果冻传媒

The meat processing industry is undergoing a significant transformation, driven by changing consumer preferences, evolving societal norms, and advances in animal husbandry. Sustainability has become a central focus, with facilities adopting more humane practices, reducing waste, and adhering to stricter food safety regulations. Additionally, consumers are demanding greater transparency and traceability, wanting to know the origins of their meat, including the farms and animals involved. These shifts have also sparked an increase in demand for convenience and variety in meat products, from pre-marinated or pre-cooked cuts to sustainable packaging options. However, these advancements add complexity to meat processing systems, making their design more challenging. For his PhD research Nick Paape explored the use of design space exploration methods to simplify this process, allowing for the automatic calculation of numerous production line configurations to identify the optimal one, significantly accelerating system design.

Meat processing systems cover the entire process from live animals entering the facility to the end products sold in supermarkets. They are characterized by the cutting up of animals into various end products (known as co-production), the unpredictable variations in product weight and quality (known as random yield), and the combination of make-to-stock and make-to-order production strategies. Novel design approaches, especially model-based design methods, can be of great value when designing meat processing systems that can effectively deal with this complexity and uncertainty. These methods provide early validation of design choices and the prediction of key performance indicators. investigated the application of model-based design methods to meat processing systems.

Comprehensive framework

This research contains contributions on the design space exploration of production systems. Design space exploration is the process of analyzing and evaluating design alternatives to identify the design most suitable for the task at hand. Three contributions are made, each focusing on one of the main aspects of design space exploration: analysis, representation, and exploration. The first contribution focuses on the automatic analysis of a production system design using discrete-event simulation. A simulation model is automatically constructed using a model library and is then subjected to a series of simulation experiments in which the system鈥檚 performance in various production scenarios is predicted. The second contribution focuses on the representation of the design space using a formal design space specification language. This language can be used to specify component types, component instances, and constraints on the design. Its main advantage is that it enables the automated generation of feasible designs. The third contribution focuses on effective exploration of the design space. A novel genetic algorithm is proposed for the optimization of production system design, which greatly reduces the number of designs that need to be simulated to find the optimal design. The genetic algorithm can be extended with a neural network surrogate model, further reducing the dependency on computationally expensive simulation. These contributions collectively offer a comprehensive framework for automated design space exploration, applicable not only to meat processing but also to other production systems. Each contribution is validated using a case study in poultry processing.

Agent-based production control

One of the challenges of using design space exploration effectively is that it requires a control structure that is easily adaptable to the physical design of the system. Various intelligent and distributed control architectures that offer such adaptability have been developed throughout the years, such as in multi-agent and holonic manufacturing. However, due to the unique characteristics of meat processing systems, these existing control structures are not suitable. One of the contributions of this PhD research is a framework for agent-based production control tailored to meat processing systems. This framework describes which agent types are responsible for what tasks, how these agents communicate, and how the control layer can be used to route products in a system with co-production and random yield.

Capture the flow

Traditionally, discrete-event simulation has been used for analyzing production systems, as it is effective at capturing the flow of material in a system. However, intelligent and distributed control architectures are becoming more prevalent, for which the behavior is better captured through agent-based simulation. Paape reviewed which simulation software is capable of a hybrid discrete-event and agent-based simulation. This review follows the following steps: first, the scope and the evaluation criteria are identified. Based on the scope, a selection of simulation tools is collected. This collection is evaluated and classified based on the chosen criteria. The most promising tool is subjected to a case study to assess its capabilities.

Title of PhD thesis: . Supervisors: Associate Prof. Michel Reniers (果冻传媒), and Dr. Joost van Eekelen (Vanderlande).