A recent perspective article published in Engineering delves into the application of artificial intelligence (AI) in process manufacturing (PM), exploring how AI can be integrated with process systems engineering (PSE) methods and tools to address various challenges in the field.
PM, a crucial activity in chemical, biochemical, and related engineering, involves converting raw materials into products. However, it faces numerous complex problems, such as continuous or batch operations, quality control, and safety hazards. AI, with its ability to provide innovative solutions, has gained significant attention. The paper focuses on the concept of hybrid AI, which combines machine learning (ML) methods with first-principles-based methods of symbolic AI, to create more powerful tools for PSE.
The authors first define four key topics within PM: chemical product design, process synthesis and design, process control and monitoring, and process safety and hazards. They then review the current state of AI applications in these areas. In chemical product design, AI is used in computer-aided molecular or mixture design, with advancements in molecular structure representation and property prediction. For process synthesis and design, hybrid AI approaches are being developed to find optimal processing routes and designs, considering sustainability and other criteria. In process control and monitoring, techniques like neural network modeling and reinforcement learning (RL) are being employed, although challenges such as system safety and stability remain. Regarding process safety and hazards, AI can help in reducing the time and effort of process hazards analysis and identifying potential risks.
Looking ahead, the paper outlines several challenges and opportunities. For chemical product design, better utilization of chemical libraries, more efficient computational algorithms, and improved handling of complexity with hybrid AI are needed. In process synthesis and design, a unified database of process flowsheets, integrating sustainability into flowsheet development, and enhancing the integration of optimization-based methods with hybrid AI are crucial. For process control and monitoring, adapting to changing operational conditions, handling limited feedback signals, incorporating diverse measurement signals, and implementing AI-augmented control algorithms are key areas of focus. In process safety and hazards, creating a database of dangerous chemicals, developing better language models, and integrating hazardous and safety issues more effectively are essential.
While AI has shown promise in PM, there is still much work to be done. Developing AI-augmented PSE tools that can efficiently transfer data to model-based process simulation and optimization techniques is necessary for failure-free decision-making in PM. This research provides valuable insights for engineers and researchers working in the field, guiding future efforts to leverage AI for more sustainable and efficient process manufacturing.
The paper "A Perspective on Artificial Intelligence for Process Manufacturing," authored by Vipul Mann, Jingyi Lu, Venkat Venkatasubramanian, Rafiqul Gani. Full text of the open access paper: https://doi.org/10.1016/j.eng.2025.01.014
