Artificial Intelligence approaches for Heating Elements: algorithmic solutions in the the IoT scenario

Zoppas Industries is leveraging machine learning to enhance its product offerings, particularly in domestic coffee machines. This thesis presents a data-driven solution to monitor and manage limescale levels in the thermoblock, a critical component responsible for maintaining precise water temperature during coffee brewing. By predicting the thermoblocks health state in terms of limescale clogging effect, the system maintains coffee quality and optimizes maintenance schedules, reducing costs for Zoppas and customers. A prerequisite for this system is capsule recognition, which consists in correctly identify the type of capsule being brewed based on the water flow signal from the thermoblock. This classification task can be separated as either binary, distinguishing between original and compatible capsules, or multiclass, where the model predicts the specific capsule type between those encountered during the training phase. The solution integrates into an IoT architecture, leveraging a cloud-based platform for data storage, visualization, and the implementation of machine learning algorithms: Statwolf. This research not only enhances performance and customer satisfaction but also aligns with Industry 5.0 principles by advancing cutting-edge algorithmic solutions alongside to traditional methods. Lastly, it presents the experimental setup designed for data acquisition, exploring the entire machine learning pipeline: from hardware for dataset generation to model predictions.