Thermal processing operations including drying, cooking, and pasteurization rely on heat (mostly obtained by burning fossil fuels) and fluid motion to raise the temperature and reduce the moisture content of food. These processing operations are widely employed for food preservation and sterilization, directly affecting the food supply chain. Thermal processing operations are also infamously known for their low energy efficiency (defined as the minimum theoretical energy required divided by the actual energy consumed) ranging from 10% to 60% in case of drying. A 10% improvement in efficiency for process heating in the food and beverage industry would reduce carbon emission by 1.1 million tons, equivalent to 2.7 billion miles driven by a gasoline-powered car annually. It has been shown in earlier studies that both yield and energy efficiency can be substantially improved via control and optimization of processing conditions. For example, the throughput or production rate will be set as high as possible while balancing the need to reduce energy loss (e.g., by lowering air temperature) and satisfying food quality and safety constraints. This project will take the next step toward the optimal control of thermal food processing -development, validation, and implementation of the approach to maintain optimality in daily operations under the inevitable influences from strong process variability.A thermal food processing operation is subject to many variations. First, there continues to be a need for scale-up, equipment changes, and food-stock switchover. Although these discrete variations are directly measurable and quantifiable, any of such changes renders existing control parameters including set processing conditions completely irrelevant. Abrand-new model for process control has to be built from the ground up requiring extensive experiments and causing a serious disruption to regular production.Therefore, amodular and physics-based modeling approachrequiring minimal production disruptions is highly desirable.Second, there are the unmeasurable disturbances and variations, such as changes in moisture contents of incoming food stocks, unmonitored ambient conditions, and aging of equipment. Unlike in the case of the discrete and measurable variations, minor tuning of the control parameters would suffice to maintain the optimal operation. This would be a straightforward task if direct feedback of the product state were available. However, it is almost impossible to acquire real-time values of the moisture content and other qualities in food, because installation of a physical sensor in each individual productunder the harsh processing environment is impractical. Therefore, anew on-line approach based on integration of a high-fidelity process model with indirect feedbackis required to accurately estimate the product state.The research objective of this project is to build and evaluate a novel cyber-physical system (CPS) for real-time management of the process variability from both measurable and unmeasurable sources in optimal control of thermal food processing.The first-principle models,representing the complex spatio-temporal dynamics of thermal food processing,can potentially help achieving such a goal. The model parameters of first-principle models, such as the thermal conductivity of food, have physical meanings. Therefore, the food model, once developed in a lab setting without causing any production disruptions, will remain valid regardless of changes in processing scale and equipment.This opens the way to modular structure built on separate models of food and equipment, so a cumulative model can be deployed rapidly via plug-and-play coupling of pre-built components. Moreover, high-fidelity simulation of thermal food processing can be accomplished via Computational Fluid Dynamics (CFD) analysis. However, the high computational cost of CFD models renders their use for any time-critical applications intractable. A key task and the major novelty of the proposed work will be overcoming this obstacle via employment of aReduced-Order Modeling (ROM)approach, in which a high-order CFD model is replaced with a lower-order model, thereby drastically lowering the computational cost with little loss in accuracy. Taking advantage of the computational efficiency of the developed ROM,estimation of unmeasurable variationsand on-line optimization of processing conditions will be implemented. The proposed CPS for smart food processing will be implemented and validated usingthe pilot-scale equipment atthe Michigan State University (MSU)for continuous hot-air drying of food as a test case. The developed procedures and algorithm from this project are expected to be broadly applicable to a wide range of thermal processing operations for food and agricultural products. Improved yield and energy efficiency of thermal food processingwill contribute to societal benefit through innovative and sustainable food production.