Title : Intelligent systems for food processing: Challenges and innovations
Abstract:
The integration of intelligent systems into food processing represents a transformative challenge. The presentation examines the challenges and innovations associated with implementing intelligent technologies such as artificial intelligence (AI), machine learning (ML), and Internet of Things (IoT) in the food processing industry. Key innovations include machine vision systems for real-time quality control, AI-driven predictive maintenance of machinery, and sensor fusion for precise monitoring of critical parameters such as temperature, humidity, and microbial activity. These systems enable automation and decision-making, reducing human error and optimizing resource utilization. The optimization strategy is based on machine learning, which helps to link controlled variables (time, temperature, airflow) to specific quality attributes to establish criteria for optimization. This knowledge is used in the design of an adaptive control system, which allows fine-tuning of processing conditions.
This general approach is illustrated with an application example in drying of medicinal herbs. The concept of intelligent drying is based on identification of critical control points (CCP) and optimization of drying conditions with respect to product quality. It includes five steps:
Determination of critical quality attributes. The set of critical quality attributes is mostly determined by the objectives of drying. Some of them, such as moisture content, shrinkage, color, are measurable in real time, while others, such as porosity, density, bioactives are not directly measurable. In the latter case the development of soft sensors, based on computer vision, electronic nose and biosensors is necessary.
Development of product observer (machine vision) to measure critical quality attributes of the product under drying. This step includes the development of relationships between image features and food quality attributes, based on image processing and pattern recognition techniques. Non-linear relationships between sensors response and quality attributes are established by AI tools, such as artificial neural networks (ANN), fuzzy logic (FL), and evolutional algorithms (EA).
Development of process observer: to measure operational conditions, such as temperature, humidity, airflow, pressure in air convective drying. Machine learning makes the control intelligent in terms of added capacity to learn directly from the drying process, accumulate this knowledge, and adjust the control strategy with respect to observable changes of quality. In general, it requires the fusion of information from process observer (temperature, humidity, airflow, pressure) with information from product observer to determine instantaneous sensitivity of quality attributes to operational conditions.
Development of control strategies: Control strategies include dynamic multi-objective optimization targeting drying objectives. Due to the integration with a product observer, control strategies provide both local and global controllability of the drying process. Implementation of this system for ginseng drying significantly enhanced efficiency, sustainability, and product quality.
