AI-Enabled Process Optimization & Control
Learning how a process behaves, and using that to control quality, yield, and throughput
What it means
Modeling approaches that transform data into decisions that improve quality, reduce variation, accelerate process and material understanding, and support control.
Why it matters
Many inspection systems still rely on manual review, fixed thresholds, or generic tools that miss subtle defects in complex production data, so deviations propagate across stations before they are caught and become costly to correct. High-fidelity simulations are often too slow to guide real-time optimization, leaving manual adjustments that are inefficient and non-optimal.
What this research enables
Value to Industry
- Defects caught earlier and less variation
- Faster quality review and shorter cycle times
- Higher yield and throughput on processes that are difficult to control
In-Situ Monitoring and Anomaly Detection
Detecting defects in high-dimensional sensor and image data streams
Operational Gap
Many inspection systems still depend on manual review, fixed thresholds, or generic tools that miss subtle defects in complex production data
Core Innovation
Real-time detection and classification of subtle, high-dimensional quality patterns that are difficult to define in advance or capture with rule-based inspection
Value to Industry
Earlier defect detection, lower warranty and scrap risk, faster quality review, and scalable inspection across products and lines
Battery Coating process Monitoring
In-situ monitoring of coating processes in battery manufacturing, detecting defects in high-dimensional sensor and image data streams.
Faculty: Yao Xie
LG ElectronicsVehicle End-of-Line Anomaly Detection
Real-time detection and classification of subtle, high-dimensional quality patterns that rule-based inspection cannot capture. Delivers earlier defect detection, lower warranty and scrap risk, and scalable inspection across products and lines.
Faculty: Kamran Paynabar
FordUltraprecision Polishing for Controlled Nuclear Fusion
Faculty: Yu Ding
Lawrence Livermore National LabQuality Feedback Control
Closing the loop from quality sensing to corrective process action
Operational Gap
Quality deviations can propagate across stations before they are detected, making them costly to correct later
Core Innovation
Connecting in-process quality sensing to predictive models and corrective control, enabling tooling, actuator, and process adjustments while production is still underway.
Value to Industry
Reduced dimensional variation, improved precision assembly, better root-cause diagnosis, shorter cycle time, and less manual trial-and-error
Boeing 787 fuselage shape control
Closing the loop from quality sensing to corrective process action, enabling tooling, actuator, and process adjustments while production is underway. Delivers reduced dimensional variation and improved precision assembly.
Faculty: Jianjun Shi
BoeingMultistage Assembly Variation Compensation
Compensating dimensional variation across multistage assembly, with applications in automotive and shipbuilding.
Faculty: Jianjun Shi
automotive and shipbuildingManufacturing Digital Twins
Turning complex physics into fast and faithful models for process optimization
Operational Gap
High-fidelity physics models capture complex manufacturing behavior, but they are often too slow for real-time process optimization; thus, process tuning still relies heavily on manual adjustments
Core Innovation
“Fast and Faithful” data-calibrated digital twins that connect physics-based simulation, sensor data, prediction, and control for real-time decisions.
Value to Industry
Faster process tuning, reduced trial-and-error, improved quality control, and practical real-time optimization. These models help manufacturers make better engineering/production decisions without waiting for slow simulations.
Adaptive model order reduction for engineering processes
"Fast and faithful" surrogate models and data-calibrated digital twins that link simulation, prediction, and control for engineering processes.
Faculty: Xiao Liu
U.S. National Science Foundation (NSF)Boeing Fuselage Assembly Digital Twin
A data-calibrated digital twin of fuselage assembly supporting faster engineering decisions and reduced trial and error.
Faculty: Jianjun Shi
BoeingAdditive Manufacturing Qualification and Control
Using in-situ sensing, digital twins, and AI to improve 3D/4D printing
Operational Gap
Current practice relies on heterogeneous sensing, post-build inspection, and limited control, making qualification and shape accuracy difficult to guarantee.
Core Innovation
AI-enabled fusion of in-situ sensing, digital twins, and physics-informed models to detect anomalies, predict functionality, and optimize 3D/4D printing processes, in real-time.
Value to Industry
Improves confidence in printed part functionality, enables quality prediction and corrective control, and supports broader adoption of complex 3D/4D printing and composite manufacturing technologies.
AI-enabled NDI and durable composite/additive structures
Faculty: Chuck Zhang
U.S. National Science Foundation (NSF) IUCRC-CHMIFunctional Qualification in additive manufacturing
Faculty: Jianjun Shi
U.S. National Science Foundation (NSF)AI-Driven Materials Characterization
Turning microstructure images into quantitative material and process understanding
Operational Gap
Manual and generic image tools struggle to quantify manufacturing-specific morphology, texture, orientation, dispersion, and dynamic evolution
Core Innovation
Methods that transform microscopy and microstructure data into measurable material and process features for classification, anomaly detection, and process insights
Value to Industry
Faster material assessment, consistent microstructure characterization, and stronger quality control in advanced manufacturing
Material Image Analysis for Automated Manufacturing
Transforming microscopy and microstructure data into measurable material and process features for classification, anomaly detection, and process insight.
Faculty: Yu Ding
U.S. National Science Foundation (NSF) Air Force Office of Scientific Research (AFOSR)Microstructure Characterization in Additive Manufacturing
Quantifying manufacturing-specific morphology, texture, orientation, and dispersion from microstructure imagery in additive processes.
Faculty: Kamran Paynabar
NovelisWork With Us
Have a Problem in One of These Areas?
Every thrust in this portfolio began with an operational gap a partner needed closed. Tell us about yours.