RII Track-4:NSF: Automated Design and Innovation of Chemical Production Processes with Intelligent Computing

Conceptual process design plays a critical role toward creating innovative chemical plants to address the outstanding challenges of energy and sustainability. Computer-aided methods are essential to rapidly screen the optimal process design among a plethora of existing technologies or even to discover new ones outside the box of current industrial practice.

However, their potential is yet to be fully exploited. Toward this direction, the vision of this project is to drive systematic innovation of chemical process designs by augmenting physical laws, artificial intelligence (AI), and quantum computing (QC). We aim to develop a novel phenomena-based process synthesis approach which opens the opportunity to re-invent unit operations leading to breakthrough process performances, while coupled with quantum machine learning to intelligently learn the path for design improvements.

The resulting methodology will be unique with the capacity to expedite the development of next-generation chemical and energy process technologies by incorporating advanced scientific computing while significantly saving human efforts. The methods and skills developed by the PI and graduate trainee, in collaboration with Cornell AI for Science Institute, will greatly strengthen the research capacity in West Virginia University (WVU) at the forefront of advanced scientific computing.

The project deliveries will also be incorporated to curriculum courses, workshops, and online learning modules tailored for the training of diverse undergraduate and graduate students.

This Research Infrastructure Improvement Track-4 EPSCoR Research Fellows (RII Track-4) project will provide a fellowship to an Assistant Professor and training for a graduate student at West Virginia University. This work would be conducted in collaboration with researchers at Cornell University. The project will develop a computer-aided approach to systematically generate novel, optimal, and sustainable chemical process designs.

It builds on a generalized chemical process representation using physicochemical phenomena which allows to synthesize process designs, conventional or intensified, by optimizing the fundamental mass and heat transfer toward thermodynamic limits. The bottom-up process synthesis using phenomenological building blocks serves as a departure from traditional unit operation-based design which may hinder the generation of creative process solutions.

Artificial intelligence and quantum computing-assisted algorithms will be developed to achieve process design, optimization, and innovation by synergizing: (i) Reinforcement learning to smartly search the process design space characterized by physics-based attainable region, (ii) Autoencoder neural network to develop a quantitative understanding on the feasible and infeasible process design space, (iii) Quantum reinforcement learning to accelerate the speed of design discovery. Thus, the methodology developed from this project will automatedly identify optimal (and potentially out-of-the-box) design solutions with substantially improved process performances, which typically rely on engineering expertise and efforts.

The application showcase will be used to increase the economic competitiveness of sustainable hydrogen production from centralized or distributed natural gas utilization. This project will transform the PI’s individual career by sparking the first-time collaboration with Cornell, opening new research directions on the QC frontier, obtaining formal training on AI and QC, and accessing state-of-the-art cloud computing platforms.

WVU and Cornell will jointly develop learning materials, conference presentations, journal papers, and competitive proposals as continuation of this project.

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.