Exploring Machine Learning-Driven Virtual Commissioning: A Proof of Concept for CTPack Machines in Food Packaging Lines

In this era of Industry 4.0, Virtual Commissioning has transformed the industrial landscape by enabling the simulation, testing and optimization of automation systems in a virtual environment prior to their physical implementation. It offers industries the ability to reduce commissioning time and costs, identify potential issues early, and increase operational efficiency. At the same time, Machine Learning (ML) has gained traction for its ability to analyze large datasets, identify patterns and make intelligent predictions. When combined with Virtual Commissioning, Machine Learning can introduce a new dimension to automation, where virtual environments can be used to optimize systems based on data-driven insights. This thesis explores the integration of Machine Learning (ML) within a Virtual Commissioning environment, focusing on a CT-Pack Delta Robot operating within a simulated production line. It investigates how the use of ML can enhance the automation of pick and place tasks by facilitating smarter decision-making and improving the adaptability of the system to diverse scenarios. The study discusses the implementation of a virtual environment in Unreal Engine 5.3, alongside the development of various simulation components. A Delta Robot with complete inverse kinematics is also implemented inside the simulation. Additionally, a dynamic and adaptable camera system is developed inside Unreal Engine to gather data for the Machine Learning model. A communication protocol is also discussed for exchanging data between the simulation and the Machine Learning model. Finally, the development of a Machine Learning algorithm is discussed for the classification of different products on a conveyor.