description

Objective
To develop methods, models, software tools, open data and best practices for data integration, management, and fusion in additive manufacturing to accelerate AM Industrialization.

Technical Idea
Our technical idea is to enable an integrated, streamlined and effective AM development and supply chain with standardized data models and common exchange formats, advanced simulation and data fusion methods, and best practices in data management. 

Enhanced common data models, metadata models and common data exchange formats can enhance cross-domain AM data interoperability required by AM industrialization. We will expand our current effort on AM common data model development to cover the broader integration needs of AM systems with other manufacturing systems and applications for production.  In addition, a data model for AM simulation and uncertainty quantification will be developed, tested, and validated using the information from various projects, such as the NIST AM Bench. The technical effort includes the development of the data model for simulations and software, a tool for model validation and repeatability analysis, and standards development. 

AM data must always be available in the appropriate quality to maximize its value. The AM data quality will be evaluated toward AM part qualification, including the use of both in-situ data and ex-situ data. ISO 8000 series standards will be the basis for quantitative AM data quality metrics definition. The research also includes the development of a standard procedure to statistically assess the impact of the veracity of the training data on the part quality prediction models or data fusion results, and a software tool that validates the methods and enables the standard procedure. This project will address AM security challenges through leading an ASTM standardization activity to apply NIST’s risk-based security frameworks to AM security guideline development. Leveraging existing security guidance for information and operational technology systems, this AM security guideline will utilize AM process knowledge, AM attack taxonomies, and how those attacks can result in sabotaged parts, technical data theft, and counterfeiting. 

Massive complex data are generated from AM development and deployment, with a multitude of modalities and high dimensions, and at various scales and sampling rates.  That is, information acquired from an individual data source exhibits limitations in AM decision makings. Instead, different data modes offer varying amounts of discriminative information that fused not only plays a key role in advancing the understanding of AM processes but also drive the engineering decision makings in the lifecycle and value chain. Multi-modal, multi-scale data fusion sets a requirement of data registration, a process of transforming different sets of data into one coordinate system. Data Registration involves aligning those datasets temporally and spatially and recording metadata needed for the data alignment. Data registration standards development involves a definition of common reference frames, a standard procedure guiding transformations of data between reference frames and common metadata definitions for data alignment.  

Research Plan
The research project will be conducted in 3 thrusts, each of which includes multiple research tasks to deliver intermediate and final products.

Data Integration for AM Industrialization

This work package starts with an exploration of integration cases of AM systems with other manufacturing functions, for example, Manufacturing Execution Systems (MES), Enterprise Resource Planning (ERP), machine learning and AI engineering for scale-up production. In addition, value chain data exchange scenarios will be studied as well.  Based on the use cases, we will evaluate the capability of the in-progress AM common data models and existing manufacturing models, for example, ISA 95 Level 3 object models, and develop new models through harmonization. An AM data integration testbed will be developed to enable the testing, validation, and demonstration of the data models. This testbed implements a multi-layer AM integration framework, including an AM emulator for in-situ monitoring data generation, an embedded real-time/near real-time control system for in-process monitoring, data streaming and control function testing, an enterprise platform to test the information exchange methods between AM systems and MES, PLM, ERP and supply chain management functions. 

Our data integration for AM industrialization will also cover the use of ICME tools and digital twin technology for process control, part quality improvement and other AM development decision making.  A data model for simulations and uncertainty quantification will be developed, tested, and validated through collaboration with other NIST projects.

The resulting models and tools will be transferred to standard development and deployment.  Established manufacturing standards will be heavily leveraged. Collaborations with the AM industry, especially small and medium enterprises (SME) are the key path.

Data Management for AM Industrialization

Input data quality (e.g., completeness, and accuracy) determines the effectiveness of data-driven AM applications. AM-specific data quality metrics will be defined based on existing data quality measurement methods and standards. With the defined data quality metrics, the impact of AM data quality on digital twin-based AM decision making will be studied respectively.  A set of tools will be developed to quantify the fitness of AM data toward AM digital twin applications. A workshop will be organized on AM data quality management and for research result transfer, for example, AM data quality management guidelines and AM metadata modeling for quality management etc. 

The scope of data management includes AM data security. Since the inherent complexity of AM designs and AM processes render them attractive targets for cyber-attack, risk-based Information Technology (IT) and Operational Technology (OT) security guidance standards are highly demanded by AM industrialization stakeholders. NIST will co-organize a small-scale workshop and collaborate with DoD to collect security requirements on data from the AM industry. Workshop findings will guide research on how to apply NIST security frameworks to adequately address the security, governance, and stewardship of data repositories appropriate for the storage and use of findable, accessible, interoperable, and reusable (FAIR) data.  The research will also support the ongoing effort of developing AM security standards with ASTM F42.08.

AM Data Registration and Fusion

Data registration is the transformation of datasets from various sensors and inspection instruments into a common reference coordinate system for integral processing, also known as data fusion. Existing efforts include the development of an ISO/ASTM standard on AM data registration. The project team will continue leading the working group for this standard. In addition, we will identify and prioritize additional research needs in data registration based on ISO/ASTM input. this next stage of data registration research will include developing an AM data registration software tool for standards conformance validation and standards adoption.  

Our AM data fusion research covers both data-driven modeling using machine learning and ICME for AM digital twin, as well as how data and models can be fused to solve AM problems, for example, process anomaly detection, part quality detection and part certification. We will work with America Makes to target the development of data fusion methods and standards toward the application in AM process/part delta-qualification. 

Major Accomplishments

AM Material Database

 

 

 

 

 

 

 

AM Data Integration Testbed

 

 

 

 

 

 

 

 

AM Data Standards

• AM Common Data Dictionary (ASTM F3940-21)
• AM Data Registration: ISO-ASTM DIS 52953
• AM Security Guideline: ASTM WK 78322
• AM Common Data Model (CDM) (https://github.com/kaggour/AM-CDM)
• AM dataset Metadata Model (ConnectSpec)