Scientific, technical, and societal advances are increasingly dependent on new insights, theories, and tools to exploit data effectively for the timely delivery of relevant and accurate information and for knowledge discovery. For the purpose of effective and efficient learning from the data to improve operational safety, manufacturing efficiency, energy efficiency, and sustainability in manufacturing, our team aims to explore machine learning (ML) and artificial intelligence (AI) for improved information extraction, pattern recognition, and decision making towards smart, data science-enhanced manufacturing.

Our team targets developing applicable and generalizable ML and AI techniques suitable for manufacturing data analytics that is featured by the large volume, high dimensionality, heterogeneity, non-linearity, and uncertainty. Current research thrusts include:

• Integrating ML models (e.g., structures, training loss) with domain knowledge to improve model credibility and generalizability;
• Self-supervised learning from big, unlabeled plant data and unsupervised continual ML model updates from continuous data streaming for machine and process monitoring on the shop floor;
• In-situ process monitoring (i.e., defect detection and quality prediction) and real-time control of additive manufacturing processes;
• Robotic automation of welding processes, by endowing robots with advanced perception, incremental learning, and critical thinking;
• Development of cost-effective edge devices, communication protocols, semantic indexing for advanced data management, processing, and learning for building digitalized manufacturing factories;
• Digital thread in life cycle analysis for improved product design and supply chain management.