1

Digital Twins in Manufacturing: Collaborating to Advance the Field��Dawn Tilbury, University of Michigan�Wenlong Zhang, Arizona State University

2

NSF IUCRC

Center for Digital Twins in Manufacturing (CDTM)

What is a Digital Twin?

Periodic Table of �Digital Twins

​

​

​

​

3

National Academies: Digital Twins for �Science, Engineering, and Medicine

250 “classifications” of digital twins

• Across lifecycle, hierarchy, �use cases

Digital Twin Definition used in the Center

CDTM proposes the following definition, �based on existing standards and prior research

​

A Digital Twin is:

• Synchronized with its physical counterpart while existing in the cyber world
• Able to provide an estimate or prediction of its physical twin, using models together with real-time data
• Impactful in its ability to positively influence its environment

4

“A Requirements Driven Digital Twin Framework: Specification and Opportunities,” J. Moyne, Y. Qamsane, E. C. Balta, I. Kovalenko, J. Faris, K. Barton, and D. M. Tilbury, IEEE Access, 8:107781–107801, June 2020.

​

Physical “thing”

Real-time Data

Digital Twin

Models & Analytics

Metrics

Confidence

Existing uses of DTs in Manufacturing

• Predictive maintenance, reducing unscheduled downtime
• Virtual metrology to predict quality, increase yield, reduce cycle time and reduce scrap
• Optimize supply chain for reduced cost
• Facilitate reconfiguration, enable “Design for Manufacturability” (DFM): reducing time and cost, increasing quality

​

5

Cost

Time

Scrap

Downtime

Variability

Quality

Throughput

Revenue

Profit

While many DT solutions are being used today,

a DT framework must accommodate existing solutions

Center Goal: Facilitate a framework that leverages existing DT solutions, while building next generation DTs

Everybody is working on Digital Twins!

US Government efforts:

• ManufacturingUSA Institutes
• ARM
• CESMII
• LIFT
• MxD
• NIST

​

​

​

​

Industry-specific efforts:

• SEMI standards
• USCAR data standards

​

General industry efforts:

• Digital Twin Consortium
• Many companies working on DTs

6

Our Center focus:

• Basic research
• DT frameworks
• Interoperability
• TRL (1-2)

​

​

Center Goals: 3 Thrust Areas

7

Fundamental Research Challenges and �Example Research Questions

• Develop structures, frameworks and tools for Digital Twins
• Provide reusable, extensible, interoperable, and maintainable solutions
• How can the outputs (estimates or predictions) of multiple Digital Twins, representing related components, be aggregated to give an estimate or prediction of the overall subsystem? Or subsystem?
• Once a Digital Twin has been developed for a component, how can it be re-used for a similar but different component? What are the engineering trade-offs to consider regarding new data collection, model training, vs. estimation/prediction accuracy?
• Many more… topics to be considered based on industry interest

​

8

Center Timeline

2022

May: Hybrid workshop

2023

Sept.: Receive planning grant

2024

Jan.: Planning workshop

June: Submit full proposal

2025

June: NSF award

July: Info workshop

September: Bylaws approved by NSF

October: Project planning workshop�December 4: Met w/ Members to select topics to begin in Jan.

2026

January: Begin topic explorations

March: Kick-Off (First IAB meeting)

​

​

May: Projects Start

Fall: Second IAB meeting

Select projects via member votes

Successful Projects with Industry

Digital Twin and Open Automation for Process Manufacturing

Requirements Framework for Full-Stack Virtual Commissioning

A Digital Twin Framework for Health State Estimation in Rotating Equipment

Predictive Health Monitoring of Commercial 3D Printers Leveraging Digital Twin Technology

Real-time Manufacturing Machine and System Monitoring using Industrial Internet of Things (IIoT)

DTFab for Simulation Generation and Control

Digital Twins for In-space Robotic Servicing and Manufacturing

… and many more

10

Testbeds for Digital Twins

11

SMART 4.0 Testbed with Additive and Subtractive Processing, Collaborative and Mobile Robots, Rotating Equipment Predictive Maintenance, and an Integrated System-level Data Collection and Controller through a partnership with Applied Dynamics International

DT Testbeds for Robots, Humans, Manufacturing Processes: partnership with FANUC & Rockwell Automation on data-driven DTs of manufacturing systems

video of U-M SMART testbed

Why become a member?

• Improve YOUR manufacturing bottom line with cutting-edge research outcomes
• Enhance YOUR competitive position over others in your domain
• Drive down YOUR R&D risks w/ ~10x leveraging
• Meet YOUR specific needs and influence technology via project selection and participation
• Collaborate in YOUR field with suppliers, customers and other leaders in a pre-competitive, technical environment
• Recruit outstanding talent --- well-trained and ready to contribute

​

12

Member Engagement

• Vote on projects
• Participate in as many projects as desired, sending as many employees as helpful to enhance their skills
• Receive monthly research updates
• Attend semi-annual advisory board meetings
• Guide research, provide insight on challenges
• OPTIONAL:
• Industry partners may share data (with appropriate agreements)
• Industry partners may co-author publications, or patents, as appropriate

13

Center Budget Model

• Every member’s contribution of $90K/year is leveraged more than 9x
• Each project is $120K/year: Graduate student stipend, benefits, tuition, faculty time, etc.
• Indirect/overhead cost on industry memberships is limited to 10%
• Industry funding goes to PROJECTS. NSF and Univ. cover Admin expenses.
• Affiliate membership for small businesses (<500 employees), $45K

​

14

Income: 6 members

Expenses: 4 projects

NSF

Univ.

Admin

Total Center

Budget = $840K/yr

Members have access to ALL project outcomes

Members can participate in any/all projects

2026 Projects

Four projects were selected by the Industry Advisory Board �(IAB) Members. The projects began May 1, 2026, and each project has a monthly progress meeting.

Project 1

Digital Twins (DTs) for Worker Assistance in Flexible Manufacturing

Project 2

Exploring the Use of Foundational Models (FMs) to Enhance Efficiency and Innovation in Design for Manufacturing

Project 3

Developing and Deploying a Defect-detection Digital Twin for Advanced Manufacturing

Project 4

Methodology for Synchronization of Physical and Simulation Models (the Diagnostic Twin)

​

15

2026 Projects

​

Project 1

Digital Twins (DTs) for Worker Assistance in Flexible Manufacturing will develop human digital twins with uncertainty quantification and integrate it with a process digital twin for automated worker assistance.

​

Problem: Humans remain an integral part in manufacturing, but they are not perfect. Robots can help, but need information on how best to work with/help humans.

Opportunity: Human Digital Twin (HDT) is a digital representation of the twinned human, emulating their state and dynamics

Project Focus Points: Human behavior and action prediction; group conventions and individual variance; error detection

​

16

2026 Projects, Continued

​

Project 2

Exploring the Use of Foundational Models (FMs) to Enhance Efficiency and Innovation in Design for Manufacturing will investigate the use of foundational models with more traditional models within a digital twin framework. Data will be integrated across the design for manufacturing pipeline to identify potential failures and key system correlations for enhanced system design.

​

Problem: Manufacturing systems need a unified digital framework that can use historical data, improve prediction, and support system-level process design and evaluation.

Opportunity: Foundation models can connect design intent, process knowledge, and operational data for scalable manufacturing decision support.

Project Focus Points: Training Methodology, Data Requirements, Validation and Evaluation

17

2026 Projects, Continued

Project 3

Developing and Deploying a Defect-detection Digital Twin for Advanced Manufacturing will deliver a physics-informed digital twin that enables inter-layer defect detection for certifiable metal additive manufacturing.

Problem: Manufacturing processes lack robust digital twins capable of detecting, predicting, and ultimately preventing defects.

Opportunity: This project establishes a framework for physics-informed learning with the model updating to develop robust digital twins for manufacturing defect detection and prediction.

Project Focus Points: Sensing observability, Surrogate updating, Demonstration

18

2026 Projects, Continued

Project 4

Methodology for Synchronization of Physical and Simulation Models (the Diagnostic Twin) will demonstrate a methodology for synchronizing a robotic digital twin with its physical counterpart in manufacturing.

Problem: Digital twins inevitably drift from the physical systems they mirror, yet current practice flags every deviation against fixed, hand-set tolerances — unable to separate normal run-to-run variation from meaningful change, or to tell whether the cause is the model, the environment, or a fault.

Opportunity: By streaming live sensor data into a purpose-scoped twin and running parallel counterfactual models, a diagnostic twin can not only detect divergence but explain it — automatically deciding whether to recalibrate the model, reinterpret the environment, or flag a fault — turning digital-twin maintenance from a manual, reactive step into a continuous, self-diagnosing capability for manufacturing.

Project Focus Points: Synchronization Methodology & Metrics, Noise-Floor / Minimum Detectable Deviation, Counterfactual Source Attribution, Time-Aligned Physical–Simulation Logging

19

2026 Members

20

Smart Manufacturing Cell Testbed

21

• RoboDrill (α-D21MiB5 ADV Plus)
• Fanuc Machine Tending Robot
• Payload: 4KG
• Reach: 500mm
• Autonomous Mobile Robots (AMR)
• LiDAR and 3D cameras
• Robotic Inspection Station (machine vision)
• Conveyor belts with vision sensors
• AS/RS
• RFID/Barcode Scanners
• Open API for data acquisition
• MTConnect-enabled for IIoT applications

PI: Farhad Ameri, Arizona State University

Laser-Powder Bed Fusion Testbed

22

PI: Zhengtao Gan, Arizona State University

Multi-material recoater system

In-situ process monitoring

Full-system Aconity MIDI+

Multiphysics-resolved Digital Twin Framework

Process Data

Model-to-Sensor Alignment

Multi-GPU Multiphysics simulation

Design& Control

Aconity MIDI+

Multi-Material L-PBF

• Suitable for metal and ceramic powders
• Multi-Material Recoater
• Powder bed preheating up to 1000oC
• Scanning speed up to 12 m/s
• Open access to nearly all SLM relevant process parameters
• 1200 W nLIGHT laser with beam shaping
• One high-speed pyrometer for one 3D Scanner
• Real-time closed-loop control

Laser-Wire Directed Energy Deposition Testbed

23

PI: Binil Starly, Arizona State University

MELTIO

Open Platform LW-DED

• Steel, Ni, Ti, Cu, Al alloys
• No size limitation for parts
• 5-axis robot for control
• 250mm printing size
• 1200 W laser
• Real-time process control/multi-sensor
• Additive + subtractive add-on possible
• Open platform – custom sensor, printbed conditions

Description and Mission Statement

• Description: The Center for Digital Twins in Manufacturing is a multi-year, multi-university research center governed by an industrial advisory board (IAB) consisting of industry and government members, academic researchers, with funding from the National Science Foundation (NSF), dedicated to research & development in the Digital Twins in Manufacturing space.�

​

• Mission Statement: Provide pre-competitive research outcomes for our manufacturing IAB membership (and the manufacturing industry in general) that increase member productivity and ROI through the advancement of digital twin technology and solutions across the manufacturing ecosystem

24

Questions, want to learn more?

Email

​

cdtm-iucrc@umich.edu

or

Center Industry Liaison: Vicki Waters, watersvg@umich.edu

​

​

​

Thank you!!

25