Industrial Agentic AI and generative modeling in complex systems

Abstract

Manufacturing, consumer, transportation, and supply chain processes present significant challenges in monitoring, control, and design due to their inherently nonlinear nature and the difficulty of measuring critical variables in real time. The convergence of major innovations from the computer science field has the potential to revolutionize the engineering and control of complex industrial systems. Digital twinning and process simulation have been a staple of computers in process engineering for decades now. However, the advent of advanced sensor systems and big data integration, combined with generative AI and agentified AI (classic and quantum) systems, allows for much more granular and autonomous process control and real-time optimization of complex systems. Advanced process modeling, Agentic AI, and generative AI models have emerged as powerful tools to address the challenges of complex nonlinear systems. We propose here an integrated systems feedback and control architecture (SIC: Sense, Infer, Control) that leverages complementary process knowledge for enhanced real-time monitoring and decision-making, fully integrated into control system functions and the accompanying sensors. In this paper, we explore this integration of generative models in agentic AI ensembles into industrial processes through the lens of four recent industrial case studies: (1) the real-time optimization of motorsports strategy, (2) the development of indirect (soft) sensors for sustainable large-scale manufacturing operations, (3) the creation of sensor data-driven personalized health and cosmetic chemical formulations, and (4) the design of biomanufacturing systems using quantum and classic Agentic AI. These examples demonstrate how agentic and generative models, combined with full-scale process simulation and digital twinning, effectively augment process control, enabling advanced solutions for process optimization, quality improvement, and sustainable operations. The proposed SIC systems architecture serves to enhance process control automation by capturing complex nonlinear patterns and leveraging easily measurable variables. Generative models bridge gaps in process understanding, sensor technologies, control, and monitoring, offering actionable insights for efficient and informed decision-making across diverse industrial applications.