CAREER: Digital Intelligence Architectures for Efficient and Secure Operation of Integrated Polygeneration Systems

Chemical manufacturing produces many important everyday products like fuels, fertilizers, medicines, plastics, and advanced materials, but it also uses a lot of energy and complex systems, making it vulnerable to supply problems, cyberattacks, and market changes. This CAREER project will focus on improving how chemical and energy systems work so they are safer, more reliable, and more efficient. It will study systems that can make multiple products at once—such as ammonia, hydrogen, and methanol—which are efficient but difficult to manage because different parts use separate computer tools and cannot easily share data. To address this, the project will develop new digital methods that allow these systems to learn and coordinate with each other without sharing all their data in one place, helping protect sensitive information and respond better to disruptions. Overall, the project will support national needs by strengthening advanced manufacturing, improving energy security, and making U.S. industries more competitive, while also helping train future engineers through hands-on learning, student projects, and outreach programs. This CAREER project will develop a secure and adaptable computational framework for distributed polygeneration systems by integrating adaptive federated learning (AFL), blockchain-based coordination, and multi-agent estimation and control. The research will address three key technical challenges in distributed chemical–energy networks: incompatibility among heterogeneous modeling environments, limited adaptability in coordinating distributed subsystems, and insecure or unreliable communication during model synchronization and decision-making. To overcome these challenges, the project will create a unified, multiscale surrogate modeling framework that combines first-principles and data-driven approaches to represent diverse subsystems in an interoperable and computationally efficient form suitable for real-time applications. It will also establish a bi-level, multi-agent architecture for distributed state estimation and coordinated control, enabling local autonomy while preserving system-wide coordination. Within this framework, AFL will support distributed model training and collaboration without requiring the exchange of raw data, thereby enhancing data privacy. A blockchain layer will provide secure, tamper-resistant aggregation of model updates, verified synchronization among agents, and transparent data provenance. In addition, transformer-based techniques for cyberattack detection and federated state reconstruction will be integrated into the control architecture to improve resilience against corrupted measurements and communication attacks. The overall framework will be validated using a benchmark polygeneration system to evaluate its scalability, interoperability, cybersecurity, and operational performance. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2543146 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Davood Pourkargar | Institution: Kansas State University, MANHATTAN, KS | Award Amount: $524,960 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543146 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543146.html